Neurobehavioral Genetics - Methods and Applications 2nd Ed - B. Jones, P. Mormede (CRC, 2007) WW

Second Edition
Neurobehavioral
Genetics
Methods and Applications
1903_Prelims.fm Page ii Friday, July 21, 2006 7:34 PM
Second Edition
Neurobehavioral
Genetics
Methods and Applications
Edited by
Byron C. Jones, Ph.D.
Department of Biobehavioral Health
The Pennsylvania State University
University Park, Pennsylvania
Pierre Mormde, D.V.M., Ph.D.

Laboratoire de Neurogntique et Stress INRA
Universit Victor Segalen
Bordeaux, France
Boca Raton London New York
CRC is an imprint of the Taylor & Francis Group,

an informa business
CRC Press
Taylor & Francis Group
6000 Broken Sound Parkway NW, Suite 300
Boca Raton, FL 33487-2742
2007 by Taylor & Francis Group, LLC
CRC Press is an imprint of Taylor & Francis Group, an Informa business
No claim to original U.S. Government works

Printed in the United States of America on acid-free paper
10 9 8 7 6 5 4 3 2 1
International Standard Book Number-10: 0-8493-1903-X (Hardcover)

International Standard Book Number-13: 978-0-8493-1903-7 (Hardcover)
This book contains information obtained from authentic and highly regarded sources. Reprinted
material is quoted with permission, and sources are indicated. A wide variety of references are
listed. Reasonable efforts have been made to publish reliable data and information, but the author
and the publisher cannot assume responsibility for the validity of all materials or for the conse-
quences of their use.
No part of this book may be reprinted, reproduced, transmitted, or utilized in any form by any
electronic, mechanical, or other means, now known or hereafter invented, including photocopying,
microfilming, and recording, or in any information storage or retrieval system, without written
permission from the publishers.
For permission to photocopy or use material electronically from this work, please access www.
copyright.com (http://www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC)
222 Rosewood Drive, Danvers, MA 01923, 978-750-8400. CCC is a not-for-profit organization that
provides licenses and registration for a variety of users. For organizations that have been granted a
photocopy license by the CCC, a separate system of payment has been arranged.
Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and
are used only for identification and explanation without intent to infringe.
Visit the Taylor & Francis Web site at
http://www.taylorandfrancis.com
and the CRC Press Web site at
http://www.crcpress.com
1903_C000a.fm Page v Thursday, July 27, 2006 8:32 PM
Dedication
Between the publishing of our first and second editions, we lost two of our colleagues
in the Behavioral Neurogenetics Initiative, Donald J. Nash and Lorraine Flaherty.
Both Don and Lori had long, distinguished careers. Don was professor of biology at
Colorado State University and Lori was director of the Genomics Institute, chief of
the Laboratory of Mammalian Genomics, and director of the Histocompatibility and
Paternity Testing Laboratory, Wadsworth Center, New York State Department of
Health. Don was president of the Southwestern and Rocky Mountain Division of the
American Association for the Advancement of Science for many years. In 2005, Lori
received the first Lifetime Achievement Award from the International Behavioral and
Neurogenetics Society. Both Don and Lori were active participants in organizing and
teaching at our International Summer School on Behavioral Neurogenetics in Europe
and North America. In fact, Don was a founding member of the Behavioral Neuro-
genetics Initiative (BNI). Both had great and infectious joie de vivre. We even miss
Dons outrageous puns. Above all, was their passion for neurobehavioral genetics
and bringing this emerging discipline to students and researchers worldwide. The
efforts of the BNI have been greatly enriched by the contributions of Lori and Don.
Donald J. Nash Lorraine Flaherty

December 20, 1930 January 8, 1946
January 13, 2002 February 22, 2006
1903_C000a.fm Page vi Thursday, July 27, 2006 8:32 PM
1903_C000a.fm Page vii Thursday, July 27, 2006 8:32 PM
In Memoriam
Gilbert Gottlieb
19292006
Dr. Gilbert Gottlieb was born in Brooklyn, NY in 1929. He earned his Bachelors
(1955) and Masters (1956) degrees from the University of Miami, and his Ph.D. in
clinical psychology at Duke University in 1960. This was no ordinary clinical
psychology degree because his dissertation was entitled, The Following Response
of Wild and Domestic Ducklings of the Same Species (Anas platyrhynchos). Thus
began an amazing career that spanned several interrelated disciplines (e.g., psychology,
zoology, embryology, genetics), the common thread being developmental science
and a unique theoretical perspective that he referred to as simply, the developmental
point of view. Instead of reducing development to a simple interaction of the
organism and its environment, Dr. Gottliebs view embraced the complexity of organ-
ismic development by considering the intricate transactional nature of events that
occur among all levels of organization (genetic, neural, behavioral, environmental)
at every point in time, such that the organism is a new organism throughout its
transactional developmental trajectory.
This view of development emerged from his early career as a research scientist
at the North Carolina Division of Mental Health (Raleigh, NC), where, in 1961, he
established the Psychology Laboratory at Dorothea Dix Hospital and pursued both
field and laboratory investigations on the acoustic basis of species identification in
ducklings. Obtaining baseline data in naturalistic contexts on the kinds of stimuli
that organisms normally encounter, as well as the behaviors in which they typically
engage, provided a foundation for his elegant and innovative experimental analyses
of developmental mechanisms. These studies, which continued for over three decades,
revealed the importance of embryonic (as well as early postnatal) experience on the
development of species-typical behavior. The empirical and conceptual foundations
of this body of work culminated in Dr. Gottliebs volume, Synthesizing Nature-Nurture
(1997, Erlbaum), for which he won the Eleanor Maccoby Book Award of the Devel-
opmental Psychology Division of the American Psychological Association.
In 1982, Dr. Gottlieb became the head of the psychology department at the
University of North Carolina at Greensboro, where he also held an Excellence
Foundation Professorship until his retirement from academia in 1995. That retire-
ment, however, was not a retirement from developmental science, because in that
year, he became a research professor and member of the Executive Committee of
the Center for Developmental Science at the University of North Carolina at Chapel
Hill. In this stimulating multidisciplinary environment, he was able to further develop
his theoretical perspectives on the transactional nature of genetic and experiential
influences on behavioral development.
Throughout his distinguished career, Dr. Gottlieb received recognition and honors
from his peers. He was the guest of the Czechoslovak Academy of Sciences in Prague
1903_C000a.fm Page viii Thursday, July 27, 2006 8:32 PM
(1967) and the USSR Academy of Sciences in Moscow (1989), as well as a Visiting
Fellow at The Neurosciences Institute in San Diego (1996). He was also elected
Fellow of the Animal Behavior Society and the American Association for the Advance-
ment of Science; and in 1997, the Society for Research in Child Development honored
him with the Distinguished Scientific Contributions to Child Development Award. He
was supported from 1962 to 2006 by grants from the National Institute of Child
Health and Human Development, the National Science Foundation, and the National
Institute of Mental Health, all of which enabled him to pursue research questions
and theoretical approaches that he recently described to me as off-the-beaten-track.
But he maintained that it is important for scientists to follow their noses when so
inspired because at the end of the trail, discoveries will be made.
Dr. Gilbert Gottlieb died on July 13, 2006. Although he did not live to see the
publication of this volume, in a taped interview that I conducted with him at his
home in Raleigh 1 month before his death, he proclaimed that the chapter contained
herein is his most important theoretical contribution to science.
David B. Miller
Department of Psychology
University of Connecticut
Gilbert Gottlieb
19292006
Photo Credit: Dr. Marc S. Gottlieb
1903_C000.fm Page ix Wednesday, July 26, 2006 7:58 PM
Preface
This second edition of Neurobehavioral Genetics: Methods and Applications is
issued about 7 years after the first edition. Since its appearance in 1999, the book
has been used as text material for the International Behavioral Neurogenetics Sum-
mer School (formerly the FrenchAmerican Summer School). In September 2005,
we held our 12th school in Moscow. This is particularly remarkable because the
initial funding, provided by the Scientific Mission of the French Embassy to the
U.S. in 1995, was the spark that launched this effort. We will always be thankful to
Dr. Jean-Marie Guastavino, attach for Science and Technology at that time, for
arranging for the generous gift from the French government. An alumna of our third
school, Dr. Yamima Osher, is author of the chapter on genegene interactions.
When we issued the first edition, we knew that the field was developing rapidly,
and, in fact, the advances in methods and our knowledge have outstripped even our
most optimistic expectations. Methods to examine gene expression, new approaches
in bioinformatics, and the rapid growth in knowledge that has followed are remark-
able. Major developments in the field of behavioral neurogenetics are the foundation
of the Complex Traits Consortium (CTC) and the equally important International
Behavioural and Neural Genetics Society (IBANGS). These new societies have
provided much needed forums for examination of brain and behavior from a systems
genetics perspective. There have been remarkable advances also in single-gene
techniques. This second edition has been produced to highlight some of these
important advances. We are particularly pleased to include a contribution by Dr.
Gilbert Gottlieb, who has been an important and tough critic of behavioral genetics.
Fortunately, he has always been kind to behavior geneticists!
Special thanks to our managing editor, Barbara Norwitz, at Taylor & Francis,
for her encouragement and guidance.
B.C.J & P.M.

University Park, PA
1903_C000.fm Page x Wednesday, July 26, 2006 7:58 PM
1903_C000.fm Page xi Wednesday, July 26, 2006 7:58 PM
Editors
Byron C. Jones, Ph.D., is a professor of biobehavioral health and pharmacology at

The Pennsylvania State University. Dr. Jones received a B.A. in psychology in 1969
from Eastern Washington University, Cheney, an M.A. in psychology from the
University of Arizona, Tucson, in 1972, and a Ph.D. in physiological psychology
from the University of Arizona, Tucson, in 1975. He subsequently received post-
doctoral training in neuropharmacology at the University of Arizona and in phar-
macogenetics at the University of Colorado, Boulder.
Dr. Jones is a member of the Society for Neuroscience, the International Behav-
ioral Neuroscience Society, the Socit des Neurosciences, International Brain
Research Organization and International Behaviour and Neural Genetics Society.
He is a member of the editorial boards for Pharmacology Biochemistry and Behavior
and Nutritional Neuroscience.
Dr. Jones is the author of more than 90 refereed papers, reviews, and chapters.
His principal interests are in the pharmacogenetics of drugs that affect the central
and peripheral nervous systems.
1903_C000.fm Page xii Wednesday, July 26, 2006 7:58 PM
Pierre Mormde, D.V.M., Ph.D., is director of research at the French National

Institute for Agricultural Research (INRA) and the head of the Laboratoire de
Neurogntique et Stress sponsored by INRA and the Universit Victor Segalen,
Bordeaux, France. Dr. Mormde received a D.V.M. degree from the Veterinary
School in Maisons-Alfort, France, and a Ph.D. degree in biology of behaviour from
the University Paul Sabatier in Toulouse, France. He subsequently pursued post-
doctoral training in molecular neuroendocrinology at the Salk Institute, San Diego,
California.
Dr. Mormde is a member of the International Behavioural and Neural Genetics
Society, the Socit des Neurosciences, and the Socit de Neuroendocrinologic
Experimentale.
Dr. Mormde is the author of more than 150 refereed papers, reviews, and
chapters, and a member of the editorial board of Psychoneuroendocrinology and
Genes, Brain and Behavior. His principal interests are in the psychobiology of stress
and drug abuse, and molecular genetics of individual variations in behavioral and
neuroendocrine reactivity to environmental constraints.
1903_C000.fm Page xiii Wednesday, July 26, 2006 7:58 PM
Contributors
Irmgard Amrein Valerie Bolivar
Institute of Anatomy Wadsworth Center
University of Zrich-Irchel Troy, New York
Zrich, Switzerland
Guntram Borck
Rachel Bachner-Melman Institut National de la Sant et de la
Sarah Herzog Memorial Hospital, Givat Recherche Mdicale (INSERM) U781
Shaul Hpital Necker-Enfants Malades
Jerusalem, Israel Paris, France
Andrew Canastar
Amsale T. Belay
Department of Psychiatry
Department of Biology
University of Colorado Health Sciences
University of Toronto
Center
Mississauga, Canada
Denver, Colorado
John Belknap Elissa J. Chesler

Portland Alcohol Research Center Mammalian Genetics and Genomics
Department of Behavioral Group
Neuroscience Life Sciences Division
Oregon Health & Science University Oak Ridge National Laboratory
Portland Veterans Affairs Medical Oak Ridge, Tennessee
Center
Portland, Oregon Laurence Colleaux
INSERM
Robert H. Belmaker Hpital Necker-Enfants Malades
Division of Psychiatry Paris, France
Ben Gurion University of the Negev
Beer Sheva, Israel Daniel Comas
Centre National de la Recherche
Jonathan Benjamin Scientifique (CNRS)
Division of Psychiatry Gif-sur-Yvette, France
Ben Gurion University of the Negev
Beer Sheva, Israel John C. Crabbe
Portland Alcohol Research Center
Wade Berrettini Department of Behavioral Neuroscience
Department of Psychiatry Oregon Health & Science University
University of Pennsylvania Portland Veterans Affairs Medical Center
Philadelphia, Pennsylvania Portland, Oregon
1903_C000.fm Page xiv Wednesday, July 26, 2006 7:58 PM
Wim E. Crusio Guillaume Isabel

Laboratoire de Neurosciences Gnes et Dynamique des Systmes de
Cognitives CNRS Mmoire, CNRS
Universit Bordeaux I Ecole Suprieure de Physique et de
Talence, France Chimie Industrielles
Paris, France
Birgit Dreier
Department of Biochemistry Jean-Marc Jallon
University of Zrich Laboratoire de Neurobiologie de 1
Zrich, Switzerland Apprentissage, Mmoire et
Communication (NAMC)
Richard P. Ebstein Universit Paris-Sud XI
Scheinfeld Center for Human Orsay, France
Behavioral Genetics
Hebrew University of Jerusalem Gert Jansen
Jerusalem, Israel Erasmus Medical Center
Rotterdam, the Netherlands
Lorraine Flaherty
Wadsworth Center Byron C. Jones
Troy, New York Department of Biobehavioral Health
The Pennsylvania State University
Philip Gorwood University Park, Pennsylvania
INSERM
Paris, France Leslie C. Jones
Interdepartment Graduate Program in
Irving I. Gottesman Neuroscience
Department of Psychiatry The Pennsylvania State University
University of Minnesota University Park, Pennsylvania
Minneapolis, Minnesota
James E. King
Department of Psychology
Gilbert Gottlieb
University of Arizona
Center for Developmental Science
Tuscon, Arizona
University of North Carolina
Chapel Hill, North Carolina
Jean-Michel Lassalle
Centre de Recherche sur la
Robert Hitzemann Cognition Animale
Portland Alcohol Research Center CNRS 5169
Department of Behavioral Universit Paul Sabatier
Neuroscience Toulouse, France
Oregon Health & Science
University Hans-Peter Lipp
Portland Veterans Affairs Medical Institute of Anatomy
Center University of Zrich-Irchel
Portland, Oregon Zrich, Switzerland
1903_C000.fm Page xv Wednesday, July 26, 2006 7:58 PM
Stephen C. Maxson Michael F. Pogue-Geile

Department of Psychology Department of Psychology
University of Connecticut University of Pittsburgh
Storrs, Connecticut Pittsburgh, Pennsylvania
Shannon McWeeney Thomas Prat

Portland Alcohol Research Center Gnes et Dynamique des Systmes de
Department of Behavioral Mmoire, CNRS
Neuroscience Ecole Suprieure de Physique et de
Oregon Health & Science University Chimie Industrielles
Portland, Oregon Paris, France
Marie-Pierre Moisan Richard A. Radcliffe

Laboratoire de Neurogntique et Stress Department of Pharmaceutical
INRA Sciences
Universit Victor Segalen University of Colorado
Bordeaux, France Denver, Colorado
Florence Molinari
Andr Ramos
INSERM
Departmento de Biologia Celular,
Hpital Necker-Enfants Malades
Embriologia e Gentica
Paris, France
Universidade Federal de Santa
Catarina
Pierre Mormde
Florianpolis, Brazil
Laboratoire de Neurogntique
et Stress INRA
Universit Victor Segalen Nicolas Ramoz
Bordeaux, France INSERM
Paris, France
Tracy L. Nelson
Department of Health and Exercise Laurent Sgalat
Science Centre de Gntique Molculaire et
Colorado State University Cellulaire CNRS
Longmont, Colorado Universit Lyon 1 Claude Bernard
Villeurbanne, France
Yamima Osher
Beer Sheva Mental Health Center Lutz Slomianka
Ben Gurion University of the Negev Institute of Anatomy
Beer Sheva, Israel University of Zrich-Irchel
Zrich, Switzerland
Jeremy L. Peirce
Department of Anatomy and Marla B. Sokolowski
Neurobiology Department of Biology
University of Tennessee University of Toronto
Memphis, Tennessee Mississauga, Canada
1903_C000.fm Page xvi Wednesday, July 26, 2006 7:58 PM
Peter Sonderegger Keith E. Whitfield

Department of Biochemistry Department of Biobehavioral Health
University of Zrich The Pennsylvania State University
Zrich, Switzerland University Park, Pennsylvania
Douglas Wahlsten Robert W. Williams

Department of Biological Sciences Department of Anatomy and
Great Lakes Institute for Environmental Neurobiology
Research University of Tennessee
University of Windsor Memphis, Tennessee
Windsor, Canada
David P. Wolfer
Alexander Weiss Institute of Anatomy
Department of Psychology University of Zrich-Irchel
University of Edinburgh Swiss Federal Institute of Technology
Edinburgh, Scotland Zrich, Switzerland
1903_bookTOC.fm Page xvii Wednesday, July 26, 2006 8:05 PM
Table of Contents
Chapter 1
A History of Behavior Genetics................................................................................ 1
Stephen C. Maxson
Chapter 2
Developmental Neurobehavioral Genetics: Development as Explanation ............. 17
Gilbert Gottlieb
Chapter 3
Some Basics, Mendelian Traits, Polygenic Traits, Complex Traits....................... 29
Byron C. Jones
Chapter 4
An Introduction to Quantitative Genetics ............................................................... 37
Wim E. Crusio
Chapter 5
From QTL Detection to Gene Identification .......................................................... 55
Marie-Pierre Moisan
Chapter 6
Gene Expression ...................................................................................................... 73
Richard A. Radcliffe
Chapter 7
Bioinformatics of Behavior ..................................................................................... 95
Elissa J. Chesler
Chapter 8
Congenic and Consomic Strains ........................................................................... 115
Lorraine Flaherty and Valerie Bolivar
Chapter 9
Animal Resources in Behavioral Neurogenetics .................................................. 129
Chapter 10
Sample Size Requirements for Experiments on Laboratory Animals ................. 149
Douglas Wahlsten
1903_bookTOC.fm Page xviii Wednesday, July 26, 2006 8:05 PM
Chapter 11
The Role of Association Studies in Psychiatric Disorders................................... 169
Nicolas Ramoz and Philip Gorwood
Chapter 12
Family and Twin Methods..................................................................................... 183
Keith E. Whitfield and Tracy L. Nelson
Chapter 13
GeneEnvironment Interactions ............................................................................ 189
Byron C. Jones and Leslie C. Jones
Chapter 14
And Now it Starts to Get Interesting: GeneGene Interactions........................... 201
Yamima Osher
Chapter 15
Schizophrenia: Study of a Genetically Complex Phenotype ................................ 209
Michael F. Pogue-Geile and Irving I. Gottesman
Chapter 16
Genetics of Major Affective Disorders ................................................................. 227
Wade Berrettini
Chapter 17
Pedigree Analyses and the Study of Chimpanzee
(Pan troglodytes) Personality and Subjective Well-Being.................................... 247
Alexander Weiss and James E. King
Chapter 18
The Elusive World of Personality Genes: Cherchez le Phenotype ...................... 263
Richard P. Ebstein, Rachel Bachner-Melman,
Jonathan Benjamin, and Robert H. Belmaker
Chapter 19
Aggression: Concepts and Methods Relevant to
Genetic Analyses in Mice and Humans................................................................ 281
Stephen C. Maxson and Andrew Canastar
Chapter 20
Genetic Analysis of Emotional Behaviors
Using Animal Models ........................................................................................... 291
Andr Ramos and Pierre Mormde
Chapter 21
Genetic Analysis of Food Search Behavior in the Fruit Fly (Drosophila
melanogaster) ........................................................................................................ 307
Amsale T. Belay and Marla B. Sokolowski
1903_bookTOC.fm Page xix Wednesday, July 26, 2006 8:05 PM
Chapter 22
Genetic and Molecular Analyses of Drosophila
Courtship Behavior................................................................................................ 319
Jean-Marc Jallon
Chapter 23
A New Era for Drosophila
Learning and Memory Studies .............................................................................. 333
Daniel Comas, Guillaume Isabel, and Thomas Prat
Chapter 24
Behavioral Genetics in the Nematode
Caenorhabditis elegans ......................................................................................... 353
Gert Jansen and Laurent Sgalat
Chapter 25
Genetics, Behavior, and Brain Dopamine Systems .............................................. 371
Robert Hitzemann, Shannon McWeeney, and John Belknap
Chapter 26
Natural Genetic Variation of Hippocampal
Structures and Behavioran Update.................................................................... 389
Hans-Peter Lipp, Irmgard Amrein, Lutz Slomianka, and David P. Wolfer
Chapter 27
Expression and Brain Structure:
Black Boxes between Genes and Behaviors......................................................... 411
Jeremy L. Peirce and Robert W. Williams
Chapter 28
Synaptic Mechanisms Involved in Cognitive Function: Cues from Mental
Retardation Genes.................................................................................................. 435
Guntram Borck, Florence Molinari, Birgit Dreier,
Peter Sonderegger, and Laurence Colleaux
Chapter 29
Pharmacogenetics .................................................................................................. 449
Byron C. Jones
Chapter 30
Alcohol Psychopharmacogenetics ......................................................................... 457
John C. Crabbe
Index...................................................................................................................... 469
1903_bookTOC.fm Page xx Wednesday, July 26, 2006 8:05 PM
1903_C001.fm Page 1 Wednesday, July 26, 2006 6:08 PM
1 A History of Behavior
Genetics
Stephen C. Maxson
CONTENTS
Summary .................................................................................................................... 2
1.1 Introduction ...................................................................................................... 2
1.2 Preliterate and Ancient History ....................................................................... 3
1.3 Two Victorian Cousins..................................................................................... 3
1.4 The First Century of Behavior Genetics ......................................................... 4
1.4.1 1910 to 1952 ........................................................................................ 4
1.4.1.1 Fruit Flies.............................................................................. 4
1.4.1.2 Rodents ................................................................................. 4
1.4.1.3 Humans ................................................................................. 5
1.4.2 1953 to 1970 ........................................................................................ 5
1.4.2.1 Fruit Flies.............................................................................. 6
1.4.2.2 Rodents ................................................................................. 7
1.4.2.3 Dogs ...................................................................................... 7
1.4.2.4 Humans ................................................................................. 7
1.4.3 1971 to 1985 ........................................................................................ 8
1.4.3.1 Nematodes............................................................................. 8
1.4.3.2 Fruit Flies.............................................................................. 8
1.4.3.3 Rodents ................................................................................. 8
1.4.3.4 Canids ................................................................................. 10
1.4.3.5 Humans ............................................................................... 10
1.4.4 1986 to 2005 ...................................................................................... 10
1.4.4.1 Nematodes........................................................................... 10
1.4.4.2 Fruit Flies............................................................................ 11
1.4.4.3 Bees..................................................................................... 11
1.4.4.4 Rodents ............................................................................... 11
1.4.4.5 Humans ............................................................................... 12
1.5 Janus: The Past and Future of Behavior Genetics ........................................ 13
1.6 A Highly Personal Note................................................................................. 14
References................................................................................................................ 14
1
2 Neurobehavioral Genetics
SUMMARY
This is a broad brush and highly selective history of behavior genetics. I begin it
long ago in prehistoric and ancient times to indicate that even then there was
knowledge of and speculation about the inheritance of behavior in animals and
humans. I then turn to the contributions of two Victorian cousins, Charles Darwin
and Francis Galton. They set the stage for the first century of behavior genetics. I
divide the history of first century of behavior genetics into four phases: 1910 to
1952, 1953 to 1970, 1971 to 1985, 1986 to 2005. Within each, I consider the history
of behavior genetics by key organisms: nematodes, fruit flies, rodents, other critters,
and humans. Throughout, individual contributors to behavior genetics are mentioned.
I conclude with some reflections on past conceptual streams of behavior genetics,
and on how these streams can and should come together in a comparative genetics
of behavior.
1.1 INTRODUCTION
The history of behavior genetics begins long before Darwins work on evolution or
Mendels work on inheritance. It ends in the very recent past. I hope that this history
will be of value in: (1) recognizing who the many contributors to the field were and
are, (2) tracing the origins and developments of conceptual themes in behavior
genetics, and (3) relating how new methods have contributed again and again to
raising questions and enabling answers to them in behavior genetics.
1.2 PRELITERATE AND ANCIENT HISTORY

Humans may have been interested in animals and their own behaviors for as long
as there have been Homo sapiens, about 150,000 years.26 For animals, evidence for
this is clear by 35,000 years ago. The archeological record shows that humans were
hunters and fishers. Both require at least practical knowledge of animal behavior.
Furthermore, in France and Spain, there are cave paintings realistically depicting
ibex, bison, horses, and other animals; these date from 35,000 years ago. Also, there
is ample evidence from anthropology that people in preliterate cultures were and
are extremely interested in the behaviors of other humanswatching, recording,
and analyzing them. We can only speculate as to whether family resemblances in
behavior were noted and theories developed to explain them.
From the beginning of the Neolithic Age (about 12,000 years ago), humans have
been domesticating animals. The first domesticated animal may have been the dog.
For each domesticated animal, there was subsequently selective breeding for behav-
ioral and other traits. In order to initially domesticate and subsequently selectively
breed animals, humans had to recognize that there is variability within a species,
that relatives resemble one another, and that breeding like to like produces like. In
other words, humans would have to have had some idea about the inheritance of
behavior and other traits.
At some point, some humans came to believe that there was inheritance of
their own behavioral and other traits. In the sixth century B.C.E., Theognis
A History of Behavior Genetics 3
suggested that well-bred parents have well-bred offspring and that poorly bred
parents have poorly bred offspring. Similarly, Plato (427346 B.C.E.) argued in
The Republic that matings of the best man to the best woman would produce
the best children and that matings of the worst man to the worst woman would
produce the worst children. For this reason, both philosophers proposed that
humans, like animals, should be selectively bredwith only the well bred or
the best having children. This was eugenics long before there was a science of
genetics.
There was also an interest in the ancient world in whether the development
of behavior was innate or acquired. For example, Aristotle (384322 B.C.E.) in
Greece argued that the brood parasitism of the cuckoo did not depend on post-
hatching experience. Similarly, Galen (129200 C.E.), a Roman physician, pro-
posed, based on an experiment, that the preference of newborn goats for mothers
milk did not depend on postnatal experience. Some of the ancient arguments for
eugenics may have depended on the belief that if a behavior is inherited, its
development is independent of experience. This mistaken idea, which confounds
inheritance and innateness, has been a persistent problem in behavior genetics,
even until today.
We now leap over almost 2000 years of history to the beginning of a science of
behavior genetics.
1.3 TWO VICTORIAN COUSINS

Charles Darwin published The Origin of Species in 1859. Darwin argued that the
evolution of both adaptations and species was due to past effects of natural selection
on heritable variations. He considered complex instincts in chapter 8 of The Origin
of Species, behavioral sex differences due to sexual selection in chapters 8 to 20 of
The Descent of Man (1871), animal and human emotions in The Expression
of Emotions in Man and Animals (1873), and animal mind in chapters 3 and 4 of
The Descent of Man (1871). In The Origin of Species, he also described numerous
examples of selective breeding for animal behavior. Selective breeding for animal
behaviors would be an important method in the field of behavior genetics.
The writings of Darwin on behavioral evolution gave rise to a stream of behavior
genetics that is largely concerned with genetics of adaptive behaviors in animals
and humans.
The other major stream of behavior genetics is concerned with the causes of
variation in human behaviorsespecially cognition, personality, psychopathology,
and addictions. It is also concerned with the genetics of animal models for these
human behaviors. This stream derives largely from the works of Francis Galton,
who was Darwins cousin. Galton initiated three methods for studying behavioral
inheritance in humans. These are the family, twin, and adoption methods. The
family method and adoption method are discussed in Galtons Hereditary Genius
(1869). The twin method is discussed in Inquiry into Human Faculties and Its
Development (1883). Galton also initiated many methods of statistical analysis.
Regrettably, he also supported eugenics. Perhaps, he also confounded inheritance
and innateness.
1.4 THE FIRST CENTURY OF BEHAVIOR GENETICS

I am going to divide the first century of behavior genetics into four periods: 1910
to 1952, 1953 to 1970, 1971 to 1986, and 1987 to 2005. Within each I will consider
the history of behavior genetics by key organisms: the nematode, C. elegans from
1973, fruit flies, rodents, other animals, and humans.
1.4.1 1910 TO 195212,17

In 1900, the Mendelian rules of inheritance for the alleles of genes were rediscovered
by Correns, DeVries, and von Teschermak, and in 1910, T. H. Morgan proposed the
chromosome theory of inheritance. In this theory, each gene is located on one of
the chromosomes of its species, and the behavior of the alleles of a gene in gamete
formation are due to what happens to chromosomes in gamete formation. Next, it
was shown by R. A. Fisher, J. B. S. Haldane, and S. Wright between 1918 and 1921
that the rules of Mendelian genetics would apply to variation and inheritance of
quantitative traits like those of concern to Darwin and Galton. The conceptual and
methodological stage was now set for this first period in behavior genetics.
1.4.1.1 Fruit Flies12
There was a great interest in looking for single-gene effects on fly behavior. This
includes the studies of George Wald and Werner Reichard on visual acuity, of Paul
Scott on phototropism, of R. S. McEwen on geotaxis and phototaxis, and of
A. Sturtevant on mate preference and choice. The objects of study, in many cases,
were genes with visible morphological effects on flies, including effects on eye color
and shape, body color, and bristle number. There were also studies by H. T. Spieth
and Dobzhansky on genetics, mating success, sexual isolation, and evolution. These
were both within- and between-species studies.
1.4.1.2 Rodents17
In the 1930s and 1940s, R. C. Tryon reported on the selective breeding of two strains
of rats that differed in maze learning as indexed by trial errors. The two strains or
lines were the maze brights, with few errors, and the maze dulls, with many
errors. Also, C. S. Hall reported in 1938 on the selective breeding of two lines of
rats that differed in emotionality as measured by open-field activity and defecation.
Subsequently, selective breeding would be widely used to study the genetics of rat
and mouse behaviors.
In 1942, two papers were published showing differences in aggressive behavior
among three inbred strains of mice. One was published by John Paul Scott, and the
other by Benson Ginsburg and W. C. Allele. In the 1940s, another focus of strain
difference in mice was sound-induced, or audiogenic, seizures. Subsequently, there
have been a great many studies of strain differences in mouse behaviors. Such strain
differences are considered to be evidence for genetic effects on phenotypic variation.
Not just mice but rabbits and dogs too were among the organisms studied at Hamilton
Station for Behavioral Research, established in 1946 at the Jackson Laboratory. Its
full-time or summer staff included John Paul Scott, John Fuller, Benson Ginsburg,
Betty Beaman, Calvin Hall, and many others.
For the maze dull and maze bright rats, there were also genotype-by-environment
interactions observed for the maze errors. In enriched environments, the error scores
of the maze dulls but not the maze brights were decreased, whereas in impoverished
environments, the error scores of the maze brights but not the maze dulls were
increased.
A genotype-by-environment interaction was also observed for aggression in
mice. Scotts most aggressive strain was Ginsburg and Alleles most pacific strain
C57BL/10. When mice of the C57BL/10 strain were transferred from cage to cage
by forceps, as Scott did, they were aggressive; and when mice of the C57BL/10
were transferred from cage to cage in a small box, as Ginsburg did, they were pacific.
These treatments had no effect on the aggression of the other two strains. Such
genotype-by-environment interactions would be found again and again for rodent
behaviors and would be an active subject of research up to the present day (see
Chapter 20 in this volume).
In 1951, Calvin S. Hall13 reviewed the literature on the genetics of rat and mouse
behaviors and proposed there were three goals for behavior genetics research: (1)
to determine whether individual differences in behavior were heritable and to what
extent, (2) to determine for each heritable behavior the number of variant genes
involved and the chromosomal location of each, and (3) to determine how each of
these genes affects the heritable behaviors. These are still among the goals of
behavior genetics in animals and humans.
1.4.1.3 Humans12
Family and twin studies were conducted for IQ (H. D. Carter, H. H. Neuman, F. N.
Freeman, and K. J. Holzinger), personality (M. N. Crook, H. D. Carter, and
L. Portenier), schizophrenia (F. J. Kallman), and depression (E. Slater). There were
also some adoption studies. The emphasis was on ascertaining whether individual
differences were due to genes, environment, or both. Although many studies were
consistent with effects of genes on trait variation, definite conclusions were limited
by sample size.
Pedigree analysis for single genes with large effects was also developed at this
time. Using pedigrees, phenylketonuria was shown by L. Penrose in 1933 to be a
single gene recessive trait. It was believed that this was an inborn error of metabolism
similar to alkaptonuria as described by A. E. Garrod in 1909.
1.4.2 1953 TO 197023,25,27

In 1952, James Watson and Francis Crick described the structure of deoxyribose
nucleic acid (DNA) and the implications of that structure for the replication, function,
and mutation of genes. Over the next 20 years, the basics of molecular genetics were
established, which included an understanding of the two primary functions of DNA.
These are its role in determining the amino acid sequence of proteins and in regu-
lating when and where a protein is made as well as how much of it. The stage was
now set for Halls third goal for behavior genetics, namely, identifying and charac-
terizing the effect of DNA and DNA variants on behavior.
In 1958, Benson Ginsburg published Genes as a Tool in the Study of Behavior.11
The first part of this paper was a reflection on the implications of molecular genetics
for the study of behavior, and the second part argued that genetics should be used
as a tool to analyze animal and human behavior. It could be used as a tool in an
evolutionary context to determine natural units of behavior, to study the development
of behavior and the respective roles of genes and environment, and to analyze the
biological mechanisms of behaviors. This approach would become one of the main
tributaries of modern behavior genetics.
The other main tributary of modern behavior genetics focused on the contribution
of genetic and environmental factors in individual differences in animal and human
behaviors and on identifying and characterizing each of the genes with effects on
these individual differences. Also, a prime interest of the research with animals was
to develop genetic models relevant to the study of individual differences in human
behaviors. In 1960, the literature on this tributary of behavior genetic research on
human and animals was reviewed and summarized in the text by J. L. Fuller and R.
Thompson. The text was simply titled, Behavior Genetics.12 Developments in quan-
titative genetics as described in D. S. Falconers Introduction to Quantitative
Genetics11 and in K. Mathers Biometrical Genetics24 were a basis for further
advances in this area of behavior genetics.
In 1961 and 1962, there were important meetings on behavior genetics at the
Center for Advanced Studies in the Behavioral Sciences at Stanford CA. These were
attended by many leaders in the field and in many ways set the agendas for the next
decade of behavior genetics research. Some of the presentations at these meetings
were published in 1967.20
1.4.2.1 Fruit Flies
During this period, there were several selective breeding projects for behaviors in
fruit flies. Jerry Hirsch and Niki Erlenmeyer-Kimling initiated bi-directional selec-
tion for geotaxis. Selective breeding of these lines has continued to the present day.
Now, crosses of the two lines cannot have fertile progeny. Hirsch and Erlenmeyer-
Kimling also tested for the contributions of individual chromosome regions to the
high and low scores. Similarly, Dobzhansky selectively bred for geotaxis and pho-
totaxis in another species of fruit fly. He reported that very little of the individual
differences in these traits was due to genetic variation. About the same time, Aubrey
Manning carried out a bi-directional study of mating speed.
During this period, Lee Ehrman initiated her studies of rare male mating advan-
tage in several species of fruit flies. Also, Speiss, Ehrman and Dobzhansky among
others continued the research on mating preference and on sexual isolation within
and between species.
About 1967 Seymour Benzer first proposed that single gene mutants with large
effects could be used to dissect the neural and other mechanisms of behaviors in
flies. Later Sydney Brenner made a similar suggestion for the nematode, C. elegans.
These contributions are described more fully in the next period (1970 to 1986) of
behavior genetics. In some ways, this was to be a large substream of the tributary
that would use genetics as a tool to analyze behavior as originally proposed by
Ginsburg in 1958.
1.4.2.2 Rodents
During this period selective breeding was used to establish new phenotypic lines of
rats and mice. These included the Maudsley Emotional and Non-emotional strains
(Peter Broadhurst) and the Roman High and Low Avoidance strains (C. Bignami)
in rats. Also, mice were selectively bred for high and low activity in an open field
(John Defries) and aggression (Geert van Oortmerssen and K. Lagerspetz).
About this time, Jan Bruel suggested in several articles that adaptive traits of
mice would have low additive genetic variation and high directional dominance
variation. This set the stage for research with diallel F1 crosses and F2 hybrids of
inbred strains to determine the additive and dominance variance of many mouse
behaviors. This program of research will be further discussed in the next period
(19711985) of behavior genetics.
Now strain differences were also being used to investigate the neural basis of
behaviors. For example, M. R. Rosenzweig, D. Krech, and E. L. Bennett had shown
maze dull and maze bright rats differed in brain cholinesterase levels. Subsequently,
Thomas Roderick selectively bred for cholinesterase levels in the brains of rats and
showed that these strains differed in learning performance.
Also, at this time, research was begun on strain differences in consumption and
effects of alcohol (Gerald McClearn). Research on genetics, alcohol, other drugs,
and behavior would become a major area of mouse behavior genetics.
1.4.2.3 Dogs
In 1965, John Paul Scott and John L. Fuller published the book, The Genetics and
Social Behavior of the Dog.34 This book detailed the findings of nearly 20 years of
research at the Jackson Laboratory on the roles of genes and environment in the
behavior of five breeds of dogs and their derived crosses. This work remains a
critically acclaimed classic on dog behavior and on methods for genetic analysis of
social behaviors.
1.4.2.4 Humans
During this period, there were more extensive adoption studies of IQ (M. Skodak
and H. M. Skeels, M. Honzig, and J. Sheilds) and schizophrenia (L. Heston,
D. Rosenthal, S. S. Kety). About this time, the large Louisville Twin Study was used
to assess the contribution of genes and environment to variation in many traits (S.
Vandenberg).
Also, during this time, many single genes with large effects on mental retardation
were identified, and the behavioral effects of many chromosomal anomalies were
also identified. This included the chromosomal basis for Downs and Turners syn-
dromes. Development of staining techniques enabled the identification of each
human chromosome.
1.4.3 1971 TO 19854,9,14

In 1967, the Institute for Behavioral Genetics was founded at the University of
Colorado. Here many of todays behavioral geneticists received doctoral or postdoc-
toral training. Subsequently, many more centers of behavior genetics would be
established in the U.S. and internationally. In 1971, the Behavior Genetics Associ-
ation was founded and its journal, Behavior Genetics, was first published. Also, new
textbooks in behavior genetics were published. These included Gerald McClearn
and John Defriess Introduction to Behavioral Genetics (1973) and Lee Ehrman and
Peter Parsons, The Genetics of Behavior (1976).10 Also a second edition of Fuller
and Thompson (now titled, Foundation of Behavior Genetics) was published in
1978.13 In 1974, J. H. F. van Abeelen edited The Genetics of Behavior;36 this had
important conceptual, methodological, and substantive reviews of behavior genetics
across animals and humans.
1.4.3.1 Nematodes3
The research on nematodes began in 1973. The nematode has several advantages
for neurobehavioral genetics. These are a small genome (100 Mb), small nervous
system of 302 identified and mapped neurons, fast reproductive cycle (three days),
and self-fertilization. The major approach has been to expose nematodes to chemical
mutagens, to select for single gene mutants with behavioral effects and then to
characterize the DNA of the mutant gene. Initially mutants were detected that showed
sensory and motor effects.
1.4.3.2 Fruit Flies18,39
In 1973, Peter A. Parsons published Behavioral and Ecological Genetics: A Study

in Drosophila.30 This summarized much of the findings to date on the genetics of
mate choice and mating in relation to fruit fly evolution and speciation. Lee Ehrman,
as well as others, continued to work in this area.
A major contributor to finding the initial mutants was Ronald Konopka. This
initial work is well summarized in Benzers biography, Time, Love and Memory, by
Jonathan Weiner.39 The long-term goal of this approach was to use genetic mutants
to dissect the neural and other biological bases of behavior. About this time, the
same approach used by was W. Pak and K. Grossfield to study vision in fruit flies
and by many others to investigate the molecular and cell biology of neurons.
1.4.3.3 Rodents40
In 1970, Gardner Lindzey and Delbert D. Thiessen edited Contributions to Behavior-

Genetic Analysis: The Mouse as a Prototype.22 Each article in it is a review of a
major area of research in mouse behavior genetics during in the 1960s, and these
articles became the basis for much of mouse behavior genetics research in the 1970s
and 1980s.
Selective breeding continued to be used with mice and rats to establish lines
differing in behavior and to determine the heritability of the behavior. Some were
ones previously described. Others were new. These later included in rats the Syracuse
high and low avoidance lines (Bob Brush). Also, David Blizard and others continued
the study of the Maudsley Emotional and Non-emotional rats. Blizard also studied
behaviors of mice selected for thyroid function. In mice, Carol Lynch selected for
lines differing in the size of thermal nests and related her work to the ecology and
evolution of mice. Also, Bob Cairns and Kathryn Hood selected mice for male
aggression, and Janet Hyde and Patricia Ebert selected mice for female aggression.
There was also selection in rats and mice for effects of alcohol and drugs on behavior
such as the post-alcohol injection long sleep and short sleep lines. Much of this work
in mice was done by John Crabbe, Andrew Smolen, Toni Smolen, Thomas Johnson,
Bruce Dudek, as well as many others. Another study selected for high and low brain
weights and related these to mouse behaviors (John Fuller and Martin Hahn).
During this period investigators continued to using F1 diallels and F2 dihybrids
to assess the adaptive significance of behaviors. The behaviors studied included
ultrasounds, open field activity, aggression, mating, and learning. Contributors
included Norman Henderson, John Hewitt, Martin Hahn, Tom McGill, and many
others.
There was also an interest in finding the individual genes and chromosomes that
contributed to variation in mouse behaviors. One approach by Del Theissen and
colleagues was to see whether coat color and morphological variants had effects on
behavior. They did, but the mechanisms for the effects seemed trivial. More relevant
was research described by D. L. Coleman on two spontaneous mutants with effects
on feeding, satiety, metabolism, and obesity. These are the Obese and Diabetes
mutant mouse lines. Another approach was to try to use the newly developed
recombinant inbred strains to chromosomally map genes with behavioral effects
(Basil Eleftheriou and Thomas Seyfried). Because there were not enough chromo-
somal markers and too few recombinant inbred strains, these initial attempts to map
mouse variants with behavioral effects were often not replicated. However, Benson
Ginsburg, Stephen Maxson, and their doctoral students successfully used reciprocal
F1s and congenic strains to show that the male specific part of the Y chromosome
had effects on aggression and mating. The findings on the Y chromosome have been
replicated many times (Pierre Roubertoux, Michele Carlier, Frans Sluyter, and Geert
van Oortmerssen).
Also, Michele Carlier, Pierre Roubertoux, and their colleagues also began a
long-term research program on the interactions of maternal factors and genotype in
the development of mouse behaviors.
During this period there were also programs focusing on pharmacology and
neuroanatomy to find the connections between genes and behaviors. For example,
Kurt Schlesinger and colleagues used drugs to manipulate neurotransmitter systems
in relation to audiogenic seizures, and Richard and Cynthia Weimer measured strain
differences in cell number and other morphological parameters of the hippocampus
in relation to learning and memory. Douglas Whalsten also explored strain differ-
ences in corpus colossal morphology in relation to behavior. There were also studies
on strain difference in the endocrine systems and behavior such as those by Thomas
McGill on male mating, by Jack Vale on male aggression, and by Bruce Svare on
female aggression.
1.4.3.4 Canids
In 1970, a study of the evolution and genetics of canids was established at the
University of Chicago by Benson Ginsburg. This included research on the social
behavior and organization of a pack of wolves and on the inheritance of threat
behaviors in dogs (beagles) and coyotes.
1.4.3.5 Humans19
Research continued on cognition, personality, and psychopatholgy with family, twin,

and adoption methods. There were major conceptual and methodological advances.
There were developments in the design and analysis of these studies by David Fulker,
John DeFries, John Loehlin, Lindon Eaves, Robert Plomin, and many others. These
focused on combining data from at least two of these three, family, twin and adoption
studies, to develop mathematical models of genetic and environmental contributions
to individual differences. Also a major adoption project was established at the
University of Colorado by John DeFries and others, and Thomas Bouchard Jr., Nancy
Segal and their colleagues began a long-term intensive study of twins raised apart.
In addition, large twin registers were begun in Australia by Nicholas Martin, in
Sweden by Nancy Pedersen, and the Netherlands by Dorett Boomsma. Also, the
Louisville Twin Study was continued (Ron Wilson).
In this period, pathways from gene to neurological, behavioral or mental effects,
such as phenyketonuria, were worked out for many gene variants with large effects.
1.4.4 1986 TO 20058,31,32,37

There were many developments in molecular genetics that had direct impact on
behavior genetics. These included: the polymerase chain reaction method, sequenc-
ing the genomes of nematodes, flies, mice, and humans, identification of DNA
markers on chromosomes of nematodes, flies, mice, and humans for gene mapping,
and transgenics and knockouts in mice. These techniques revolutionized all of
biology including neural and behavior genetics. Their impact is discussed in detail
elsewhere in this volume. Here, I consider some of what was happening in behavior
genetics during these years.
Also, in this period, the International Behavioral and Neural Genetics Associa-
tion or IBANGS (1996) and the International Society for Psychiatric Genetics or
ISPG (1992) were established. Genes, Brain and Behavior is the journal of IBANGS,
and Neuropsychiatric Genetics is the journal of ISPG.
1.4.4.1 Nematodes2
The genome (100 Mb) of C. elegans was sequenced in 1998. During this period,
there was continued screening for mutants with effects on olfactory and touch
sensory systems, chemotaxis, thermotaxis, biological rhythms, social behavior,
learning and memory and behavioral effects of alcohol, cocaine, and nicotine. Many
genes with effects on these systems were identified and characterized. In many cases,
the effects of the gene can be or are being traced from DNA to protein to neural
organization and function to behavior.
1.4.4.2 Fruit Flies21,35,37
The genome (160 Mb) of the fruit fly was sequenced in the year 2000. During this
period single gene mutants were used to dissect the molecular and neural basis on
the circadian clock (Jeff Hall), mating (Jeff Hall, Charalambos P. Kyriacou, and
Jean-Marc Jallon), learning and memory (Tim Tully), and effects of alcohol and
drugs on behavior (F. E. Wolf and U. Huberlein). There was also an interest in
natural genetic variation in fly behavior. One example of this is Marla Sokolowskis
studies of the forager gene. Another example is the identification of the genes
involved in Hirschs long-term selection for the high and low geotaxis lines (Ralph
Greenspan). This gene identification used a combination of gene expression chips
and mutants. Greenspan has also investigated and reviewed the complex interactions
of genes with effects on fruit fly behaviors.
1.4.4.3 Bees33
The genome (200 Mb) of honeybees is being sequenced. Recent research (Gene
Robinson and others) on honeybees provides an excellent example of the association
between behavioral development and changes in gene expression and neural struc-
ture. Worker bees engage in nursing behavior from birth to about 2 to 3 weeks of
age. After that they engage in foraging behavior. This behavioral change is associated
with the increase or decrease in expression of many genes in the brain. Two of these
genes are period and forager, which were first found in fruit flies.
1.4.4.4 Rodents7,16,37,38
The mouse genome (3500 Mb) was sequenced in the year 2002. During this period,
the main focus of research has been to identify the genes that cause variation in
behavior. There have been four approaches for this. The first is to chromosomally
map quantitative trait loci (QTLs). One such study used crosses of the lines selected
for open field activity (Jonathan Flint and others.) Another approach used crosses
of inbred strains to find QTLs for aggression (Pierre Roubertoux and others). Some
studies combined recombinant inbred strains and strain crosses to find QTLs for
effects of alcohol and other drugs (John Crabbe, John Belknap, Keri Buck, Thomas
Johnson, Byron Jones, and others). In 1991, John C. Crabbe and R. Adron Harris
edited The Genetic Basis of Alcohol and Drug Actions.6 There has also been devel-
opment of large, recombinant, inbred strain sets that have been used to find QTLS
for brain structures and expressions of genes in mouse brains (Robert Williams and
others). Another approach has used specific gene knockouts and/or transgenics
(Crawley). This has been used with many behaviors including sensation, motor skills,
emotionality (Rene Hen), feeding (C. H. Vaughn and many others), aggression
(Randy Nelson), mating (Sonoko Ogawa and Emilie Rissman), learning and memory
(J. Z. Tsien and Jeanne Wehner), and effects of alcohol and drugs (Tamara Philips,
Kari Buck, and many others). Also, there are now tissue or temporal specific
knockouts and transgenics.
A related approach is based on spontaneous or induced mutations. As previously
described, Obese and Diabetes are two spontaneous mutations with effects on feed-
ing, satiety and metabolism. Recently, it was shown that the obese mouse has a
coding mutation for the polypeptide hormone leptin, and the diabetic mouse has a
coding mutation in a leptin receptor (J. M. Friedman and others). The study of these
two spontaneous mutants detailing the pathway form DNA to behavior opened wide
the door to research on the hormonal and neural mechanisms of feeding and satiety.
Also, male mice have been exposed to chemical mutagen and the F1 or F2 progeny
of these mice screened to detect mutations with neural or behavioral effects. One
of the first induced mouse mutants to be detected was Clock which affects circadian
rhythms (Joseph Takahashi and others). Other genes with effects on mouse circadian
rhythms were found to have fruit fly homologs, such as Period and Timeless. The
study of the Clock mutant and variants for Period and Timeless from DNA to behavior
opened up further research on the neural mechanism of the circadian clock in
mammals. There are now centers for mouse mutagenesis and behavior at the Jackson
Laboratory, Northwestern University, and the University of Tennessee Health Sci-
ences Center. Each has high throughput screens for a range of mouse neural and
behavioral traits.
Another approach is to look for differences in gene expression among selected
strains of mice. This was done for hippocampal gene expression in SAL (short attack
latency) and LAL (long attack latency) strains (D. E. Feldker and others).
Other research in the period has used inbred or selected strain-based genetic
correlation to relate brain and behavioral phenotypes. One example is the study of
variation in the size of the hippocampal mossy fibers and related cognitive or
emotional behaviors (Wim E. Crusio and Hans-Peter Lipp). Another interest has
been in the standardization of behavioral tests (Norman Henderson, Martin Hahn,
and Stephen Maxson) and phenotypic stability across laboratories (John Crabbe,
Douglas Wahlsten, and Bruce Dudek). Furthermore, during this period mouse models
for Turners, Klinefelters, and Downs syndromes were developed.
One series of fascinating studies has related the genetics of species differences
in the behavior of voles focused on oxytocin and vasopressin systems and aggression,
parenting and monogamy (Larry J. Young).
1.4.4.5 Humans1,5,28,29
The human genome (3500 Mb) was sequenced in the year 2000. During this period
path analysis and model fitting were fully developed to analyze the components of
individual differences in behavior (David Fulker, John DeFries, Lindon Eaves,
Nicholas Martin, Stacy Cherny, and many others). This approach allowed the com-
bined used of family, twin and adoption studies. It was widely applied to cognition
(Robert Plomin, Dorett Boomsma, and Sandra Scarr), personality (John Loehlin),
personality disorders (Richard Rose, Irwin Waldman, Laura Baker, and Liz DiLalla),
sexual orientation (Michael Bailey and Dean Hamer), psychopathology (Peter
McGuffin and Irving Gottesman), and addictions (C. R. Cloninger, Kenneth Kendler,
Matt McGue and Andrew Heath). There was also an interest in the relative roles of
within family environments, between family environments, genotypeenvironment
correlations and genotypeenvironment interactions (Robert Plomin and David
Rowe) as applied to the same range of human behaviors.
A goal of this period was also to chromosomally map the genes that are involved
in the individual differences for a behavioral domain such as cognition. The two
approaches to this are linkage analysis and association analysis (Eric Lander and
Pak Sham).
The former often uses the affected sib pair method and the latter often uses the
transmission disequilibrium test. Both methods make use of nongenic DNA markers
and association analysis uses not only nongenic DNA variants, but also DNA variants
of known genes. The main problem to date has been replication of findings. For this
reason, meta-analysis across studies is often used.
Animal models may also be used in the search for genes with behavioral effects
in humans. Animal models can indicate chromosome regions or genes to focus on.
1.5 JANUS: THE PAST AND FUTURE OF BEHAVIOR

GENETICS
I believe that there have been at least three conceptual divides in the history of
behavior genetics.
The first of these divides traces back to Darwin and Galton. The Darwinian
stream has been interested in the genetics of adaptive behavior in animals and
humans. Its findings are very relevant to biology in general. The Galtonian stream
has been interested in the genetics of human variation, especially cognition, person-
ality, personality disorders, psychopathology, and addictions. Its findings are very
relevant to education, employment and medicine.
The second divide has been between those who are interested in using genetics
as a tool in the study of behavior and those who are interested in using genetics to
analyze the causes of individual differences. The former investigates genetic variants
as a way to dissect the developmental and physiological mechanisms of behavior
and to find related general principles. The latter is interested in disentangling the
genetic and environmental causes of behavior and in finding the genes that cause
species behaviors to vary.
The third divide has been between those who study single gene induced muta-
tions or natural variants with large effects and those who study the effects of natural
variants of many genes with small effects. In some ways, this divide overlaps that
described in the previous paragraph.
These three divides have fragmented behavior genetics. Here I suggest one
approach to unifying behavior genetics in its second century. This is a comparative
genetics of behavior. In part, this will be based on genetic studies of many more
species than the big four of nematodes, fruit flies, mice, and humans. In the first
two periods of the first century, a wide range of animals was studied. There should
be a return to this practice. A comparative genetics of behavior will also depend on
the growing number of projects to sequence the DNA of representatives of major
taxa. This now includes not only nematodes, flies, mice and humans but also jellyfish,
bees, zebrafish, chickens, pigs, cattle, dogs, cats, monkeys, chimpanzee, and others.
Such a comparative genetics should also be gene centered comparing the behavioral
effects of homologous genes across species. For example, variants of the gene for
MAOA (monoamine oxidase A) have been shown to have similar effects on aggres-
sion in mice, monkeys and humans and depend for their effects on specific aspects
of the social environment during development.
1.6 A HIGHLY PERSONAL NOTE

I began my career in behavior genetics in 1960. My mentor was Benson Ginsburg.
Over the years, I have been privileged and honored to know most of those mentioned
in this brief history. All are colleagues and many are friends. I met many of the
founders of behavior genetics at the 1961 meeting in Stanford. Others, I have met
over the years at annual meetings of the Behavior Genetics Association. I have
attended 30 of the 35 meetings. Some I have met at meetings of IBANGs or the
Society for Neuroscience. There are many detailed stories that I could tell about
each one. But I have tried here to take a wider view of our mutual history. I am sure
that there are many behavioral geneticists whom I have not mentioned, especially
those working with nematodes, fruit flies, other critters, and humans and perhaps
even some of those working with rodents. For this, I apologize to one and all.
Regardless, there are so many of you today that I could not mention each and every
one of you. Just a list of all the names would fill this history to its page limit. This
is one measure of how successful behavior genetics has been in its first 100 years.
REFERENCES
1. Benjamin, J., Ebstein, R. P., and Belmaker, R. H. Molecular Genetics and Human
Personality. American Psychiatric Publishing Inc., Washington, DC, 2002.
2. Bonn, M. and Villu, Maricq A. Neuronal substrates of complex behaviors in C.
elegans. Ann. Rev. Neurosci., 28: 451501, 2005.
3. Brenner, S. The genetics of behaviour. Br. Med. Bull., 29: 269271, 1973.
4. Broadhurst, P. L. and Fulker, D. W. Behavioral genetics. Ann. Rev. Psychol., 25:
389415, 1974.
5. Carey, G. Human Genetics for the Social Sciences, Sage Publication, London, 2003.
6. Crabbe, J. C. and Adron Harris, R., Eds., The Genetic Basis of Alcohol and Drug
Actions, Plenum Press, New York, 1991.
7. Crawley, J. N. What is Wrong with my Mouse: Behavioral Phenotyping of Transgenic
and Knockout Mice. John Wiley & Sons, New York, 2000.
8. Crusio, W. E. and Gerlai, R. T. Handbook of Molecular-Genetic Techniques for Brain
and Behavior Research. Elsevier, Amsterdam, 1999.
9. Defries, J. C. and Plomin, R. Behavioral genetics. Ann. Rev. Psychol., 29: 473515,
1978.
10. Ehrman, L. and Parsons, P. The Genetics of Behavior. Sinauer Associates, Sunderland,
MA, 1976.
11. Falconer, D. S. Introduction to Quantitative Genetics. The Ronald Press, New York,
1960.
12. Fuller, J. and Thompson, W. R. Behavior Genetics. John Wiley & Sons, New York,
1960.
13. Fuller, J. and Thompson, W. R. Foundation of Behaviour Genetics, The C.V. Mosby
Co., St. Louis, 1978.
14. Fuller, J. and Simmel, E. C. Behavior Genetics: Principles and Applications.
Lawrence Erlbaum Associates Publishers, Hillsdale, NJ, 1983.
15. Ginsburg, B. E. Genetics as a tool in the study of behavior. Perspectives in Biology
and Medicine, 1: 397424, 1958.
16. Goldowitz, D., Wahlsten, D. and Wimer, R. E. Techniques for the Genetic Analysis
of Brain and Behavior: Focus on the Mouse. Elsevier, Amsterdam, 1992.
17. Hall, C. S. The genetics of behavior. In: Handbook of Experimental Psychology,
Stevens S. S., Ed. John Wiley & Sons, New York, 1951, chap. 9.
18. Hall, J. C., Greenspan, R. J. and Harris, W. A. Genetic Neurobiology. MIT Press,
Cambridge, MA, 1982.
19. Henderson, N. D. Human behavior genetics. Ann Rev. Psychol., 33: 401440, 1982.
20. Hirsch, J., Ed. Behavior-Genetic Analysis. McGraw-Hill, New York, 1967.
21. Kyriacou, C. P. Single gene mutations in Drosophila: What can they tell us about the
evolution of sexual behavior? Genetica, 116: 197203, 2002.
22. Lindzey, G. and Thiessen, D. D. Eds., Contributions to Behavior-Genetic Analaysis:
The Mouse as a Prototype, Appelton Century Crofts, New York, 1970.
23. Lindzey, G., Loehlin, J., Manosevitz, M., and Theissen, D. Behavioral genetics. Ann.
Rev. Psychol., 22: 3994, 1971.
24. Mather, K. Biometrical Genetics, Methuen, London, 1949.
25. McClearn, G. E. Behavioral genetics. Ann. Rev. Genet., 4: 417468, 1970.
26. McClearn, G. E. and DeFries, J. C. Introduction to Behavioral Genetics. W.H. Free-
man, San Francisco, 1973.
27. McClearn, G. E. and Meredith, W. Behavioral genetics. Ann. Rev. Psychol., 17:
515550, 1966.
28. McGue, M. and Bouchard Jr., T. J. Genetic and environmental influences on human
behavioral differences. Ann. Rev. Neurosci., 21: 124, 1998.
29. McGuffin, P., Owen, M. J., and Gottesman, I. J. Psychiatric Genetics and Genomics.
Oxford University Press, New York, 2002.
30. Parsons, P. A. Behavioral and Ecological Genetics: A study in Drosophila, Oxford
University Press, New York, 1973.
31. Pfaff, D. W., Berrettini, W., Joh, T. H. and Maxson, S. C. Genetic Influences on
Neural and Behavioral Function, CRC Press, Boca Raton, FL, 2000.
32. Plomin, R., DeFries, J. C., Craig, I. W., and McGuffin, P. Behavioral Genetics in the
Postgenomic Era. American Psychological Association, Washington, DC, 2002.
33. Robinson, G. E. and Ben-Shahar, Y. Social behavior and comparative genomics: new
genes or new gene regulation. Genes, Brain, Behav., 1: 187203, 2003.
34. Scott, J. P. and Fuller, J. L. The Genetics and Social Behavior of the Dog, University
of Chicago Press, 1965.
35. Sokowlowski, M. B. Drosophila genetics meets behavior. Nat. Genet., 2: 879880,
2001.
36. van Abeelen, J. H. F., Ed. The Genetics of Behavior, North-Holland, Amsterdam, 1974.
37. Wahlsten, D. Single-gene influences on brain and behavior. Ann. Rev. Psychol. 50:
599624, 1999.
38. Weiner, J. Time, Lover, and Memory: A Great Biologist and His Quest for the Origins
of Behavior. Alfred A. Knopf, New York, 1999.
39. Whener, J. M., Radcliffe, R. A., and Bowers B. J. Quantitative genetics and mouse
behavior. Ann. Rev. Neurosci., 24: 845867, 2001.
40. Wimer, R. E. and Wimer, C. E. animal behavior genetics: a search for the biological
foundations of behavior. Ann. Rev. Psychol. 36: 171218, 1985.
2 Developmental
Neurobehavioral Genetics:
Development as
Explanation
Gilbert Gottlieb
CONTENTS
Introduction.............................................................................................................. 17
2.1 Preformation and Epigenesis ......................................................................... 17
2.2 Development as Explanation ......................................................................... 19
2.3 A Currently Acceptable Developmental Neurobehavioral
Genetics: Interim Solution ............................................................................. 24
2.4 Summary and Conclusions ............................................................................ 26
Acknowledgments.................................................................................................... 26
References................................................................................................................ 26
INTRODUCTION
From the very dawn of human history, there must have been people who wondered
how we come to benot just in the grand religious sense that ancient texts like the
Hebrew bible attempt to answer, but also in the more mundane and practical sense
of wondering about the mechanisms involved in human (and all animal) development
from egg and sperm to full-grown adult. By the time of Aristotle in the fourth century
B.C.E., there were two main schools of thought on how we become. In fact, the
proper scientific name for the study of individual development was derived from
the Greek language: ontogeny.
2.1 PREFORMATION AND EPIGENESIS

One school of thought about ontogeny, the preformationist school, held that a very
tiny version of the complete, adult individual was prepackaged either in the female
egg or in the male sperm. (There were actually two schools of preformationist thought:
ovists and spermists.) In other words, according to preformationist thinkingwhether
17
ovist or spermistall the parts and organs that an individual will ever have are present
from the outset; development is merely the growth of these preexisting parts until
they reach their full, adult size. Aristotles view, on the other hand, gained from his
personal observations of the developing chick embryo, was that individual develop-
ment happens through epigenesis; that is, through a series of successive transforma-
tions of some sort of early, homogeneous mass that brings the parts and organs of
an organism into being. In this view, what happens during development is not just
the quantitative growth of preexisting parts, but the gradual qualitative differentiation
of parts that then grow as well. Today we still do not completely understand all of
the mechanisms that make development happen, but we do know that it is correct to
say that individual developmentwhether in terms of psychology, behavior, physi-
ology, or anatomyis epigenetic and not preformative.
Nevertheless, some scientists continue to operate under a modern-day version
of preformationist thinking when they say that genes govern individual development.
The only difference between the ancient idea of the unfolding of preformed parts
and the modern idea of the unfolding of the genetic code is that the unfolding, itself,
is now understood to be a transformative processin other words, epigenetic. All
biological scientists understand that successive transformations of the forms and
functions of tissues occur during development. This is what is meant by epigenetic
development. But many scientists seem to think that these transformations are
controlled by little, preformed, inherited, autonomous packets of information that
we know today as genes. This is a modern-day version of preformationism. As the
noted biologist J. S. Haldane13 observed, the idea of genes as the controllers of
individual development only substitutes an extremely complicated molecular struc-
ture for the ancient idea of the original miniature adult(p. 147). I call the idea of
epigenetic transformation that is controlled by genes, predetermined epigenesis.
According to this way of thinking, further developmental processes may occur due
to environmental influences, but genes and the environment are viewed as separate
causal systems rather than as parts of one, integrated system. This idea that there
are separate causal systems is the crux of what has become known as the naturenur-
ture dichotomy.
The advocates of a genetically predetermined epigenesis have often recognized
that the naturenurture dichotomy is a shaky construct because, although based on
a recognition that environment does make a difference in development, it lacks an
adequate theory of how, when, and where this happens.22 In the end, only lip service
is paid to the idea of geneenvironment interactions, and genetic predeterminists
say it is genes primarily that make development happen and that the environment
plays only a passive, permissive, or supportive role. The formative drive, they say,
is governed by genes acting as autonomous agents.
However, as several excellent popular books have already described16,18 recent
research demonstrates that genes are not really unchanging little packets of unfold-
ing information standing on the sidelines of epigenetic processes and stoically
directing the developmental game. Instead, as M.-W. Ho has written,14 they get
variously chopped, rearranged, transposed, and amplified in different cells at
different times (p. 285)and themselves must be activated in order to make any
contribution at all to development. With the supposed role of genes as independent
Developmental Neurobehavioral Genetics 19
agents of development thus brought into question, the idea of a naturenurture

dichotomy has gotten to be on even shakier ground than before. That is, given
that genes need nongenetic inputs to be activated, the naturenurture dichotomy
is undermined at its most basic level.
The aim of this chapter is to replace the naturenurture dichotomy with a
thoroughly integrative, coactional theory of epigenesis that takes into account the
role of genes and environments as well as two other domains of influence in the
formative process: behavior and the nervous system. The various processes that work
together across these four levels of development are what produce a fully formed,
behaviorally mature organism. Because according to this theory there is no single
factor or influence that causes developmental outcomes to be the way they are, I
call the point of view probabilistic epigenesis. This is the theory that any devel-
opmental outcome occurs not as a matter of predetermined necessity but as a result
of thoroughly interacting developmental processes and that, if the operating param-
eters of any of these processes are changed during development, the outcome will
also change to one degree or another. The coordination of the necessary influences
from the four levels does not occur with absolute fidelity, thus giving a probabilistic
character to the process as well as to the outcomes of development. That is why
epigenesis is probabilistic, not fixed.
2.2 DEVELOPMENT AS EXPLANATION

The most difficult concept to get across in the understanding of psychology, animal
behavior, neuroscience, or neurobehavioral genetics is development-as-explanation.
Development is viewed by many as just one more thing we study as psychologists,
animal behaviorists, neuroscientists, or geneticists. So, I sometimes get a blank stare
or a quizzical look when I propose the idea of development as explanation. For
example, some of my colleagues study neuroscience or the brain because in the
usual theoretically reductionistic scheme of things, they feel that will allow them to
understand the mind; others, for the same reason, study genetics because they feel
that will allow them to understand normal and abnormal behavior. My theoretically
nonreductionistic scheme (Figure 2.1) for developmental analysis includes the ner-
vous system and genetics, as well as behavior and the physical, social, and cultural
aspects of the environment.
These four levels of analysis are mutually influential, so the arrows of influence
are top-down as well as bottom-up. This is a fully bidirectional coactional system,
with traditionally trained developmental psychologists working at the Behav-
iorEnvironment level, developmental neuropsychologists and neuroscientists at
the Neural ActivityBehaviorEnvironment levels, and biologically trained per-
sons beginning now to work on the Genetic ActivityNeural Activity levels. (For
an outstanding example of the latter, see Rampon et al.21) Viewed in this way,
developmental understanding or explanation is a multilevel affair involving at least
culture, society, immediate social and physical environments, anatomy, physiology,
hormones, cytoplasm, and genes. Hierarchical, multilevel, or developmental systems
analysis is methodologically reductionistic in the sense that biological factors are
included to make investigations more complete. However, it is not theoretically
DEVELOPMENTAL ANALYSIS
ENVIRONMENT
(Physical, Social, Cultural)
BEHAVIOR
NEURAL ACTIVITY
GENETIC ACTIVITY
Individual Development
FIGURE 2.1 A developmentalpsychobiological systems model to illustrate the comprehensive
analysis required for developmental explanation. (Modified from Gottlieb8 with permission.)
reductionistic in the sense that psychological understanding or explanation does not

come from lower levels (or reside in the lower levels), but that both higher and lower
levels of analysis are necessary to explain developmental outcomes. The term coac-
tion is used to emphasize the multilevel systems nature of developmental analysis.
Any outcome (genetic activity, neural activity, behavior) is an epigenetic out-
come of prior developmental processes in the sense that it is an emergent property.
The only way to understand the emergence of that outcome, whether genetic, neural,
or behavioral, is to study the coactional influences within and between the four levels
of analysis that gave rise to the emergence of that outcome. Since all outcomes are
emergent, they have a developmental history that explains their emergence. That is
why developmental analysis is explanation.
When my model (Figure 2.1) is unpacked to reveal all of its otherwise hidden
mysteries, as has been done in Figure 2 by Johnston and Edwards,15 all the reciprocal
intermediate steps between genetic activity, neural activity, behavior and the envi-
ronment can be visualized. (The issue of time, which is crucial for developmental
analysis, is treated elsewhere [pp. 2930] in Johnston and Edwardss article.)
The significance of Figure 2.2 is that it raises the kinds of developmental questions
raised by findings that might otherwise be misinterpreted as evidence of a direct
link between genes and behavior. Even within the same species there is not just one
way to get from genes to behavior.23
Figure 2.2 extends and explicates the important features of Figure 2.1 in four
ways, as follows from Johnston and Edwardss discussion of time. First, it includes
both neural and non-neural components. Certainly, the most thoroughly studied
examples of non-neural contributions to structural, physiological, and behavioral
development involve hormones; however, development also involves bones, muscles,
horns, feathers, and other bodily structures; and all of these need to be taken into
account. For example, in Thelens25 analysis of infant locomotion, gross morpho-
logical changes play a critical role.
Second, the immediate consequences of genetic activity are confined to the
cell. Genetic effects on behavioral development must take into account the various
BEHAVIOR
SENSORY
Patterned
STIMU- Neural Activity
LATION
Neural Non-neural
Connectivity Structures
Individual
Nerve Cell
Activity Neural Non-neural
Growth Growth
Extracellular
Cell Membrane Biochemistry
Intracellular
Biochemistry
Protein PHYSICAL
Synthesis INFLUENCES
GENETIC
ACTIVITY
FIGURE 2.2 Completely unpacked model of developmental neurobehavioral genetics, show-

ing all the coacting factors involved in the developmental construction of behavior and the
coactions among them. The model includes both neural and non-neural elements, the latter
encompassing such influences as hormones (which constitute part of the extracellular bio-
chemistry), bones, muscles, feathers, and so forth. Sensory stimulation is shown to be influ-
enced not only by behavior (as the animal moves about in its environment, both producing
and modifying the stimulation it receives), but also by the connectivity of its nervous system
(which partly determines its sensitivity to sources of stimulation) and by the current state of
neural activity. The elliptical arrow depicts the effects of spontaneous neural activity. All
enduring experiential effects on development that have their immediate impact on patterns of
neural activity act by modifying events at the cellular level, including patterns of genetic
activity. Note that there is no direct connection between genetic activity and behavior; all
genetic effects on behavior are mediated through the cell membrane and subsequent coactions
among cells and neural networks. Solid lines with arrows represent causal relationships
between coacting factors. Dotted lines connecting patterned neural activity to individual nerve
cell activity indicate that the latter is nested within the former; the relationship between the
two is not causal. (From Johnston and Edwards15 with the permission of the authors and the
American Psychological Association.)
coactions that follow from protein synthesis and its consequences for events at the
cell membrane, coactions among cells, and so on. The model treats genes as an
integral part of the developing system, rather than placing them outside the system,
and sees genes as influencing behavior indirectly, not directly. When a particular
gene has been implicated in the development of some behavior, the model would
accommodate the identification of various roles that the genes activity might play
in development. For example, in a review of the effects of single-gene mutations on
the development of touch receptors in nematodes, Chalfie5 proposed four develop-
mental roles (generation, specification, function, and maintenance) for the 18 genes
that have been implicated so far in touch receptor development. Although Chalfies
taxonomy deals with anatomy rather than behavior, similar taxonomies might be
proposed for the development of behavior (e.g., Gottliebs delineation of the three
roles of experience in behavioral and neural development: induction, facilitation,
and maintenance10). The JohnstonEdwards model (Figure 2.2) has the advantage
that it provides an explicit representation of the intervening coactions implied by
such taxonomies, even though their coactions may not always be specified. Thus, it
indicates the kinds of developmental questions raised by findings that might other-
wise be taken as evidence of a direct link between genes and behavior.
Third, when experience has more than a transient effect on behavior, the effect
is almost certainly mediated through changes in genetic activity. The model
implies that all instances in which experience has been shown to affect behavioral
development must involve some change in genetic activity. Developmental theory
holds that there can be no genetic effects on behavior independent of the envi-
ronment and there are probably no environmental effects on behavior independent
of genetic activity. The model (Figure 2.2) helps to make this statement more
precise by showing the pathway by which experience activates genes through the
agency of neural activity, once again supplying a mediating pathway for findings
that might otherwise be interpreted as evidence of a direct link between genes
and behavior.
Fourth, Johnston and Edwardss model in Figure 2.2 recognizes that nervous
system activity needs to be considered at two levelsin terms of neural activity,
often involving networks of cells in different anatomical regions, and in terms of
the activity of individual nerve cells within which the genes are located. The dotted
lines connecting these two boxes in Figure 2.2 indicate that the activity of individual
cells is nested within the patterns of activity of cell networks. Individual cell activity
neither causes nor is caused by the patterns of activity in cell networks: rather, there
are two levels at which neural activity must be analyzed. Such a dual level of analysis
does not mean that an account of genetic activity must be given individual cell by
individual cell. Rather, ways of describing the developing nervous system both in
terms of populations of cells with similar patterns of genetic activity and in terms
of networks of cells that participate in behavior must be found. For example, Bren-
nan, Hancock, and Keverne have shown that the immediate-early genes c-fos and
zif-268 (but not c-jun) show distinct patterns of both transient induction and persistent
induction in the accessory olfactory bulb (AOB) of female mice immediately after
mating.2 The induction of c-fos is seen only in the granule cells of the AOB, whereas
the zif-268 induction occurs in both the granule and mitral cells. The AOB is known
to be involved in changes in female olfactory responsiveness to male pheromones

following mating, and the differential patterns of gene expression in these two cell
types indicate the complexity of the relationships that are likely to exist between
neural networks and genetic activity in the AOB.
A further caution to those like myself, who wish to establish correlative links
between genetic activity and behavior, is posed by the phenomenon of RNA-editing.
The role of genetic activity in producing protein is often depicted in a much too
simple unilinear formula: DNA RNA Protein, in which DNA establishes the
nucleotide structure of mRNA which then translates into the structure of the partic-
ular protein resulting from the process. It may come as news to many biologists that
Erwin Chargraff 6 could have this to say as early as 1978: One of the obnoxious
dogmas to which it [molecular biology] has given risethe so-called Central
Dogma: DNA makes RNA; RNA makes proteinsis no longer valid. We now know
that stretches of DNA are composed of exons (coding portion) and introns (noncod-
ing portion). Before the structure of the protein is composed, nearly 40% of human
genes are alternatively spliced (i.e., mRNA is edited), not only removing the silent
introns but replacing some of the coding exons! This enormously complicates the
identification of which genes are actually involved in the making of particular
proteins. As stated on p. 1346 of the report The Sequence of the Human Genome:26
as was true at the beginning of genome sequencing, ultimately it will be necessary
to measure mRNA in specific cell types to demonstrate the presence of a gene.
Meanwhile, while we continue to try to correlate genotypes with events at the neural
and behavioral levels, we need to remind ourselves of the uncertainty involved. Since
much of the genetic analysis in humans and other species involves single nucleotide
polymorphisms (SNPs), which are markers for as yet unidentified genes, RNA-
editing adds yet a further complication in linking particular genes with particular
proteins. It is estimated that one human gene may produce up to five different proteins
as a result of alternative splicing.20
Before leaving the topic of development as explanation, for the sake of com-
pleteness it is essential to mention an often-overlooked factor in developmental
neurobehavioral genetics: the role of the absence of genetic activity in development.
When discussing mental disorders it is common to hear talk of mutations. What is
sometimes not distinguished in the absence of genetic activity (deletions) is a frank
mutation of a gene in which the usual genetic structure is somehow scrambled,
repeated, and/or translocated, thereby resulting in an aberrant protein rather than the
mere absence of the normal protein. Talk of disease genes is particularly mislead-
ing when the culprit is in fact merely the lack of genetic activity and thus the absence
of a functional protein essential to normal development as occurs not only in the
well-known case of phenylketonuria (PKU) mental retardation but a host of neurop-
sychiatric disorders such as Prader-Willi syndrome, fragile X syndrome, Williams
syndrome, and lissencephaly, among others (see Table 1 in Brodsky and Lombroso,3
p. 3). If we are to be able to clarify the basic molecular mechanisms in normal and
abnormal development, it is necessary to distinguish the role of mutated proteins
from the functioning of normal proteins. To discourse about nonexistent disease
genes when solely deletions are involved only serves to confuse matters in an
already very complex situation.
2.3 A CURRENTLY ACCEPTABLE DEVELOPMENTAL

NEUROBEHAVIORAL GENETICS: INTERIM
SOLUTION
Figure 2.2 portrays a finished or compleat developmental neurobehavioral genet-
ics. It is an ideal to be sought and will take many years to achieve. In the meantime
we need a model that can be utilized in our day-to-day efforts, so I offer the following,
which is based on the notion of the ubiquity of gene-environment coaction docu-
mented earlier.9
The failure of replications of gene phenotype associations are legion (extensive
review in Wong, Buckle, & Van Tol27), the reason being that the environmental (or
what I shall call the life experience) component is left out when one merely tries to
relate a gene (or genes) to an outcome (e.g., genes for schizophrenia, genes for
aggression, genes for novelty-seeking). I am proposing that it is essential to link the
gene(s) with a life experience to get a consistent (replicable) result, a proposal that is
now receiving some attention in the clinical literature.17 For example, in a study of the
association of a certain genotype (TT genotype) with low concentrations of high
density lipoprotein (HDL) cholesterol, it was found that only those TT bearers whose
fat intake was at least 30% of their total consumed energy manifested low HDL; the
other persons in the sample with the TT genotype did not manifest low HDL.19 The
crucial environmental or life-experience factor here is ingestion of a certain level of
particular nutrients. Likewise, in a behavioral example of individuals with a genotype
that was associated with low levels of the enzyme monoamine oxidase A (MAOA), it
was those who had experienced severe maltreatment in their younger years that were
prone to violence in adulthoodthose having the same genotype who experienced no
maltreatment were unlikely to be violent in adulthood.4 Eighty-five percent of the
males having the low-activity MAOA genotype who were severely maltreated devel-
oped some form of antisocial behavior; 15% did not develop antisocial behavior.
Obviously, even with the inclusion of the crucial life-experience factor, we are still
talking about probabilities (epigenesis continues to be probabilistic).
Some of us take it as a given that genes, in and of themselves, cannot produce
any neural or behavioral outcome and that geneenvironment interaction is a require-
ment of normal as well as abnormal development. Thus, my probabilistic epigenetic
model of developmental outcomes assumes that individuals of the same genotype can
have different neural and behavioral outcomes according to the dissimilarity of their
relevant life experiences, broadly construed. I think this is the basis for the lack of
replications among studies that look only at genotypes and attempt to correlate a
particular genotype with a certain neural or behavioral outcome without looking for
the presence or absence of intervening life experiences that may be crucial to the
presence or absence of the outcome. Take the much-studied inhibitory neurotrans-
mitter serotonin. Low levels of serotonin are associated with depression and alcohol
abuse in humans. However, correlates of low serotonin are not behaviorally specific
(i.e., low serotonin is involved in a number of psychiatric disorders). In rhesus
monkeys, low concentrations of serotonin metabolites (collected from cerebral spinal
fluid) are associated with higher levels of impulsive aggression and risk taking (e.g.,
taking long leaps in Suomis study24). Rhesus infants who develop the least secure
attachment with their mothers are also the most likely to have deficits in their central
serotonin metabolism. Because there is a positive correlation between maternal and
infant serotonin level, a genetic deficit could be involved, but it is possible that
aberrant maternal care may make a necessary contribution to the serotonin deficit.
To shed light on the genetic and interactive aspect, Bennett et al.1 genotyped the
monkeys in Suomis laboratory for a known polymorphism (long and short allele) in
the serotonin transporter gene (5-HTT). The short allele confers low transcriptional
efficiency to the 5-HTT gene promoter (relative to the long allele), so low 5-HTT
expression may result in lower serotonergic function. However, evidence for this in
humans is inconsistent because the necessary life experience correlates have not been
examined. In the case of rhesus monkeys, when attempting to correlate the genetic
polymorphism to serotonin metabolism, serotonin concentration did not differ as a
function of long or short 5-HTT status for mother-reared monkeys, whereas, among
peer-reared monkeys, individuals with the short allele had significantly lower sero-
tonin concentrations than those with the long allele.1 Thus, the lowered serotonin
metabolism was not simply a consequence of having the short allele but required the
life experience of peer rearing in this instance. This result supports my idea that the
inconsistencies in the human literature are likely due to unknown but influential
differences in the experiential histories of the populations under study.
Thus, the notion that the short allele of the 5-HTT gene is inevitably associated
with a central nervous system (CNS) deficit or defect is not true: the neural outcome
depends on the developmental rearing history of the animal, as well as the particular
genotype of the animal itself, what has elsewhere been termed relational causality.11
The present finding most likely also explains why there are inconsistencies in the
human literature in finding anxiety-, depression-, and aggression-related personality
traits associated with variations in the serotonin transporter gene.1 The association,
or lack thereof, does not simply reflect genetic causality but developmental-relational
causality.
In sum, the chances of linking genotypes to behavioral (and other) outcomes
will be vastly improved when crucial intervening life experiences are routinely
included in developmental behavioral genetic investigations. This follows from the
empirical work reviewed earlier indicating that gene-environment coactions are the
rule in developmental investigations.9
While gene-environment coactions are a step up from simple single-gene
outcome associations, single-genesingle-experience coactions are going to continue
to be prone to lack of replication because complex behavioral outcomes are no doubt
backed by multiple genes (epistasis) and possibly more than one crucial life
experience. The ultimate solution will be to actually employ more than one relevant
gene and more than one relevant life experience if we are eventually to achieve
highly replicable findings. For example, the Caspi et al.4 study of the low MAOA
genotype and maltreatment alluded to earliera single-gene life experience study
has been replicated at least once7 and has not been replicated at least once.12 So, in
my opinion, the interim solution proposed in the present section of this essay will
be best implemented using more than one genetic marker and, whenever possible,
more than one life experience if we aspire to a mature developmental neurobehavioral
genetics in which replication is the measure of that maturity.
2.4 SUMMARY AND CONCLUSIONS

With the implementation of the psychobiological models presented herein for the
analysis of behavioral development at all levels of inquiry, the analytic apparatus is
at hand to pursue a developmental neurobehavioral genetics, or development as
explanation. Also, adopting a multigenemultilife experience approach should help
to overcome the problem of replicability in the field of behavioral genetics.
ACKNOWLEDGMENTS
The authors research and scholarly activities are supported, in part, by National
Institute of Mental Health Grant P50-MH-52429 and National Science Foundation
Grant BCS-0126475. Some material from my 2003 Human Development article18
is included in this chapter. I thank Brenda Denzler for editorial assistance with the
introduction of this chapter.
REFERENCES
1. Bennett, A.J. et al. Early experience and serotonin transporter gene variation interact
to influence primate CNS function, Molecular Psychiatry, 7: 118122, 2002.
2. Brennan, P.A., Hancock, D., & Keverne, E.B. The expression of the immediate-early
genes c-fos, egr-1, and c-jun in the accessory olfactory bulb during the formation of
the olfactory memory in mice. Neuroscience, 49: 277284, 1992.
3. Brodsky, M., & Lombroso, P.J. Molecular mechanisms of developmental disorders.
Development and Psychopathology, 10: 120, 1998.
4. Caspi, A. et al. Role of genotype in the cycle of violence in maltreated children.
Science, 297: 851854, 2002.
5. Chalfie, M. Touch receptor development and function in Caenorhabditis elegans,
J. Neurobiology, 24: 14331441, 1993.
6. Chargraff, E., Hericlitian Fire: Sketches from a Life Before Nature, Rockefeller
University Press, New York, 1978, pp. 106107.
7. Foley, D.L. et al. Childhood adversity, monoamine oxidase a genotype, and risk for
conduct disorder. Archives of General Psychiatry, 61: 738744, 2004.
8. Gottlieb, G. Individual Development and Evolution: The Genesis of Novel Behavior,
Erlbaum, Mahwah, New Jersey, 2002.
9. Gottlieb, G. On making behavioral genetics truly developmental, Human Develop-
ment, 46: 337355, 2003.
10. Gottlieb, G. The roles of experience in the development of behavior and the nervous
system, in Neural and Behavioral Specificity, G. Gottlieb, Ed., Academic Press, New
York, chap. 2, 1976.
11. Gottlieb, G., & Halpern, C.T. A relational view of causality in normal and abnormal
development. Development and Psychopathology, 14: 421435, 2002.
12. Haberstick, B.C. et al. Monoamine oxidase A (MAOA) and antisocial behavior in
the presence of childhood and adolescent maltreatment. Neuropsychiatric Genetics,
135B:5964, 2005.
13. Haldane, J.S. The Philosophical Basis of Biology, Doubleday, Garden City, New York,
1931.
14. Ho, M.-W. Environment and heredity in development and evolution, in Beyond Neo-
Darwinism: An Introduction to the New Evolutionary Paradigm, M.-W. Ho & P.T.
Saunders, Eds., Academic Press, London, 267289, 1984.
15. Johnston, T.D., & Edwards, L. Genes, interactions, and the development of behavior.
Psychological Review, 109: 2634, 2002.
16. Keller, E.F. The Century of the Gene, Harvard University Press, Cambridge, 2000.
17. Merikangas, K.R., & Risch, N. Genomic priorities and public health, Science, 302:
599601, 2003.
18. Moore, D. The Dependent Gene: The Fallacy of Nature vs. Nurture, Henry Holt,
New York, 2002.
19. Ordovas, J.M. et al. Dietary fat intake determines the effect of a common polymor-
phism in the hepatic lipase gene promoter on high-density lipoprotein metabolism:
Evidence of a strong dose effect in the genenutrition interaction in the Framingham
study, Circulation, 106: 23152321, 2002.
20. Peters, R.J.G., & Boekholdt, S.M. Gene polymorphisms and the risk of myocar-
dial infarctionan emerging relation, New England Journal of Medicine, 347:
19631965, 2002.
21. Rampon, C. et al. Enrichment induces structural changes and recovery from nonspatial
memory deficits in CA1 NMDAR1-knockout mice, Nature Neuroscience, 3: 238244,
2000.
22. Rutter, M. et al. Testing hypotheses on specific environmental causal effects on
behavior, Psychological Bulletin, 127: 291324, 2001.
23. Schaffner, K.F. Genes, behavior, and developmental emergentism: One process indi-
visible? International Journal of Behavioral Development, 24: 514, 1998.
24. Suomi, S.J. A biobehavioral perspective on developmental psychopathology, in Hand-
book of Developmental Psychopathology, A.J. Sameroff, M. Lewis, & S.M. Miller,
Eds., Kluwer Academic/Plenum, New York, chap. 13, 2000.
25. Thelen, E. Motor development: A new synthesis, American Psychologist, 50: 7995,
1995.
26. Venter, J.C. et al. The sequence of the human genome, Science, 291: 3041351, 2001.
27. Wong, A.H.C., Buckle, C.E., & Van Tol, H.H.M. Polymorphisms in dopamine recep-
tors: What do they tell us? European J. Pharm., 410: 183203, 2000.
1903_C003.fm Page 29 Thursday, July 27, 2006 9:04 PM
3 Some Basics, Mendelian

Traits, Polygenic Traits,
Complex Traits
Byron C. Jones
CONTENTS
3.1 Introduction and Levels of Investigation....................................................... 29

3.2 Fundamental Questions. What Causes What?............................................... 30
3.3 Phenotypes and Phenotypic Selection ........................................................... 30
3.4 GenotypePhenotype Relationships .............................................................. 31
3.5 Mendels Laws ............................................................................................... 31
3.5.1 Law of Segregation ............................................................................ 31
3.5.2 Law of Independent Assortment........................................................ 32
3.6 Linkage........................................................................................................... 32
3.7 Mendelian Traits ............................................................................................ 32
3.7.1 Recessively Transmitted Traits .......................................................... 32
3.7.2 Dominant Traits ................................................................................. 33
3.8 Complex Traits and Additive Genetic Effects............................................... 33
3.9 Interactions ..................................................................................................... 34
3.9.1 GeneEnvironment Interactions......................................................... 34
3.9.2 GeneGene Interactions..................................................................... 35
3.10 Summary ........................................................................................................ 35
References................................................................................................................ 35
3.1 INTRODUCTION AND LEVELS OF

INVESTIGATION
To some in the functional school of psychology of the early twentieth century, the
theory of evolution gave reason to believe that behavior, as much as morphology
and biochemistry, served as means of fitness. Thus, instinct became as valid a
criterion for fitness as length of a limb, rate of a biochemical reaction, etc. This led
to a circularity of reasoning that gave the ammunition to the behaviorist school of
psychology to eschew all things innate as concerns behavior. Accordingly, the
environment was the sole source of behaviors, including their differences in expres-
sion and development. For many years, the environmentalist view of acquisition
29
and development of behavior dominated the discipline of psychology in America.

Even today, the mere mention of genes as influencing behavior elicits skepticism.
In Europe, behavioral biologists took a more heredity-friendly view of behavior and
its development, and a great deal of debate ensued between the groups. A modern
synthesis of the best of the thinking on both sides has lent credence to the notions
that (1) genes do in fact influence behavior, (2) genes interact with the environment
and with each other, and (3) the actions of genes and their interactions are accessible
for study.
Conrad H. Waddington was a developmental biologist who, from the 1930s until
the 1960s, helped to elucidate how genes and their environments could influence
morphology and biochemical pathways. He asserted that genes exert multiple effects
(pleiotropy) and that the immediate environmental conditions, nutrients, hormones,
etc., affected gene action and, hence, development. His concept was, epigenesis +
genes = epigenetics.10 This useful framework also applies in the examination of
neurobehavioral genetics. As we will see in the coming chapters, phenotypes
(observable, measurable characteristics), including behaviors, arise as a function of
the genotype, the environment, and geneenvironment and genegene interactions.
Another useful conceptualization is that of G.E. McClearn who makes the distinction
between the fixed genotype (that which is inherited from the parents) and the
momentary-effective genotype, i.e., the summed action of genes that are expressed
at a particular time in the life of an organism.6
Neurobehavioral genetics is one of the few disciplines that crosses the (artificial)
levels of analysis. Thus, we can investigate neurobiology at anatomical and cell-
molecular levels, gene expression to behavior. In later chapters, we shall introduce
genetic correlational analysis and complex traits analysis, multiple gene techniques,
and techniques that place single genes under scrutiny through amplification and
nullification.
3.2 FUNDAMENTAL QUESTIONS. WHAT

CAUSES WHAT?
Periodically, the popular media may report that some researcher has identified THE
GENE for X (fill in the X for yourself). While this may be true for relatively rare,
single-gene disorders, such as cystic fibrosis or Huntingtons disease, most of
the neurobehavioral maladies result from the action of several genes, acting in
concert with the environment. Human neurobehavioral diseases such as drug abuse,
alcoholism, depression, and most others fall into the latter category. Moreover,
multiple genes may operate independently on a particular phenotype or through
networksanother of Waddingtons insights.10
3.3 PHENOTYPES AND PHENOTYPIC SELECTION

Phenotypes are any measurable characteristic of an organism. How phenotypes are
operationally defined, however, can be a daunting task. Take, for example, aggres-
sion. Aggression is one of the favorite subjects of animal researchers. We are
Some Basics, Mendelian Traits, Polygenic Traits, Complex Traits 31
interested in the genetic and environmental underpinnings, yet aggression per se is

not a phenotype. It is a low-level construct that collects several different behaviors
and describes the relationship between pairs (at least) of individuals. The individual
behaviors that constitute the larger construct, however, may be suitable phenotypes.
A commonly used measure of aggression is the latency to attack. This is the time
that it takes a resident mouse to lunge at an intruder mouse after the experimenter
has introduced the intruder to the resident mouses home cage. Another commonly
used measure is the number of attacks. The reader can see that there is an inherent
danger in naming social behavior phenotypes, i.e., the behavior of the intruder may
indeed influence the behavior of the subject and thus the phenotype is partially
confounded and less amenable to genetic analysis.
3.4 GENOTYPEPHENOTYPE RELATIONSHIPS

Sometimes, observing the phenotype reveals all about the genotype concerning that
characteristic. Take, for example, the domestic cat with a calico coat color. The
calico consists of orange mixed with black, sometimes in distinct spots, sometimes
combined with white and sometimes in a complete mlange (tortoise shell). Black
and orange alleles of the same gene, located on the X chromosome are codominant
and both are expressed. So, by observing calico, one knows that the cat carries one
allele of each. One can also discern that barring the XXY karyotype of Kleinfelters
condition, the cat is also female (the Y chromosome does not have a comparable
locus, so the tomcat coat color is dependent upon the allele contained on its sole X
chromosome). It is the rare case, however, that the behavioral phenotype reveals all
about the genotype. Certain kinds of movement may reveal an individual afflicted
with Huntingtons disease, although there are possible conditions, other than the
disease, that may produce similar movement patterns. The lesson here is that most
of the neurobehavioral phenotypes of interest in health and disease are influenced
by multiple genes, combining algebraically, some increasing the phenotype, and
others decreasing, so that merely observing an animals or humans behavior or
neurobiological functions tells us rather little about the genotype. There are other
methods, described throughout this volume, that must be employed for that job.
3.5 MENDELS LAWS

Gregor Mendels major contribution to understanding heredity and genetics is that
he proposed that inheritance involves acquisition of matter from parents and that
the matter could come in different forms. He then proposed two main laws that form
the basis of particulate inheritance and basic mathematical modelswhat we call
today classical genetics.
3.5.1 LAW OF SEGREGATION

Mendel proposed that trait characteristics are determined by inherited factors (par-
ticulate inheritance, if you will) that, for each trait, there is an alternative form. His
examples from the garden included green vs. yellow peas, smooth vs. round peas,
short vs. long pea vine length, etc. Moreover, these factors come in pairs that may
be of the same type or different types (what we call homogeneous or heterogeneous).
Finally, during the time of gamete formation, these factors separate to be recombined
with like or different types at fertilization. Some of the factor types (what we call
alleles) would dominate over the alternative form, so that if an individual plant
inherited the factor for wrinkled peas from one parent and the factor for smooth
peas from the other parent, the observed type would be smooth. Mendel then termed
smooth the dominant trait. From these observations, Mendel produced the familiar
mathematical rules to predict outcomes when homogeneous or heterogeneous
matings were performed. Thus, mating pure wrinkled with pure smooth would, in
the first filial generation, produce all smooth. Mating these first filial (F1) plants
among one another would produce a mix of smooth and wrinkled in the ratio of 3
smooth to 1 wrinkled for the phenotype, whereas the genotypes would be
1(SS):2(SW):1(WW).
3.5.2 LAW OF INDEPENDENT ASSORTMENT

Mendel knew nothing about chromosomes or genes, so it is understandable that he
proposed that all characteristics of an organism were inherited independently, i.e.,
as if there were physically separated particles in the gametes. The fact that many
traits are linked together is one of the great contributions to genetic analysis and, in
fact, is called linkage analysis.
3.6 LINKAGE
During gametogenesis, or more accurately, meiosis, both of the chromosome
homologs replicate to give 4 copies of the chromosomethe tetrad. The inner
homologous chromosomes can break from physical pressure and reassemble such
that a piece of a maternal homolog, containing several genes, splices with the
remainder of the paternal homolog and vice versa. So, genes for various character-
istics can be linked. The lineup of genes (alleles) on a single strand homolog is
referred to as the haplotype. Now, under certain conditions, a phenotype may inform
the researcher about the genetics of a linked phenotype.
3.7 MENDELIAN TRAITS

3.7.1 RECESSIVELY TRANSMITTED TRAITS
There are a few human traits that follow apparent single-gene segregation. Sickle cell
anemia (named for the characteristic shape of erythrocytes) is well known among
individuals of recent sub-Saharan African descent. It is a recessive condition, mean-
ing that both parents must contribute the sickling allele (substitution of valine for
glutamic acid at position 6 of the hemoglobin beta-polypeptide). An individual
carrying both alleles for sickling has about a 50% chance of living to age 20 without
therapy. Remarkably, those individuals carrying sickling and normal alleles (het-
erozygotes) are resistant to malaria, one of the big killers in Africa and other parts
of the world. Thus, the heterozygous configuration confers fitnessthe ability to

pass ones genes on and helps to explain why this allele is preserved in the population.
In neurobehavioral genetics, there are a number of associated recessive traits,
although most are rare. One of the best known of these is phenylketonuria or PKU.
An estimated 1 in 10,000 individuals among Caucasians (especially from northern
Europe) is afflicted and an estimated 1 in 50 individuals is a carrier of the defective
allele(s). Affected individuals lack the ability biochemically to convert the amino
acid, phenylalanine to tyrosine. Far from being simple, there are numerous mutations
in the gene for phenylalanine hydroxylase (PAH, the critical enzyme) that can render
the entire enzyme inactive or with substantially reduced activity. As a result, instead
of tyrosine (the base amino acid for the catecholamines) the neurotoxin, phenylpyru-
vic acid is formed and causes severe neurological damage and consequent mental
retardation if left untreated. Not surprisingly, the affliction ranges from mild to
severe, depending on the type of mutation, but all forms can be treated by eliminating
phenylalanine from the diet. The diet is essential during most of development from
childhood to adulthood, and there is some indication that it can be abandoned in
adulthood, however the timing is in dispute. A recent study reports good results
from children staying on the diet for 10 years.3 Other forms of PKU caused by a
deficiency with the PAH cofactors (pterins) are not so easily treated.
3.7.2 DOMINANT TRAITS

These diseases are less common even than recessive trait disorders. Two examples
of lethal dominant genetic diseases are Huntingtons disease and familial fatal
insomnia. Both require only one copy of the allele to produce the disease thus making
the probability of contracting the disease when one parent is afflicted 50% compared
to 25% when both parents are carriers as is seen with recessively transmitted traits.
Most lethal dominant disorders are expressed during embryogenesis; however, in
the case of Huntingtons disease and familial fatal insomnia, the symptoms of the
disease are manifest after the affected individual has attained reproductive age. In
1983, the genetic marker associated with Huntingtons disease was located on human
chromosome 4 and in 1993, the gene together with its polymorphisms was named
IT15.8 The important polymorphisms consist of cytosine-adenine-guanadine (CAG)
repeating sequences with unaffected individuals having 30 or fewer repeats and
affected individuals showing 36 to more than 100 repeats. The number of repeats
correlates with disease onset and severity such that the longer the repeat sequence,
the earlier age of the onset of the disease.
3.8 COMPLEX TRAITS AND ADDITIVE

GENETIC EFFECTS
Most of the phenotypes of interest in the study of brainbehavior relations are called
complex traits. Complex traits analysis of these phenotypes implies that they are
influenced by multiple genes, the environment and their interactions. As evidence
for this, when we examine their frequency distributions, we often observe that
the shape of the distribution is Gaussian, whereas with single-gene influenced traits,
Response
0 1 2 3 . . . . . . . . . . . . . . . . . . . . . . . . .n
Generation Number
FIGURE 3.1 Theoretical representation of progression of phenotype-based bi-directional
selection.
we would expect to see a bimodal distribution. More than 30 years ago, McClearn
and Kakihana produced lines of mice selected for high sensitivity and low sensitivity
to the hypnotic effects of ethanol.7 These lines were named long-sleep and short-
sleep, respectively. Figure 3.1, is a theoretical representation of how bi-directional
selection of a complex trait should progress across generations. Thus alleles that
confer increased or decreased sensitivity would accumulate across generations. At
the end of more than 40 generations of bi-directional selection, it was concluded
that at least four polymorphic genes are involved in sensitivity to the hypnotic effects
of ethanol.5 Thus the genes work together in additive fashion in their influence on
the phenotype.
3.9 INTERACTIONS
Genes do not operate in isolation. As Waddington9 proposed, how genes operate in
embryogenesis depends in large part on the extra- (or epi-) genetic milieu, e.g.,
nutrients, toxins, etc. We also know that genes can influence one another to enhance
or suppress expression.
3.9.1 GENEENVIRONMENT INTERACTIONS

In the example of selection for hypnotic sensitivity to ethanol above, one of the
questions addressed the stability of the selected phenotype. In factand relevant to
the earlier discussion about phenotypeslong-sleep and short-sleep mice differ not
only in the time spent sleeping, but in the blood ethanol concentrations (BEC) at which
they wake up. Long sleep (LS) wake up at 200300 mg/dl BEC and short-sleep (SS)
wake up at 500600 mg/dl BEC. Thus, it is clear that the difference between the LS
and SS mice is based on differential target tissue (brain) sensitivity. Jones and
colleagues4 investigated the stability of the selected phenotypic difference and per-
formed a simple experiment. Normally, the LS and SS mice are housed in groups of
24 with sexes separated except for breeding. How mice are housedgrouped or
individuallyhas a great impact on the central nervous system,11 so Jones et al.

subjected LS and SS male mice to isolation housing following weaning at 21 days of
age until 2 months of age. The hypothesis was that isolation housing would decrease
the sensitivity of LS mice to alcohol but have no effect on the SS mice. In fact, as
predicted, the isolate-housed LS mice slept for a shorter period of time and their BEC
was on average 20% higher than their group-housed cohorts. Surprisingly, the isolated
SS mice also slept for a shorter period of time, but their BEC was not different from
the group-housed animals. The conclusions were (1) the genetic-based mechanisms
underlying brain sensitivity to ethanol were altered by isolate housing in LS but not
in SS mice and (2) the mechanisms responsible for eliminating ethanol from the blood
were changed by isolate housing in the SS but not LS mice.
In 1987, Cloninger published a seminal paper on alcohol-related disorders
typologies1 in humans. According to Cloninger, one type of alcohol-related disorder
has a genetic component but is environment limited. That is, manifest only under
highly stressful conditions such as a bad marriage, employment or other personal
difficulties.
3.9.2 GENEGENE INTERACTIONS

Gene actions can influence the actions of other genes. This effect is called epistasis.
Some of the most salient instances of these interactions can be seen in animals that
have been genetically modified, especially mice that have had genes inactivated or
knocked out. The phenotype observed in such animals is dependent upon the
genetic background on which the mutation is placed.2
3.10 SUMMARY
Basic knowledge about what genes are, what they do, and how they influence
neurobiology and behavior has grown exponentially over the past few years. Com-
plex traits analysis is the new look in genetics and holds great promise for helping
us to see brainbehavior relationships from a systems approach.
REFERENCES
1. Cloninger, C.R. Neurogenetic adaptive mechanisms in alcoholism, Science, 1987,
236:410416.
2. Crawley, J.N. Unusual behavioral phenotypes of inbred mouse strains, Trends.
Neurosci., 1996, 19:181182.
3. Griffiths, P., Paterson, L., and Harvie, A. Neuropsychological effects of subsequent
exposure to phenylalanine in adolescents and young adults with early-treated phe-
nylketonuria, J. Intellect. Disabil. Res., 1995, 39:365372.
4. Jones, B.C., Connell, J.M. and Erwin V.G. Isolate housing alters ethanol sensitivity
in long-sleep and short-sleep mice, Pharmacol. Biochem. Behav., 1990, 35:469472.
5. Markel, P.D., Fulker, D.W., Bennett, B., Corley, R.P., DeFries, J.C., Erwin, V.G., and
Johnson, T.E. Quantitative trait loci for ethanol sensitivity in the LS x SS recombinant
inbred strains: interval mapping, Behav. Genet., 1996, 26:447458.
6. McClearn, G.E. Combining molecular and quantitative genetics: decomposing the

architecture of lifespan development. In: Understanding Human Development: Dia-
logues with Lifespan Psychology. U. M. Staudinger, and Lindenberger, U., eds.,
Springer, New York, 2003, 376.
7. McClearn, G.E., and Kakihana, R. Selective breeding for ethanol sensitivity: Short-
sleep and long-sleep mice. In: Development of Animal Models as Pharmacogenetic
Tools. G. E. McClearn, R.A. Deitrich, and V.G. Erwin, eds., 1981, DHHS Publication
No. (ADM) 811133. Washington, DC: U.S. Government Printing Office.
8. The Huntingtons Disease Collaborative Research Group, A novel gene containing a
trinucleotide repeat that is expanded and unstable on Huntingtons disease chromo-
somes, Cell, 1993, 72:971983.
9. Waddington, C.H. The genetic control of development, Symp. Soc. Exp. Biol., 1949,
145:145154.
10. Waddington, C.H. Principles of Development and Differentiation. The Macmillan
Company, New York, 1966, 14, ff.
11. Wilmot, C.A., VanderWende, C. and Spoerlein, M.T. Behavioral and biochemical
studies of dopamine receptor sensitivity in differentially housed mice. Psychophar-
macology, 1986, 89:364369.
4 An Introduction to
Quantitative Genetics
Wim E. Crusio
CONTENTS
Summary .................................................................................................................. 37
4.1 Introduction .................................................................................................... 38
4.2 Phylogenetic Aspects of Causation ............................................................... 38
4.2.1 Natural Selection and Genetic Architecture...................................... 38
4.2.2 Theoretical Background..................................................................... 39
4.2.3 Examples ............................................................................................ 43
4.3 Phenogenetic Aspects of Causation............................................................... 43
4.3.1 The Correlational Approach: Brain Lesions
and the Locality Assumption............................................................. 43
4.3.2 Genetic Correlations .......................................................................... 44
4.3.3 Theoretical Background..................................................................... 45
4.3.4 Examples ............................................................................................ 48
4.3.4.1 Rearing Behavior in an Open Field and Hippocampal
Mossy Fibers in Mice......................................................... 48
4.3.4.2 Nerve Conduction Velocity and IQ in Man....................... 48
4.3.4.3 Localization of QTL........................................................... 49
4.4 Some Common Misapplications.................................................................... 49
4.5 Conclusion...................................................................................................... 51
Acknowledgments.................................................................................................... 52
References................................................................................................................ 52
Appendix.................................................................................................................. 54
SUMMARY
This chapter provides a brief overview of quantitative-genetic theory. Quantitative
genetics provides important tools to help elucidate the genetic underpinnings of
behavioral and neural phenotypes. This information can then provide substantial
insights into the previous evolutionary history of a phenotype, as well as into
brainbehavior relationships.
The most often employed crossbreeding designs are the classical Mendelian
cross and the diallel cross. The information rendered by the former is limited to the
37
two parental strains used and cannot be broadly generalized. The principal usefulness
of this design is for testing whether a given phenotype is influenced by either one
gene or by more genes. The diallel cross renders more generalizable information,
the more so if many different strains are used, such as estimates of genetic correla-
tions. To estimate the latter, correlations between inbred strain means may provide
a helpful shortcut.
Some commonly encountered mistakes in the interpretation of the results of
quantitative-genetic studies are presented and explained.
4.1 INTRODUCTION
Behavior is an animals way of interacting with its environment and is therefore a
prime target for natural selection. Furthermore, as behavior is the output of an animals
nervous system, this indirectly leads to selection pressures on neuronal structures. In
consequence, each species behavior and nervous system have co-evolved in the
context of their natural habitat and can be properly comprehended only when their
interrelationships are regarded against that background.4 To arrive at a profound under-
standing of neurobehavioral traits, one will therefore have to consider problems of
causation. Van Abeelen39 distinguished between the phenogenetic and the phylogenetic
aspects of causation. Both aspects deal with the genetic correlates of neurobehavioral
traits, the first in a gene-physiological, the second in an evolutionary sense. In other
words, neurobehavioral geneticists try to uncover the physiological pathways under-
lying the expression of a trait and to evaluate its adaptive value for the organism. As
I have shown before,9 quantitative-genetic methods may be employed with profit to
address problems related to both aspects of causation.
4.2 PHYLOGENETIC ASPECTS OF CAUSATION

4.2.1 NATURAL SELECTION AND GENETIC ARCHITECTURE
Selection pressures mold a populations genetic makeup, which subsequently will
show traces of this past selection. Therefore, information about the genetic archi-
tecture of neurobehavioral traits might permit us to deduce the probable evolutionary
history of these traits.2 With genetic architecture we mean all information pertaining
to the effects of genes influencing a particular phenotype in a given population at a
given time, including information concerning the presence and size of certain genetic
effects, the number of genetic units involved, etc. Generally, however, information
about the presence and nature of dominance suffices.7
With very few exceptions, natural, nonpathological variation in neurobehavioral
phenotypes is polygenically regulated. If dominance is present, we may envisage
two different situations: either (uni)directional or ambidirectional dominance. In the
first case, dominance acts in the same direction for all genes involved (e.g., for high
expression of the trait), whereas in the latter case it acts in one direction for some
genes and in the opposite one for others. In its most extreme form, ambidirectional
dominance may lead to situations where an F1 hybrid is exactly intermediate between
its parents, despite the presence of strong dominance effects.
An Introduction to Quantitative Genetics 39
Mather30 distinguished between three kinds of selection: stabilizing, directional,

and disruptive. Stabilizing selection favors intermediate expression of the phenotype,
directional selection favors either high or low expression, whereas with disruptive
selection more than one phenotypic optimum exists. Disruptive selection will lead
to di- or polymorphisms, which may be in stable equilibrium or may even lead to
breeding isolation and incipient speciation.37 The most common example of a dimor-
phism is the existence of two sexes, whereas a possible example of speciation as a
consequence of disruptive selection is the explosive adaptive radiation and speciation
found among fish species belonging to the family Cichlidae in the great East African
lakes.20
Selection acts in favor of those genotypes that not only produce the phenotype
selected for, but also are capable of producing progeny that differs little from this
phenotype. In the end, this results in a population whose mean practically coincides
with the optimum. Stabilizing and directional selection have therefore predictably
different consequences for the genetic architecture of a trait. Genes for which a
dominant allele produces a phenotypic expression opposite to the favored direction
will become fixed very rapidly for the recessive allele under directional selection.
A similar rapid fixation will then occur for genes for which dominance is absent,
i.e., where the heterozygote is intermediate between the two homozygotes. In con-
trast, selection against recessive alleles is much slower. The result will be that after
only a relatively short period of directional selection, the first two types of genes
will not contribute to the genetic variation within the population anymore. Those
genes where the dominant allele produces the favored phenotypical expression will
remain genetically polymorphic for a much longer time, conserving genetic variance.
Thus, directional selection leads to situations where dominance is directional, in the
same direction as the selection. Stabilizing selection leads to situations where dom-
inance is either absent or ambidirectional. Furthermore, directional selection gener-
ally results in lower levels of genetic variation than ambidirectional selection does.30
The genetic architecture of a trait may be uncovered by using appropriate quantitative
genetic methods.
4.2.2 THEORETICAL BACKGROUND

At this point, a brief excursion into the field of quantitative genetics is necessary.
For the sake of simplicity, considerations of possible interactions between genes
(epistatic interactions) will be omitted from the present treatment. Similarly, we will
assume that sex-linked genes as well as pre-and postnatal effects are absent. Pertinent
references and more technical details for cases where these simplifying assumptions
do not hold true may be found in Crusio.7
Classical Mendelian analysis studies characters influenced by a limited number
of genes, two or three at the most, that are easy to separate into discrete phenotypic
classes. Many characters, however, show continuous gradations of expression and
are not separable into discrete classes (so-called quantitative traits). The simulta-
neous actions of a large number of genes (polygenes), combined with phenotypic
deviations caused by variations in the environment (environmental variance, VE),
may explain such patterns of phenotypic variation.19 The genetic contribution to a
phenotype can then be divided into two main sources: additive-genetic effects and
dominance deviations. In the case of one single gene, with alleles A and a, we may
denote the phenotypical values of the three possible genotypes as follows*:
AA = m + da Aa = m + ha aa = m da (4.1)
The parameter da is used to represent half the difference between the homozy-
gotes, ha designates the deviation of the heterozygote from the midparental point m.
Note that in quantitative genetics capital letters are used to indicate increaser alleles,
which are not necessarily also the dominant ones. Hence, da is positive by definition,
whereas ha may attain all possible values. If we now consider two inbred strains A
and B in a situation where many genes affect the phenotype, we may denote the
average phenotype of strain A by
m + S(d+) + S(d) (4.2)
(shortened to m + [d] for ease of representation), where S(d+) indicates the summed
effects of those genes that are represented by their increaser alleles and S(d) indicates
the same for decreaser alleles. Parameter m is a constant, reflecting the average
environmental effects both strains have in common as well as genetic effects at loci
where the strains are fixed for the same alleles. The average phenotype of strain B
will then equal m [d]. Similarly, the phenotypic value of an F1 hybrid between A
and B may be written as
m + S(h+) + S(h) (4.3)
(shortened to m + [h]). It must be noted that [h] is the sum of the dominance
deviations of many genes. If these effects are balanced in opposite directions, [h]
can be low or zero, even with dominance present. The same applies to [d], of course.
Because variations in a phenotype can be thought of as the summed effects of
variations in genotype and environment, plus the interaction and covariation between
these two factors, we may express the phenotypic variance P of a population as:
P = G + E + G*E + 2cov(g,e) (4.4)
In the controlled situation of animal experiments in the laboratory (but not in

the field) the covariance between genotype (g) and environment (e) can be mini-
mized. Further, the absence of genotypeenvironment interaction means that all
individuals, regardless of phenotype, will have similar sensitivities to environmental
variations. As a result, genetically homogeneous groups (for example, inbred strains
and their F1 hybrids) should exhibit similar variances. If such is not the case, the
effects of (G*E) may often be removed by choosing an appropriate measurement
scale,6 leaving
* In this chapter, we will follow the notation of Mather and Jinks.31 In the appendix I present a table
comparing this notation with the one followed by Falconer and Mackay.16
P=G+E (4.5)
The genetic component of the variance (G) can, of course, be divided into
components due to additive-genetic variation (D) and dominance deviations (H).
We may demonstrate the partitioning of genetic variance into its additive-genetic
and dominance components by the example of an F2 cross between two inbred
strains. When only one gene with two alleles influences the phenotype, the expected
genetic composition of the F2 population will be AA, 25%; Aa, 50%; and aa, 25%.
From the foregoing, this leads to a phenotypic mean of
da + ha da = ha (4.6)
(m is set at zero by a simple shift of the measurement scale). The sum of squares
of deviations from the mean then equals
da2 + ha2 (4.7)
In the absence of epistasis and linkage, the contribution of a number of genes

(k) to the F2 variance becomes
k k

i =1
di2 +
i =1
hi2 (4.8)
shortened to
D+H (4.9)
for ease of representation. The total phenotypical variance of an F2 is thus
VP = D + H + E (4.10)
By using groups with different genetic compositions it will be possible to obtain

estimates for the three parameters D, H, and E. From these parameters we may then
estimate the proportion of the phenotypical variance due to additive-genetic effects,
the heritability in the narrow sense
h2n = ( D)/VP (4.11)
and the proportion of the phenotypical variance due to all genetic effects, the
heritability in the broad sense
h2b = ( D + H)/VP (4.12)

When investigating populations other than an F2 between two inbred strains

(such as a diallel cross), allele frequencies need not be identical. Using u to indicate
the frequency of the increaser allele and v as the frequency of the decreaser allele
(with u = 1 v) we may amend the definitions of D and H as follows:
D=
i =1
4uividi2 (4.13)
and
H=
i =1
4uivihi2 (4.14)
(Formally, this definition of H should be called H1, to distinguish it from H2,

the other one of the two diallel forms of H.31)
It should be noted that because D and H represent summations of the squared
effects of single genes, they can only be zero if additive-genetic effects or dominance,
respectively, are absent. This is in obvious contrast to [d] and [h].
The crossbreeding designs employed most often in neurobehavioral genetic
studies are the classical Mendelian cross and the diallel cross (for other possible
designs, see Chapter 10 and Crusio7). The former consists of two inbred strains and,
at least, their F1 and F2, often supplemented with backcrosses of the F1 with both
parentals. The latter consists of a number of inbred strains (at least 3) that are crossed
in all possible combinations. The two designs render different types of information
on the genetic architecture of a trait. The classical cross permits very detailed genetic
analyses and the detection of very small genetic effects, but on a very restricted
sample of two inbred strains, only. Furthermore, it is very hard and almost always
outright impossible to distinguish between directional and ambidirectional domi-
nance when using this design, because a significant parameter [h] only indicates
that dominance is present and, at least, not completely balanced. In fact, a classical
Mendelian cross is, generally speaking, only useful if one wants to establish whether
the difference between two inbred strains is determined by either one gene or by
more genes. In contrast, the analysis of a diallel cross renders information on a larger
genetic sample and is therefore much more generalizable, but this carries a price in
that the information obtained is less detailed. A great advantage, however, is the
possibility to distinguish between ambi- and unidirectional dominance. To uncover
the genetic architecture of a trait, the diallel cross will be nearly always the design
of choice.
When results from different crosses are available, it should be realized that
comparing them is not very informative. Obviously, different results may be obtained
depending on the genetic makeup of the parental strains used, especially if the
crossbreeding design employed has a low generalizability (e.g., the classical cross).
However, such results may be combined in order to provide a more complete and
generalized picture of the genetic architecture of a trait. For instance, if one cross-
breeding experiment indicates dominance in the direction of, say, high expression
of the trait, but another cross indicates dominance in the opposite direction, then
this constitutes prima facie evidence for ambidirectional dominance. In fact, the
presence of directional dominance may only be inferred if all available evidence
indicates that dominance is acting in the same direction.
4.2.3 EXAMPLES
Crusio and van Abeelen14 addressed the question of what exactly is the adaptive
value of various mouse exploratory behaviors carried out in novel surroundings. As
one result of exploration is the collection of new, or the updating of previously
acquired, information, we argued that, if an animal enters a completely novel envi-
ronment, it is obviously of prime interest to collect as much information as possible
in a short time. On the other hand, high exploration levels will render the animal
more vulnerable to predation. Taking all together, we hypothesized an evolutionary
history of stabilizing selection for exploration. This hypothesis was subsequently
confirmed by the results of several crossbreeding experiments11,14 that revealed
genetic architectures comprising additive genetic variation and/or ambidirectional
dominance for most behaviors displayed in an open field. The above reasoning is,
of course, not specific for the mouse species. Indeed, Gerlai et al.21 found similar
genetic architectures for exploration in an open field in a diallel cross between inbred
strains of paradise fish, Macropodus operculatus.
4.3 PHENOGENETIC ASPECTS OF CAUSATION

4.3.1 THE CORRELATIONAL APPROACH: BRAIN LESIONS
AND THE LOCALITY ASSUMPTION
When a behavioral neuroscientist wants to investigate the function of some brain

system, he or she often will do so by manipulating the system in question. Brain
lesions or pharmacological interventions to impair the functioning of the structure
of interest are among the most-often-used techniques. Some time ago, Farah18
reviewed the problems connected with the use of the so-called locality assumption,
that more or less equalizes the function of an impaired structure with the defects
exhibited by the damaged brain and which is almost always invoked to interpret the
results of interventionist studies. In an elegant way, Farah18 provided evidence that
this reasoning may lead to false conclusions. An additional disadvantage of inter-
ventionist studies is related to the fact that large interindividual differences in brain
structure exist. This heritable variation of the brain is an aspect that many neurosci-
entists tend to ignore, most likely at their own peril. For example, widely divergent
behavioral effects of septal,15,17 or limbic-system lesions,1 and pharmacological inter-
ventions in the hippocampus40 have been reported in mice, depending on which
particular inbred strain was being used.
It appears that, in the field of neurobehavioral genetics, an alternative
approach that does not suffer from these drawbacks exists: using genetic methods
exploiting naturally occurring individual differences as a tool for understanding

brain function. No brain is like another and every individual behaves differently.
The assumption that there is a link between the variability of the brain and
individual talents and propensities seems quite plausible. This approach differs
from the usual one in neuropsychology in two important aspects. First, no subjects
are studied that, by accident or by design, have impaired or damaged brains.
Rather, all subjects fall within the range of normal, nonpathological variation
(provided animals carrying deleterious neurological mutations are excluded).
Second, instead of comparing a damaged group with normal controls, we study
a whole range of subjects and try to correlate variation at the behavioral level
with that at the neuronal level.
This noninvasive strategy is reminiscent of the phrenological approach propa-
gated by Franz Josef Gall (17581828); Lipp has coined the name microphrenol-
ogy for it.29 It appears that, as long as variation in one neuronal structure is
independent of that in another, there will be no need for a locality assumption to
interpret results of experiments carried out along these lines. Especially when used
in combination with methods permitting the estimation of genetic correlations,7,8
this strategy yields a very powerful approach.
4.3.2 GENETIC CORRELATIONS

A weakness inherent in correlational studies is that a phenotypical correlation
between characters does not necessarily reflect a functional relationship. On the other
hand, if two independent processes, one causing a positive relationship, the other
causing a negative relationship, act simultaneously upon two characters, the effects
may cancel each other so that no detectable correlation can emerge.
These problems can largely be avoided by looking at the genetic correlations,
that is, at correlations between the genetic effects that influence certain characters.
Such correlations are caused either by genes with pleiotropic effects or by a linkage
disequilibrium. With linkage disequilibrium we mean situations where certain allele
combinations at closely linked genes are more frequent than might be expected based
on chance. This will occur, for instance, in F2 crosses between two inbred strains
(or populations derived from them, such as Recombinant Inbred or Recombinant
Congenic Strains).
By using inbred strains that are only distantly related, the probability that a
linkage disequilibrium occurs may be minimized so that a possible genetic
correlation will most probably be caused by pleiotropy, that is, there exist one
or more genes that influence both characters simultaneously. Thus, for these
characters, at least part of the physiological pathways leading from genotype to
phenotype must be shared and a causal, perhaps also functional, relationship
must exist. It is this special property that makes the genetic-correlational
approach such a uniquely valuable addition to the behavioral neuroscientists
toolbox. It should perhaps be noted at this point (as stated by Carey5 and to be
seen easily from the equations below) that the inverse need not be true; that is,
the genetical correlation can still be low or even zero although pleiotropic genes
are present.
4.3.3 THEORETICAL BACKGROUND

In the univariate analysis, we may partition the phenotypic variation into its
components E, D, and H, the environmental, additive-genetic, and dominance con-
tributions. In the bivariate analysis, the covariation between two traits x and y is
partitioned into its equivalent components Exy , Dxy , and Hxy . We may define the latter
two parameters as:
Dxy =
i =1
4uividxidyi (4.15)
and
Hxy =
i =1
4uivihxihyi (4.16)
where dxi and dyi are the additive-genetic effects of the ith gene on characters x and
y, respectively, and hxi and hyi are the respective dominance deviations due to the ith
gene. Evidently, only genes that have effects on both of the characters x and y
contribute to the genetic covariance terms, whereas all genes that affect either x or
y contribute only to the respective genetic variance terms. Combining these compo-
nents of the covariance with the components of the variance obtained in the univariate
analyses we may estimate genetic correlations as follows:
rD = Dxy/ DxDy (4.17)
and
rH = Hxy/ HxHy (4.18)
As is the case with normal correlations, genetic correlations are bound by

1 and 1. If a genetic correlation equals unity, then all genes affecting character x
also affect character y with gene effects on both characters being completely pro-
portional. A genetic correlation will become zero only in case no gene at all affects
both characters simultaneously or in some balanced cases. For instance, if the effects
of genes on character x are uncorrelated to the effects on character y (so that some
genes influence both characters in the same direction, whereas others do so in
opposite directions5). Obviously, similar observations can be made about rE, the
correlation between environmental effects on two phenotypes x and y.
In principle, every breeding design allowing the partitioning of variation also
enables one to partition covariation. However, in practice some designs turn out to
be not very well-suited to estimating genetical correlations. Especially the classical
cross is very problematic in this respect and although many examples exist in the
literature in which authors claim to have analyzed genetic correlations with this
design, none really have done so. The problem appears to be due mainly to the
frequent occurrence of the phenomenon, first observed by Tryon,38 that the variance
of a segregating F2 population is not significantly larger than those of nonsegregating
populations (in extreme cases, it will even be smaller). Several possible explanations
have been brought forward. First, Hall23 attributed the Tryon effect to an insufficient
degree of inbreeding of Tryons selected (but not inbred) strains. Of course, this
would enlarge the genetic variation within the parental and F1 generations, but one
would still expect the F2 to have a somewhat larger variance. In addition, the Tryon
effect has since also been observed in crosses between highly inbred strains. A
second explanation was presented by Hirsch.26 He argued that most phenotypes are
influenced by more than one gene. If we take the rat, with a karyotype of 21
chromosome pairs, as an example and, for simplicity, treat these chromosomes as
major indivisible genes, one can see that this organism can produce 221 different
kinds of gametes, leading to 321 (= 1.05 1010) different possible genotypes. In
reality, this number will be even larger because chromosomes are not indivisible.
Obviously, no experiment can take from an F2 generation a sample large enough to
have all these genotypes represented and Hirsch assumed this sampling effect to
lower the observed F2 variances below expected levels. Tellegen36 quite correctly
countered that as long as the sampling from the F2 is random, an unbiased estimate
of the population variance should be obtained. In addition, it can easily be seen that
even if Hirschs reasoning were correct, the F2 variance would still be expected to
significantly exceed that of nonsegregating generations, being the sum of environ-
mentally induced variation and, in his reasoning, at least some genetic variation.
Bruell3 had observed that the amount of variance caused by segregation in the
F2 increases if gene effects are larger and decreases if more genes influence
the phenotype studied (cf. Equations 4.13 and 4.14). If environmental influences
on the phenotype are large, an extremely large sample would be needed to detect
the difference in variance between the F2 and F1 populations at a sufficient level of
significance in situations with many genes and relatively small gene effects. As
sample sizes are limited by considerations of time, money, and space, while envi-
ronmental influences on behavioral characters are usually very pronounced, one
should normally expect F2 variances not to differ significantly from F1 variances.
Due to sampling error, they may then even be smaller, although usually not signif-
icantly so. Homeostatic processes may be responsible if the latter situation occurs.27
The problem therefore boils down to one of statistical power. It should be recalled
here that larger sample sizes are needed to obtain accurate estimates of the variance
of a population than for estimating its mean. By analogy, this also goes for covari-
ances. In sum, unless rather huge sample sizes are used, classical crosses will gen-
erally lack the statistical power needed to accurately estimate genetical correlations.
Another reason that the classical cross is less suited for bi- and multivariate studies
is the fact that results are not generalizable, but based on a restricted sample of two
inbred strains only: even if genetic correlations are estimated correctly with this design
(which happens only rarely; see Hayman24 for appropriate statistical methods), there
exists a non-negligible probability that they would be due to a linkage disequilibrium
instead of pleiotropy. Of course, it is exactly the latter property that researchers employ
when localizing genes. Note, however, that this is a special case in which one
character, the molecular marker, is completely determined by the genotype (see also
the following section).
A more suitable method is the diallel cross, for which a bivariate extension is
available,8 whereas an interesting shortcut is offered by using a panel of inbred
strains. Correlations between strain means either permit the estimation of additive-
genetic correlations,25 or provide a direct lower-bound estimate of additive-genetic
correlations (if the traits to be correlated have been measured in different individuals
from these strains).
From the foregoing, it may easily be seen that the variance of the means of a
set of inbred strains equals
D + E/n (4.19)
where n is the harmonic mean of the number of subjects per strain. The covariance
between the means obtained for two characters x and y can then be expressed as
Dxy + Exy/n (4.20a)
in case both characters are being measured on the same individuals or as
Dxy (4.20b)
in case characters x and y are being measured on different individuals from the same
strains. The correlation between the strain means in these two situations will now
equal
(Dxy + Exy/n)/ (Dx + Ex /n )(Dy + Ey /n ) (4.21a)
or
Dxy/ (Dx + Ex /n )(Dy + Ey /n ) (4.21b)
respectively. Especially if environmental effects are small and large numbers of

subjects are being used, the correlation between inbred strain means will approach
the genetical correlation. In addition, it can easily be seen that Equation 4.21b will
always render a lower-bound estimate of the genetical correlation, even if n is small
or if environmental effects are large. It should be realized that Equation 4.21a may
render a significant correlation even in the complete absence of any genetical effects.
In the latter case, Equation 4.21a reduces to the environmental correlation. By using
the within-strain variation and covariation as estimates of the environmental vari-
ances and covariances, respectively, Equations 4.19 and 4.20 render unbiased esti-
mates of environmental and genetical correlations for both cases.25
Recombinant inbred strains (RISs) have sometimes also been used to estimate
genetic correlations between phenotypes, using the above-described methods for
correlations using ordinary inbred strains. It should be realized, however, that there
is a considerable risk that any genetic correlations thus found will be due to a linkage
disequilibrium because RISs have been derived from an F2 between two inbred
strains. As was the case with the classical cross itself, this property is of course of
interest for researchers hoping to localize quantitative trait loci (QTL; but see Section
4.3.4.3). In fact, a genetic correlation obtained with RISs implies that both characters
under study are influenced by linked genes, i.e., that map to the same chromosomal
segments, at least in part. Only if previous evidence of the existence of a genetic
correlation has been obtained with other methods (such as a screen of normal,
unrelated inbred strains), can a genetic correlation obtained from an RI study be
considered evidence of the localization of a QTL caused by a common gene.
Except in the case of correlations between inbred strain means, testing the
significance of genetic correlations is often problematic and the power of available
tests is not well known. Fortunately, when the environmental and genetic correlations
are used as input for further, multivariate, analyses, the possible significance or lack
thereof of an individual correlation is no longer very important.
4.3.4 EXAMPLES
4.3.4.1 Rearing Behavior in an Open Field and Hippocampal
Mossy Fibers in Mice
Crusio et al.12 carried out a diallel cross study in which 5 different inbred strains
were crossed in all possible ways. In 150 male mice from the 25 resulting crosses
they measured the rearing-up frequency during a 20 min session in an open field
and the extent of the hippocampal intra- and infrapyramidal mossy fiber (IIPMF)
projection. They obtained a marginally significant phenotypical correlation of 0.138
(df = 148, 0.05 < P < 0.10). Ordinarily, one would take such a result as evidence
that variation in the size of the IIPMF is not related to behavioral variation. However,
a quantitative-genetic partitioning of the covariation showed that the genetic corre-
lation was quite sizable: 0.479. The low phenotypical correlation was explained by
the modest heritability for rearing (0.25 vs. 0.53 for the IIPMF10) and by the fact
that the (low) environmental correlation had a sign opposite to that of the genetic
correlation. The genetic relationship between rearing, on the one hand, and the
IIPMF, on the other hand, was confirmed by the finding that a line selected for high
rearing frequency had larger IIPMF projections than a line selected for low rearing
frequencies,13 exactly as would be predicted from a positive genetic correlation
between these characters. From these data, it was concluded that the IIPMF plays
an important role in the regulation of open-field rearing.9
4.3.4.2 Nerve Conduction Velocity and IQ in Man
Lately, human behavior geneticists have also started to use genetic correlations to
uncover brainbehavior relationships, especially the very active group around Dorret
Boomsma at the Free University of Amsterdam (the Netherlands). In a recent
experiment they used the twin method (see Chapter 12) to examine the possible
existence of a genetic correlation between speed-of-information processing (SIP)
and IQ.35 It was postulated that SIP, as derived from reaction times on experimental
tasks, measures the efficiency with which subjects can perform basic cognitive
operations underlying a wide range of intellectual abilities. Phenotypic correlations
generally range from 0.2 to 0.4. Rijsdijk et al.35 showed that genetic correlations
also fell in this range at ages 16 and 18 years, whereas environmental correlations
were essentially zero. A common, heritable biological basis underlying the SIPIQ
relationship is thus very probable.
4.3.4.3 Localization of QTL
Currently, RISs are widely used as a tool to localize QTL. Unfortunately, problems
of statistical power (often also due to multiple testing) lead to many false positives
and negatives, as illustrated by the studies of Mathis et al.32 and Gershenfeld et al.22
In the first study, a number of QTLs associated with open-field behavior were
identified using a large set of RIS between the inbred mouse strains C57BL/6J
and A/J. In the second study, an F2 generation between these same inbred strains
was studied, again leading to the identification of a number of QTLs for the same
behavioral phenotypes. None of the QTL found was common to both studies,
illustrating the problem of both types of statistical errors inherent with the use of
recombinant inbred strains. Nevertheless, there have been a few instances where
QTL have been replicated within and across laboratories.28,33 Recently, recombinant
strain sets have become much larger through the development of new, large sets as
well as of additional strains for existing sets. This development may expect to lead
to a vast increase of power and a sizeable reduction of type I and type II errors.
4.4 SOME COMMON MISAPPLICATIONS

A final point of caution is warranted here. Quantitative-genetic methods are not only
used, but unfortunately, also regularly abused. One problematic issue already
addressed above concerns whether information obtained at one level (on the com-
ponents of means, say) can render information about another level (on the compo-
nents of variance or covariance, for instance). This is sometimes the case, sometimes
not. A few other frequently made mistakes are mentioned here.
1. IF SOME GENETIC EFFECTS ARE FOUND FOR ONE CHARACTER BUT NOT FOR ANOTHER,
THIS IMPLIES THAT THESE CHARACTERS ARE INFLUENCED BY DIFFERENT GENES.
WRONG! THE EQUATIONS GIVEN ABOVE SHOULD ALREADY MAKE IT ABUNDANTLY
CLEAR THAT THIS IS NOT TRUE. AN EXAMPLE MAY ILLUSTRATE THIS.
Albinism in mice, for instance, is a character that is completely recessive as far as coat
color is concerned. A quantitative-genetic analysis would indicate the significant pres-
ence of both d and h of equal size, in such a way that the heterozygote would completely
resemble the non-albino homozygote. However, if we now would perform a quantita-
tive-genetic analysis of the phenotype activity of the enzyme tyrosinase, we would
find that dominance is completely absent for this character, the heterozygote being
completely intermediate between the two homozygotes. Still, as we know very well,
only one and the same gene is involved here, which acts as a recessive on the level of
coat color, but shows intermediate inheritance on the level of the activity of the
responsible enzyme.
In fact, only the presence of a genetic correlation between two phenotypes provides
evidence that (a) gene(s) is (are) simultaneously influencing both. As was pointed out
above, the reverse need not be true.
2. IF TWO CHARACTERS ARE CORRELATED BETWEEN TWO PARENTAL STRAINS BUT NOT
IN THEIR F2 , THEY SEGREGATE INDEPENDENTLY AND, HENCE, ARE INFLUENCED BY
DIFFERENT GENES (IN OTHER WORDS, THERE IS NO GENETIC CORRELATION BETWEEN
THEM). WRONG! IN FACT, THIS OBSERVATION MAY BE TRUE, BUT IN ONLY ONE
EXCEPTIONAL SITUATION: IF AT LEAST ONE OF THE CHARACTERS WE ARE DEALING
WITH (FOR INSTANCE, A MOLECULAR-GENETIC MARKER) IS COMPLETELY DETERMINED
BY THE GENOTYPE. THIS LATTER CONDITION IS CALLED COMPLETE PENETRANCE OR,
IN QUANTITATIVE-GENETIC TERMS, THE HERITABILITY IN THE BROAD SENSE IS SAID
TO EQUAL 1. FOR BEHAVIORAL AND NEURAL PHENOTYPES, THIS CONDITION ALMOST
NEVER OCCURS. A FEW FURTHER OBSERVATIONS SHOULD BE MADE.
First, in all situations (including the above one), the phenotypical correlation within
an F2 generation will be a function of the heritabilities of both characters and the sizes
and signs of the genetic and environmental correlations between them. If, to simplify
the equation, we suppose dominance effects to be absent for both characters x and y, then
rP = hxhyrD + exeyrE (4.22)
where rP is the phenotypical correlation, hx and hy are the square roots of the narrow-
sense heritabilities of characters x and y, ex and ey are the square roots of the environ-
mentalities (the proportion of the phenotypical variance due to the environment;
ex2 = 1 hx2), and rE is the environmental correlation.
Equation 4.22 has a number of important implications. For instance, the size and sign
of rP evidently do not render any information at all about the size and sign of rD. It
can easily be seen that a significant phenotypical correlation may even be completely
absent (rP not significantly different from zero) in case rD and rE have opposite signs
and the absolute value of hxhyrD comes close to that of exeyrE. In recent years, experi-
ments attempting to determine the locations of polygenes, or QTL, have become ever
more popular. In such experiments one character, the molecular-genetic marker whose
possible linkage to a putative QTL is being tested, will have a heritability of 1 and
zero environmentality. In such a case, Equation 4.22 reduces to
rP = hxrD (4.23)
In this case, the additive-genetic correlation rD is equivalent to the square root of the
proportion of the genetic variance explained by the QTL under study. This in turn, is
a function of the distance between the genetic marker and the QTL and the relative
effect of the QTL compared to other genes. An important implication of this equation
is that there exists an upper bound to the correlation between a behavioral or neural
phenotype, on the one hand, and a marker locus, on the other hand, even if this marker
locus would not just be linked to the hypothetical QTL but actually be identical with
it and even if there is only one single gene influencing the phenotype. In the latter case,
rD would reduce to 1 and the phenotypical correlation would be expected to equal the
heritability. Equation 4.23 explains why, especially when heritabilities are low or
numbers of QTL are large (and also because of the lack of statistical power of estimates
of variance and covariance in F2 populations referred to in Section 4.3.3), the power
to reliably detect QTL is often very low, leading to many false positives and negatives.34
Second, note that in the above erroneous statement the words between two parental
strains were used. The correlation within such strains has obviously no bearing at all
on the eventual presence or absence of genetic correlations. As all individuals within
inbred strains have the same genotype, any correlations occurring within such a strain
are of environmental origin, of course. This is not to say, of course, that at any given
point in time, two individuals belonging to the same inbred strain may not have different
profiles of gene expression. If such is the case, however, then such differences in gene
expression itself must be due to environmental influences (assuming, of course, that
both individuals are at similar stages of development).
3. TO DETERMINE THE HERITABILITY OF A CHARACTER ONE SHOULD CARRY OUT A

SELECTION STUDY.
This proposition is perhaps formally not incorrect (heritabilities can be derived from
selection studies), but it contains in fact two conceptual errors. The first one is that
there is some information to be gained from a heritability coefficient. Actually, except
as an intermediate step in estimating genetic correlations, heritabilities do not have any
intrinsic value. The only interesting facts about heritabilities are whether they differ
significantly from zero (meaning that there is significant genetic variation) or from
unity (meaning that there is significant environmental variation). There is one further
use of heritability estimates, which is that their size predicts the eventual effects of
selection pressures (whether artificial or natural16). The second conceptual error derives
from this fact: once a selection study has been carried out and selection has led to the
successful establishment of divergent lines, knowledge about heritability become more
or less useless: it is not interesting any more to determine whether selection might be
successful! Thus, there is only a single situation in which one would perform a selection
experiment to estimate heritabilities: when one wishes to estimate genetic correlations
and for some reason inbred strains are not available.
4.5 CONCLUSION
In conclusion, if the above pitfalls are avoided, quantitative-genetic experiments can
render valuable information on the genetic architecture of a trait. In addition, they can
provide information about the multivariate genetic structure of complexes of traits.
Because of this last property, quantitative genetics may serve as a valuable additional
tool in the neuroscientists arsenal and may greatly enhance our understanding of the
genetic and neural mechanisms underlying individual differences in behavior.
ACKNOWLEDGMENTS
This chapter is dedicated to the memory of my teacher, mentor, and dear friend,
Hans van Abeelen (19361998).
REFERENCES
1. AMMASSARI-TEULE M, FAGIOLI S, ROSSI-ARNAUD C: Genotype-dependent
involvement of limbic areas in spatial learning and postlesion recovery. Physiol Behav,
1992, 52:505510.
2. BROADHURST PL, JINKS JL: What genetical architecture can tell us about the
natural selection of behavioural traits. In JHF van Abeelen (Ed): The Genetics of
Behaviour. North-Holland, Amsterdam, 1974, pp 4363.
3. BRUELL JH: Dominance and segregation in the inheritance of quantitative behavior
in mice. In EL Bliss (Ed): Roots of behavior. Hoeber and Harper, New York, 1962,
pp 4867.
4. CAMHI JM: Neural mechanisms of behavior. Curr Opinion Neurobiol, 1993,
3:10111019.
5. CAREY G: Inference about genetic correlations. Behav Genet, 1988, 18:329338.
6. CRUSIO WE: HOMAL: A computer program for selecting adequate data transfor-
mations, J Hered, 1990, 81:173.
7. CRUSIO WE: Quantitative Genetics. In D Goldowitz, D Wahlsten, R Wimer (Eds):
Techniques for the Genetic Analysis of Brain and Behavior: Focus on the Mouse.
Techniques in the Behavioral and Neural Sciences, Volume 8, Elsevier, Amsterdam,
1992, pp 231250.
8. CRUSIO WE: Bi- and multivariate analyses of diallel crosses: A tool for the genetic
dissection of neurobehavioral phenotypes. Behav Genet, 1993, 23:5967.
9. CRUSIO WE: Natural selection on hippocampal circuitry underlying exploratory
behaviour in mice: Quantitative-genetic analysis. In E Alleva, A Fasolo, H-P
Lipp, L Nadel, L Ricceri (Eds): Behavioural Brain Research in Naturalistic and
Seminaturalistic Settings. NATO Advanced Study Institutes Series D, Behavioural
and Social Sciences, Kluwer Academic Press, Dordrecht, Pays-Bas, 1995,
pp 323342.
10. CRUSIO WE, GENTHNER-GRIMM G, SCHWEGLER H: A quantitative-genetic
analysis of hippocampal variation in the mouse. J Neurogenet, 1986, 3:203214.
11. CRUSIO WE, SCHWEGLER H, VAN ABEELEN JHF: Behavioral responses to
novelty and structural variation of the hippocampus in mice. I. Quantitative-genetic
analysis of behavior in the open-field. Behav Brain Res, 1989, 32:7580.
12. CRUSIO WE, SCHWEGLER H, VAN ABEELEN JHF: Behavioral responses to
novelty and structural variation of the hippocampus in mice. II. Multivariate genetic
analysis. Behav Brain Res, 1989, 32:8188.
13. CRUSIO WE, SCHWEGLER H, BRUST I, VAN ABEELEN JHF: Genetic selection
for novelty-induced rearing behavior in mice produces changes in hippocampal mossy
fiber distributions. J Neurogenet, 1989, 5:8793.
14. CRUSIO WE, VAN ABEELEN JHF: The genetic architecture of behavioural
responses to novelty in mice. Heredity, 1986, 56:5563.
15. DONOVICK PJ, BURRIGHT RG, FANELLI RJ, ENGELLENNER WJ: Septal
lesions and avoidance behavior: Genetic, neurochemical and behavioral consider-
ations. Physiol Behav, 1981, 26:495507.
16. FALCONER DS, MACKAY TFC: Introduction to Quantitative Genetics. 4th Edition.
Longman Group Ltd, Harlow, Essex, 1996.
17. FANELLI RJ, BURRIGHT RJ, DONOVICK PJ: A multivariate approach to the
analysis of genetic and septal lesion effects on maze performance in mice. Behav
Neurosci, 1983, 97:354369.
18. FARAH MJ: Neuropsychological inference with an interactive brain: A critique of
the "locality" assumption. Behav Brain Sci, 1994, 17:43104.
19. FISHER RA: The correlation between relatives on the supposition of Mendelian
inheritance. Trans R Soc Edinb, 1918, 52:399433.
20. FRYER G, ILES TD: The Cichlid Fishes of the Great Lakes of Africa: Their Biology
and Evolution. Oliver and Boyd, Edinburgh, 1972.
21. GERLAI R, CRUSIO WE, CSNYI, V: Inheritance of species-specific behaviors in
the paradise fish (Macropodus opercularis): A diallel study. Behav Genet, 1990,
20:487498.
22. GERSHENFELD HK, NEUMANN PE, MATHIS C, CRAWLEY JN, LI X, PAUL
SM: Mapping quantitative trait loci for open-field behavior in mice. Behav Genet,
1997, 27:201210.
23. HALL CS: The genetics of behavior. In SS Stevens (Ed): Handbook of Experimental
Psychology, John Wiley & Sons, New York, 1951, pp 304329.
24. HAYMAN BI: Maximum likelihood estimation of genetic components of variation.
Biometrics, 1960, 16:369381.
25. HEGMANN JP, POSSIDENTE B: Estimating genetic correlations from inbred
strains. Behav Genet, 1981, 11:103114.
26. HIRSCH J: Behavior-genetic, or "experimental," analysis: The challenge of science
versus the lure of technology. Am Psychol, 1967, 22:118130.
27. HYDE JS: Genetic homeostasis and behavior: Analysis, data, and theory. Behav
Genet, 1973, 3:233245.
28. JONES BC, TARANTINO LM, RODRIGUEZ, LA, REED CL, MCCLEARN GE,
PLOMIN R, ERWIN VG: Quantitative-trait loci analysis of cocaine-related behav-
iours and neurochemistry. Pharmacogenet, 1999, 9:607617.
29. LIPP HP, SCHWEGLER H, CRUSIO WE, WOLFER D, LEISINGER-TRIGONA
M C, HEIMRICH B, DRISCOLL P: Using genetically-defined rodent strains for the
identification of hippocampal traits relevant for two way avoidance learning: A non-
invasive approach. Experientia, 1989, 45:845859.
30. MATHER K: The Genetical Structure of Populations. Chapman & Hall, London,
1973.
31. MATHER K, JINKS JL: Biometrical Genetics, 3rd ed., Chapman & Hall, London,
1982.
32. MATHIS C, NEUMANN PE, GERSHENFELD H, PAUL SM, CRAWLEY JN:
Genetic analysis of anxiety-related behaviors and responses to benzodiazepine-related
drugs in AXB and BXA recombinant inbred mouse strains. Behav Genet, 1995,
25:557568.
33. MCCLEARN GE, TARANTINO LM, RODRIGUEZ LA, JONES BC, BLIZARD
DA, PLOMIN R: Genotypic selection provides experimental confirmation for an
alcohol consumption quantitative trait locus in mouse. Molec Psychiat, 1997,
2:486489.
34. MELCHINGER AE, UTZ FH, SCHN CC: Quantitative trait locus (QTL) mapping
using different testers and independent population samples in maize reveals low power
of QTL detection and large bias in estimates of QTL effects. Genetics, 1998,
149:383403.
35. RIJSDIJK FV, VERNON PA, BOOMSMA DI: The genetic basis of the relation
between speed-or-information-processing and IQ. Behav Brain Res, 1998, 95:7784.
36. TELLEGEN A: Note on the Tryon effect. Am Psychol, 23:585587.
37. THODAY JM: Disruptive selection. Proc R Soc Lond B, 1972, 182:109143.
38. TRYON RC: Genetic differences in maze learning ability in rats. Yrbk Natl Soc Stud
Educ, 1940, 39:111119.
39. VAN ABEELEN JHF: Ethology and the genetic foundations of animal behavior. In
JR Royce, LP Mos (Eds): Theoretical Advances in Behavior Genetics. Sijthoff &
Noordhoff, Alphen aan den Rijn, 1979, pp 101112.
40. VAN ABEELEN JHF: Genetic control of hippocampal cholinergic and dynorphiner-
gic mechanisms regulating novelty-induced exploratory behavior in house mice.
Experientia, 1989, 45:839845.
APPENDIX
Comparison of the Notation Systems Used by Mather and Jinks31 and

Falconer and Mackay16
Description of Statistic Mather and Jinks Falconer
Additive-genetic effect d a
Dominance deviation h d
Additive-genetic variance D 2VA
Dominance variance H 4VD
Environmental variance E VE
Genotypic variance G VG
Phenotypic variance P VP
Additive-genetic correlation rD rA
Dominance correlation rH rD
Environmental correlation rE rE
Phenotypic correlation rP rP
5 From QTL Detection

to Gene Identification
Marie-Pierre Moisan
CONTENTS
Summary .................................................................................................................. 55
5.1 Introduction .................................................................................................... 56
5.2 High-Resolution Mapping.............................................................................. 56
5.2.1 Congenic Strains and Substrains ....................................................... 56
5.2.2 Alternative Strategies ......................................................................... 59
5.3 Positional Cloning.......................................................................................... 62
5.3.1 Pure Positional Cloning ..................................................................... 63
5.3.2 Positional Candidate Approach ......................................................... 65
5.3.2.1 Mapping of Candidate Genes in the QTL ......................... 65
5.3.2.2 Combining QTL Mapping and Expression Profiling ........ 66
5.3.2.3 Haplotype Mapping ............................................................ 66
5.4 Conclusion...................................................................................................... 67
References................................................................................................................ 67
SUMMARY
Although quantitative trait loci (QTL) mapping analysis has become very popular,
the final goal to positionally clone and identify the relevant gene(s) remains difficult
for the genetic dissection of complex traits. This chapter reviews the different
approaches that have been used thus far. First, various breeding strategies and
haplotype mapping have been developed in order to narrow down the QTL interval.
Once high resolution mapping of the QTL is achieved, positional cloning of the
relevant gene can be performed. Until a few years ago, a contig of DNA clones
covering the QTL interval was constructed and several techniques were used in
order to extract the coding sequences of the contig. This is now obsolete in human,
mouse and rat species for which genomes are sequenced. Progress in genome
annotation and techniques based on microarrays gene expression analysis are used
to pull out candidate genes from the fine-mapped QTL interval. Finally, sequence
analysis provides candidate mutations that then need to be validated by functional
studies.
55
5.1 INTRODUCTION
Recent progress in genetic strategies and technologies has allowed for the dissection
of complex traits in mice, rats and humans. Behavior geneticists took part in this
new research field and a high number of QTL for behavioral traits has now been
reported in mice10,12,13,24,26,40,54,61,62 and in rats.23,42,43,52,57
Many of these studies are conducted in inbred animals. The advantages are
multiple: (1) the environment is controlled, thus its effect on the phenotype is
minimized, (2) problems of genetic heterogeneity (different genes causing the same
phenotype) are avoided, (3) an adequate number of progeny and type of crosses are
available permitting robust statistical testing. The first behavioral QTL mapping
studies used recombinant inbred strains (RISs). The main advantage of using these
strains is that behavioral measures are conducted on groups of animals of each strain
thus strain means and variances become the unit of analysis. This is important for
phenotypes such as behavioral traits that are easily influenced by environmental
noise. However, because only a limited number of RISs are available from the same
panel, mapping resolution is often very low (unless the QTL effect is strong);
therefore, QTL detection using RISs is usually replicated on an F2 population for
validation. Although many successful F2 and backcross QTL mapping analyses have
been reported, very few causative gene variants have been identified (see Korstanje34
for a 2002 report), as the route from QTL detection to gene identification is long
and uncertain. This review will attempt to report various strategies that have been
used thus far for complex traits with their successes and limitations and the future
prospects of this continuously evolving discipline.
5.2 HIGH-RESOLUTION MAPPING

Once a QTL is detected with a significant log of odds (LOD) score (consensus
guidelines require LOD > 4.3),35 the next step is to saturate the chromosomal area
with polymorphic DNA markers in order to narrow the confidence interval as much
as possible. However at this stage in an F2 or backcross population, the chromosomal
region containing a QTL is still very large, typically 5 to 20 cM. Therefore additional
experimental strategies must be undertaken to improve the mapping resolution before
moving to a candidate gene approach or positional cloning of the relevant gene.
5.2.1 CONGENIC STRAINS AND SUBSTRAINS

An approach that has been used for many complex traits such as diabetes, hyper-
tension epilepsy and behavior is the construction of congenic strains in which a
mapped QTL is isolated from the others present in the parental strain, and then
substrains are produced to obtain a higher resolution of its localization.
To obtain a congenic strain, DNA of a QTL region is moved from one parental
strain (donor strain) to another (recipient strain). This is achieved first by an F1
mating of the two parental strains followed by successive backcrosses to the recipient
strain (Figure 5.1). Each backcross results in a 50% loss of donor strain alleles, on
average. Thus, progressively the genome of the hybrids become of the recipient type
From QTL Detection to Gene Identification 57
recipient strain donor strain
m1 m1 M1 M1
m2 m2
X M2 M2
selection m1 M1
m2 M2
F1
M1
selection M2 N2
m1 M1
m2 M2 N3
>6
m1 M1 heterozygous congenic
m2 M2
intercross
M1 M1
homozygous congenic
M2 M2
FIGURE 5.1 Congenic strains. A QTL contained in the interval delimited by the markers
M1 and M2 is introgressed from a donor strain into a recipient strain by marker-assisted
selection. N2 is the first backcross generation.
except at the QTL because at each backcross generation animals are genotyped with
genetic markers flanking the QTL. Heterozygous individuals that have retained the
donor strain alleles (for example M1 and M2) at the QTL are selected and mated
to the recipient strain thus producing the next backcross generation. To speed up the
congenic process, individuals of each generation are also genotyped with a set of
genetic markers covering the rest of the genome. This time one looks for animals
that are homozygous for the recipient alleles at every loci except the QTL. This
strategy is referred to as marker-assisted selection (MAS). Using this strategy, full
congenics could be obtained in 4 to 6 generations of backcrosses instead of eight
backcrosses if no counter-selection on the other chromosome is done.29 Once het-
erozygous full congenics (M1m1, M2m2) are obtained, they are intercrossed in order
to obtain homozygous animals (M1M1, M2M2), theoretically one fourth of the
progeny. Finally, two homozygous congenic animals are selected, based on their
phenotypic differences with the recipient strains, and bred brothersister to fix the
donor strain alleles in the recipient strain background. Alternatively, a heterozygous
congenic male is mated to several females of the recipient strain in order to obtain
identical congenic animals to be then intercrossed. This procedure takes about two
years to complete in a mouse or rat.
Theoretically, a pure congenic strain differs from the parental strain only in the
QTL area; thus, rigorous phenotypic comparisons between congenic and parental
strains can be made such that: (1) the existence of the QTL is verified (congruent
variation in phenotype), (2) the impact of a given QTL on the phenotype can be
measured and, (3) interaction between various QTLs can be tested.
Once the existence of a QTL has been verified by the development of a congenic
strain, the QTL interval may be further refined by constructing substrains
(Figure 5.2). A congenic strain that shows the expected difference in phenotype from
the recipient strain is crossed to the latter and the offsprings are backcrossed to the
recipient strain, producing a large segregating population. Using a dense genetic
map, individuals recombinant at various places throughout the QTL region are then
selected. Again, homozygous animals for the desired crossover chromosome must
be obtained (by mating 2 heterozygotes) and these homozygotes are bred to fix the
genotype in the congenic substrain. Theoretically phenotypic analysis of these var-
ious congenic substrains should lead to a refined localization of the QTL.
There are, however, several limitations to this strategy:
1. There is always an unknown amount of unselected genome that may

remain associated within the congenic strain and bias the analysis.
2. The effects of such isolated QTL are reduced compared to their effect in
the original F2 or backcross. Indeed, the genetic background of the recipient
strain is very important and some QTL may require epistatic interactions
recombinant substrains recipient

congenic
strain strain
M1 M1 M1 M1 M1 . m1 . m1
.
M3 . m3 M3 . m3 M3 M3 . m3
. . .
. . .
M4 . m4 . m4 . m4 M4 M4 . m4
. . .
M5 M5 . m5 . m5 M5 . m5 . m5
. .
QTL
.
. . . .
M6 M6 . m6 M6 . m6 . m6 . m6
. . .
M2 M2 M2 M2 m2 . m2 . m2
Phenotype + + - + + - -
FIGURE 5.2 Fine mapping using congenic substrains. A QTL was localized between markers
M1 and M2 in a congenic strain. Phenotypic analysis of congenic substrains recombinant
within the M1-M2 interval has allowed to fine-map the QTL to the M5-M6 chromosomal
segment.
with other QTL to be expressed (e.g., Idd18, a QTL contributing to the

development of insulin-dependent diabetes mellitus in the NOD mouse,50).
In some cases, the phenotypic effects of a single QTL may be too small
to be detected (e.g., epilepsy QTL El3),25 or it is reduced to a stage that
is hardly reliable (e.g., the congenic substrain for El2 locus shows a within-
strain variance that is close to that of interstrain crosses).25,36
3. Many congenic mapping studies have led to the dissection of a QTL
(localized by segregation analysis) into several linked QTLs.50,51 Some-
times these linked QTLs have an opposite effect on the phenotype,21,63
thus increasing the difficulty of identifying the gene(s) involved.
4. A dense genetic map of polymorphic DNA markers must be available in
the QTL area in order to follow properly the introgression of the donor
strain chromosomal segment and also to construct substrains.
5. When constructing congenic substrains, the number of animals that must
be screened to isolate recombinants over short distances may be exces-
sively large.
Despite these limitations, many QTLs have been fine-mapped by means of

congenic strains. As examples, congenic strains with the expected phenotypic effect
were obtained for epilepsy,36 for alcohol and drugrelated phenotypes,8,22 or for
saccharin preference.5,6,38
5.2.2 ALTERNATIVE STRATEGIES

Recombinant congenic strains (RCS) is an approach related to congenics except that
nonlinked genes controlling a complex trait are separated before their localization
by linkage analysis. Such strains are obtained by 2 generations of backcrossing
followed by 20 generations of inbreeding. In this manner, each RCS of a panel
contains, on average, 87.5% genes of a common background and 12.5% genes of a
common donor strain. For example, the Ccs series, developed by Demant and Hart18
contains 20 homozygous RCS derived from the parental strains BALB/c and STS.
Using this panel of RCS, P. Demants group was able to fine map a colon tumor
susceptibility gene (Scc1) to a 2.4 cM region thus opening the way to its positional
cloning. Scc1 was first mapped by linkage analysis to a 31 cM interval in an F2 cross
between BALB/c and a colon tumor susceptible RCS, Ccs-19. To achieve a higher
resolution, Css-19 mice were crossed again to BALB/c in order to obtain a new
segregating population. Sixty-eight F2 hybrids were genotyped with closely linked
markers covering the Scc1 interval. F2 animals who showed crossings-over between
Ccs-19 and BALB/c alleles within the critical interval were selected. These recom-
binants were then backcrossed to BALB/c producing mice that had crossings-over
only on one chromosome. These heterozygous recombinants were then tested for
colon tumor susceptibility, which is the dominant phenotype. As in congenic sub-
strains, phenotypic comparison of the various recombinants allowed for fine mapping
of Scc1. Two additional backcross generations of some recombinants were performed
to produce new types of recombinants and to finally reach the 2.4 cM resolution.41
Chromosomal substitution strains or consomic strains are strains in which an

entire chromosome is introgressed into the isogenic background of another inbred
strain using marker-assisted selection. Although such a panel takes several years to
develop, many advantages are linked to it: (1) QTL are assigned to chromosomes
by simply phenotyping the panel of strains with substituted chromosomes, (2) fine-
mapping of the QTL is obtained by crossing the consomic strain containing the QTL
with the parental recipient strain, with a much higher power than intercrossing the
2 parental strains of interest, (3) congenic strains over a narrow region are rapidly
developed from a consomic strain containing the QTL of interest and (4) replicative
and longitudinal studies are possible. A panel of 44 consomic rat strains from the
BN and Dahl salt-sensitive parental strains is under development at the Medical
College of Wisconsin.15 The Physgen project has the goal to characterize the panel
of consomic rats using 214 phenotypes specific to heart, lung, kidney, vasculature,
and blood function in response to environmental stressors (hypoxia, exercise, salt
intake) and to link these traits to the genomes of the mouse and human by compar-
ative mapping (http://pga.mcw.edu). Panels of consomic strains have also been
constructed in mice.45
Interval-specific congenic strains (ISCS) are described as a means to map QTL
to a 1 cM interval. The strategy is a shortcut to congenic mapping, in that various
recombinant animals throughout the QTL interval are selected from an F2 population
and they are then backcrossed to the recipient strain with marker-assisted selection
to retain the specific donor strain interval. Counter-selection for the rest of the
genome is done only for chromosomal intervals known to contain other interacting
QTL. This strategy was used by Bennet and colleagues9,22 thus yielding resolution
for an alcohol-related QTL interval to 3.7 cM.
Advanced intercross lines (AIL) are proposed as a resource for fine genetic mapping.
These lines are produced from a standard F2 population by randomly and sequentially
intercrossing each generation inter se (Figure 5.3). Consequently, animals are recom-
binants over shorter and shorter chromosomal segments thus increasing the mapping
resolution. This strategy has been used to finely map trypanosomiasis resistance QTL
in mice.32 Recently, a new set of recombinant inbred (RI) strains was generated from
two independent advanced intercrosses between C57BL/6J and DBA/2J mouse strains.49
Compared to standard RI strains, this AIL-based panel has approximately twice as many
recombinations as classical panels generated from an F2 population.
Heterogeneous stocks derived from inbred mouse strains have been used to fine-
map QTL. This approach takes advantage of the fact that the genealogy of the
heterogeneous stock is known, coming from an eight-way cross between eight known
inbred strains, and the high number of filial generations of one current stock (>58
generations) allows for high resolution of QTL. To fine-map an anxiety-related QTL,
Talbot et al.56 had estimated genotypephenotype associations by ANOVA in ~750
mice from the heterogeneous stock to detect QTL. Then, multiple regression was
used to establish the direction of effect of each allele and to infer the strain origins
of the QTL. Finally, haplotyping of each strain (see below) in the QTL interval
allowed to refine the QTL to 0.8 cM. The same strategy was used to fine-map
ethanol-induced locomotor activity.19 To learn about other genetic strategies, the
reader is directed to a review by A. Darvasi.17
strain A strain B
F1 F1
F2
n intercrosses
Fn
FIGURE 5.3 Advanced intercross lines (AIL). A standard F2 is produced, then each gener-
ation is intercrossed with itself. Inbreeding is avoided as much as possible.
A few years ago, a community of geneticists working on complex traits formed

the Complex Traits Consortium (CTC). Part of their mission was to discuss new
resources needed for QTL fine-mapping and identification. As a result of their
conversations, 1000 new lines of advanced, multi-parental recombinant inbred mice
are in the process of development under the operational name, Collaborative Cross.
These lines are being developed from an initial panel of eight inbred strains that
have been outcrossed for four generations and then followed by inbreeding, now in
progress. The main advantage of the resource will be the high mapping resolution
and cumulative, integrated data from the various participating laboratories.58 Funding
for this endeavor has been obtained from the Ellison Foundation, and progress
of the Collaborative Cross and other CTC projects can be monitored at
www.complextrait.org.
Should one of the aforementioned strategies work, the next step is positional
cloning of the gene(s) involved.
5.3 POSITIONAL CLONING

Once a QTL interval is isolated, if possible fixed in a homozygous fashion in a
strain, and its size sufficiently reduced, physical mapping is undertaken prior to
cloning of the relevant gene. For sequenced genomes, this step is no longer necessary,
as all the information can be found in databases. However, I shall briefly describe
the methodologies that were used before whole genome sequences were available
as they are still used in nonsequenced species of interest for complex traits.
Physical mapping implies coverage of the chromosomal area defined by linkage
studies with a set of overlapping DNA clones. In order to reduce the number of
clones to analyze, large-insert vectors are used such as YACs (yeast artificial chro-
mosomes),11 PACs (P1-derived artificial chromosomes)31 or BACs (bacterial artificial
chromosomes).55 These vectors can integrate DNA inserts of several hundred kilo-
bases (kb) (PACs and BACs) up to 2 megabases (Mb) for YACs; however, the latter
often display chimerism, that is integration of two or more noncontiguous DNA
inserts.
The first step involves screening of genomic libraries of these large-insert vectors
using the DNA markers contained in the QTL interval. At this stage, even nonpoly-
morphic markers such as sequence tag sites (STSs) can be used. STSs are short
sequences of DNA obtained by random sequencing of genomic DNA. From their
sequence, oligonucleotides are derived so that STSs are easily amplifiable by PCR
and their position can be specifically determined.46 To walk across the interval,
both ends of the first isolated clones are sequenced and new STSs are derived to
extract overlapping clones from the same libraries.
Clones are then ordered by analyzing the STS they have in common such that
at the end, a set of contiguous clones called a contig is constructed (Figure 5.4).
High throughput technology is available to rapidly construct contigs. Thus, large-
insert libraries are arranged in super pools ready for PCR analysis. The STS data-
bases and their localization are available via Internet; robots are used to replicate
these libraries rapidly and efficiently, and to extract DNA.
Contigs are constructed that cover a region as large as 10 Mb which roughly
corresponds to 510 cM depending on the species. Such big contigs are used to find
new polymorphic markers that can then be tested in congenic substrains to reduce
the QTL interval (see for example Denny et al.20). However, some chromosomal
regions may contain as much as one gene every 30 kb, thus positional cloning of
gene(s) responsible for the mutant phenotype from the contig is started once the
interval is restricted to around 1 cM.
In species for which the genome is sequenced, there is no need to construct a
contig of DNA clones in the QTL interval. New markers are found by simply
Genetic map
0.28 cM
M1 M2 M3 M4 M5 M6 M7 M8 M9
Physical map
S1 S2 S3 S4 S5 S6
M4 M5
YACs
BACs
10kb
FIGURE 5.4 Construction of a contig. M1-M9 are genetic markers. S1-S6 are sequence tag
sites.
consulting the genome sequence corresponding to the QTL interval in databases

[National Center for Biotechnology Information (NCBI) genome browser, Ensembl].
5.3.1 PURE POSITIONAL CLONING

Identification of genes responsible for a phenotype requires the isolation of expressed
sequences from cloned DNA. Again, when the genome has not been sequenced,
various techniques have been developed and good reviews on the subject are avail-
able.44,47,48 I shall describe them only briefly.
Evolutionary conservation: Because coding sequences are more conserved than
are noncoding sequences, candidate DNA fragments from the contig are hybridized
to Southern blots of genomic DNA from multiple species (also called zoo blot).
DNA fragments that hybridize to several species are selected for further investiga-
tions as these hybridizations indicate that these clones contain an evolutionary
conserved sequence.
CpG islands: This strategy is based on the observation that around 60% of genes
are associated in their 5untranslated region with short stretches of DNA with a high
frequency of the dinucleotide CpG (i.e., cytosine connected by a 3-5 phosphodiester

bond to a guanine). These unmethylated CpG-rich sequences are called CpG islands.
They are detected easily by cleavage with rare-cutting restriction enzymes of the
contig DNA clones.
cDNA selection: In this approach candidate DNA fragments are hybridized to
amplified cDNA libraries. After removal of nonspecific binding, the cDNA inserts
that hybridize specifically to the cloned genomic DNA are eluted and amplified
again by PCR. Two or three rounds of this process are performed before cloning
the selected cDNAs. Several variations in the cDNA selection protocol have been
used: the cloned genomic DNAs are either immobilized on nylon membrane or the
hybridization is performed in solution. A biotin-streptavidin system and magnetic
beads that capture the cDNA inserts have also been designed.
Exon trapping: This technique relies on splice sites flanking coding sequences and
necessary for RNA splicing. Candidate DNA fragments are cloned into a vector that
contains all the necessary sequences to produce mRNA. This recombinant vector is
then transfected for expression in cells and the mRNA produced from the clones is
identified by RT-PCR. If the DNA insert contains an exon with splice sites this exon
will be trapped into this chimeric mRNA which can then be cloned and sequenced.
Direct sequencing: The availability of high-throughput automated sequencing
machines, combined with advances in computer algorithms, has favored the devel-
opment of sequence-based methods for searching exons in cloned genomic DNA.
Shotgun sequencing approaches (that is cloning of an entire DNA population from
randomly generated fragments) from the contig clones is often used. Programs
devised to find potential coding sequences include GRAIL59 and BLAST.3,27
So far, most positionally cloned genes have been identified by a combination of
all of these techniques.48 For the positional cloning of the mouse circadian Clock
gene, a >4 Mb contig was constructed.33 The techniques that have been used to
identify the gene include: (1) the search for CpG islands, (2) cDNA selection of
suprachiasmatic nucleus cDNA using one BAC, and (3) shotgun sequencing of
subclones from two BACs of the contig. Selected candidate sequences issued from
these approaches were used as probes on Northern blots of RNA from hypothalamus
and eye from the mutant and wild type mice. Using one of the clones, RNA
differences of abundance and size were detected between the two types of mice.
The complete cDNA sequence of this gene was obtained with 11 overlapping cDNA
clones isolated with the approaches mentioned above. This gene named Clock spans
almost 100,000 bp of genomic DNA and contains 24 exons. Further evidence for
the action of Clock gene in the circadian rhythm in mutant mice was provided by a
BAC rescue experiment. A BAC clone from the critical region of the contig was
microinjected into mutant Clock mouse oocytes. Its expression in the adult transgenic
mouse permitted in vivo complementation of the loss-of-rhythm phenotype;4 i.e.,
the mice showed recovery of the circadian activity cycle. Similarly, the mouse
saccharin preference locus was positionally cloned after reducing the QTL interval
by congenics to 0.7 cM and constructing a BAC contig of the region. The 0.7 cM
QTL interval had a physical size of 194 kb and contained 12 predicted genes. Among
them, the Tas1r3 gene stood out as a good candidate as its protein is a G-protein
coupled receptor with homology to taste receptor protein, T1R1.6
5.3.2 POSITIONAL CANDIDATE APPROACH

5.3.2.1 Mapping of Candidate Genes in the QTL
In a 1995 review, Francis Collins14 made the now verified prediction that the posi-
tional candidate strategy would overtake pure positional cloning. This involves
combining fine scale mapping and survey of candidate genes within the interval. In
species not yet sequenced, efforts are underway to improve the transcript maps in
which coding sequences are localized precisely. Besides cDNA sequences already
identified and mapped, short, PCR-amplifiable partial cDNA sequences called EST
(expressed sequence tag, a subtype of STS) have been generated for this purpose
and stored in a public database (dbEST, NCBI). Fine mapping of these coding
sequences has been possible with the advent of radiation hybrid panels. Radiation
hybrid mapping is a somatic cell genetic approach in which diploid cells of the
species studied are exposed to a lethal dose of x-rays, breaking each chromosome
into small fragments. These irradiated cells are then rescued by fusion to non-
irradiated hamster cells. Each independent radiation hybrid clone retains between
15 and 20% of the irradiated cell and the retention of these random fragments of
the species chromosomes is stable without selection. The mapping strategy employs
the frequency of x-ray breakage between two markers as a statistical measure of
distance between these markers. Therefore any sequence that can be distinguished
from hamster DNA can be mapped using this approach without need for polymor-
phisms.16 These tools have been widely used in humans, mice and rats before their
genome was sequenced and remains very useful for other species.
Thus, the first step after establishing a contig, is a computer exercise in which
a query is made to select ESTs mapped in a given chromosomal subregion. These
ESTs are then placed on the contig simply by PCR amplification of the contig DNA
clones using primers specific for a given EST. A considerable amount of information
can be extracted from an EST; for example, homologies with known cDNA
sequences can be obtained (for example using the BLAST program). In turn, these
data may point to a gene already cloned or to a gene family, providing important
clues about the function of the gene and thus its relevance to the disease or mutant
phenotype. If an EST does not display homology to any known sequence, it could
be used as a hybridization probe to examine its tissue expression and to clone a
longer fragment of the cDNA in question. As outlined by Ballabio,7 the features of
candidate genesincluding sequence domains, expression pattern, imprinted expres-
sion, sequence instability (such as nucleotide triplet expansion) and developmental
expressionare compared to the disease or phenotype features in order to select
the strongest positional candidates. Finally, variants of that gene are screened for by
comparing affected individuals to controls.
An example of this approach is provided by the cloning of genes for early-onset
familial Alzheimers disease. The AD3 locus was mapped to chromosome 14 by
linkage analysis and a contig of >5 Mb was constructed around this region. Expressed
sequences encoded within the critical interval were then isolated by hybridization
of brain cDNA to the contig YAC clones. After selection based upon map location
and evolutionary conservation (see above), 19 cDNA clones were kept. The
sequences of these cDNA clones were compared in affected and normal post-mortem
brain tissue by reverse-transcriptase PCR, leading to the identification of a novel
gene, presenilin-1 (PS-1) within the AD3 locus.53 Shortly after this study, the pres-
enilin-2 gene was identified on chromosome 1. This finding was based on the
location of an EST within the candidate region, defined by linkage studies, that has
a high sequence homology with PS-1.37
5.3.2.2 Combining QTL Mapping and Expression Profiling
The availability of genome sequences, together with the advent of DNA microarrays
or chips, has favored the search for positional candidates through differential
expression analysis. The first time this strategy combining QTL mapping and large-
scale expression analysis was used, it was in a rat model of insulin-resistance.2 cDNA
molecules obtained from reverse-transcribed RNA extracted from adipose tissue (the
target tissue) of the two parental strains and a congenic strain for the detected QTL
were hybridized pairwise to a DNA chip of 10,000 arrayed rat cDNA. Around 70
genes were found to be differentially regulated between the parental and congenic
strains. Among them, there was a biological candidate that was then mapped to the
expected chromosome region. Further analysis of this candidate showed that it maps
at the peak of the QTL and displays functional sequence variation between the
parental strains. Similarly, -synuclein has been proposed as the causative gene of
a QTL associated to alcohol preference in a rat model following QTL mapping
combined to microarray expression analysis in various brain regions.39 In addition
to its strengths, this approach has several limitations: (1) the causative gene may not
necessarily vary in expression between the parental strains (or subject and control
in human studies); (2) the difference in expression may be too small to be detected
by microarray analysis. Indeed, this is of particular concern for behavioral traits in
which the brain is the target tissue and a small difference in gene expression may
have an important functional impact; (3) it is not always obvious which target tissue
to choose in which the causative gene will be differentially regulated; and (4) the
gene under investigation is not represented on the chip; this problem should be
resolved with progress in genome annotation and availability of pangenomic chips.
5.3.2.3 Haplotype Mapping
An haplotype is a set of linked alleles on one chromosome or a segment of a

chromosome. Single nucleotide polymorphisms (SNPs) are the polymorphisms most
often used to define a haplotype because they are stable over many generations
and can be traced back to a common ancestral population. Recent studies have
suggested that genomes are composed of haplotype blocks with regions of high
linkage disequilibrium (the nonrandom association of alleles) separated by a region
of a high number of recombination events. Identification of such haplotype blocks
is useful for association studies in humans by comparing the frequency of haplotypes
in affected vs. control individuals.60 In mouse and rat inbred strains, the situation is
even more favorable because all inbred strains are derived from a limited number
of ancestral populations. Thus, genomes of inbred strains appear to be haplotype
blocks mosaics derived from few original strains. This genome structure is exploited
to reduce the number of high-priority candidates genes in a QTL interval by con-
centrating on genes that are in divergent haplotype blocks between the two strains
that have served to detect the QTL. Even more rewarding is to compare haplotypes
of a QTL region detected in various crosses using different strains. Such a multiple
crossmapping strategy has been applied and combined to expression profiling and
sequence analysis, leading to the identification of a strong candidate gene for a QTL
associated with basal locomotor activity.30
5.4 CONCLUSION
Although still challenging, identification of genes underlying QTL is becoming
easier. Completion of genome sequences in human and model organisms, advances
in genomic and statistical techniques, and production of new genetic resources in
mice and rats are the main components of this progress. Nowadays, the tendency is
to set up a large-scale project integrating advanced genetic resources, in-depth
phenotyping and high-throughput genomics, such as in the Physgen project in rat
or the Collaborative Cross* in mice. Two tasks demanding a great deal of effort
remain to complete a QTL gene identification study. The first is to prove the
involvement of the candidate gene and then to find the causative mutation(s). The
best approach in model organisms is to introduce the variant allele in the control
strain by means of knock-in technology. So far, however, this is only possible in one
strain of mice and not yet in rats. Thus, as stated and detailed by others,1,28 accu-
mulation of evidence converging to a candidate gene, including functional in vitro
and in vivo studies, is generally accepted.
REFERENCES
1. Abiola O, Angel JM, Avner P, Bachmanov AA, Belknap JK, Bennett B, Blankenhorn
EP, Blizard DA, Bolivar V, Brockmann GA, Buck KJ, Bureau JF, Casley WL, Chesler
EJ, Cheverud JM, Churchill GA, Cook M, Crabbe JC, Crusio WE, Darvasi A, de
Haan G, Dermant P, Doerge RW, Elliot RW, Farber CR, Flaherty L, Flint J, Gersh-
enfeld H, Gibson JP, Gu J, Gu W, Himmelbauer H, Hitzemann R, Hsu HC, Hunter
K, Iraqi FF, Jansen RC, Johnson TE, Jones BC, Kempermann G, Lammert F, Lu L,
Manly KF, Matthews DB, Medrano JF, Mehrabian M, Mittlemann G, Mock BA,
Mogil JS, Montagutelli X, Morahan G, Mountz JD, Nagase H, Nowakowski RS,
O'Hara BF, Osadchuk AV, Paigen B, Palmer AA, Peirce JL, Pomp D, Rosemann M,
Rosen GD, Schalkwyk LC, Seltzer Z, Settle S, Shimomura K, Shou S, Sikela JM,
Siracusa LD, Spearow JL, Teuscher C, Threadgill DW, Toth LA, Toye AA, Vadasz
C, Van Zant G, Wakeland E, Williams RW, Zhang HG, Zou F. The nature and
identification of quantitative trait loci: a community's view. Nat Rev Genet 4:911916,
2003.
* At printing, the Collaborative Cross is operative and had received generous support from the Ellison
Foundation. For further information consult http://atlas.utmem.edu/pipermail/ctc/2005-October/
000284.html.
2. Aitman TJ, Glazier AM, Wallace CA, Cooper LD, Norsworthy PJ, Wahid FN, al
Majali KM, Trembling PM, Mann CJ, Shoulders CC, Graf D, St. Lezin E, Kurtz TW,
Kren V, Pravenec M, Ibrahimi A, Abumrad NA, Stanton LW, Scott J. Identification
of Cd36 (Fat) as an insulin-resistance gene causing defective fatty acid and glucose
metabolism in hypertensive rats [see comments]. Nat Genet 21:7683, 1999.
3. Altschul SF, Lipman DJ. Protein database searches for multiple alignments. Proced
Nat Acad Sci USA 87[14], 55095513, 1990.
4. Antoch MP, Song EJ, Chang AM, Vitaterna MH, Zhao Y, Wilsbacher LD, Sangoram
AM, King DP, Pinto LH, Takahashi JS. Functional identification of the mouse circa-
dian clock gene by transgenic BAC rescue. Cell 89, 655667, 1997.
5. Bachmanov AA, Li X, Li S, Neira M, Beauchamp GK, Azen EA High-resolution
genetic mapping of the sucrose octaacetate taste aversion (Soa) locus on mouse
Chromosome 6. Mamm Genome 12:695699, 2001.
6. Bachmanov AA, Li X, Reed DR, Ohmen JD, Li S, Chen Z, Tordoff MG, de Jong
PJ, Wu C, West DB, Chatterjee A, Ross DA, Beauchamp GK. Positional cloning of
the mouse saccharin preference (Sac) locus. Chem Senses 26:925933, 2001.
7. Ballabio A. The rise and fall position cloning? Nat Genet 3:277281, 1993.
8. Bennett B. Congenic strains developed for alcohol- and drug-related phenotypes.
Pharmacol Biochem Behav 67:671681, 2000.
9. Bennett B, Beeson M, Gordon L, Carosone-Link P, Johnson TE. Genetic dissection
of quantitative trait loci specifying sedative/hypnotic sensitivity to ethanol: mapping
with interval-specific congenic recombinant lines. Alcohol Clin Exp Res
26:16151624, 2002.
10. Berrettini WH, Ferraro TN, Alexander RC, Buchberg AM, Vogel WH. Quantitative
trait loci mapping of three loci controlling morphine preference using inbred mouse
strains. Nat Genet 7:5458, 1994.
11. Burke DT, Carle GF, Olson MV. Cloning of large segments of exogenous DNA into
yeast by means of articicial chromosome vectors. Science 236:806812, 1987.
12. Caldarone B, Saavedra C, Tartaglia K, Wehner JM, Dudek BC, Flaherty L. Quanti-
tative trait loci analysis affecting contextual conditioning in mice. Nat Genet
17:335337, 1997.
13. Carr LG, Foroud T, Bice P, Gobbett T, Ivashina J, Edenberg H, Lumeng L, Li TK.
A quantitative trait locus for alcohol consumption in selectively bred rat lines. Alcohol
Clin Exp Res 22:884887, 1998.
14. Collins FS. Positional cloning moves from perditional to traditional. Nat Genet
9:347350, 1995.
15. Cowley AW, Jr., Roman RJ, Jacob HJ Application of chromosomal substitution
techniques in gene-function discovery. J Physiol 554:4655, 2004.
16. Cox DR, Burmeister M, Price ER, Kim S, Myers RM. Radiation hybrid mapping:
a somatic cell genetic method for constructing high-resolution maps of mammalian
chromosomes. Science 250:245250, 1990.
17. Darvasi A. Experimental strategies for the genetic dissection of complex traits in
animal models. Nat Genet 18:1924, 1998.
18. Demant P, Hart AA. Recombinant congenic strainsa new tool for analyzing
genetic traits determined by more than one gene. Immunogenetics 24:416422,
1986.
19. Demarest K, Koyner J, McCaughran J, Jr., Cipp L, Hitzemann R. Further character-
ization and high-resolution mapping of quantitative trait loci for ethanol-induced
locomotor activity. Behav Genet 31:7991, 2001.
20. Denny P, Lord CJ, Hill NJ, Goy JV, Levy ER, Podolin PL, Peterson LB, Wicker LS,
Todd JA, Lyons PA. Mapping of the IDDM locus Idd3 to a 0.35-cM interval containing
the interleukin-2 gene. Diabetes 46:695700, 1997.
21. Dukhanina OI, Dene HY, Deng AY, Choi CR, Hoebee B, Rapp JP. Linkage map and
congenic strains to localize blood pressure QTL on rat chromosome 10. Mamm
Genome 8:229235, 1997.
22. Ehringer MA, Thompson J, Conroy O, Yang F, Hink R, Bennett B, Johnson TE,
Sikela JM. Fine mapping of polymorphic alcohol-related quantitative trait loci can-
didate genes using interval-specific congenic recombinant mice. Alcohol Clin Exp
Res 26:16031608, 2002.
23. Fernandez-Teruel A, Escorihuela RM, Gray JA, Aguilar R, Gil L, Gimenez-Llort L,
Tobena A, Bhomra A, Nicod A, Mott R, Driscoll P, Dawson GR, Flint J. A quantitative
trait locus influencing anxiety in the laboratory rat. Genome Res 12:618626, 2002.
24. Flint J, Corley R, DeFries JC, Fulker DW, Gray JA, Miller S, Collins AC. A simple
genetic basis for a complex psychological trait in laboratory mice. Science
269:14321435, 1995.
25. Frankel WN, Johnson EW, Lutz CM. Congenic strains reveal effects of the epilepsy
quantitative trait locus, El2, separate from other El loci. Mamm Genome 6:839843,
1995.
26. Gershenfeld HK, Paul SM. Mapping quantitative trait loci for fear-like behaviors in
mice. Genomics 46:18, 1997.
27. Gish W, States DJ. Identification of protein coding regions by database similarity
search. Nat Genets 3:266272, 1993.
28. Glazier AM, Nadeau JH, Aitman TJ. Finding genes that underlie complex traits.
Science 298:23452349, 2002
29. Gould KA, Dietrich WF, Borenstein N, Lander ES, Dove WF. Mom1 is a semi-
dominant modifier of intestinal adenoma size and multiplicity in Min/+ mice. Genetics
144:17691776, 1996.
30. Hitzemann R, Malmanger B, Reed C, Lawler M, Hitzemann B, Coulombe S, Buck
K, Rademacher B, Walter N, Polyakov Y, Sikela J, Gensler B, Burgers S, Williams
RW, Manly K, Flint J, Talbot C. A strategy for the integration of QTL, gene expres-
sion, and sequence analyses. Mamm Genome 14:733747, 2003.
31. Ioannou PA, Amemiya CT, Garnes J, Kroisel PM, Shizuya H, Chen C, Batzer MA,
DeJong PJ. A new bacteriophage P1-derived vector for the propagation of large human
DNA fragments. Nat Genet 6:8489, 1994.
32. Iraqi F, Clapcott SJ, Kumari P, Haley CS, Kemp SJ, Teale AJ. Fine mapping of
trypanosomiasis resistance loci in murine advanced intercross lines. Mamm Genome
11:645648, 2000.
33. King DP, Zhao Y, Sangoram AM, Wilsbacher LD, Tanaka M, Antoch MP, Steeves
TDL, Vitaterna MH, Kornhauser JM, Lowrey PL, Turek FW, Takahashi JS. Positional
cloning of the mouse circadian clock gene. Cell 89:641653, 1997.
34. Korstanje R, Paigen B. From QTL to gene: the harvest begins. Nat Genet 31:235236,
2002.
35. Lander E, Kruglyak L. Genetic dissection of complex traits: guidelines for interpreting
and reporting linkage results. Nat Genet 11:241247, 1995.
36. Legare ME, Bartlett FS, Frankel WN. A major effect QTL determined by multiple
genes in epileptic EL mice. Genome Res 10:4248, 2000.
37. Levy-Lahad E, Wasco W, Poorkaj P, Romano DM, Oshima J, Pettingell CY, Jondro
PD, Schmidt SD, Wang K, Crowley AC, Fu YH, Guenette SY, Galas D, Nemens E,
Wijsman EM, Bird TD, Schellenberg GD, Tanzi RE. Candidate gene for the chro-
mosome 1 familial Alzheimers disease locus. Science 269:973977, 1995.
38. Li X, Inoue M, Reed DR, Huque T, Puchalski RB, Tordoff MG, Ninomiya Y,
Beauchamp GK, Bachmanov AA. High-resolution genetic mapping of the saccharin
preference locus (Sac) and the putative sweet taste receptor (T1R1) gene (Gpr70) to
mouse distal Chromosome 4. Mamm Genome 12:1316, 2001.
39. Liang T, Spence J, Liu L, Strother WN, Chang HW, Ellison JA, Lumeng L, Li TK,
Foroud T, Carr LG. alpha-Synuclein maps to a quantitative trait locus for alcohol
preference and is differentially expressed in alcohol-preferring and -nonpreferring
rats. Proc Natl Acad Sci U S A 100:46904695, 2003.
40. Melo JA, Shendure J, Pociask K, Silver LM. Identification of sex-specific quantitative
trait loci controlling alcohol preference in C57BL/6 mice. Nat Genet 13:147153,
1996.
41. Moen CJ, Groot PC, Hart AA, Snoek M, Demant P. Fine mapping of colon tumor
susceptibility (Scc) genes in the mouse, different from the genes known to be somat-
ically mutated in colon. Proc Natl Acad Sci U S A 93:10821086, 1996.
42. Moisan MP, Courvoisier H, Bihoreau MT, Gauguier D, Hendley ED, Lathrop M,
James MR, Mormde P. A major quantitative trait locus influences hyperactivity in
the WKHA rat. Nat Genet 14:471473, 1996.
43. Moisan MP, Llamas B, Cook MN, Mormede P. Further dissection of a genomic locus
associated with behavioral activity in the Wistar-Kyoto hyperactive rat, an animal
model of hyperkinesis. Mol Psychiatry 8:348352, 2003.
44. Monaco AP. Isolation of genes from cloned DNA. Current opinion in Genetics and
Development 4:360365, 1994.
45. Nadeau JH, Singer JB, Matin A, Lander ES. Analysing complex genetic traits with
chromosome substitution strains. Nat Genet 24:221225, 2000.
46. Olson M, Hood L, Cantor C, Botstein D. A common language for physical mapping
of the human genome. Science 245:14341435, 1989.
47. Parimoo S, Patanjali SR, Kolluri R, Xu H, Wei H, Weissman SM. cDNA selection
and other approaches in positional cloning. Analyt Biochem 228:117, 1995.
48. Parrish JE, Nelson DL. Methods for finding genes a major rate-limiting step in
positional cloning. GATA 10:2941, 1993.
49. Peirce JL, Lu L, Gu J, Silver LM, Williams RW. A new set of BXD recombinant
inbred lines from advanced intercross populations in mice. BMC Genet 5:7, 2004.
50. Podolin PL, Denny P, Armitage N, Lord CJ, Hill NJ, Levy ER, Peterson LB, Todd
JA, Wicker LS, Lyons PA. Localization of two insulin-dependent diabetes (Idd) genes
to the Idd10 region on mouse chromosome 3. Mamma Genome 9:283286, 1998.
51. Podolin PL, Denny P, Lord CJ, Hill NJ, Todd JA, Peterson LB, Wicker LS, Lyons
PA. Congenic mapping of the insulin-dependent diabetes (Idd) gene, Idd10, localizes
two genes mediating the Idd10 effect and eliminates the candidate Fcgr1. J Immunol
159:18351843, 1997.
52. Ramos A, Moisan MP, Chaouloff F, Mormede P. Identification of female-specific
QTLs affecting an emotionality-related behavior in rats. Mol Psychiatry 4:453462,
1999.
53. Sherrington R, Rogaev EI, Liang Y, Rogaeva EA, Levesque G, Ikeda M, Chi H, Lin
C, Li G, Holman K, Tsuda T, Mar L, Foncin JF, Bruni AC, Montesi MP, Sorbi S,
Rainero I, Pinessi L, Nee L, Chumakov I, Pollen D, Brookes A, Sanseau P, Polinsky
RJ, Wasco W, DaSilva HAR, Haines JL, Pericak-Vance MA, Tanzi RE, Roses AD,
Fraser PE, Rommens JM, St George-Hyslop PH. Cloning of a gene bearing missense
mutations in early-onset familial Alzheimer's disease. Nature 375:754760, 1995.
54. Shimomura K, Low-Zeddies SS, King DP, Steeves TD, Whiteley A, Kushla J,
Zemenides PD, Lin A, Vitaterna MH, Churchill GA, Takahashi JS. Genome-wide
epistatic interaction analysis reveals complex genetic determinants of circadian behav-
ior in mice. Genome Res 11:959980, 2001.
55. Shizuya H, Birren B, Kim UJ, Mancino V, Slepak T, Tachiiri Y, Simon M. Cloning
and stable maintenance of 300-kilobase-pair fragments of human DNA in Escherichia
coli using an F-factor-based vector. Proc Nat Acad Sci USA 89:87948797, 1992.
56. Talbot CJ, Nicod A, Cherny SS, Fulker DW, Collins AC, Flint J. High-resolution
mapping of quantitative trait loci in outbred mice. Nat Genet 21:305308, 1999.
57. Terenina-Rigaldie E, Moisan MP, Colas A, Beauge F, Shah KV, Jones BC, Mormede
P. Genetics of behaviour: phenotypic and molecular study of rats derived from high-
and low-alcohol consuming lines. Pharmacogenetics 13:543554, 2003.
58. Threadgill DW, Hunter KW, Williams RW. Genetic dissection of complex and quan-
titative traits: from fantasy to reality via a community effort. Mamm Genome
13:175178, 2002.
59. Uberbacher EC, Mural RJ. Locating protein-coding regions in human DNA sequences
by a multiple sensor-neural network approach. Proc Natl Acad Sci USA
88:1126111265, 1991.
60. Wall JD, Pritchard JK. Haplotype blocks and linkage disequilibrium in the human
genome. Nat Rev Genet 4:587597, 2003.
61. Wehner JM, Radcliffe RA, Rossmann ST, Christensen SC, Rasmussen DL, Fulker
DW, Wiles M. Quantitative trait locus analysis of contextual fear conditioning in
mice. Nat Genet 17:331334, 1997.
62. Yoshikawa T, Watanabe A, Ishitsuka Y, Nakaya A, Nakatani N. Identification of
multiple genetic loci linked to the propensity for "behavioral despair" in mice.
Genome Res 12:357366, 2002.
63. Zhang QY, Dene HY, Deng AY, Garrett MR, Jacob HJ, Rapp JP. Interval mapping
and congenic strains for a blood pressure QTL on rat chromosome 13. Mamm Genome
8:636641, 1997.
6 Gene Expression
Richard A. Radcliffe
CONTENTS
Introduction.............................................................................................................. 73
6.1 The Gene ........................................................................................................ 74
6.2 Transcription: Basic Mechanism of mRNA Biogenesis ............................... 76
6.3 Regulation of Transcription ........................................................................... 80
6.4 Transcriptional Regulation through Signal Transduction ............................. 83
6.5 RNA Interference ........................................................................................... 85
6.6 Gene Silencing ............................................................................................... 86
6.7 Sources of Variation in Gene Expression...................................................... 87
6.8 Measuring Gene Expression .......................................................................... 88
References................................................................................................................ 92
INTRODUCTION
Gene expression, or transcription, is an important element of The Central Dogma,
the phrase used by Francis Crick1 to describe what was and still is thought to be the
basic and only flow of genetic information in a cell: DNA is transcribed to messenger
RNA (mRNA) which is then translated to protein. In this scheme, only nucleic acids
actually contain information that is encoded in the specific sequence of nucleotides
that make up the DNA or RNA. Cells can construct proteins from the information
contained in nucleic acids, but the transfer of information is one way; cells do not
have the capacity to reproduce nucleic acid sequence information from proteins
alone. While it is known that RNA performs more than just information functions,2
proteins are still thought to be the primary working molecules in a cell carrying out
diverse functions related to energy management, immune response, structure, trans-
port and storage, signal transduction, cell division, etc. Thus, implicit in the Central
Dogma are mechanisms for the tightly controlled regulation of mRNA and thus
protein amount, one of several mechanisms through which the cell maintains strict
control over the final activity of a protein. On a grander scale, the integrated
transcriptional activity of the entire genome (the entire DNA sequence in an organ-
ism) to a great extent determines the nature of a cell and its interactions with other
cells in an organism or with the environment.
Transcription is an essential mechanism throughout the development of a mul-
ticellular organism. The coordinate induction or repression of the thousands of genes
found in the organisms genome results in a wide array of cell phenotypes through
73
the process of cell differentiation. Mature cells, including highly specialized post-
mitotic cells such as neurons, continue to make use of transcription throughout their
life. One reason for this is that active transcription is required to replace proteins
all of which have a finite duration, at least to the extent that is known, and therefore
must be replaced throughout the life of the cell. Cells also need to adapt to a dynamic
environment necessitating variable levels of some proteins. Environmental adapta-
tion through transcription is perhaps most highly developed in neurons of the central
nervous system (CNS) in which one of the most important and interesting functions
of the CNS takes place: learning and memory.3 The CNS is constantly being bom-
barded with sensory information, some of which is critical and must be dealt with
immediately, but most of which is extraneous and can be ignored. In either case,
processes of learning and memory aid the organism in distinguishing into which
category the incoming information belongs. These processes cannot take place in
the absence of a functioning transcription apparatus.
The fundamentals of transcription are essentially the same in all eukaryotic cells,
but the timing and nature of the regulatory events that determine which genes are
actively being transcribed vary from gene to gene. This chapter describes the fun-
damentals of transcription as well as the basic principles of transcriptional regulation.
Also included is a discussion of sources of variation in mRNA levels and a brief
description of some standard methods for measuring mRNA amounts from cell or
tissue samples.
6.1 THE GENE

The molecular mass of a eukaryotic chromosome is composed of approximately
50% DNA, with the other half made up of proteins. DNA with its associated proteins
is called chromatin. Histones, the majority of proteins found in chromatin, compact,
organize, and protect the DNA. There are four histone subunits that assemble into
a flat, round octamer around which DNA wraps approximately 1.65 times. This
complex of histone proteins and the associated approximately 147 base-pair (bp; a
nucleotide and its compliment in native double-stranded DNA) length of DNA is
called a nucleosome. Nucleosomes are connected to one another by a length of linker
DNA (2060 bp) and are found along the entire length of DNA except in areas
where the DNA is actively engaged by various nonhistone proteins. Nonhistone
proteins are much less abundant than histones and are involved in the regulation and
execution of transcription, DNA repair, recombination, and replication.4,5
The protein fraction of chromatin is very important for its proper functioning,
but it is the nucleotide sequence in DNA that is the critical information-storage
component, with the gene as the elemental unit of information (see Figure 6.1). A
gene is generally considered to be the coding region portion of a particular DNA
sequence. This is the sequence that is transcribed into mRNA and that contains the
specific information from which the primary structure of the gene product is deter-
mined; i.e., the protein. Just as important, however, is the regulatory region of a
gene which is composed of noncoding DNA sequence. The regulatory region inter-
acts with a large number of proteins that dictate when and where the gene will be
transcribed and can include sequence found many thousands of bp away from the
Gene Expression 75
5GTGTATAAAATGATTGTT 5GTACCATTCCGAATGTAT
3CACATATTTTACTAACAA 3CATGGTAAGGCTTACATA
TATA Box Initiator
Regulatory Region Core Promoter

(100s -1000s bp) (~80 bp)
5
DNA 3 -40+ +1 40

Start Site
Coding Region
(100s -1000s bp)
1
H3 C
5 5 Exon 1 Exon 2 Exon 3

Pre-mRNA 3 HO
P-P-P- AAAAA
Intron Intron 3 UTR
2
H3 C
5 5
Mature mRNA 3 HO
P-P-P- AAAAA
FIGURE 6.1 Basic structure and transcription of a hypothetical gene. Two binding elements
are shown, the TATA box and the initiator, along with their complementary DNA strand
(sequence on either side of the element is arbitrary). During transcription, the 5' end of the
nascent mRNA is capped with methyl-guanine and the 3' end is polyadenylated (step 1).
Introns are removed from the pre-mRNA in the final processing step toward the mature mRNA
molecule (step 2). In a very small number of cases, the mRNA is further processed with RNA
editing (not shown). Transcription takes place in the nucleus followed by transport of the
mature mRNA out of the nucleus to other cellular compartments where translational occurs.
coding region. The regulatory region can be further subdivided into the promoter
(the location on the DNA where the transcriptional machinery binds) and the regu-
lator binding sites (individual sites on the DNA to which proteins bind; also called
binding elements). Regulator binding sites that promote transcription are often clus-
tered in regions known as enhancers. Similarly, clusters of binding sites that suppress
transcription when bound by regulatory proteins are called silencers.6
The basis for the specificity of interactions between nucleic acids or between
nucleic acids and proteins is in the linear sequence of the four nucleotide bases that
are found in double-stranded DNA or single-stranded RNA: adenine, guanine,
cytosine, and thymine (uracil in RNA). This was the fundamental discovery made
by Watson and Crick some 50 years ago.7 It is the unidirectional order of these bases
(specified with relation to the 5' and 3' phosphodiester linkages between ribose or
deoxyribose sugar molecules in the RNA or DNA backbone) and their two- and
three-dimensional relationships to one another that contain specific information
about the structure of proteins and their temporal and spatial expression. Many
different types of regulatory and enzymatic proteins are able to read the nucleic
acidbased information in an integrated fashion and efficiently accomplish accurate

expression of the gene product. A crucial aspect of the information-carrying capacity
of nucleic acids is that it is easily transmittable in a manner that accurately preserves
the sequence information. Thus, when a cell divides, the daughter cells each contain
an exact copy of DNA. Similarly, during transcription, the mRNA is synthesized in
a precise rendition of the coding portion of the gene. This is a result of the comple-
mentary nature of DNA, often referred to as WatsonCrick base-pairing, wherein
an adenine (A) is always paired to a thymine (T; uracil in mRNA U) and a guanine
(G) is always paired to a cytosine (C).
6.2 TRANSCRIPTION: BASIC MECHANISM OF MRNA

BIOGENESIS
Synthesis of mRNA proceeds through three basic phases: initiation, elongation, and
termination. Initiation occurs when the histone-bound DNA is made accessible, the
transcriptional machinery binds to the appropriate location on the DNA, and synthesis
of the complementary strand of RNA commences. During the elongation phase, the
nascent RNA is lengthened and proofreading functions take place to ensure that
the message is error-free. The last phase of transcription, termination, occurs when the
new RNA molecule is released thus causing transcription to end. This is followed by
detachment of the transcriptional machinery from the DNA. Often transcription is
taking place simultaneously at multiple sites on the same gene such that many
identical mRNA molecules are being synthesized simultaneously (see Figure 6.2).
Single-stranded mRNA is synthesized from and complementary to the 35
strand of the DNA; thus, the mRNA sequence is identical to the 53 DNA strand.
Transcription starts at the location on the DNA that corresponds to what will be the
5 terminus of the single-stranded mRNA molecule, often called the transcript.
Approximately 40 nucleotides on either side of the start site constitute the core
promoter. The core promoter contains the minimal amount of sequence information
required for the initiation of transcription. There are four specific sequence motifs
located in the core promoter; any given gene typically includes only two or three
of these elements. Additional elements located within 200 bp of the start site are
referred to as promoter-proximal elements. Classes of proteins called transcription
factors are able to recognize the promoter sequences and by doing so facilitate the
binding and activation of the RNA-synthesizing enzyme RNA polymerase. Some
important transcription factors and their binding element consensus sequences are
shown in Table 6.1. Additional elements, some as far as 100,000 bp upstream of the
start site, may also contribute to the induction or suppression of transcription. Thus,
certain combinations of transcription factors are required to initiate transcription and
it is the cooperative activity of the specific combination of factors that are present
in a cell that determines the temporal and spatial expression of a given gene.810
Eukaryotic cells have three distinct polymerases for the synthesis of RNA: RNA
Pol I, II, and III. Pol II is responsible for the transcription of all protein-coding
genes; Pol I and III transcribe specialized RNA molecules including transfer RNA
(tRNA) and ribosomal RNA (rRNA) and will not be considered here. Pol II forms
Gene Expression 77
General transcription Pol II

factors
Mediator complex
Preinitiation
complex
Nucleosome
modifiers
1
Activators
100s -1000s bp
2 3
PPP PPP
Pre-mRNA
5
5
FIGURE 6.2 The molecular machinery of transcription. General transcription factors facil-
itate the binding of Pol II onto the promoter; this is the preinitiation complex. In vivo, many
other components are needed to initiate transcription including the mediator complex, nucleo-
some modifiers, and activators (step 1). Phosphorylation of the Pol II tail by TFIIH causes it
to be released from the preinitiation complex and begin transcription (step 2). Sequence signals
in the transcript cause it to be released from Pol II and polyadenylated, thus terminating
transcription (step 3).
a pre-initiation complex with numerous general transcription factorsproteins that

aid Pol II in the recognition of and binding to the core promoter (see Table 6.1). It
is thought that the general transcription factors assemble on the core promoter in a
step-wise fashion in such a way as to denature the DNA and ultimately allow the
binding of Pol II in the appropriate location. Pol II wraps around a single strand of
the DNA duplex (the 35 strand) and uses the information contained in the DNA
sequence to consecutively add nucleotides to a growing RNA chain that shows
WatsonCrick complementarity to the template DNA. The preinitiation complex is
the minimal requirement for transcription to proceed on naked DNA in vitro. In
vivo, however, nucleosomal DNA is generally inaccessible to nonhistone proteins.
Thus, initiation of transcription can only take place when an inaccessible target
region of the genome is made accessible by the activity of other transcription factors
TABLE 6.1
Common Promoter Elements
Binding Element
Consensus Transcription
Promoter Location Positiona Sequenceb Factorc
Initiator Core promoter element +1 PyPyAN(T/A)PyPy TBP
TATA box Core promoter element 35 to 20 TATAAA TBP
CAAT box Promoter-proximal element 200 to 70 CCAAT CBF, NF1,
C/EBP
GC box Promoter-proximal element 200 to 70 GGGCGG SP1
a Number of nucleotides from the start site with negative numbers indicating in the upstream direction;
i.e., toward the 5 end.
bThe consensus sequence is the ideal sequence for the most effective interaction with its factor; binding
elements can often tolerate slight differences from the consensus. Py = pyrimidine (C or T); N = any
(A, C, T, G).
c TBP: TATA binding protein; CBF: CAAT binding protein; NF1: nuclear factor 1; C/EBP:
CAAT/enhancer binding protein.
that facilitate the binding of the preinitiation complex to the core promoter. These
factors, which include the mediator, a complex of more than 20 subunits, regulatory
proteins, and nucleosome-modifying proteins, are engaged in various activities
including covalent modification of histones and Pol II recruitment (see below).10,11
Once the preinitiation complex has formed at the core promoter, Pol II begins
transcription and attempts to separate itself from the complex. Several short tran-
scripts are synthesized, but the polymerase cannot enter into the elongation phase
until its C-terminal tail, which is composed of a long string of Tyr, Ser, Pro, and
Thr residues, is phosphorylated by the kinase activity of the factor TFIIH and other
kinases. Once the tail has been phosphorylated to a certain level, affinity of Pol II
for the general transcription factors decreases considerably allowing the polymerase
to move away from the core promoter and begin synthesizing a full-length transcript
during the elongation phase. Phosphorylation and dephosphorylation of the C-
terminal tail are important regulatory mechanisms of Pol II activity.1113
As Pol II enters into the elongation phase, the general transcription factors are
replaced with a new set of factors that bind favorably to the phosphorylated form
of the C-terminal tail. The new factors fulfill three basic functions: (1) promote
continued RNA synthesis by stimulating Pol II activity; (2) proofread and correct
the nascent RNA to ensure exact complementarity to the DNA template; and (3)
process the emerging RNA molecule with several types of covalent modifications
primarily for stabilization. Many of these factors bind to the C-terminal tail of
Pol II. As the RNA is synthesized, its 5 end emerges from the polymerase near
the C-terminal tail. Thus, the factors are in the ideal position for gaining access to
the newly synthesized RNA.1113
Gene Expression 79
P-TEFb is a central factor recruited to the C-terminal tail early in the elongation
phase. Along with other factors, it stimulates the elongation of the RNA molecule
by Pol II. P-TEFb has inherent kinase activity and it phosphorylates additional
residues on the Pol II tail. This activity aids in the recruitment of the elongation
factors hSPT5, TAT-SF1, and TFIIS, all of which promote the continued elongation
of the transcript. Pol II does not synthesize RNA at a constant rate; rather, the rate
is dependent on the specific sequence being transcribed and some sequences will
cause the polymerase to come to a near complete stop. One of the functions of TFIIS
is to smooth the progress of Pol II through these more involved sequences so that
it progresses at a faster overall rate. Combined with the nuclease activity of Pol II,
TFIIS also aids in the detection and replacement of mismatched nucleotides. This
proofreading capability is not as efficient as that which occurs during DNA synthesis.
Approximately 1 in 107 mismatched nucleotides escape detection during DNA rep-
lication, but transcription is somewhat less accurate to approximately 1 in 104 to 105
errors are made during transcription. Of course, the consequences of a DNA mutation
are potentially much more severe than the occasional single mutant protein translated
from a mismatched mRNA sequence.1113
When the new transcript reaches a length of 20 to 40 nucleotides, the 5 end is
modified with the addition of a methylated guanine, known as capping. Guanine is
added in an unusual 5-5 triphosphate linkage and then a methyl group is added to
the nitrogen in the 7 position of the guanine moiety (see Figure 6.2). The cap serves
as a signal for other processes including recruitment of the translation machinery.
As the transcript reaches full-length, a second processing step takes place, polyade-
nylation, that coincides with termination of transcription. CPSF and CstF are proteins
that have been recruited to the phosphorylated tail of Pol II to separate the nascent
mRNA from the RNA that is still being made and to facilitate polyadenylation.
Transcription of the poly-A signal sequence promotes release of CPSF and CstF
from the Pol II onto the new transcript. These factors cause the mRNA to be cleaved
from the still growing RNA strand and recruit poly-A polymerase which enzymati-
cally adds approximately 200 adenine molecules (from ATP) to the 3 end of the
transcript. Pol II continues to synthesize complementary RNA sometimes as long
as several hundred nucleotides, but eventually Pol II dissociates from the DNA and
releases the aberrant RNA molecule which is rapidly degraded. The signals that
cause dissociation of the polymerase may have to do with either the absence of a
5 guanine cap and/or with the transfer of CPSF and CstF from the tail of Pol II,
but the details of this have not been entirely worked out.12,13
For most eukaryotic genes, the primary transcript, or pre-mRNA, requires one
additional processing step, splicing, before it is mature and ready for the translation
of protein. The pre-mRNA usually consists of alternating segments known as exons
and introns that specify coding and noncoding portions of the gene, respectively.
All of the exons in a gene (more than 300 in some cases) must be joined in a
contiguous fashion in order for the transcript to be appropriately translated into the
protein product. Thus, introns are removed and exons are spliced together in a
reaction catalyzed by a large protein-RNA complex known as the spliceosome. The
spliceosome is directed by specific sequences at the exon/intron border which guide
the splicing reaction. Some primary transcripts have the capacity to be spliced in
different configurations. This alternative splicing provides a method by which a

single gene can produce many, often related, protein products. Alternative splicing
can be either constitutive or regulated; i.e., alternate transcripts are made on a regular
basis or they are made as dictated by the needs of the cell.12,13
RNA editing is another process that can change one or more nucleotides in a
transcript such that the protein product is changed either in sequence or in size.14 RNA
editing does not occur for all transcripts and it typically takes place in a cell or tissue
specific manner. An important neuronal gene that undergoes RNA editing is the Glur2
subunit of the AMPA-type glutamate receptors, ligand-gated cation channels. A spe-
cific adenosine is deaminated to inosine in Glur2 pre-mRNA.15 The adenosine is part
of a codon (a nucleotide triplet that codes for an amino acid) that normally codes for
a glutamine residue, but the inosine-containing codon codes for arginine. This amino
acid switch has profound effects on the function of the channel, especially with regard
to calcium conductance. The editing appears to be critical because mutant mice that
lacked the ability to make the amino acid conversion were found to have severe
neurological problems and die within just a few weeks after birth.16
Messenger RNA is transported out of the nucleus into the cytoplasm where it
is directed to the appropriate location for translation, but only capped, polyadeny-
lated, fully spliced mature mRNA is transported. The transport signal consists of
the various proteins that have accumulated on the mRNA since the beginning of its
synthesis. Once out of the nucleus, the proteins are shed and transported back into
the nucleus where they are recycled for continued transcription. Some proteins are
translated and then transported or diffused to their final destination, but in many
cases, the mRNA is transported to specific sites and translated locally. This is of
special interest in the nervous system in which mRNAs related to cytoskeletal
structure, signaling, and even transcriptional regulation have been detected in den-
drites and axons where the machinery for translation is also found and in which
local translation does appear to take place.17
6.3 REGULATION OF TRANSCRIPTION

Regulation of transcription is one method through which the cell controls protein
amount, although it is important to bear in mind that there is not always a direct
correspondence between mRNA levels and protein levels and also that there are
numerous other mechanisms for controlling the amount and activity of a protein.18
Transcriptional regulation acts primarily at the level of initiation. Regulation also
occurs during elongation, termination, and RNA splicing, but these will not be
discussed here. The importance of gene regulation lies in the notion that for a cell
to function properly, only a specific subset of proteins is required at any particular
time. Thus, mRNA for a given gene is typically expressed in a highly regulated
tissue- and time-dependent manner. In any particular cell, some genes are exclusively
transcribed during certain periods of early development and others only later. Still
other mRNAs are expressed throughout the life of a cell or only in response to
intrinsic or extrinsic signals. Whether a particular mRNA is transcribed is dictated
by dynamic and concerted interactions between DNA and a wide variety of tran-
scription factors and other regulatory proteins.
Gene Expression 81
Transcription factors can either facilitate or suppress transcription and are either
constitutive or inducible. Constitutive transcription factors are present in the cell at
all times and are typically activated through signal transduction pathways. Inducible
transcription factors are not usually present in the cell, but are transcribed and then
translated following some kind of intra- or extracellular signal. Positive transcription
factors are known as activators; negative factors are called repressors. Transcription
factors bind with high specificity to specific DNA sequences in the promoter or
elsewhere in the regulatory region of genes and interact with other proteins to
influence transcription. The proteinprotein interactions are of two general types:
recruitment, in which a protein facilitates the cooperative binding of one or more
other proteins to a specific genomic location; and allostery, in which one protein
triggers a conformational change in another eliciting some kind of functional mod-
ification in the second protein. Often both types of interactions take place. In many
cases, DNA-binding proteins bind to adjacent sites on the DNA, but this is not always
true. Interacting proteins may bind at very long distances from one another and are
able to make physical contact because of DNAs ability to loop back on itself. Indeed,
there are specific DNA-binding proteins that facilitate the bending of DNA so that
distant sites can be brought in close proximity to one another.19,20
Transcription factors typically have distinct DNA-binding domains and protein-
binding domains. The DNA-binding domains have high affinity for specific DNA
nucleotide sequences and often bind to DNA as dimers with an -helix functioning as
the DNA recognition site. Other parts of the DNA-binding domain interact with the
DNA backbone to align the DNA recognition helix in the proper position in the major
groove of DNA. Some of the more common DNA-binding domains found in transcrip-
tion factors include homeodomain proteins, zinc-containing DNA-binding domains
(includes zinc finger proteins), leucine zippers, and helix-loop-helix proteins.21
One of the important functions of activator proteins is to recruit Pol II to the
promoter. Typically this is not a direct interaction with Pol II, but rather recruitment
takes place indirectly through interactions with preformed polymerase-associated
protein complexes, such as mediator or TFIID, which are often already associated
with Pol II. Often the concerted action of many activators recruits the transcriptional
machinery. For a simple example, consider two distinct activators that contact medi-
ator at different sites. Either one alone provides insufficient binding energy, but the
combined binding energy of the two is able to effectively recruit the complex.
Transcriptional initiation often requires the presence of many activators that work
together in such an integrated manner. Each gene does not have its own specific
activator; rather, the combined activity of a group of activators present in the right
combination and at the right time is required for gene expression. This kind of
combinatorial control is a key element for the appropriate cell-, time-, and response-
dependent expression of many genes.8
The DNA in chromatin is generally inaccessible for transcription because pro-
moters are tightly bound by histone proteins. Some activators recruit specific proteins
that alter nucleosomes so that the transcriptional machinery can access the gene.
This occurs in one of two ways: nucleosome remodeling or through covalent mod-
ifications. Activators recruit nucleosome remodeling complexes that change the
nature of DNA-histone interactions. These complexes can free the promoter either
by moving the position of the histone relative to the promoter or by causing the
promoter to become less tightly bound to the histone. Either situation promotes the
binding of Pol II and its associated proteins. The second type of alteration involves
reversible covalent modifications to histone proteins. Activators recruit histone acety-
lases that add acetyl groups to the N-terminal tails of histones. The tails are rich in
lysine residues whose positive charge normally facilitates tight binding of the histone
to the negatively charged phosphate portion of the DNA backbone. Acetylation
effectively neutralizes the tails positive charge causing the nucleosome structure to
relax facilitating the binding of the transcriptional machinery to the promoter. In
addition, certain proteins and protein complexes such as TFIID have specific domains
known as bromodomains that bind with high affinity to acetylated histones further
promoting the initiation of transcription. Remodeling and histone modification typ-
ically act in combination to promote transcription, sometimes the activity of one
enhancing that of the other.5,22
Histone tails can also be modified with the addition of methyl groups by histone
methylases. This can occur in an analogous manner as acetylation, but histone
methylation is usually associated with inhibition of transcription. Thus, chromo-
domain-containing proteins interact with methylated histones to prevent transcription
by inhibiting the binding of transcriptional activators. Methylation also recruits
proteins with methylase activity that propagate methylation locally to unmodified
histones (see below). The covalent modification of histone tails is a dynamic process
and it is believed that it is the overall pattern of histone modifications that is an
important factor in transcriptional regulation.5,22
As a result of the integrated activity of many activators and the proteins with
which they interact, only certain genes are transcribed at any given time. Transcrip-
tional repressors also contribute to the transcriptional status of a gene. In general,
repressors behave very similarly to activators, i.e., they bind with high affinity to
specific DNA sequences and also bind to other proteins, but with the end result
being the inhibition of initiation. A repressor can bind to a specific DNA site allowing
it to interact with the transcriptional machinery thereby preventing initiation through,
for example, allosteric modulation. Repressors can also inhibit transcription more
indirectly by preventing the binding of activators either to DNA or to the proteins
with which the activators interact. Repressors can also prevent transcription through
the recruitment of histone deacetylases. These enzymes remove acetyl groups from
the N-terminal tails of histones thereby restoring nucleosome structure.23
Messenger RNA is ultimately degraded through the sequential activity of specific
decay enzymes found localized to discrete cytoplasmic bodies. The initiating step
of decay is the deadenylation of the mRNA. From there, it proceeds through one of
two known pathways that end in its digestion by the activity of either 53 or
35 exonucleases. As might be expected, the decay of mRNA is highly regulated,
a consequence of which is that the mRNA for constitutively expressed genes tends
to have a longer half-life than for genes that are expressed in response to external
signals. The signal for degradation is generally, but not exclusively, found in the 3
untranslated region (3UTR) of the mature mRNA. There are specific elements in
the 3UTR whose bound regulatory proteins recruit the decay enzymes. Conversely,
certain regulatory proteins are able to stabilize an mRNA, prolonging its life in the
Gene Expression 83
cell. The fate of a transcript is determined by the localization, levels, or regulated

activity of the decay enzymes and/or the regulatory proteins, some of which are
influenced by external signals. Thus, the amount of any given transcript is the
combined result of its transcription rate and its decay rate both of which are highly
regulated processes.24
6.4 TRANSCRIPTIONAL REGULATION THROUGH

SIGNAL TRANSDUCTION
Whether a gene is transcribed is dependent on the coordinate activity of many
proteins: activators, repressors, mediator, TFIID, Pol II, etc. How are all of these
proteins brought together at the right time and in the right place? In prokaryotes,
environmental signals can directly influence the activity of activators or repressors.
For example, the sugar lactose binds to the Lac repressor preventing its binding to
the lacZ gene, the product of which is -galactosidase. -galactosidase cleaves
lactose into glucose and galactose, the preferred energy source. Thus, in the presence
of lactose, lacZ is derepressed and -galactosidase is synthesized for the metab-
olism of lactose, but in the absence of lactose, the Lac repressor binds to lacZ
preventing transcription. This efficient scheme ensures that the cell only makes the
enzyme when it is needed. At some level, all cells respond to environmental signals,
but this sort of simple, direct mechanism is generally not the case in multicellular
eukaryotic organisms. Rather, transcription is often mediated indirectly through
signal transduction pathways.
Signal transduction pathways are activated by the binding of a signal molecule
ligand to a cell surface receptor, a protein or protein complex that is embedded in
the cell membrane with one surface exposed on the extracellular side and a second
surface exposed on the intracellular side. Signal molecules include neurotransmitters,
steroid hormones, neuropeptides, modified amino acids, and other kinds of molecules
that are released from other cells in response to external or internal signals. The
binding of the signal molecule on the extracellular surface causes a conformational
change in the receptor which then triggers a cascade of intracellular events ultimately
leading to some sort of transcriptional regulatory activity. The signal transduction
pathway can be simple or it can be very elaborate and complex. The cAMP response
element binding (CREB) protein system is presented as a simple example of signal-
mediated activation of transcription and as an illustration of many of the principles
described above (see Figure 6.3).25
CREB is a constitutive transcription factor and the prototypical member of the
CREB family that includes CREM (cAMP response element modulator) and ATF-1
(activating transcription factor 1). CREB itself has at least five splice variants that can
act as either activators or repressors, dependent on the exon configuration. CREB
variants that are activators all contain three important domains: bZIP which mediates
DNA binding to the cAMP response element (CRE) and also mediates CREB dimer-
ization; Q2/CAD (constitutive active domain) which interacts with the transcriptional
machinery; and the kinase inducible domain (KID) which contains a serine residue (Ser-
133) whose phosphorylation state is a critical determinant of CREBs activity level.26,27
1
R AC
G G ATP Cell membrane
GDP GTP GTP cAMP

2
GTP
3
PKA
4
CREB Nuclear membrane
ADP ATP
P 5
FIGURE 6.3 Activation of CREB through stimulation of a typical signal transduction pathway.
The process is initiated by the binding of a small molecule ligand to a cell surface receptor
(R) embedded in the cell membrane (step 1). A subsequent conformational change in the
receptor causes guanosine-diphosphate (GDP) to be replaced with guanosine-triphosphate
(GTP) on the trimeric G-protein (G) and its GTP-containing subunit to become dissociated
(step 2). The /GTP complex activates the enzyme adenylyl cyclase (AC) which catalyzes the
formation of cyclic-AMP (cAMP) from ATP. cAMP triggers the dissociation of the catalytic
subunit of protein kinase A (PKA) which then diffuses into the nucleus and phosphorylates
CREB (steps 3 and 4). Phosphorylated CREB now binds to CRE on the DNA and promotes
transcription through its interactions with the preinitiation complex and CREB binding protein
(CBP) (step 5).
Unphosphorylated CREB resides in the nucleus of many different cell types. In

this state, it is thought to be able to stimulate a very low level of basal transcription
of CRE-containing genes through its interaction with the transcriptional machinery.
Stimulus-induced phosphorylation of Ser-133 increases CREB-mediated transcrip-
tion manyfold. It has become clear that numerous pathways can lead to the phos-
phorylation of Ser-133, but the first described and perhaps most prominent is through
activation of adenylyl cyclase (AC). Ligand binding to any one of a number of
membrane-bound G-protein-coupled receptors activates AC through the G-protein
mediated hydrolysis of GTP. Activated AC catalyzes the formation of cAMP from
ATP. cAMP interacts with protein kinase A (PKA) stimulating the release of its
catalytic subunit which then migrates into the nucleus and phosphorylates Ser-133.
Phosphorylated CREB has very high affinity for CREB binding protein (CBP) and
its paralogue p300. Binding of either of these two proteins to CREB promotes
Gene Expression 85
transcription through two mechanisms: stabilization of the preinitiation complex and

the intrinsic histone acetylase activity of CBP. CREB activity is terminated by the
removal of the Ser-133 phosphate through the action of a phosphatase, itself activated
through signal transduction pathways.26,27
CREB-mediated transcription peaks at about 1 hour after stimulation and then
slowly declines to basal levels after about 4 hours. The rate-limiting step is the
movement of the PKA catalytic subunit into the nucleus. Some of the first genes to
be transcribed following stimulation are the so-called immediate early genes (IEGs).
Many of these are themselves transcription factors including c-fos, ATF-3, JunD,
Nurr1, and CREB itself.26,27
The basic mechanism of CREB activation described above is well supported,
but there are unresolved questions related to CREB signal discrimination. In addition
to AC signaling, CREB can be activated by many signals and in response to a wide
variety of physiological stimuli. Moreover, a potentially large number of genes are
CREB-responsive, but under a given stimulus or in a particular cell type, only a
small set of specific genes are transcribed in response to CREB activation. What
accounts for this specificity? This is likely determined by the presence of other
activators or repressors and possibly also by the phosphorylation state of other sites
on CREB. Herein is an example of combinatorial control whose complexity belies
what was once thought to be a simple model of transcriptional regulation.26,27
6.5 RNA INTERFERENCE

RNA interference (RNAi) through microRNA-mediated mechanisms is a novel form
of post-transcriptional gene regulation. RNAi has profound implications not only
from a basic biological perspective, but also for its exploitation as an experimental
tool and its potential therapeutic application in disease. MicroRNAs (miRNAs) are
small RNA molecules (approximately 22 nucleotides) that are complementary to the
sequence of mRNA targets and are thought to be an evolutionary by-product of a
cellular defense mechanism directed against viral pathogens.28 MicroRNAs are
thought to be synthesized in a manner that is similar if not identical to mechanisms
of mRNA synthesis.29 Approximately 300 miRNAs have been definitively identified
in plants, nematodes, fruit flies, fish, rodents, and humans, and it has been estimated
that there are as many as 255 human miRNA genes.30 The few miRNA targets that
have been characterized are involved in developmental patterning and timing,31 but
in general little is known of miRNA targets, especially in animals. Dysfunction of
miRNA processes have been implicated in a limited number of human pathologies,32
but miRNA research is an emerging field and the ubiquity of this observation has
yet to be determined.
Many miRNA genes are polycistronic (tightly clustered in a single genomic
region and transcribed together), but quite a few are embedded within the introns
of their target genes.29 Clustered miRNAs are often, but not always functionally
related. Similarly, functionally related miRNA genes are sometimes physically
distant to one another within the genome. Some miRNAs are known to be synthe-
sized by Pol II and the normal mRNA transcriptional machinery suggesting that
regulation of miRNA expression occurs in conjunction with or similarly to that
of mRNA.29 However, little is currently known about the regulation of miRNA

expression. miRNAs are initially synthesized as a single-stranded RNA precursor
of about 600 nucleotides in length known as pri-miRNA. A portion of each of the
5 and 3 ends are cleaved off by the Drosha endonuclease leaving behind an
approximately 75-nucleotide intermediate (now called pre-miRNA) that hybridizes
to itself by WatsonCrick base-pairing into a stem-loop structure in which the
miRNA segment forms the double-stranded stem. This structure is transported out
of the nucleus and into the cytoplasm where it interacts with another endonuclease
known as Dicer and possibly with members of the Argonaute protein family. Dicer
cleaves off the loop of the pre-miRNA to complete the processing of the now mature
approximately 22-nuleotide double-stranded miRNA molecule.33
Following Dicer processing, miRNA becomes incorporated into the ribonucle-
oprotein RNA-induced silencing complex (RISC; note that many authors refer to this
complex as miRNP). RISC contains numerous proteins including one of the Argo-
naute family members around which the miRNA binds very tightly. Only one strand
of the miRNA duplex is incorporated, the strand that shows complementarity to the
target mRNA; the other strand is identified and degraded by RISC by unknown
mechanisms. RISC binds the target mRNA and inhibits its further activity by one
of two methods. If there is perfect complementarity between the miRNA and the
target transcript, the mRNA is degraded. If there are mismatches between the two
RNA molecules, translation is prevented. The degree of complementarity is probably
not the sole determinant of the specific mechanism of interference, but either situ-
ation concludes with the same result: protein expression is inhibited.33
6.6 GENE SILENCING

Large sections of DNA can be turned off through the concerted activity of meth-
ylases, deacetylases, and other proteins. Gene silencing is associated with protein-
dense genomic regions known as heterochromatin. One such area is found at the
end of chromosomes the telomere. A complex of proteins that includes Sir 1, 2,
3, and 4 (silent information regulators) is recruited by proteins that bind to specific
sequences in the telomere. The Sir proteins bind to unacetylated histones and then
deacetylate nearby histones thus propagating a growing heterochromatin complex.
Transcription is inhibited as a result of the inability of the transcription machinery
or activators to bind to the complex. Eventually histones are encountered that have
been modified in certain ways that prevent the binding or deacetylation activity of
the Sir proteins. Heterochromatin is thought to make up approximately 50% of the
genome and most of it is located at the telomeres or centromeres, but is also found
elsewhere throughout the genome.34,35
Gene silencing can also occur through covalent modifications on the DNA itself.
During development or in other situations, an initiating signal induces certain genes
to be completely shut off. This is accomplished through methylation of the DNA.
DNA methyltransferase adds a methyl group onto the ring structure of the nucleotide
cytosine. Proteins that recognize methylcytosine bind with high affinity and recruit
histone remodeling and modifying proteins whose activity results in the complete
inhibition of transcription on that gene.36
Gene Expression 87
During cell division, it is usually necessary to preserve the status of silenced

genes in daughter cells. This implies that the 50% of newly synthesized histones
and DNA must be properly modified, but in the absence of the original initiating
signal. Methylated histones will remain in the correct position within the chromatin
of daughter cell DNA distributed approximately equally between the two cells. These
modified histones subsequently recruit chromodomain-containing proteins which
include histone methylase. Methylation then proceeds on the nearby unmethylated
histones. A similar mechanism occurs for methylated DNA. A specific enzyme,
maintenance methylase, recognizes the partially methylated DNA and completes
local methylation of the new strand. This triggers protein recruitment and gene
silencing as described above. This is the basis of a phenomenon known as imprinting
in which a gene allele is silenced selectively on either the maternal or paternal
chromosome and is transmitted to the offspring. Preservation of silenced genes or
alleles during mitosis, meiosis, or fertilization is a form of DNA modification that
is heritable in the absence of DNA mutation and that does not require the original
initiating signal. This general phenomenon is referred to as epigenetic regulation.37,38
6.7 SOURCES OF VARIATION IN GENE EXPRESSION

Transcription can be influenced by a wide array of stimuli including exposure to
exogenous substances, external sensory stimulation, and in response to pathological
conditions. Transcriptional regulation by these sources facilitates an organisms
response and adaptation to an ever-changing internal and external environment. For
example, acute exposure to many psychoactive substances such as alcohol, cocaine,
morphine, and other drugs of abuse, stimulates the phosphorylation of CREB leading
to a cascade of gene expression events.39 Most notably, there is an increase in the
transcription of inducible transcription factors, also known as immediate early genes
(IEGs) including c-Fos, FosB, c-Jun, JunB, JunD, Egr1, Egr2, Zif268, and others.
The increased mRNA is often translated into increased protein which then induces
late-response target genes. This phenomenon has been exploited in model organisms
to map regions of the brain in which the drugs are active, as well as to learn something
about the neurobiology of psychoactive drug responses. Different drugs do not all
show similar patterns of response across IEGs, time, or brain region. For example,
cocaine induces c-Fos in a number of brain regions including a very robust induction
in the striatum.40 This is not too surprising because the striatum is a major dopamine
output area and cocaine stimulates increased dopamine activity through its action as
an indirect dopamine agonist. However, alcohol, which stimulates dopamine release
in the striatum does not or only weakly induces c-Fos transcription in this brain
structure.41,42 On the other hand, Egr1 is significantly induced in the striatum after
either drug.39,43 It is believed that these kinds of IEG cascades in the reward and
motivation circuitry of the brain contribute to the pathological state of addiction;44
however, a clear understanding of how these events do so has not yet been achieved.
Within a heterogeneous population, variation in transcription is also attributable
to genetic factors; i.e., sequence polymorphisms in a gene. It is easy to imagine that
a DNA sequence difference in a promoter or binding element might cause differential
transcription factor binding affinity, and therefore result in a differential rate or
magnitude of transcription. Sequence variation can also affect other pretranslation

processes such as splicing and mRNA stability. These kinds of genetic influences on
gene expression are undoubtedly an important source of genetic variation for behavior
and other complex traits. Indeed, gene expression itself can be considered as a
complex or quantitative trait just as can virtually all measures of behavior or physi-
ology. This simply means that the expression level of a given gene in a heterogeneous
population is continuously distributed due to influences from both environmental and
genetic sources. Like other complex traits, the genetic component of the expression
variance is often under the control of multiple genes. As such, it is possible to map
quantitative trait loci (QTL) for the expression of a gene (see Chapter 4). QTLs are
regions of the genome that contain two or more DNA sequence variants or alleles
(within a population) that mediate the genetic component of variation for the trait of
interest; in this case, gene expression. An expression QTL can be either cis- or trans-
acting. In the former case, the QTL is tightly linked to the gene itself, suggesting
that a polymorphism within a cis regulatory element of the gene impacts its own
expression, while in the latter case, the QTL is found physically distant from the gene
and modulates expression through any one of a number of possible interactions. Single
gene mutations can affect the expression of many genes through trans effects,4547
but the discovery of a cis-QTL in a more genetically complex population is especially
informative because this immediately points to a candidate gene for any other trait
that has had a QTL mapped to the same genomic region. Standard mRNA quantifi-
cation methods can be used to map gene expression for a single gene at a time or
gene expression microarrays can be used to map QTLs for thousands of genes
simultaneously (see below). This general strategy was used in combination with other
methods to identify a specific polymorphism in the regulatory region of mouse Kcnj9
(a G-protein inwardly rectifying potassium channel) that appeared to be responsible
for differential regulation of the gene. Perhaps more importantly, a compelling argu-
ment was made that this Kcnj9 polymorphism was, in fact, a QTL for the behavioral
trait of basal locomotor activity.48
6.8 MEASURING GENE EXPRESSION

Measurement of the expression level of a gene can begin to reveal clues about its
general regulation and about its involvement in a cells response to external pertur-
bations. Thus, techniques have been developed to accurately determine the amount
of a given transcript from a cell or tissue sample. The most common methods rely
on the properties of WatsonCrick base-pairing making it very easy to design probes
that are highly selective for a specific transcript. However, members of gene families
typically have at least partial sequence similarities and therefore care must be taken
in some situations to ensure that only the gene of interest is being measured rather
than the gene in addition to one or more related genes with similar sequences. The
methods are generally not difficult to perform, but certain precautions must be taken
in handling any kind of RNA because of the ubiquity of ribonucleases (RNA-specific
degrading enzymes) in both prokaryotic and eukaryotic organisms. Several com-
monly used methods of mRNA quantification are briefly described.
Gene Expression 89
The standard workhorse method of quantification for a single mRNA species is

the Northern blot. Total RNA is isolated from cultured cells or a tissue sample using
one of several methods.49 The RNA sample can be further fractionated into purified
polyadenylated RNA (i.e., pure mRNA), but this is not always necessary. Following
isolation, mRNA of different molecular weights (i.e., lengths) is separated by poly-
acrylamide gel electrophoresis (PAGE) using standard methods. The separated bands
are transferred to a nylon membrane and incubated in the presence of a nucleic acid
probea strand of radiolabeled DNA that is complementary to the transcript of
interest. The membrane is then exposed to x-ray film. If all goes well, a single band
is revealed on the film from which the amount of target mRNA in the sample is
quantified using image densitometry. Typically the membrane is stripped and the
process repeated with a probe for a so-called housekeeping gene to control for
loading on the gel and other factors that could potentially introduce technical error
to quantification of the band. Housekeeping genes are genes that are thought to
remain at a constant level across experimental conditions and include the actins,
glyceraldehyde-3-phosphate dehydrogenase (G3PDH), the tubulins, cyclophilin,
hypoxanthine phosphoribosyltransferase (HRPT), and others, as well as the 28S and
18S ribosomal RNAs.50 However, the consistency of these genes expression level
must be verified when possible because they often do change across experimental
conditions. Northern blotting can be reasonably quantitative, but the many steps
involved and the compressed dynamic range of densitometric readings reduces
quantitative precision somewhat.
The Taqman-based quantitative reverse-transcriptase polymerase chain reaction
(RT-PCR) is becoming increasingly popular for the quantification of single-gene
mRNA concentration. It is highly selective for the gene of interest, extremely
quantitative, and can measure very small amounts of transcript; in theory, just a
single mRNA molecule.5153 PCR primers are designed to bracket a fairly short
(3050 nucleotides) sequence within a region unique to the target mRNA. A probe
of 18 to 22 nucleotides in length is designed to be complementary to the sequence
between the primers. The 5 end of the probe is covalently tagged with a fluorescing
reporter dye (e.g., FAM) and a quenching dye is added to the 3 end (e.g., TAMRA).
Standard PCR is performed in the presence of the RNA sample, the primers, the
probe, and a polymerase that has three important enzymatic capabilities: reverse
transcriptase activity (synthesis of a single-stranded DNA molecule that is comple-
mentary to the single-stranded mRNA template), DNA polymerase activity, and
53 exonuclease activity. The reaction is conducted in a specialized thermocycler
that is capable of measuring fluorescence at the end of each PCR cycle. Normal
PCR amplification takes place with the mRNA as the initial template. During the
hybridization step of each cycle, the probe binds to newly synthesized DNA product
and is then digested by the exonuclease activity of the polymerase during the
extension step. This effectively separates the reporter dye from the quencher dye
allowing the reporter to be detected by the thermocycler. Thus, fluorescence increases
as the product increases with a rate of increase that is dependent on the amount of
mRNA template initially present. The number of PCR cycles it takes to reach a
certain level of detected fluorescence (the threshold cycle) is a quantitative measure
of how much target mRNA was present.54 If a clone of the target transcript is
available, extremely accurate quantification can be achieved using a standard curve.

Otherwise, the assay will include a primer/probe set for a housekeeping gene that
is amplified in the same reaction vessel and the threshold cycle of the target is
expressed as a ratio to that of the housekeeping gene mRNA. There are other RT-
PCR-based expression assays,55 but the Taqman version possesses the combined
benefits of high specificity, sensitivity, and quantitative precision.
In some cases, it is desirable to not only quantify a transcript of interest, but
also to determine its ultrastructural localization. With either of the above two meth-
ods, localization is only as good as the dissection procedure. This is often unaccept-
able in a brain where fine structure is quite complex. Thus, the method of in situ
hybridization has been devised to measure mRNA from a section of an intact tissue
sample that can be examined at high resolution, down to the level of individual cells.
Typically, the tissue is frozen at the time of sacrifice, sliced on a cryostat, and the
slices mounted onto glass microscope slides. The tissue is fixed in such a way as
to preserve the structural and chemical nature of the tissue and to make it permeable
to nucleic acid probes. RNA or DNA probes complementary to the target transcript
are labeled with a radioisotope during synthesis. Other labels can be used (e.g.,
fluorescent), but these typically are not as sensitive as radioisotopes. The mounted
tissue is incubated in the presence of the probe during which time it hybridizes
selectively to the target transcript. The slide is exposed to an x-ray film and selected
areas from the subsequent image can be quantified using densitometry techniques.
This method can be considered only semiquantitative because of the many sources
of error that can be introduced during the procedure.56
Over the past decade, methods have been developed to quantify thousands of
different transcripts from the same RNA sample making it possible to measure the
level of every single transcript from a single cell culture or tissue sample. Thus,
control cells can be compared to various kinds of experimental groups (e.g., drug-
treated, diseased, cells in a particular stage of growth, etc.) and the genes that respond
to the experimental condition can then be identified. Often referred to as gene
expression profiling, these methods have been employed as an unbiased approach
applied with many goals in mind including the elucidation of biochemical or neu-
ronal pathways, the identification of coregulated networks of gene expression, the
determination of patterns of recognition as predictors of drug responses or pathologic
states, and in gene discovery applications as described above. A variety of method-
ologies have been developed including differential display,57 serial analysis of gene
expression (SAGE58), and the solid platform microarray of which several different
types have been devised and is by far the most commonly used technique for
expression profiling. There are a number of commercially available microarrays and
many individual laboratories possess the instrumentation for producing their own
in-house microarrays. Perhaps the most well-developed commercial microarray sys-
tem is that offered by Affymetrix. The Affymetrix system will be used to illustrate
the general principles of expression profiling using microarray methodologies.
Total RNA is isolated using standard methods and then amplified and biotin-
labeled through a series of enzymatic reactions and purification steps.49,59 This biotin-
labeled complementary RNA (cRNA) is fragmented and hybridized to the Affymetrix
GeneChip microarray (Affymetrix, Santa Clara, CA). Affymetrix uses a lithographic
Gene Expression 91
manufacturing process to grow short oligonucleotides (nucleic acid sequences of

25 nucleotides in length) in a specific location, or array cell, on a solid support.
Each cell contains hundreds of thousands of identical oligonucleotides. Each organ-
isms mRNAs are represented by 22 such spots that consist of a series of 11 perfect
match (PM) and mismatch (MM) pairs, known collectively as a probe set. The
PM sequence is perfectly complementary to a unique sequence within the target
mRNA, while its corresponding MM is identical but with a single altered nucleotide.
The theory is that the biotin-labeled material will bind with high affinity and speci-
ficity to only the PM sequences, but not the MM sequences because of the mismatched
nucleotide. Thus the MM probes control for background and nonspecific binding. In
reality, some of the PM/MM pairs do not always perform as desired, but the system
generally works well when the entire probe set is considered. Following hybridization,
the microarray is bathed in a solution containing phycoerythrin-streptavidin. Strepta-
vidin binds with very high affinity to the hybridized biotinylated cRNA molecules
and phycoerythrin is a broad spectrum fluorophore. The fluorescence intensity of
each individual cell is determined with the use of a scanning laser and the intensity
corresponds directly to the amount of labeled cRNA bound to the oligonucleotides
in the cell. One of various analytical methods is used to combine the fluorescence
values from all cells in a probe set to generate a single value of transcript abun-
dance.6062 Many thousands of probe sets can be manufactured onto a single array
meaning that a substantial proportion if not all of an organisms transcripts can be
represented on a single microarray. Most other types of microarrays use only a single
cell per mRNA target, do not use the PM/MM concept, and are manufactured differ-
ently, but otherwise are very similar in basic design and overall theory of operation.63
The wet laboratory procedures for expression profiling are relatively straight-
forward, but the analysis can be complex in design and interpretation. Typically, an
experiment will go through three levels of analysis. First, procedures are used to
determine if a gene was actually present in the sample taking into account such
factors as the intensity of a given cell and nonspecific background fluorescence.
Second, statistical analyses are conducted to identify those genes that have changed
as a result of the experimental condition(s). This step can be problematic because
of multiple testing issues. Strictly speaking, the number of genes being examined
on a microarray is equivalent to the number of hypotheses being tested. Therefore,
the nominal P value in a standard statistical test would have to be extremely low for
an expression difference to be considered significantly different and many true
differences would be rejected because of this. On the other hand, many false positives
would be accepted as true if the statistical criteria are not stringent enough. Currently,
methods such as the BenjaminiHochberg False Discovery Rate (FDR64) are widely
accepted to control for error rates in microarray experiments. Rather than a strict
cutoff, the FDR uses a sort of sliding scale that allows the experimenter to dictate
the extent of error in an experiment. Finally, with a set of genes in hand, higher
order analyses such as clustering and biological pathway analyses are used to identify
coregulated genes and the biological networks that they fit into.65 The computational
tools and bioinformatics resources for these kinds of analyses can still be considered
fairly immature, but are constantly being improved. It will be the next generation
of tools and the scientists trained to use them that will start to really tap into the
power of this experimental strategy.
REFERENCES
1. Crick, F.H.C., On protein synthesis, in Symposium of the Society for Experimental
Biology XII, Academic Press, New York, 1958, 153.
2. Cech, T.R., The efficiency and versatility of catalytic RNA: implications for an RNA
world, Gene, 135, 33, 1993.
3. Kaczmarek, L., Molecular biology of vertebrate learning: is c-fos a new beginning?,
J. Neurosci. Res., 34, 377, 1993.
4. Belmont, A.S., Dietzel, S., Nye, A.C., Strukov, Y.G., and Tumbar, T., Large-scale
chromatin structure and function, Curr. Opin. Cell Biol., 11, 307, 1999.
5. Wolffe, A.P., Chromatin structure and the regulation of transcription, in Transcription
Factors, Locker, J., Ed., BIOS Scientific Publishers Ltd., Oxford, 2001, chap. 3.
6. Watson, J.D. et al., Molecular Biology of the Gene, 5th ed., Cold Spring Harbor
Laboratory Press, San Francisco, 2004.
7. Watson, J.D. and Crick, F.H.C., Genetical implications of the structure of deoxyri-
bonucleic acid, Nature, 171, 964, 1953.
8. Courey, A.J., Regulatory transcription factors and cis-regulatory regions, in Tran-
scription Factors, Locker, J., Ed., BIOS Scientific Publishers Ltd., Oxford, 2001,
chap. 2.
9. Garvie, C.W. and Wolberger, C., Recognition of specific DNA sequences, Mol. Cell,
8, 937, 2001.
10. Young, B.A., Gruber, T.M., and Gross, C.A., Views of transcription initiation, Cell,
109, 417, 2002.
11. Pugh, B.F., RNA polymerase II transcription machinery, in Transcription Factors,
Locker, J., Ed., BIOS Scientific Publishers Ltd., Oxford, 2001, chap. 1.
12. Maniatis, T. and Reed R., An extensive network of coupling among gene expression
machines, Nature, 416, 499, 2002.
13. Howe, K.J., RNA polymerase II conducts a symphony of pre-mRNA processing
activities, Biochim. Biophys. Acta, 1577, 308, 2002.
14. Schmauss, C. and Howe, J.R., RNA editing of neurotransmitter receptors in the
mammalian brain, Sci. STKE, 2002, pe26, 2002.
15. Sommer, B. et al., RNA editing in brain controls a determinant of ion flow in
glutamate-gated channels, Cell, 67, 11, 1991.
16. Brusa, R. et al., Early-onset epilepsy and postnatal lethality associated with an editing-
deficient GluR-B allele in mice, Science, 270, 1677, 1995.
17. Job, C. and Eberwine, J., Localization and translation of mRNA in dendrites and
axons, Nat. Rev. Neurosci., 2, 889, 2001.
18. Anderson, L. and Seilhammer. J., A comparsion of selected mRNA and protein
abundances in human liver, Electrophoresis, 18, 533, 1997.
19. Lefstin, J.A. and Yamamoto, K.R., Allosteric effects of DNA on transcriptional
regulators, Nature, 392, 885, 1998.
20. Ptashne, M. and Gann, A., Transcriptional activation by recruitment, Nature, 386,
569, 1997.
21. Fairall, L. and Schwabe, J.W.R., DNA binding by transcription factors, in Transcrip-
tion Factors, Locker, J., Ed., BIOS Scientific Publishers Ltd., Oxford, 2001, chap. 4.
Gene Expression 93
22. Fry, C.J. and Peterson, C.L., Chromatin remodeling enzymes: who's on first?, Curr.
Biol., 11, R185, 2001.
23. Maldonado, E., Hampsey, M., and Reinberg, D., Repression: targeting the heart of
the matter, Cell, 99, 455, 1999.
24. Wilusz, C.J. and Wilusz, J., Bringing the role of mRNA decay in the control of gene
expression into focus, Trends Genet., 20, 491, 2004.
25. Shore, P. and Sharrocks, A.D., Regulation of transcription by extracellular signals,
in Transcription Factors, Locker, J., Ed., BIOS Scientific Publishers Ltd., Oxford,
2001, chap. 6.
26. Mayr, B. and Montminy, M., Transcriptional regulation by the phosphorylation-
dependent factor CREB, Nat. Rev. Mol. Cell Biol., 2, 599, 2001.
27. Lonze, B.E. and Ginty, D.D., Function and regulation of CREB family transcription
factors in the nervous system, Neuron, 35, 605, 2002.
28. Waterhouse, P.M., Wang, M.B., and Lough, T., Gene silencing as an adaptive defence
against viruses, Nature, 411, 834, 2001.
29. Bartel, D.P., MicroRNAs: genomics, biogenesis, mechanism, and function, Cell, 116,
281, 2004.
30. Lim, L.P. et al., Vertebrate microRNA genes, Science, 299, 1540, 2003.
31. Ambros, V., The functions of animal microRNAs, Nature, 431, 350, 2004.
32. Nelson, P. et al., The microRNA world: small is mighty, Trends Biochem. Sci., 28,
534, 2003.
33. Meister, G. and Tuschl, T., Mechanisms of gene silencing by double-stranded RNA,
Nature, 431, 343, 2004.
34. Gartenberg, M.R., The Sir proteins of Saccharomyces cerevisiae: mediators of tran-
scriptional silencing and much more, Curr. Opin. Microbiol., 3, 132, 2000.
35. Gottschling, D.E., Gene silencing: two faces of SIR2, Curr. Biol., 10, R708, 2000.
36. Martienssen, R.A. and Colot, V., DNA methylation and epigenetic inheritance in
plants and filamentous fungi, Science, 293, 1070, 2001.
37. Richards, E.J. and Elgin, S.C., Epigenetic codes for heterochromatin formation and
silencing: rounding up the usual suspects, Cell, 108, 489, 2002.
38. Tilghman, S.M., The sins of the fathers and mothers: genomic imprinting in mam-
malian development, Cell, 96, 185, 1999.
39. Ryabinin, A.E., ITF mapping after drugs of abuse: pharmacological versus percep-
tional effects, Acta Neurobiol. Exp., 60, 547, 2000.
40. Graybiel, A.M., Moratalla, R., and Robertson H.A., Amphetamine and cocaine induce
drug-specific activation of the c-fos gene in striosome-matrix compartments and
limbic subdivisions of the striatum, P. Nat. Acad. Sci. U.S.A., 87, 6912, 1990.
41. Hitzemann, B. and Hitzemann, R., Genetics ethanol and the Fos response: a compar-
ison of the C57BL/6J and DBA/2J inbred mouse strains, Alcohol. Clin. Exp. Res.,
21, 1497, 1997.
42. Ryabinin, A.E. et al., Differential sensitivity of c-Fos expression in hippocampus and
other brain regions to moderate and low doses of alcohol, Mol. Psychiatr., 2, 32, 1997.
43. Jouvert, P. et al., Differential rat brain expression of EGR proteins and of the tran-
scriptional corepressor NAB in response to acute or chronic cocaine administration.
Neuromol. Med., 1, 137, 2002.
44. Nestler, E.J., Molecular basis of long-term plasticity underlying addiction, Nat. Rev.
Neurosci., 2, 119, 2001.
45. Monti, J. et al., Expression analysis using oligonucleotide microarrays in mice lacking
bradykinin type 2 receptors, Hypertension, 38, E1, 2001.
46. Yoshihara, T. et al., Differential expression of inflammation- and apoptosis-related

genes in spinal cords of a mutant SOD1 transgenic mouse model of familial amyo-
trophic lateral sclerosis, J. Neurochem., 80, 158, 2002.
47. Cadet, J.L. et al., Temporal profiling of methamphetamine-induced changes in gene
expression in the mouse brain: evidence from cDNA array, Synapse, 41, 40, 2001.
48. Hitzemann, R. et al., A strategy for the integration of QTL, gene expression, and
sequence analyses, Mamm. Genome, 14, 733, 2003.
49. Mannhalter, C., Koizar, D., and Mitterbauer, G. Evaluation of RNA isolation methods
and reference genes for RT-PCR analyses of rare target RNA, Clin. Chem. Lab. Med.,
38, 171, 2000.
50. Thellin, O. et al., Housekeeping genes as internal standards: use and limits, J. Bio-
technol., 75, 291, 1999.
51. Giulietti, A. et al., An overview of real-time quantitative PCR: applications to quantify
cytokine gene expression, Methods, 25386, 2001.
52. Applied Biosystems, User Bulletin #2, Relative Quantitation of Gene Expression:
ABI PRISM 7700 Sequence Detection System, 2001.
53. Bustin, S.A., Absolute quantification of mRNA using real-time reverse transcription
polymerase chain reaction assays, J. Mol. Endocrinol., 25, 169, 2000.
54. Livak, K.J. and Schmittgen, T.D., Analysis of relative gene expression data using
real-time quantitative PCR and the 2(-Delta Delta C(T)) Method, Methods, 25, 402,
2001.
55. Ponchel, F. et al., Real-time PCR based on SYBR-Green I fluorescence: an alternative
to the TaqMan assay for a relative quantification of gene rearrangements, gene
amplifications and micro gene deletions, BMC Biotechnol., 3, 18, 2003.
56. Ziolkowska, B. and Przewlocki, R., Methods used in inducible transcription factor
studies: focus on mRNA, in Immediate Early Genes and Inducible Transcription
Factors in Mapping of the Central Nervous System Function and Dysfunction, Kac-
zmarek, L., and Roobertson, H.A., Eds., Elsevier Science, Amsterdam, 2002, chap. 1.
57. Bartlett, J.M., Differential display: a technical overview, Methods Mol. Biol., 226,
217, 2003.
58. Powell, J., SAGE. The serial analysis of gene expression, Methods Mol. Biol., 99,
297, 2000.
59. Affymetrix, GeneChip Expression Analysis Technical Manual, 2004.
60. Zhang, L., Miles, M.F., and Aldape, K.D., A model of molecular interactions on short
oligonucleotide microarrays, Nat. Biotechnol., 21, 818, 2003.
61. Irizarry, R.A. et al., Exploration, normalization, and summaries of high density
oligonucleotide array probe level data, Biostatistics, 4, 249, 2003.
62. Han, E.S. et al., Reproducibility, sources of variability, pooling, and sample size:
important considerations for the design of high-density oligonucleotide array exper-
iments, J. Gerontol. A Biol., 59, 306, 2004.
63. Holloway, A.J. et al., Options availablefrom start to finishfor obtaining data from
DNA microarrays II, Nat. Genet., 32 Suppl., 481, 2002.
64. Benjamini, Y. and Hochberg, Y. Controlling the false discovery rate: a practical and
powerful approach to multiple testing, J. Roy. Stat. Soc. B, 57, 289, 1995.
65. Slonim, D.K., From patterns to pathways: gene expression data analysis comes of
age, Nat. Genet., 32 Suppl., 502, 2002.
7 Bioinformatics of Behavior
Elissa J. Chesler
CONTENTS
Summary .................................................................................................................. 96
7.1 What Is Bioinformatics? ................................................................................ 96
7.2 Database Design............................................................................................. 97
7.2.1 Data Storage beyond the Spreadsheet: Laboratory Information
Management Systems (LIMS) ........................................................... 97
7.2.2 Data Modeling.................................................................................... 98
7.2.2.1 The Flat Format .................................................................. 98
7.2.2.2 The Hierarchical Format..................................................... 98
7.2.2.3 The Relational Database Management System (RDBMS) 98
7.3 Natural Keys in Biological Databases ....................................................... 99
7.3.1 Genome Location as Reference....................................................... 100
7.3.2 Genes as References ........................................................................ 100
7.3.3 Genetic Reference Populations........................................................ 101
7.3.3.1 Genetic Correlations and Reference Populations ............ 102
7.3.3.2 GRPs for Mapping QTLs ................................................. 103
7.3.4 Gene Sets as References .................................................................. 104
7.3.4.1 Gene Ontology.................................................................. 104
7.3.4.2 The Interactome ................................................................ 104
7.4 Applications ................................................................................................. 104
7.4.1 Sequence Analysis Scoring Matrices and Phylogeny ..................... 105
7.4.1.1 Scoring Matrices............................................................... 106
7.4.1.2 Motif Search and Alignment............................................ 106
7.4.1.3 Structure Analysis............................................................. 106
7.4.2 QTL Candidate Gene Selection....................................................... 107
7.4.3 Microarray, Proteomic, and Other High-Throughput Gene
Set Analysis...................................................................................... 108
7.4.4 Text Mining...................................................................................... 109
7.4.5 Integrating the Genome and the Phenome for Systems-Level
Bioinformatics.................................................................................. 109
7.5 Toward a Bioinformatics of Behavior ......................................................... 110
Acknowledgments.................................................................................................. 111
References.............................................................................................................. 111
95
SUMMARY
This chapter is intended to provide a brief introduction to biological databases for
two major purposesfirst, to familiarize readers with the structure and design of
databases for use in their own laboratories, and second, to illustrate examples of
public biological databases and approaches that have grown from early bioinformatic
methods.
7.1 WHAT IS BIOINFORMATICS?

Bioinformatics is a rapidly developing field that originated with the need for inte-
gration, analysis, and dissemination of rapidly accumulating nucleic acid and protein
sequence data. With the massive influx of data from genome sequencing efforts,
tools for the compilation, storage and assembly were devised. Phylogenic methods
were developed to examine the relationships among species through the comparison
of sequence data. Though biologists now had the string of letters, we only knew a
few of the words and sentences. Annotation was required to parse the genome
sequence and determine the identity and location of genes, their start and stop codons,
introns, exons, enhancers, promoters, and other features. The mouse genome
sequence was largely completed in 2002,1 but the accumulation and annotation of
nucleic acid sequence data continues today. The location of polymorphisms and,
increasingly, their distribution patterns among strains are being evaluated. The
genome, now largely anchored to the C57BL/6J strain, is being generalized across
members of the species, and the study of haplotype blocks24 and linkage disequi-
librium is bringing new understanding to the interaction and selection pressures on
genomes within a species. Newly discovered genome features such as micro-RNAs5,6
and other functional noncoding structures are being annotated.
As behavioral geneticists we attempt to perform a second, much higher-order
annotation of the genome. A major goal of neurobehavioral genetic analysis is to
find genome locations, genes and polymorphisms that influence the variability in
behavioral traits. This brings up an important distinction between the search for
genes underlying behavioral variation and the genomic analysis of behavior. The
former endeavor is the attempt to identify the actual polymorphisms that are segre-
gating in a natural population that are responsible for the heritable variation in a
behavioral phenotypeidentifying the gene in a Mendelian sense, as a mode of
inheritance. The latter effort uses genomic technology such as genome wide
mutagenesis and microarray methods to determine which genes and gene products
are involved in the effector pathway, regardless of whether the pathway members
are polymorphic. Both of these efforts have been fruitful in identifying novel gene-
to-behavior relations.
A third emerging application in bioinformatics will be discussed in the present
chapterthe integration of data resources to address neurobehavioral questions. In
much the same way that functional MRI has aided in the identification of the shared
biological substrates for cognitive processes, we can now use genomic technology
to address hypotheses concerning our understanding of the relationships among
behavioral traits, at the same time identifying sources of individual differences in
Bioinformatics of Behavior 97
the underlying processes. The approach involves the integration of systems-level

trait exploration with knowledge of the genome. Both of these endeavors make use
of the tremendous array of biological databases that have been developed over the
past few decades.
It is not possible to present all of these databases or even a large fraction of
them in this chapter. Those most relevant to readers of this volume have been
compiled at the Neuroscience Database Gateway (NDG) by the Brain Information
Group (BIG) of the Society For Neuroscience, http://big.sfn.org/NDG/site/.7 Some
basic underlying concepts and examples are presented, with the intention of provid-
ing the reader with a framework for understanding the utility of these dynamic
resources. Readers are urged to engage in deep exploration of these databases,
starting with genes or traits of interest. Most are equipped with online tutorials and
extensive documentation. The process of operant conditioning is also quite an effec-
tive approach to training. Click and you shall receive.
7.2 DATABASE DESIGN

7.2.1 DATA STORAGE BEYOND THE SPREADSHEET: LABORATORY
INFORMATION MANAGEMENT SYSTEMS (LIMS)
The integration of massive amounts of biological data requires a shift in data
management practices for most laboratories. It is more effective to archive raw data
from a laboratory in a relational database system than a set of spreadsheets. Though
spreadsheets can be very sophisticated and often have sufficient storage capacity for
the average laboratory, their applications are limited. Most users continue to employ
them for the relative ease with which formulae can be entered and summary statistics
computed. However, this is a dubious benefit. Beyond the most simple descriptive
statistics, analysis using the most common spreadsheet software is not recom-
mended,8 and it is best to store data in a format amenable to more sophisticated
analyses. Spreadsheets do not provide a log of changes, can be erroneously modified
with ease, readily become desynchronized with one another and with external
resources, and are difficult to share without spawning a proliferation of multiple
documents. In contrast, many simultaneous users in multiple sites can share a single
simple database management system (DBMS).
DBMSs are often referred to as laboratory information management systems
(LIMS) when customized for scientific laboratory applications. Though costly com-
mercial systems are available, most typical laboratories will find that a simple
database system such as Filemaker Pro (Filemaker, Inc.) or Access (Microsoft)
will be sufficient. These can be used to track individual subject data (all the way
from breeding and animal colony climate conditions) through bio-behavioral assay
and higher order statistical results derived from them. This type of archiving allows
the ability to store, retrieve and analyze large sets of data from multiple laboratories
for retrospective meta-analysis9 in addition to simple laboratory sample tracking and
error checking. Using open data base connectivity (ODBC), most statistical packages
can query these databases for customization of data tables and views needed for
routine analyses.
7.2.2 DATA MODELING

A brief introduction to some essential concepts in data modeling is presented as a
framework for understanding data storage and database design. The transition from
simple spreadsheets to database systems can be achieved as a progression through
an orderly series of growth from paper archives and spreadsheets to high-capacity
open-source relational databases as the needs of the laboratory grow and change.
7.2.2.1 The Flat Format
Most readers are familiar with the flat data model, in which each database is a self-
contained table consisting of rows and columns. This is the typical entry format for
statistical analysis tools, in which each row is an experimental unit, subject, or record
and each column is an attribute or measured variable of that subject. It is critical to
most applications that the data in a single column be of a single typecharacter or
numeric. Users of common business spreadsheets have the luxury of entering data
of arbitrary types into these cells, rendering import into more sophisticated databases
and statistical packages a nightmare! A frequent example is the technician entered
comment, e.g., 3.45?, 52+, 6.39.6, or 43 gm. Using a rigorous data storage
technique, users can create a series of comment codes that can be later used as
filtering or analytic criteria. Other common violations of the flat file format, such
as entering data from multiple groups in separate areas of the spreadsheet rather
than creating a separate column to indicate the grouping variable, should be avoided.
7.2.2.2 The Hierarchical Format
The hierarchical format is a method for data storage in which each object is a subset
of a larger class of objects. Most users are familiar with this in the form of file
storage systems on Windows, Mac, and Unix operating systems, or even a simple
physical file cabinet. The limitation of the utility of such a tree-based system for
managing diverse, interrelated sets of data becomes readily obvious. Two folders
labeled QTL analysis, and microarray analysis do not provide a unique location
for a new paper on quantitative trait loci (QTL) analysis of microarray data. Most
individual flat files that comprise the various studies conducted in a single laboratory
each have different data structures and often get filed in such a fashion. As a result,
it is difficult to integrate them and analyze a novel, synthetic view of the data.
7.2.2.3 The Relational Database Management System (RDBMS)
Relational databases are the most versatile system for storing large complex sets of
data. In a relational database, data are stored in multiple tables and operations are
performed on tables to result in new tables. The database can be assembled from a
set of flat-format tables through a design process called normalization. Each level
of normalization represents a dramatic reduction in the redundancy of data storage.
The purpose of this type of data model is to avoid a common problem in data storage
systems, that if there are two copies of a datum, one of them will be wrong. This
may seem paradoxical at first, but is a problem that most readers have experienced
TABLE 7.1
Transitioning from Flat Tables to First Normal Form
Flat Format
MOUSE_ID Latency_Trial_1 Latency_Trial_2
A9483 3.42 2.59
A7842 5.43 5.52
First-Normal Form
MOUSE_ID Trial Latency
A9483 1 3.42
A9483 2 2.59
A7842 1 5.43
A7842 2 5.52
firsthand. Using a relational database eliminates this problem because changes made
to one datum will automatically be propagated throughout the entire database. In
the first normal form, each field contains different information. For an example, see
Table 7.1. If latency measures on a maze task were obtained twice for each individual
mouse, the record can be entered in columns with field names Mouse_ID, latency_1,
latency_2. This is a common data entry format for many statistical packages.
However, in the first normal form this would correspond to two records each with
column field names Mouse_ID, latency, score. There are four other levels of
normalization. The end result is a set of tables to which each possesses a primary
key (the unique value for each record in the table) and a set of foreign keys that
refer to records in other tables. Accession numbers are a common example of primary
keys. To create a purely normalized set of tables is a challenging effort that may
exceed the needs of most users, and ideally a balance is struck between efficiency
(obtained by a lack of redundant data) and complexity (obtained by a proliferation
of tables).
7.3 NATURAL KEYS IN BIOLOGICAL DATABASES

The concept of the key is critical to understanding the use and construction of
biological databases. Many of the challenges to data integration emerge from the
instability and lack of uniqueness of keys. Biological data are often collected in
many parallel efforts, in arbitrary forms devised by individual labs using diverse
experimental paradigms. There is an often-expressed hope that somehow these data
will be magically integrated so that biological hypotheses can be meaningfully
extracted and analyzed. To quote J. R. Platt, We speak piously of taking measure-
ments and making small studies that will add another brick to the temple of science.
Most such bricks just lie around in the brickyard.10 Computational integration of
biological data requires that normalized data structures can be created. When data
have a great degree of integration, they can be brought together into a single-flat
table. This is rarely possible from the many diverse parallel experimental efforts that
occur. But genome sequence location provides a natural key. Historically, the
assumption that biological data could be brought together post hoc in such a fashion
probably emerged because of the fortuitous properties of sequence datait anneals
spontaneously. Numerous alignment algorithms have been developed to pull this
data together optimally, but in general, any sequence data can be combined with
other overlapping sequence data to identify and compare individuals, strains, and
species. This is not the case with typical experimental observations, for which
additional data often translate into additional data structure complexity. The unaided
human brain must be used to perform the only integration possible. Other natural
keys are present in much of the data collected by mouse geneticists, the gene and
the mouse strain. To the extent that these natural keys can be well defined and remain
stable, they act as references through which data can be aggregated indefinitely and
integrated analytically.
7.3.1 GENOME LOCATION AS REFERENCE

Several browsers allow the examination of and comparison of genomic features by
location, including Ensembl, www.ensembl.org, (Figure 7.1), and the University of
CaliforniaSanta Cruz Genome Browser, www.genome.ucsc.org. They are excellent
resources that allow efficient viewing of information about genes within a region,
of particular interest to those who have identified a QTL. These displays overlay
multiple aligned data types. They offer highly customizable views to reveal positional
genes, haplotype variation, SNPs, available mutants, gene expression data, conser-
vation and synteny across species, to name a few examples. Most of them can be
queried by BLAST search of DNA or protein sequence, feature name (e.g., gene
symbol or microsatellite name), or position. The data are displayed in many rows
or tracks, and have tremendous flexibility and customization of the display. Users
can identify the location and strain distribution of single nucleotide polymorphisms
using the relatively new SNPview http://snp.gnf.org/ database, or Mouse SNP Selec-
tor at http://zeon.well.ox.ac.uk/rmott-bin/strains.cgi for examining mouse haplotypes
across many strains. Most of these viewers contain links to copious information
about genome features, and even allow the comparison across species of syntenic
regions of the genome.
7.3.2 GENES AS REFERENCES

Using the gene as a reference one can simultaneously analyze genes, mRNA (espe-
cially by transcriptomic assay) and proteins as they vary under various psychological
manipulations. This method can be used to ascribe functional roles to genes. Efforts
to knock out every gene are proposed.11 These mice will be screened on a battery
of neurological and behavioral assays, allowing further annotation of the genome
with higher order biological function. However, there are many challenges to the
use of the gene as a natural key. Genes are surprisingly ill defined. They have a complex
anatomy, including enhancers, promoters, multiple exons and splice-variants, poly-
morphisms, post-translational modifications, methylation states and multiple
sequence records. The nomenclature of genes, and particularly gene products, has
a complex history, with field specific language applying overlapping terms to
groups of genes. A few notable efforts to integrate data at the level of the gene are
FIGURE 7.1 Genome sequence as reference. A screen image of Ensembl Contig View shows
the location of genes, cross-species synteny, microsatellite markers, microarray probes, and
polymorphisms in the region of a QTL for seizure susceptibility.
EntrezGene http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=gene (a replacement

for LocusLink), and MGI12 http://www.informatics.jax.org/. A database of keys to
many gene-centered databases, GeneKeyDB http://genereg.ornl.gov/gkdb/, allows
users to cross-reference sources readily, using WebGestalt, a Web-based gene set
analysis toolkit http://watson.lsd.ornl.gov/genereg/webgestalt/.
7.3.3 GENETIC REFERENCE POPULATIONS

Genetic reference populations (GRPs) including the recombinant inbred strains,
consomics and standard inbred strain panels consist of isogenic populations of related
EFERENCE
CR
ACTGAT TI P
A
ACTATT
GE N
NE
L
000100111011111
001101110010101
Bioinformatics
000010111101101
011101110110111
Knowledge
FIGURE 7.2 Using genetic reference populations to meet the challenge of integrating data
across levels of biological scale. There is tremendous hope and expectation that compiling
biological data using bioinformatics methods will rapidly lead to knowledge. A strong exper-
imental design can greatly facilitate the process. Using a genetic reference population,
sequence polymorphisms, gene annotation, strain variation, molecular, behavioral, physiolog-
ical and neuroanatomical data can be integrated with great facility.
individuals that can be repeatedly sampled through time and space (Figure 7.2). In
this approach, the mouse strain becomes the natural key, allowing the analytic
integration of many diverse data types. A single matrix can be constructed of
attributes by mouse strain, and the data integration is performed using correlational
analyses, though various attributes, for example genotype at a QTL, can be ascribed
a causal role. Where causality cannot be obtained analytically, a new experiment
can be performed to assess the antecedence, necessity and sufficiency of the putative
cause. Two resources for the analysis of these populations are WebQTL
(www.genenetwork.org)1315 and the Mouse Phenome Database.16
7.3.3.1 Genetic Correlations and Reference Populations
A major application of the genetic reference population is that diverse trait data can
be aggregated and correlated indefinitely, allowing the identification of traits that
share pleiotropic regulation. The presence of a genetic correlation implies that the
traits share a common heritable source of variance, which is often the result of
multiple effects of a single polymorphism or mutation; however, numerous other
sources of apparent pleiotropy exist.17 While individual studies allow the determi-
nation of heritability of a phenotype and possible sources of genetic variation, i.e.,
QTLs, the aggregation of multiple studies allows for analysis of the stability of the
phenotype under different environmental conditions, and the interpretation of his-
torical data in the context of novel technologies. For example, in WebQTL, users
can examine relationships among the earliest behavioral phenotypes collected in the
BXD RI strains in the early 1980s to microarray based measures of gene expres-
sion,14,15 a technology that had not even been invented at the time of the initial data
collection! The tremendous advantage of reference populations is that results
obtained can be integrated analytically. WebQTL features a variety of analytic tools,
ranging from simple descriptive statistics, to complex and highly flexible multivariate
data analysis, with results referenced to external genomic resources.15
7.3.3.2 GRPs for Mapping QTLs
Some reference populations have the additional advantage that they can be used for
genetic mapping. These isogenic lines have some known genetic variation that can
be assembled into high precision genome-wide databases and associated with phe-
notypic data.
The recombinant inbred strains are the first and arguably still the best genetic
reference population for mapping. The largest set of these strains, the BXD RIs were
developed in the late 1970s and expanded in the 1990s by B. A. Taylor18 and recently
expanded again to a size of nearly 80 strains.19 The high density of recombinations
obtained results in a dense genotypic map that has been refined and incorporated
into WebQTL.13 These genotypes arise from just two progenitors that were allowed
to randomly segregate. This is an important contrast to the standard inbred strains,
which experienced numerous bottlenecks in their breeding history.20
The standard inbred strains are a tantalizing population for use as a genetic
mapping panel. Early approaches used limited single nucleotide polymorphism data
from a few strains to identify QTLs using these lines of mice,21 but were challenged
due to low statistical power, high error rates, and concerns about the genetic archi-
tecture of the population.22,23 As additional mouse SNP data became available3,4,24
this controversial approach has been revisited. However, the genetic architecture of
these strains is highly complex2 giving rise to concerns about the effects of population
admixture due to the breeding history of the strains that remain.25,26 This has resulted
in the need to restrict these analyses to a small fraction of the existing standard
inbred strainsfewer than that available in the larger RI sets. While the use of
mouse single nucleotide polymorphisms is an excellent approach for refinement of
the QTL interval and selection of additional informative strains for multiple cross
mapping,27,28 the use of these strains for genome wide scans is not yet sufficiently
accurate. The interest in this approach has led to the construction of a variety of
large SNP data sets. Gradually, this data is being incorporated into existing public
genome browsers, and will be a major boon to efforts at QTL candidate gene
identification. Construction of a large, powerful, randomly bred and genetically
diverse recombinant inbred strain panel is underway.29
The consomic lines30 are also a genetic reference population, but they have low
resolution and other limitations. They can be used to map sources of genetic variance
to a single chromosome, allowing for determination of the location of genes with
main effects on phenotypic variation. However, a significant problem with the
consomic lines is that most complex traits are under regulation of multiple genetic
loci, and phenotypes may be a consequence of epistasis, a situation that is not
modeled well in these lines. Occasional evidence of coadaptive alleles is observed,
such that having a pair of genes from common progenitors at two homozygous loci
results in a different phenotype than having either recombinant pair of alleles. In
consomic analysis, it would appear that these two loci have main effects on pheno-
type such that the introgressed chromosome has a main effect on the phenotype
relative to the background chromosome. The chief advantage of the consomic lines
is that they are closer to congenic status than other cross-progeny.
7.3.4 GENE SETS AS REFERENCES

Increasingly, researchers are noting the limitations of gene-at-a-time data presenta-
tion of analysis. This is particularly true as large gene sets are identified and need
to be interpreted as in microarray analysis. Furthermore, it is becoming clear that it is
necessary to examine and catalog the interactions between genes and gene products,
in addition to indexing the genes themselves. This shift in emphasis from the nodes
to the edges between them is evident in the proliferation of interactome databases.
7.3.4.1 Gene Ontology
A major effort is being made to identify meaningful sets of genes that share a
common biological structure, function or compartment.31 Numerous tools exist for
the examination of representation of category members among a gene set identified
through other biological analysis. These tools also allow users to compare submitted
lists of genes with known biological pathways databases KEGG,32 PubMeds GRIF,
and other categorization schemes. Some of these are Web-based, including DAVID33
and the Gene Ontology Tree Machine34 http://genereg.ornl.gov/gotm/ (Figure 7.3),
whereas others are standalone executables including EASE.35 A major advantage of
the standalone tools is that they allow users to store and compare gene sets generated
locally. For gene sets associated with very subtle behavioral phenotypes, the custom
development of gene sets will be necessary because few categorization schemes
reflect in-depth characterizations of behavior at the present time.
7.3.4.2 The Interactome
Multiple databases have been developed to identify genes that physically interact.
This includes DIP,36,37 BIND,38,39 and MINT.40 While the categorization of the phys-
ical interactions between gene products is interesting, it is important to note that in
the case of behavioral analysis, the interactions of gene products are often indirect,
as in the interaction between neurotransmitter synthesis enzymes and receptors.
While these gene products may be highly correlated and coregulated, and may co-
occur in various tissues, the interaction between them is not physical and would not
be encapsulated in the interactome databases.
7.4 APPLICATIONS
A few applications and examples merit further elaboration, though it is impossible
for any text to remain current on this topic. As with any biological result, the user
should make use of documentation to know and understand the source of the data
and the analytic criteria by which results are presented. Regardless of the professional
appearance of the interface, the quality of the underlying data and the validity of
molecular function
biological process cellular component
binding
cell
cellular physiological
process process
nucleic translation intracellular
acid regulator
binding activity cell
growth cytoplasm
and/or
maintenance
translation cytoplasmic
factor vesicle
activity,
nucleic transport
acid coated
binding vesicle
intracellular protein
transport transport
clathrin-coated
translation vesicle
initiation intracellular
factor protein synaptic
activity transport vesicle
FIGURE 7.3 Directed acyclic graphs (DAGs) from WebGestalts GO analysis of genes that
are co-expressed in the brains of isogenic mice indicate covariance among translation initiation
factors, transport mechanisms and synaptic vesicle related genes.
the analytical approach should be evaluated critically. Biological databases are not
perfect animals, and are known to contain errors and inaccuracies. Often, the research
community is able to deposit data with little verification or review. The more
frequently used data are going to be more accurate. This does not diminish the value
of biological databases, but is a caveat to their use as one of many biological tools.
7.4.1 SEQUENCE ANALYSIS SCORING MATRICES AND PHYLOGENY

Many of the tools for sequence analysis and phylogeny, comparison of sets of
sequence data to examine evolutionary questions, have been integrated in a single
interface at University of California San Diegos Biology Workbench41 http://work-
bench.sdsc.edu/. This is an excellent tool for rapid simultaneous retrieval of sequence
data from multiple sources performing multiple sequence alignments, identifying
sets of related sequences and simultaneously manipulating them. Biology Work-
bench has many applications; for example, it can be used to examine sequence
variation between strains in relation to probes and primers and use a variety of other
tools for sequence analysis. The increasing sophistication of genome browsers has
reduced the need for these tools, but significant advantages of Biology Workbench
are that many parameters can be adjusted, and results need not be organized with
respect to a single feature in the genome. Numerous example applications and
tutorials are available at the University of Illinois Biology Student Workbench

http://peptide.ncsa.uiuc.edu/.
7.4.1.1 Scoring Matrices
Sequence comparison, or alignment has much in common with its biological

counterpart, the annealing of DNA. Just as one could regulate hybridization strin-
gency depending on the purpose of the experiment, one could adjust the sensitivity
and specificity of a sequence alignment analysis. Querying sequence data with other
sequence data requires a scoring matrix and penalties for gaps and mismatches of
DNA sequence. Different matrices exist for DNA and amino acid sequence. Low
stringency matrices are used to identify homologs, orthologs and paralogs. This
can be very useful for identification of ESTs. This type of search can be used to
identify novel genes with similarity to known genes. For example, one can retrieve
the protein sequence of an unknown gene identified as a QTL candidate or upreg-
ulated expressed sequence and other databases for sequences using a low-penalty
scoring matrix to find similar, but not exact matches.
7.4.1.2 Motif Search and Alignment
A major application for multiple sequence analysis methods is the identification and
search for common motifs. A popular system for this is MEME, which can be used
to compare several sequences for the identification of motifs, and MAST, which can
then be used to search for the conserved motif throughout the entire genome
(http://meme.sdsc.edu/meme/website/intro.html). These tools can be powerfully
applied to gene sets in which one suspects common functional motifs, and have been
applied to upstream DNA to identify novel transcription factor binding sites that
may be shared by a coregulated gene set. Another application of the search for motifs
is to examine biologically important and conserved sequences.
7.4.1.3 Structure Analysis
Protein structure data can be obtained for many gene products and visualized using
NCBIs Cn3D tool, obtained at http://www.ncbi.nlm.nih.gov/Struc-
ture/CN3D/cn3d.shtml. This user-friendly tool allows users to examine structure
and amino acid sequence relations from structural data retrieved from the Molecular
Modeling DataBase, which contains data from the Protein Data Bank
(http://www.rcsb.org/pdb/).4244 For example, Figure 7.4 shows the ovine serotonin
acetyltransferase, highlighted to reveal residues that are different in the mouse. Trait
relevant mis-sense polymorphisms are likely to influence protein structure or func-
tion. In this example, the ligand binding site is relatively conserved between the two
species, but other regions are much more variable. Using actual structure or results
from prediction, in concert with knowledge of sequence polymorphisms, one can
rapidly predict the molecular consequences of allelic variation and prioritize analysis
of polymorphic loci. The rapid development of structure prediction algorithms will
FIGURE 7.4 Structural analysis of Aanat, the serotonin n-acetyl transferase enzyme. con-
served amino acid sequence between mouse and sheep using Cn3D. Protein structure was
retrieved from Entrez Structure, and mouse amino acid sequence was retrieved via Ensembl.
Sequences were aligned and differences were highlighted using Cn3D.
greatly increase the number of structures available and will soon make this approach
readily applicable.45
7.4.2 QTL CANDIDATE GENE SELECTION

Identifying the naturally occurring polymorphisms responsible for trait variation,
i.e., genes for a trait is as relevant today as ever. Bioinformatics has allowed the
efficient aggregation of data about the genomic features within a QTL interval, and
by including knowledge of candidate polymorphisms alongside information about
the genes, the challenge of going from QTL to QTG is reduced. As mentioned
before, a wealth of information is incorporated in sequence-based browsers, and
these are now becoming a routine first stop in the search for QTL candidate genes.
This process is now tremendously enhanced by viewers that incorporate additional
data sources and overlay them with historic sequence annotation. For example,
the Genome Institute of Novartis has incorporated a large set of tissue specific
expression data that can be examined directly within the genome browser. New tools
specifically devised for QTL candidate analysis inclucing WebQTLs interval map
viewer (Figure 7.5) and the protypical PLAD http://proto.informatics.jax.org/proto-
types/plad/ bring together data on the genes and polymorphisms within a QTL
interval. This organized framework for the analysis of candidate genes allows users
to integrate bioinformatic resources with trait-specific information to identify can-
didates for refined positional analysis, rather than resorting to purely genetic recom-
bination based approaches to refine QTL.
LRS ? Frequency of the Peak LRS ? Red indicates the currently visible
Additive Effect ? SNP Density ? magnified region of chromosome 1.
Mtap2
50.00 Mb 80.00 Mb
Significant LRS = 17.02
Suggestive LRS = 10.22
17.1
15.0
12.5
10.0 +0.20
7.5 +0.15
5.0 +0.10
2.5 +0.05
0.00
50 52 54 56 58 60 62 64 66 68 70 72 74 76 78 80 (-0.05)
(-0.10)
(-0.15)
(-0.20)
17.1
15.0
61.3
12.5
50.0
10.0 0.20
40.0
7.5 0.15
30.0
5.0 0.10
20.0
2.5 0.05 10.0
0.00 0%
-0.05
-0.10
-0.15
17.2 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 x
15.0
39.0
12.5
0.20 30.0
10.0
7.5 0.15
20.0
5.0 0.10
10.0
2.5 0.05
0.00 0%
-0.05
-0.10
-0.15
-0.20
FIGURE 7.5 From QTL to candidate gene. QTL analysis of motor protein Kif5a mRNA
abundance. Mtap2 lies in the significant regulatory locus on Chr. 1, Creb is near the peak
regulatory locus. The bottom panel is a whole genome scan from WebQTL. The middle panel
is a view of Chr. 1, and the top is WebQTL analyzer view, which shows the physical map
position of the QTL, the location of genes in the region, and the SNP density within genes
and across the region. The heavy curve indicates the likelihood ratio statistic across the
genome. The bars indicate the frequency of the QTL peak at a given location over 2000
bootstraps. Horizonal lines indicate suggestive and significant linkage thresholds based on
the permutation analysis. The light line indicates the additive effect of the DBA/2J allele. In
this example, the DBA/2J allele decreases expression of the kinesin 5a transcript.
7.4.3 MICROARRAY, PROTEOMIC, AND OTHER HIGH-THROUGHPUT

GENE SET ANALYSIS
The accumulation, interpretation and comparison of results from the numerous high-
throughput genomic assays can be daunting, but multiple parallel approaches can be
used to understand and analyze the result. It is always critical to keep in mind the
experimental context of the data set and the goals and interpretation of the analytic
method. Without careful attention to these issues, one can easily become mired in a
whirlpool of data. From a list of annotated genes, many questions can be answered
using external annotation ranging from the simple gene set analysis tools, to more
complex approaches aimed at transcriptional (e.g., MEME, MAST) and genetic
(WebQTL) coregulation. Identifying genes that are hypothesized to have these specific
relationships in common is an important first step to defining a meaningful result.
Not all transcripts that are co-expressed are regulated by the same transcription regu-
latory pathways. New tools, including MOTIF http://motif.genome.jp/ and PAINT
http://www.dbi.tju.edu/dbi/tools/paint/index.php?op=FnetBuilder, can be used to
search for known transcription factor motifs among sets of coregulated transcripts.
Microarray analysts often then identify known pathways and systems in which these
genes reside. Numerous tools are available for this type of annotation, and several
incorporate the actual expression results. These include GenMAPP, 46,4 7
DAVID/EASE,33,35 and several commercial tools.
7.4.4 TEXT MINING

Often, data relations are not established well enough for direct analytic integration.
A whole family of tools has been developed to mine this type of free-text information.
The simple PubMed query interface to the Medline database presents lists of related
documents. More advanced tools examine sets of literature for network relations
extracted via natural language processing (Chilibot.net),48 latent semantic relation-
ships in Semantic Gene Organize (SGO),49 and other implicit literature relationships
(Arrowsmith).50 Commercial tools including Ref Viz (ISI ResearchSoft) and Path-
wayAssist (Ariadne Genomics) are also available for text analysis. Tools such as
PubGene allow users to examine the network of literature co-occurences around a
gene. These tools allow users to develop hypotheses concerning novel genegene
or genetrait relationships that have been observed using other bioinformatics or
genomic approaches. All of these tools are subject to the problems of gene annota-
tion, the redundant and high-entropy set of symbols used to refer to genes and gene
products. A heroic effort is being undertaken by the National Library of Medicine
to annotate the existing literature with standard gene nomenclature. This effort, called
Gene Reference Into Function (GRIF) attempts to connect higher-order biological
literature with gene products that it refers to. To date, no automated process has
been able to match this manual curation effort, so while the data are highly accurate,
the database itself will take many years to fully populate.
7.4.5 INTEGRATING THE GENOME AND THE PHENOME

FOR SYSTEMS-LEVEL BIOINFORMATICS
For applications in behavioral genetics, the most exciting developments in bioinfor-

matics have been those that allow low-level genomic information to be integrated
with higher order phenotypic data. This integration can occur by collecting all data
in a reference population, or using a reference gene set or microarray platform.
Exploring the phenome space will provide insight into the relationships among
biological traits. WebQTL13,15 and the Mouse Phenome Database51 are two tools that
allow users to examine genetic correlations of traits across levels of biological scale,
and organ systems. These tools will allow users to explore relationships between
seemingly diverse constructs, such as memory formation and addiction, both related
through some form of neural plasticity. As data are accumulated in the recombinant
or standard inbred strains, these databases will expand dramatically in their utility.
7.5 TOWARD A BIOINFORMATICS OF BEHAVIOR

Three of the major neurobehavioral genetics questions that can be addressed using
bioinformatics are the forward genetics question, Is there a naturally occurring
polymorphism present in this population that influences this behavioral phenotype,
the reverse genetics question, Does manipulation of this gene or gene product
influence a phenotype, and a third question, which makes use of genetic approaches
to answer neurobehavioral questions, Are these traits related to one another through
a shared biological substrate? From this latter question, we can ask very high-level
questions about the relatedness of psychological processes through the black box of
genes and gene products. In this approach, the genetics itself becomes a tool for
understanding the relationships among phenomenologically defined psychobiologi-
cal constructs, i.e., those observed behavioral and personality characteristics that we
attempt to explore by careful examination of operationally defined phenotype mod-
els. By ascribing biological substratese.g., gene setsto these constructs using
bioinformatics approaches, one can discover the naturally occurring categories or
ontology of central nervous system processes.
This type of analysis can be performed using genetic correlations, which require
all data to be obtained in a single reference population, or, the approach can use
genes as references. These genes can be identified across species and across exper-
imental paradigms, in many lines of mice. By identifying the molecular components
of behavioral processes and examining their overlap, we can define the shared and
unique molecular substrates of these processes. This approach can be used to answer
fundamental questions about behavior and mouse models for behavioral disorders.
For example, one can use these genes and gene networks to define mouse behavioral
correlates of human behavioral processes and address challenges of translational
genomics. Do two processes share a common molecular network? Is gene expression
in the molecular network for spatial water maze learning induced by pretraining
exposure to the maze apparatus and other stress-related paradigms? Which genes
are also expressed in less stressful maze learning tasks and thus more specific to
the processes that subserve memory? Are any of them related to cognitive decline
in humans? Identification of the common processes will form an empirically derived
framework of nervous system function.
Building the network from genetic polymorphism to genes to phenotypes, and
then translating that information to humans have been a challenging process, but
compelling successes have been demonstrated. The powerful tools of mouse genetics
allow us to perform thorough molecular dissection of phenotypes. Effectively har-
nessing humanmouse genome relations will allow us to translate these polymor-
phisms, gene and phenotypes to other species, including our own. As in the past,
we can gain much from the experimental precision and control in the mouse,
generating novel and highly informative hypotheses of genetic mediation of human
behavior.
ACKNOWLEDGMENTS
Jeremy L. Peirce and Susan E. Bergeson provided helpful comments on this chapter.
REFERENCES
1. Waterston, R.H. et al., Initial sequencing and comparative analysis of the mouse
genome. Nature, 2002. 420(6915): p. 52062.
2. Yalcin, B. et al., Unexpected complexity in the haplotypes of commonly used inbred
strains of laboratory mice. Proc Natl Acad Sci U S A, 2004. 101(26): p. 97349.
3. Wiltshire, T. et al., Genome-wide single-nucleotide polymorphism analysis defines
haplotype patterns in mouse. Proc Natl Acad Sci U S A, 2003. 100(6): p. 33805.
4. Wade, C.M. et al., The mosaic structure of variation in the laboratory mouse genome.
Nature, 2002. 420(6915): p. 5748.
5. Lagos-Quintana, M. et al., Identification of novel genes coding for small expressed
RNAs. Science, 2001. 294(5543): p. 8538.
6. Ambros, V., microRNAs: tiny regulators with great potential. Cell, 2001. 107(7):
p. 8236.
7. Smaglik, P., Internet gateway planned for neuroinformatics data. Nature, 2000.
405(6787): p. 603.
8. McCullough, B.D. and B. Wilson, On the accuracy of statistical procedures in
Microsoft Excel 2000 and Excel XP. Computational Statistics & Data Analysis, 2002.
40: p. 713721.
9. Chesler, E.J. et al., Identification and ranking of genetic and laboratory environment
factors influencing a behavioral trait, thermal nociception, via computational analysis
of a large data archive. Neurosci Biobehav Rev, 2002. 26(8): p. 90723.
10. Platt, J.R., Strong inference. Certain systematic methods of scientific thinking may
produce more rapid progress than others. Science, 1964. 146(3642): p. 347353.
11. Austin, C.P. et al., The knockout mouse project. Nat Genet, 2004. 36(9): p. 9214.
12. Bult, C.J. et al., The Mouse Genome Database (MGD): integrating biology with the
genome. Nucleic Acids Res, 2004. 32(Database issue): p. D47681.
13. Wang, J., R.W. Williams, and K.F. Manly, WebQTL: web-based complex trait analysis.
Neuroinformatics, 2003. 1(4): p. 299308.
14. Chesler, E.J. et al., Genetic correlates of gene expression in recombinant inbred
strains: a relational model system to explore neurobehavioral phenotypes. Neuroin-
formatics, 2003. 1(4): p. 34357.
15. Chesler, E.J. et al., WebQTL: rapid exploratory analysis of gene expression and
genetic networks for brain and behavior. Nat Neurosci, 2004. 7(5): p. 4856.
16. Bogue, M.A. and S.C. Grubb, The Mouse Phenome Project. Genetica, 2004. 122(1):
p. 714.
17. Hodgkin, J., Seven types of pleiotropy. Int J Dev Biol, 1998. 42(3): p. 5015.
18. Taylor, B.A. et al., Genotyping new BXD recombinant inbred mouse strains and
comparison of BXD and consensus maps. Mamm Genome, 1999. 10(4): p. 33548.
19. Peirce, J.L. et al., Genetic architecture of the mouse hippocampus: identification of
gene loci with selective regional effects. Genes Brain Behav, 2003. 2(4): p. 23852.
20. Beck, J.A. et al., Genealogies of mouse inbred strains. Nat Genet, 2000. 24(1):
p. 235.
21. Grupe, A. et al., In silico mapping of complex disease-related traits in mice. Science,
2001. 292(5523): p. 19158.
22. Darvasi, A., In silico mapping of mouse quantitative trait loci. Science, 2001.
294(5551): p. 2423.
23. Chesler, E.J., S.L. Rodriguez-Zas, and J.S. Mogil, In silico mapping of mouse quan-
titative trait loci. Science, 2001. 294(5551): p. 2423.
24. Pletcher, M.T. et al., Use of a dense single nucleotide polymorphism map for in silico
mapping in the mouse. PLoS Biol, 2004. 2(12): p. e393.
25. Mhyre, T. et al., Heritability, correlations, and in silico mapping of locomotor behavior
and neurochemistry in inbred strains of mice. Genes, Brain, Behavior, 2005. 4:
20928.
26. Churchill, G.A., Researchers didnt scurry around fields looking for mice, says Jack-
son Lab scientist, BioMed Central, July 7, 2004.
27. Hitzemann, R. et al., Multiple cross mapping (MCM) markedly improves the local-
ization of a QTL for ethanol-induced activation. Genes Brain Behav, 2002. 1(4): p.
21422.
28. Shifman, S. and A. Darvasi, Mouse inbred strain sequence information and yin-yang
crosses for quantitative trait locus fine mapping. Genetics, 2005. 169(2): p. 84954.
29. Churchill, G.A. et al., The Collaborative Cross, a community resource for the genetic
analysis of complex traits. Nat Genet, 2004. 36(11): p. 11337.
30. Singer, J.B. et al., Genetic dissection of complex traits with chromosome substitution
strains of mice. Science, 2004. 304(5669): p. 4458.
31. Ashburner, M. et al., Gene ontology: tool for the unification of biology. Nat Genet,
2000. 25: p. 2529.
32. Ogata, H. et al., KEGG: Kyoto Encyclopedia of Genes and Genomes. Nucleic Acids
Res, 1999. 27(1): p. 2934.
33. Dennis, G., Jr. et al., DAVID: Database for Annotation, Visualization, and Integrated
Discovery. Genome Biol, 2003. 4(5): p. P3.
34. Zhang, B. et al., GOTree Machine (GOTM): a web-based platform for interpreting
sets of interesting genes using gene ontology hierarchies. BMC Bioinformatics, 2004.
5(1): p. 16.
35. Hosack, D.A. et al., Identifying biological themes within lists of genes with EASE.
Genome Biol, 2003. 4(10): p. R70.
36. Xenarios, I. et al., DIP, the Database of Interacting Proteins: a research tool for
studying cellular networks of protein interactions. Nucleic Acids Res, 2002. 30(1):
p. 3035.
37. Xenarios, I. et al., DIP: the database of interacting proteins. Nucleic Acids Res,
2000. 28(1): p. 28991.
38. Bader, G.D. et al., BINDThe Biomolecular Interaction Network Database. Nucleic
Acids Res, 2001. 29(1): p. 2425.
39. Alfarano, C. et al., The Biomolecular Interaction Network Database and related tools
2005 update. Nucleic Acids Res, 2005. 33 Database Issue: p. D41824.
40. Zanzoni, A. et al., MINT: a Molecular INTeraction database. FEBS Lett, 2002.
513(1): p. 13540.
41. Subramaniam, S., The Biology Workbencha seamless database and analysis envi-
ronment for the biologist. Proteins, 1998. 32(1): p. 12.
42. Berman, H.M. et al., The Protein Data Bank. Nucleic Acids Res, 2000. 28(1): p.
23542.
43. Berman, H.M. et al., The Protein Data Bank. Acta Crystallogr D Biol Crystallogr,
2002. 58(Pt 6 No 1): p. 899907.
44. Bhat, T.N. et al., The PDB data uniformity project. Nucleic Acids Res, 2001. 29(1):
p. 2148.
45. Berman, H.M. and J.D. Westbrook, The impact of structural genomics on the protein
data bank. Am J Pharmacogenomics, 2004. 4(4): p. 24752.
46. Dahlquist, K.D. et al., GenMAPP, a new tool for viewing and analyzing microarray
data on biological pathways. Nat Genet, 2002. 31(1): p. 1920.
47. Doniger, S.W. et al., MAPPFinder: using Gene Ontology and GenMAPP to create a
global gene-expression profile from microarray data. Genome Biol, 2003. 4(1): p. R7.
48. Chen, H. and B.M. Sharp, Content-rich biological network constructed by mining
PubMed abstracts. BMC Bioinformatics, 2004. 5(1): p. 147.
49. Homayouni, R. et al., Gene clustering by latent semantic indexing of MEDLINE
abstracts. Bioinformatics, 2005. 21(1): p. 10415.
50. Smalheiser, N.R. and D.R. Swanson, Using ARROWSMITH: a computer-assisted
approach to formulating and assessing scientific hypotheses. Comput Methods Pro-
grams Biomed, 1998. 57(3): p. 14953.
51. Grubb, S.C., G.A. Churchill, and M.A. Bogue, A collaborative database of inbred
mouse strain characteristics. Bioinformatics, 2004. 20(16): p. 28579.
8 Congenic and Consomic

Strains
Lorraine Flaherty and Valerie Bolivar
CONTENTS
Introduction............................................................................................................ 115
8.1 Congenic Strains .......................................................................................... 116
8.2 Consomic Strains ......................................................................................... 119
8.3 Knockout/Congenic Strains ......................................................................... 119
8.4 Mapping QTLs with Knockout/Congenic Mice.......................................... 123
References.............................................................................................................. 126
INTRODUCTION
Genetically defined mouse strains have become invaluable in biomedical research
because of their ability to clarify genetic vs. environmental influences on a wide
variety of traits including behavior. They have provided the means to test multiple
genetically identical or nearly identical animals at one time, thus reducing the effects
of genetic variation to a minimum. Some of the most useful types include inbred,
co-isogenic, congenic, consomic, recombinant inbred, knockout, and transgenic
strains.
The most commonly used strain of mouse is the inbred strain. Since these strains
are made by brothersister mating mice for more than 20 generations, any two mice
within an inbred strain will be genetically identical at 99.99% of their genetic loci.
A co-isogenic strain is similar to an inbred strain except that it differs from its inbred
partner by only one locus.1 These co-isogenic strains are usually the result of
spontaneous mutations that occur within an inbred stock and have been very useful
in analyzing the effects of single genes on a particular behavior or trait. However,
the investigator must either wait for mutations to occur spontaneously or induce
them by use of mutagenesis techniques, both requiring much time and effort. Because
of the usefulness of these co-isogenic strains, in the 1940s, George Snell constructed
a breeding scheme to develop a new type of strain, called the congenic strain, which
approximates a co-isogenic strain. These congenic strains have become extremely
valuable resources in analyzing single gene effects. Below we describe how these
strains are made and some of their properties. We also describe a different use for
congenic strains in mapping genes that influence quantitative traits such as behavior.
115
Finally we describe a variation of the congenic strain, called the consomic strain,
which selects for genetic differences covering an entire chromosome instead of a
single genetic locus.
8.1 CONGENIC STRAINS

A congenic strain is made by repeatedly backcrossing one strain to another and
selecting for an allele at a particular locus, called the differential locus, at every
generation (see description below for details on construction of such strains). This
process is called introgression where the donor genetic material is said to be intro-
gressed into the recipient genome. There are several different breeding schemes that
can be used to construct a congenic strain.2 The most common is illustrated in Figure
8.1. Two inbred strains are chosen to begin this processa donor strain and a recipient
strain. The donor strain contains the donor allele at the differential locus. The recipient
strain is usually chosen based on its breeding ability and characteristics. Usually it is
a strain that is readily available and used as a standard in a particular aspect of biology,
like behavior. In most cases, it is the C57BL/6 (B6) strain although other strains such
as DBA/2 or BALB/c are also used. The first cross is a simple cross between two
inbred strains, leading to an F1 (standing for first filial generation), also indicated as
N1 (standing for the first backcross generation). This F1 is then crossed to the recipient
inbred strain to yield an N2 generation. At the N2 generation and each subsequent
generation, the mice are typed at the differential locus and mice are selected which
possess the donor allele, symbolized by a in Figure 8.1. At the end of 10 or more
generations of selection and backcrossing, the mice are intercrossed and a strain that
is homozygous at the differential locus is maintained by brothersister matings. In the
Figure 8.1 example, the strain would be called B.G-a. This would stand for the recipient
strain and donor strain separated by a period and followed by a hyphen and the selected
allele. In some cases, less than 10 generations of backcrossing is used. However, here,
the genetic background will contain more contaminant or passenger genetic material
from the donor strain (Tables 8.1 and 8.2).
Making congenic strains that isolate quantitative trait loci (QTLs) is more diffi-
cult, especially when behavioral phenotypes are concerned. Since there is often a
large variation in the behavioral phenotypes of animals even with the same genotype,
the scoring of an individual animal often does not reflect its exact genotype at the
locus in question. Thus, relying on phenotype alone for selection of individuals within
the congenic strain breeding scheme will often be misleading. Progeny testing is often
necessary to confirm this genotypephenotype relationship. Thus, many geneticists
studying complex traits have chosen to make congenic strains by the alternative way
selection for flanking DNA markers framing their locus of interest.3,4,5 Since most of
the mappings of these QTLs give a log of odds (LOD) probability map for the location,
selection of flanking markers is usually based on a confidence interval for the QTL
(see Abiola6 for a discussion of QTL mapping). The distance between these flanking
markers varies considerably and depends on the sharpness of the LOD score peak as
well as the judgment of the investigator making the strain. In general, the rule of
thumb is to pick a wide enough interval (>20 cM on each side of the peak LOD
score) so that the investigator is assured that the QTL will not fall outside
Congenic and Consomic Strains 117
Strain G Strain B
Locus F
Selection for allele at Locus F
a b
Differential chromosome
a b
Cross I. X
Other 19 chromosome
b
(GXB)F1 (or N1)
a b
Backcross to Strain B
b X b
Cross 2-10. and at each generation
select for the a allele at
the differential locus
a
b
N2
b
N10
a a
Cross 11. b b Intercross at N10 to

make homozygous
at differential locus
a
B . G - a N10F1
FIGURE 8.1 Derivation of a congenic strain of mice. Selection for the a allele at locus F
of the G strain (donor) at every generation.
TABLE 8.1
Congenic and Consomic Strains of Mice and Amount of Contaminant Genetic
Material*
Average Amount of
Strain: Recipient Differential Genetic Unlinked Contaminant
Strain (donor) Region (average) (cM) Genetic Region (cM)
Consomics B6 (A) B6 (Spretus) 160 cM 3 cM
Congenics (>N10) B6 (various) 20 cM 3 cM
Speed congenics (N5) B6 (various) 40 cM 16 cM
Knockout/congenics
B6 (129) 20 cM 3 cM
(>N10)
Genome tagged mice
B6 (DBA/2, CAST) 4050 cM 48 cM
(N4-N5)
*References for calculations are 1, 2, 8, 17, 22.
TABLE 8.2
Amount of Genomic Material That Is Retained as a Function of the
N Number of Backcross Generations
Congenic
Consomic
N (other 19 chrs) Unlinked Linked
2 1282 Mb* 1282 Mb 140.6 Mb
4 321 321 82.0
8 20 20 46.7
10 5.0 5.0 37.5
16 0.1 0.1 23.4
20 <0.01 <0.01 18.7
*Numbers assume a genome size of 2700 Mb, a recombinant map size of 1600 cM and
an average chromosome size of 135 Mb. References are 1, 2.
of it. Additional recombinational events in the interval can then be used to divide the
region for further analysis. Congenic strains where selection is based on flanking
markers are often called interval-specific congenic strains.7
Another breeding scheme that has become more popular in recent years, because
of the advent of high-throughput genotyping, is marker-assisted selection and results
in quasi-congenic strains, commonly called a speed congenic.8,9 Speed congenic
strains start the same way as conventional congenics. An F1 is made between
two inbred strains of mice and the donor allele is selected in each generation. However,
at every subsequent generation, founder mice are selected that (1) retain the differential
allele and (2) have a minimum of unlinked genetic material from the donor
strain. Here, the unlinked contaminant genetic material is detected by use of a genome
scan of DNA markers spaced at regular intervals on the 19 nonsyntenic chromosomes.
The optimal spacing of these markers should be at approximately 20 cM intervals.10,11
At least 20 mice from each generation must be typed in a genome scan to obtain mice
with a minimal number of unlinked genes. With this technique, one can make the
equivalent of an N10 congenic in 4 to 5 generations, thus, reducing the time for
completion by at least one half (Table 8.1). However, the creation of these speed
congenics also requires more technician time and genotyping costs because of the
number of mice and loci that must be typed. For a more thorough discussion of this
technique, see Markel et al.8
8.2 CONSOMIC STRAINS

More recently, consomic strains of mice have been made by similar techniques.
These strains are also called chromosome substitution strains. Here, whole chromo-
somes, instead of single genes have been used for selection. These are made in a
similar fashion to congenic strains except that markers spread evenly along the
selected or differential chromosome are used for selection (Figure 8.2). These con-
somic strains are useful for preliminary mapping studies of quantitative traits since
only 20 of them will be necessary to map a trait to a particular chromosome.
The idea of studying the influence of one chromosome at a time and the con-
struction of consomic strains began with studies of the Y chromosome. Since there
is no recombination (or extremely little) between the Y and X chromosomes, a Y
consomic is simply made by mating two strains together and selecting for the male
at every generation.12 However, with the advent of a high density DNA marker map,
consomics of other chromosomes became possible, including a set of consomics
made by Jean-Louis Guenet and his group.13 These consomics involved B6 intro-
gressed with Mus spretus. A full set of consomics involving the B6 and A strains
has recently been made in the laboratory of Joseph Nadeau and are now commer-
cially available through the Jackson laboratory.14 This new consomic panel has
already been analyzed in two behavioral paradigms15,16 and it promises to be a very
valuable tool to the behavioral geneticist. A consomic strain is designated by first
giving the name of the recipient strain, followed by a hyphen, followed by a number,
indicating the transferred chromosome, and then by the name of the donor strain in
a superscript. Thus, a B6 strain with an introgressed Chromosome 3 from the A
strain would be designated B6-3A or B6-3<A>, if a superscript font is not available.
8.3 KNOCKOUT/CONGENIC STRAINS

When homologous recombination is used to inactivate a gene, the process is called
gene ablation and the strain is commonly referred to as a knockout strain. This
process involves the elimination of a functional gene in embryonic stem cells either
by replacement with a nonfunctional one or by its deletion. Since most embryonic
stem cells are derived from the 129 strain, the knockout mice are generally on a 129
background. However, most investigators prefer a B6 strain background and cross
Strain G Strain B
Selection for markers spaced
abcdefghi evenly along chromosome, e.g., Chr 3
Differential chromosome
Cross I. X
Other 19 chromosomes
abcdefghi
(GXB)F1 (or N1)
abcdefghi abcdefghi
Backcross to Strain B
Cross 2-10.
X and at each generation select
for markers spaced along
chromosome
abcdefghi
N2
abcdefghi
N10
abcdefghi abcdefghi
Cross 11. Intercross at N10 to

make homozygous for
entire chromosome
abcdefghi
abcdefghi
B-3GN10F1
FIGURE 8.2 Derivation of a consomic or chromosome substitution strain. Selection for

Chr. 3 of the G strain at every generation.
these 129 knockout strains to B6. This leads to a congenic strain with a section of
129 introgressed into the B6 background. Thus, when these knockout/congenic
strains are tested for a phenotype that differs from the B6 strain, this phenotype
Target locus
B6
Differential
From Chromosome
129 ES cell X
Other 19
chromosomes
Target locus
Flanking 129 differential

Chromosomal region
Possible unlinked contaminant 129

genetic segment
FIGURE 8.3 Derivation of a knockout/congenic strain. Striped areas indicate chromosomal

regions from 129. Ablated gene indicated with an arrow.
could either be due to the gene that is ablated or to the flanking 129 chromosomal
segment (Figure 8.3).17 This consideration is important for two reasons. First, one
should be cautioned when using knockout/congenic strains to understand that the
observed phenotype is not always due to the targeted locus. There have been several
reviews, covering the cautions attached to the use of knockout/congenics.1821 This
situation is particularly poignant when considering phenotypes that are quantitatively
different in B6 vs. 129 and when the phenotype is not predicted from the tissue
distribution of gene expression and/or the known function of the ablated gene.
Second, the knockout/congenic strains can be used, in a positive way, to scan the
genome for genes that influence quantitative traits such as behavior.17 Here, the
investigator can use these strains as a panel of congenic strains with different
segments of 129 from a wide variety of chromosomes introgressed into the B6
background. Thus, by testing these strains, the investigator will actually be testing
for the QTL location somewhere in the flanking regions surrounding the targeted
loci (see below). In this respect, there are more than 50 knockout/congenic strains
available through commercial sources. Since the average length of the differential
129 chromosomal segment in these strains is about 20 to 30 cM (see Table 8.3 for
a representative list), this assortment of knockout/congenic strains offers the
researcher a set of strains that will cover most of the mouse genome. Moreover,
often these strains provide a set of overlapping 129 chromosomal segments that will
map a particular trait. Here the region of overlap is often small and may define the
boundaries of the genetic region determining the trait. For example, in Figure 8.4,
the QTL could be narrowed to a short segment by the overlap between B and C.
TABLE 8.3
Length of Differential Chromosomal Segments in a Sampling of
Knockout/Congenic Strains*
Length of 129
Position in cM Chromosomal
Targeted Gene N Number Chr. of Target Gene Stretch**
Apoe 10 7 4 1622
Cd3z 8 1 87 910
Fcgr3 6 1 92 2535
Igl-5 6 16 10 1632
Il6 11 5 17 916
Il10 10 1 70 3341
Lck 12 4 59 814
Selp 10 1 86 2935
Tap1 10 17 19 35
*Information on position of targeted gene and number of backcross generations was taken from
www.jax.org.
**Numbers indicate minimum to maximum lengths of chromosomal segment as defined by typed

flanking DNA markers.
Trait
A yes
B yes
C yes
D no
F no
FIGURE 8.4 Mapping QTLs with overlapping differential chromosomes. Striped areas indi-
cate the chromosomal stretch on the differential chromosome that is derived from 129. The
arrows indicate the region that maps the trait.
Along this same line of reasoning, Iakoubova et al.22 constructed a similar

assortment of quasi-congenic strains of mice which they have called genome-tagged
mice. There are two sets of these mice described by Iakoubova and colleagues
one with small sections of DBA/2 introgressed into the recipient B6 strain and
one with small sections of CAST/Ei introgressed into B6. Because of the way that
these strains were derived, they often have more unlinked contaminant passenger
genetic material than a standard congenic strain (Table 8.1). So far, these are not
available commercially; however, they have been used to map several genes affecting
behavior.23,24
8.4 MAPPING QTLS WITH KNOCKOUT/CONGENIC

MICE
The major use of knockout/congenic strains is to study a given gene on a defined
genetic background. However, these same strains can be used to map genes affecting
quantitative traits such as behavior.17 It must be remembered that these strains differ
not only at the target locus, but also at a short chromosomal stretch immediately
surrounding the target locus. Because these strains are readily available, they serve
as an extensive series of congenic strains that can be used to study short chromosomal
regions that surround the indicated ablated target loci.
Thus, by comparing a series of the knockout/congenic strains with their inbred
partner, B6, we were able to map a number of behavioral traits to precise chromo-
somal locations in or around the target locus. By further matings, we were able to
prove that at least some of these traits were not due to the ablated target but rather
due to the flanking genetic region.17
The data in Figure 8.5 illustrate this point. We screened 15 knockout/congenic
mouse strains in our exploratory activity assay. This assay consisted of a 5-minute
test in a darkened activity monitor.25 The strains were selected based on their
chromosomal position and the likelihood that they may affect behavior. We also
tested these same mice for habituation to the open field by continuing this test system
for 2 more consecutive days and measuring the change in activity from day 1 to day
3.25 We have successfully used these assays in the past to establish exploratory
activity and habituation differences among inbred stains/substrains of mice2528 and
between neurological mutant mice and controls.2931
As shown in Figure 8.5A, we found that a number of these strains behaved
differently from B6 in first day exploratory activity. Based on the performance of
these knockout/congenic strains, there are genes affecting this activity on Chrs. 1,
9, 14, and 15. By use of additional backcross tests, we have now confirmed some
of these map locations on Chrs. 1 and 15. Interestingly, genes affecting habituation
scores also seem to be affected by some of these same regions (Figure 8.5B).
Some strains were significantly less active than B6, i.e., behaved more like 129,
including B6-Il10. This intermediate activity of the knockout/congenic is explainable
by an additive mode of inheritance where the knockout/congenic strain should have
an activity between B6 and 129. If the mode of inheritance is either dominant or
recessive, the knockout/strain should have an activity approximately equal to one of
its parents. This may be the case for B6-Il7r in first day exploratory activity.
The high activity of the two strains, B6-Lipc and B6-Tcrd, is more difficult to
explain. Their activity does not correspond to either a dominant, recessive or additive
mode of inheritance. There are three possible explanations. First, the increased
activity may be due to the effects of the ablated gene itself. In the case of Tcrd
(T cell receptor d), this explanation seems unlikely since this gene has no known
influence on behavior or brain function. A second explanation could be that a
mutation, either linked or unlinked to the differential locus, has occurred in the
production of this B6 knockout/congenic strain in a gene that is important to this
exploratory behavior. Again, this seems like an unlikely explanation because of the
high incidence of this phenomenon with these knockout/congenics (Bolivar and
2500
2000
* * A
1500
Total distance traveled * *
1000
on Day 1 *
500
700 * B
500
Habituation score 300

(Day 1 minus Day 3)
* * *
100
-100
-300
129 Il10 B2m Cd8a Apoe Lipc Il7r Tap1 Fmr1
B6 Fcgr3 Il6 Cd4 Apoa1 Tcrd Igl5 Tnfsf5
FIGURE 8.5 Exploratory activity and habituation of inbred and knockout/congenic mouse
strains. The knockout mouse strains tested included B6.129P2-Il10, B6.129P2-Fcgr3,
B6.129P2-B2m, B6.129S2-Il6, B6.129S2-Cd8a, B6.129S2-Cd4, B6.129P2-Apoe, B6.129P2-
Apoa1, B6.129P2-Lipc, B6.129P2-Tcrd, B6.129S7-Il7r, B6.129S2-Igl-5, B6.129S2-Tap1,
B6.129S2-Tnfsf5 and B6.129P2-Fmr1. Knockout mice are all on a B6 background and are
indicated by the name of the ablated target gene. Two inbred strains were also tested: B6/J
and 129P2 (129). At least 24 male mice of each strain were tested. Results are shown + s.e.m.
There was a significant effect of strain for first day activity (F16,491 = 12.206, p < .0001; Panel
A) and habituation (F16,491 = 8.885, p < .0001; Panel B). Tukeys posthoc tests were then used
to make pairwise comparisons. An * indicates that a knockout/congenic strain is significantly
different from B6.
Flaherty, unpublished results). Finally, and most likely, there could be B6 modifier
genes that are interacting with 129 flanking genes to produce a heightened effect.
This explanation seems the most reasonable based on the high frequency of genetic
background effects on behavioral traits.
For intersession habituation (Figure 8.5B), we have also found the B6-Il7r has an
abnormally low habituation score. Since the Il7r gene is located on Chr. 15, this
agrees with our previous data showing a strong gene affecting habituation on
this same chromosome.26 Moreover, we have subsequently confirmed the location
of a gene influencing habituation on Chr. 15 by testing a Chr. 15 consomic strain,
B6-15A.16 This strain was significantly less active in the open field (Figure 8.6A),
and habituated less (Figure 8.6B) than B6 mice.
Thus, both congenic and consomic strains have been successful tools for the
behavioral geneticist in investigating single gene effects as well as in mapping genes
1600
1400 A
1200 *
1000
Total Distance
800
Traveled (cm)
600
400
200
700
600 B
500
Habituation score 400

(Day 1 minus Day 3) 300
200
*
100
B6 B6-15A
FIGURE 8.6 Exploratory activity and habituation of B6 and B6-15A strains. Twenty-four
male mice of each strain were tested. Results are shown + s.e.m. B6 mice were significantly
more active on the first day of testing (T46 = 3.747, p = .0005; Panel A) and displayed more
habituation (T46 = 4.203, p = .0001; Panel B) than B6-15A mice. An * indicates a significant
difference between the two groups.
that influence behavioral traits. The advantage of using these strains for mapping is
that they can provide a convenient starting place for fine mapping studies. Here, the
creation of subcongenic strains that are derivatives of congenic strains and made by
further recombinational events inside the introgressed differential chromosomal
region can isolate smaller chromosomal segments that are small enough to be
convenient for positional cloning efforts. These new subcongenics can then be used
to make an overlapping map that isolates the behavioral QTL to a shorter segment
of chromosome (see Figure 8.3 for an example of mapping using overlapping
differential chromosomal regions). Since multiple and genetically identical subcon-
genic mice can be made, the testing of these subcongenic mice will help to resolve
weak effects that cannot be detected with more conventional intercross techniques.
Congenics, consomics, and knockout/congenics are readily available from

commercial sources. They are valuable tools to the behavioral geneticist and have
been successfully used to study genetic background effects as well as map QTLs.
However, to use these strains correctly, the investigator should have a proper under-
standing of their construction and genetic origin.
REFERENCES
1. Silver, L.M. Mouse Genetics Concepts and Applications, Oxford University, New
York, 1995.
2. Flaherty, L. Congenic strains. in The Mouse in Biomedical Research, Vol. 1, eds.
Foster, H.L., Small, J.D. & Fox, J.G., 21522, Academic Press, New York, 1981.
3. Bennett, B. Congenic strains developed for alcohol- and drug-related phenotypes.
Pharmacol Biochem Behav, 67: 67181, 2000.
4. Radcliffe, R.A., Bohl, M.L., Lowe, M.V., Cycowski, C.S. & Wehner, J.M. Mapping
of quantitative trait loci for hypnotic sensitivity to ethanol in crosses derived from
the C57BL/6 and DBA/2 mouse strains. Alcohol Clin Exp Res, 24: 133542, 2000.
5. Morel, L., Yu, Y., Blenman, K.R., Caldwell, R.A. & Wakeland, E.K. Production of
congenic mouse strains carrying genomic intervals containing SLE-susceptibility
genes derived from the SLE-prone NZM2410 strain. Mamm Genome, 7: 3359, 1996.
6. Abiola, O. et al. The nature and identification of quantitative trait loci: a community's
view. Nat Rev Genet, 4: 9116, 2003.
7. Darvasi, A. Interval-specific congenic strains (ISCS): an experimental design for
mapping a QTL into a 1-centimorgan interval. Mamm Genome, 8: 1637, 1997.
8. Markel, P. et al. Theoretical and empirical issues for marker-assisted breeding of
congenic mouse strains. Nat Genet, 17: 2804, 1997.
9. Wakeland, E., Morel, L., Achey, K., Yui, M. & Longmate, J. Speed congenics: a
classic technique in the fast lane (relatively speaking). Immunol Today, 18: 4727,
1997.
10. Visscher, P.M. Speed congenics: accelerated genome recovery using genetic markers.
Genet Res, 74: 815, 1999.
11. Servin, B. & Hospital, F. Optimal positioning of markers to control genetic back-
ground in marker-assisted backcrossing. J Hered, 93: 2147, 2002.
12. Hudgins, C.C., Steinberg, R.T., Klinman, D.M., Reeves, M.J. & Steinberg, A.D.
Studies of consomic mice bearing the Y chromosome of the BXSB mouse. J Immunol,
134: 384954, 1985.
13. Santos, J. et al. A new locus for resistance to gamma-radiation-induced thymic
lymphoma identified using inter-specific consomic and inter-specific recombinant
congenic strains of mice. Oncogene, 21: 66803, 2002.
14. Nadeau, J.H., Singer, J.B., Matin, A. & Lander, E.S. Analysing complex genetic traits
with chromosome substitution strains. Nat Genet, 24: 2215, 2000.
15. Singer, J.B., Hill, A.E., Nadeau, J.H. & Lander, E.S. Mapping quantitative trait loci
for anxiety in chromosome substitution strains of mice. Genetics, 169: 85562, 2005.
16. Singer, J.B. et al. Genetic dissection of complex traits with chromosome substitution
strains of mice. Science, 304: 4458, 2004.
17. Bolivar, V.J., Cook, M.N. & Flaherty, L. Mapping of quantitative trait loci with
knockout/congenic strains. Genome Res, 11: 154952, 2001.
18. Crusio, W.E. Flanking gene and genetic background problems in genetically manip-
ulated mice. Biol Psychiatry, 56: 3815, 2004.
19. Wolfer, D.P., Crusio, W.E. & Lipp, H.P. Knockout mice: simple solutions to the
problems of genetic background and flanking genes. Trends Neurosci, 25: 33640,
2002.
20. Gerlai, R. Gene targeting: technical confounds and potential solutions in behavioral
brain research. Behav Brain Res, 125: 1321, 2001.
21. Gerlai, R. Afraid of complications in gene targeting? Anxiety plays a role! Trends
Neurosci, 25: 136, 2002.
22. Iakoubova, O.A. et al. Genome-tagged mice (GTM): two sets of genome-wide con-
genic strains. Genomics, 74: 89104, 2001.
23. Liu, D. et al. Mapping behavioral traits by use of genome-tagged mice. Am J Geriatr
Psychiatry, 12: 15865, 2004.
24. Liu, D. et al. Identifying loci for behavioral traits using genome-tagged mice. J
Neurosci Res, 74: 5629, 2003.
25. Bolivar, V.J., Caldarone, B.J., Reilly, A.A. & Flaherty, L. Habituation of activity in
an open field: A survey of inbred strains and F1 hybrids. Behav Genet, 30: 28593,
2000.
26. Bolivar, V. & Flaherty, L. A region on chromosome 15 controls intersession habitu-
ation in mice. J Neurosci, 23: 94358, 2003.
27. Cook, M.N., Bolivar, V.J., McFadyen, M.P. & Flaherty, L. Behavioral differences
among 129 substrains: implications for knockout and transgenic mice. Behav Neu-
rosci, 116: 60011, 2002.
28. Bothe, G.W.M., Bolivar, V.J., Vedder, M.J. & Geistfeld, J.G. Genetic and behavioral
differences among five inbred mouse strains commonly used in the production of
transgenic and knockout mice. Genes Brain Behav, 3: 14957, 2004.
29. Bolivar, V.J., Ganus, J.S. & Messer, A. The development of behavioral abnormalities
in the motor neuron degeneration (mnd) mouse. Brain Res, 937: 7482, 2002.
30. Bolivar, V.J., Manley, K. & Messer, A. Exploratory activity and fear conditioning
abnormalities develop early in R6/2 Huntingtons disease trangenic mice. Behav
Neurosci, 117: 123342, 2003.
31. Bolivar, V.J., Manley, K. & Messer, A. Early exploratory behavior abnormalities in
R6/1 Huntingtons disease transgenic mice. Brain Res, 1005: 2935, 2004.
9 Animal Resources in
Behavioral Neurogenetics
CONTENTS
9.1 Introduction .................................................................................................. 130

9.1.1 Requirements for a Suitable Animal Model ................................... 130
9.1.2 The Genome Club ........................................................................ 131
9.1.3 The Basic Principle of Animal Tools Is Inbreeding ....................... 131
9.2 Inbred Strains ............................................................................................... 132
9.2.1 Hybrid Crosses................................................................................. 132
9.2.1.1 F1 Hybrids ........................................................................ 132
9.2.1.2 F2 Hybrids ........................................................................ 133
9.3 Advanced Intercross Lines (AILs) .............................................................. 133
9.4 Strains That Differ Only by a Known Part of Their Genome .................... 133
9.4.1 Co-Isogenic Strains .......................................................................... 134
9.4.2 Congenic Strains .............................................................................. 134
9.4.3 Consomic Strains or Chromosome Substitution Strains ................. 134
9.4.4 Conplastic Strains ............................................................................ 135
9.4.5 Selected Strains ................................................................................ 136
9.5 Strains Made from (Multiple) Inbred Strains.............................................. 136
9.5.1 Recombinant Inbred Strains (RIs) ................................................... 136
9.5.1.1 Seven Major RI Sets Are Currently Available ................ 136
9.5.2 Recombinant Congenic Strains (RCs)............................................. 137
9.5.3 Segregating Inbred Strains (SIs)...................................................... 138
9.5.4 Mixed Inbred Strains (MIs) ............................................................. 138
9.6 Genetically Engineered Mutant Mice.......................................................... 138
9.6.1 Transgenic Mice............................................................................... 138
9.6.2 Mice with Targeted Mutations (Knockouts).................................... 139
9.6.3 Mice with Chemically Induced Mutations ...................................... 139
9.6.4 Chromosomal Aberration Strains .................................................... 139
9.7 Other Strains ................................................................................................ 140
9.7.1 Outbred Strains or Outbred Colonies .............................................. 140
9.7.2 Wild-Derived Inbred Mice............................................................... 140
9.8 Origins of the Oldest Inbred Strains of Mice ............................................. 140
9.9 Strains and Substrains of Laboratory Mice................................................. 141
9.10 International Supply of Mouse and Rat Research Models ......................... 142
129
9.10.1 The Jackson Laboratory and Charles River Laboratories............... 142

9.10.1.1 Genetic Quality................................................................. 143
9.10.1.2 Animal Health Quality ..................................................... 143
9.10.2 The Complex Trait Consortium (CTC) ........................................... 143
9.11 Other Species of Interest for Behavioral Neurogenetics Research............. 143
9.11.1 Drosophila Melanogaster ................................................................ 143
9.11.2 Caenorhabditis Elegans ................................................................... 144
9.12 Useful Web Sites.......................................................................................... 144
9.12.1 Web Sites for the Mouse and the Rat ............................................. 145
9.12.2 Rodents Suppliers ........................................................................... 145
9.12.3 Web Sites for Drosophila ................................................................ 145
9.12.4 Web Sites for Caenorhabditis elegans ............................................ 145
References.............................................................................................................. 146
9.1 INTRODUCTION
Genetics is a differential approach of the intra-specific variation, either normal or
pathologic, based on several types of experimental methods. Most of the behavioral
and neural phenotypes are complex traits that require the development of specific
tools. Several animal species are currently used in the field of behavioral neuroge-
netics, from humans to rotifers, through dogs, rodents, birds, fishes and flies.
Whereas in animals it is possible to create strains according to some specific
criteria, and to use them as genuine tools to devise genetic experiments for obvious
ethical reasons, in humans experimental approaches are restricted to natural exper-
iments. These basically include family studies, twin studies, and adoption studies.
On the other hand, in animals, various true experimental tools were developed. They
are based on inbreeding, selection experiments, crossing experiments, spontaneous
mutations, experimental mutagenesis or, more recently, on genetic engineering.
9.1.1 REQUIREMENTS FOR A SUITABLE ANIMAL MODEL

Genetic approaches have needed the development of specific animal models that
constitute genuine experimental tools and that must therefore respond to some
criteria.
The requirements for a suitable animal model in the field of behavioral neuro-
genetics are particularly demanding. The candidate must be a short-lived species
since a high number of generations are needed. It must also have a high reproduction
rate and yield abundant progeny to ensure large samples of subjects in a restricted
period of time. As far as psychological and cognitive abilities are concerned, a large
behavioral repertoire and advanced cognitive abilities are required. The size and the
accessibility of its central nervous system must allow the use of the most powerful
and specialized techniques of modern neurobiology. Although not mandatory, it is
also suitable for its genetic code to have been completely deciphered. Finally, the
biology of the species must allow that experimental approaches such as selective
breeding, inbreeding, or genetic engineering to be implemented.
Animal Resources in Behavioral Neurogenetics 131
9.1.2 THE GENOME CLUB

Actually, not one species meets all these requirements, so different species are used
chosen at different levels of the animal scale. After humans and mice, the rat is the
third mammal species (quite recently) to join this elite club in biology that almost
exclusively admits species the genome of which has been entirely deciphered. The
sequence of the genome of the chimpanzee is expected to be completed by the end
of the year 2004, whereas the genome analysis of the rhesus macaque monkey and
other species of interest, such as the dog, the frog, the chicken, and the cow, are
making rapid progress. The genome of the nematode worm, Caenorabditis elegans,
a species of growing interest in the field, has already been deciphered. Nevertheless,
for technical reasons, the mouse remains the most widely used species, particularly
when knockout and transgenic approaches are needed.
9.1.3 THE BASIC PRINCIPLE OF ANIMAL TOOLS IS INBREEDING

Inbreeding is the production of offspring by closely related parents (whereas outbreed-
ing is the production of offspring by unrelated parents). A strain is considered inbred
after 20 generations of brother sister mating. At this stage, it remains only 1%
residual heterozygosity (excluding any genetic drift), so that all loci of an inbred animal
can be considered homozygous and that all animals from the same inbred strain are
isogenic, i.e., they carry the same alleles at each locus. Although the system by which
homozygosis can be approached more rapidly and conveniently in experimental ani-
mals such as mice is that of brother sister mating, other breeding schemes are also
acceptable. Self-fertilization occurring only in a limited group of plants, backcrosses
between progeny and one parent, double first cousin, first cousin, and so on can be
used. Figure 9.1 represents the percentage of homozygosis reached in successive
Consequences of inbreeding
100
90
Portion of the genome
80
70
60
50
40
30 Individual homozygosity
Portion of the genome that is fixed
20
10
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Generations of inbreeding
FIGURE 9.1 Effect of inbreeding during 20 generations on the portion of the genome of an
animal that is homozygous at each generation (solid line) and the portion that is fixed
identically in two animals chosen as parents for the next generation. (Adapted from Silver.2)
generations under brother sister mating. Most inbred strains of mice and rats have
greatly surpassed 20 generations of inbreeding, which ensures homozygosis at nearly
all loci. Nevertheless, not all species support inbreeding; birds generally become sterile
after two generations of inbreeding and it seems that Drosophila resists inbreeding.1
9.2 INBRED STRAINS

Numerous inbred strains are available in rodents. Among the 478 strains of mice
listed in the Mouse Genome Informatics database by M.F.W. Festing,3 100 are
commercially available, the others are either extinct or preserved as frozen embryos.
A rat strain list can be obtained from the Ratmap database. It is clear from their
listings that there are also lots of inbred strains of rats with many substrains, although
little is widely known about most of them.
BALB/c (C), C57BL/6 (B6), DBA/2 (D2), for the mouse and Wistar kyoto
(WKY), Lewis (LEW), Fisher (FISH), Spontaneous Hypertensive Rat (SHR) are
examples of widely used inbred strains.
The characteristic features of inbred strains are their genetic and phenotypic
uniformity, although the latter is questionable, due to inbreeding depression that
often results in less buffered homeostasis of physiological or behavioral traits.
Their main applications involve inbred strains comparisons, the production of
F1 and F2 hybrid generations or of even more complete crossing schedules (Men-
delian, diallel). They can also be used to generate controlled and reproducible
genetically heterogeneous populations (e.g., four-way, or eight-way crosses).
Among the strengths and weaknesses of inbred strains, it is worth pointing out
that they are good for mapping and gene identification. They provide excellent
genetic background for studying epistasis and constitute a powerful method for
detecting pleiotropism. They are also good to give evidence of gene environment
interactions. Overall, their genotype is invariant over time, so the database of inbred
strains is cumulative and rapidly expanding. But the other side of the picture is that
inbreeding often leads to a reduction in viability and fertility and also in develop-
mental homeostasis, which makes them more sensitive to environmental short-term
variations and increases the phenotypic variance.
9.2.1 HYBRID CROSSES

Hybrid crosses result from crosses between two (and in some cases more than two)
inbred strains. The aim of hybrid crosses is to restore hybrid vigor (F1), to maintain
lethal (targeted) alleles at the heterozygous state (F1), to increase the number of
recombination events (F2, F3, etc.) and to generate reproducible, highly heterozy-
gous, segregating populations.
9.2.1.1 F1 Hybrids
Crossing two inbred strains produces F1 hybrids. For instance, B6D2 F1 results
from a cross between a C57BL/6 female (B6) and a DBA/2 male (D2), whereas
D2B6 F1 is the reciprocal F1 cross.
The characteristic features of F1 hybrids are their genetic and phenotypic uni-
formity: although partly heterozygous and partly homozygous, depending on the
loci where the parents carry different or identical alleles, they are all isogenic. F1
hybrids present advantages that predispose them to specific applications. First of all,
F1 hybrids between two inbred strains often display hybrid vigor or heterosis that
results in improved development and longevity, better resistance to diseases and
stress, larger size and weight, increased prolificacy, etc.). They often also display
better behavioral responses to cognitive tests.4 Secondly, they are useful as hosts for
tissue transplants (tumors, skin, ovaries) from either parental strain and as recipients
for some deleterious mutations (transgenic mice). The comparison of reciprocal
crosses between the two F1s allows detection of maternal effects resulting from
cytoplasmic and/or pre- and postnatal maternal environmental influences. Last, they
can be repeatedly produced, as long as the inbred parental strains remain available
(samples of parental strains can be preserved for years as frozen embryos).
9.2.1.2 F2 Hybrids
Mating of F1 F1 mice produces an F2 generation. B6129SF2/J is the F2 of a

C57BL/6 129S/SvImJ strain.
The F2 is highly heterozygous and shows allelic assortment among all loci for
which the parents differ, due to allelic segregation and recombination events during
meiosis. Applications of F2 hybrids include their use as heterogeneous population
for selection experiments, for genetic mapping or to originate a recombinant inbred
strains series (RIs, see below). They are also often used as physiological controls
for mice carrying targeted mutations maintained on a mixed, e.g., C57BL/6 by 129
F2, background (B6; 129).
9.3 ADVANCED INTERCROSS LINES (AILS)

AILs are made by producing an F2 generation between two inbred strains and then
intercrossing in each subsequent generation, avoiding siblings mating.5 The purpose
is to increase the possibility of recombination among tightly linked genes. They can
be used to create RIs with large numbers of recombination events, starting inbreeding
with breeding pairs from the F8, for instance. Pri:B6,D2-G# is an AIL stock created
at Princeton from the inbred strains C57BL/6 DBA/2. The G number will increase
with generations.
AILs constitute a good resource for fine QTL mapping with increased precision.
9.4 STRAINS THAT DIFFER ONLY BY A KNOWN PART

OF THEIR GENOME
There are four categories of strains: co-isogenic strains, congenic strains, consomic
strains, and conplastic strains that correspond to different kinds of inbred strains
differing at only a small part of their genome. Selected strains can enter this category,
provided that they have been submitted to an inbreeding procedure after the selection
plateau has been reached.
9.4.1 CO-ISOGENIC STRAINS

They are two inbred strains that differ only at a single locus. This condition can be
due to spontaneous mutation occurring in a strain, but it can also result from targeted
mutation in embryonic stem (ES) cells maintained on the same substrain from which
ES cells were derived, or from chemically or radiation induced mutants on an inbred
background. C57BL/6 CJ, for instance, is an albino mutation that occurred spon-
taneously in the C57BL/6J strain.
Co-isogenic strains allow studying the effects of an individual mutation. Unfor-
tunately, the probability that a mutation of interest for a particular research occurs
spontaneously is extremely low.
9.4.2 CONGENIC STRAINS

As shown by Figure 9.2, congenic strains are two inbred strains produced by repeated
backcrosses to an inbred or (background) strain, with selection for a particular marker
(or mutation) from the donor strain.6,7 After a minimum of 10 backcross generations,
the strain can be regarded as congenic. This is a practical way to drive experimentally
the construction of strains that are close to the co-isogenic condition. B6.AKR-H2k
is a mouse strain with the genetic background of C57BL/6 but differing from that
strain by the introduction of a differential allele (H2k) derived from the AKR/J strain.
The two congenic B6.AKR-H2k and C57BL/6 strains are expected to be identical
at all loci except for a segment of chromosome carrying the transferred locus. The
size of this segment decreases with successive backcrosses.
There are many advantages of maintaining a specific mutation on a defined
inbred background. First, the increase of genetic and phenotypic homogeneity due
to inbreeding results in reduced variability. Secondly, it allows identification of
modifier genes present in a specific strain. Finally, the background inbred strain
provides an accurate control.
Congenic strains are useful for avoiding genetic background effects, allowing
more precise analysis of single single-gene effects, and also for confirming the
position of a QTL, for mapping studies and for identifying a QTL.
One pitfall of congenic strains is contamination on another chromosome.
9.4.3 CONSOMIC STRAINS OR CHROMOSOME SUBSTITUTION

STRAINS8
They are produced by repeated backcrossing of a whole chromosome of a donor
strain onto the genetic background of a host inbred strain for a minimum of 10
generations. For example, in the C57BL/6J-Chr 19 SPR consomic strain, a M.
Spretus chromosome 19 has been backcrossed onto C57BL/6.
Consomic strains can be used directly to initiate mapping of a single locus
responsible for a strain characteristic, or of multiple linked or unlinked loci that
contribute in an additive fashion to the phenotype. They can be also useful to confirm
weak QTLs identified in linkage crosses.
Recipient Donor
strain strain
F1
BC1
Selection
BC2
Selection
BC 10
Heterozygous congenic
Intercross
Homozygous congenic
FIGURE 9.2 Construction of congenic strains.
9.4.4 CONPLASTIC STRAINS

These are strains in which the nuclear genome from one strain has been crossed
onto the cytoplasm of another (the mitochondrial donor always is the female parent
during backcrossing). For instance, crossing C57BL/6 male mice with BALB/c
females, followed by repeated backcrossing of female offspring to male C57BL/6,
results in a conplastic strain. C57BL/6-mtBALB/c is a conplastic strain with the
nuclear genome of C57BL/6 and the cytoplasmic (mitochondrial) genome of
BALB/c.
Conplastic strains are used to study cytoplasmic inheritance and to detect point
mtDNA mutations. Inherited mtDNA is 99.9% maternal and represents less than 1%
of the total cellular DNA in humans and 0.2% in the mouse.9
9.4.5 SELECTED STRAINS

Starting with a polymorphic population, bidirectional selection proceeds by selective
breeding for a given phenotypic trait in two opposite directions over successive
generations, until the selection plateau has been reached. Selection accumulates
increasing and decreasing alleles in the high and the low line, respectively. Generally
followed by inbreeding, it results in two well-contrasted strains with stable differ-
ences. Long sleep (LS) and short sleep (SS) strains result from bidirectional selection
for sleep time in response to alcohol injection in mice.10 After 18 generations of
selective mating, LS mice slept for an average of two hours, whereas most SS mice
were not even knocked-out (10 min average sleep time).
Selected strains are good for estimating the number of genes and mapping QTLs
that control the selected trait and selection-correlated traits. They are powerful
for detecting pleiotropy in selected genes. They are also good to detect gene
environment interactions when changes in line differences occur in association with
environmental changes but are ill-suited for detecting epistasis because the selected
genes are unknown.
9.5 STRAINS MADE FROM (MULTIPLE)

INBRED STRAINS
This category includes recombinant inbred strains (RIs), recombinant congenic
strains (RCs), segregating inbred strains (SIs) and mixed inbred strains (MIs).
9.5.1 RECOMBINANT INBRED STRAINS (RIS)

RIs are issued from n breeding pairs chosen at random in the F2 of two inbred
strains and followed by 20 or more generations of brother sister mating.11,12
As shown in Figure 9.3, BXD1, BXD2, BXD32 are RI strains of a series
originating from a cross between a C57BL/6 male (B) by a DBA/2 female (D2).
This series includes the 26 strains out of 32 that survived the inbreeding procedure.
Mice from the n inbred strains of the series, each of them carrying a random
sample of the genes of the two parental strains, are homozygous and isogenic
within a strain.
9.5.1.1 Seven Major RI Sets Are Currently Available
Originally designed to detect major gene effects, they are more and more used to
detect and locate quantitative trait loci (QTLs) on chromosomes, with increased
precision as the number of strains in a series increases. Thirty-six BXD RI strains
are now available from the Jackson laboratory and the number of strains in the BXD
series will soon be greatly increased. The University of Tennessee Health Science
C57BL/6 (B) DBA/2 (D)

female male
Parental strains (fully inbred)
F1 (isogenic)
F2
(heterozygous)
20 G. brother x
sister mating
BXD RI set
(inbred,
isogenic)
BXD 1 BXD 2 BXD 3 BXD 32
FIGURE 9.3 Construction of the BXD RI series.
Center in Memphis is actively breeding 50 lines of new BXD strains. Forty-six of

these lines are likely to be inbred fully in the next 2 years. In order to increase the
power and precision of mapping, these 46 newly generated BXD RI strains originated
from an advanced intercross (AI: 9 to 14 generations of intercrossing before inbreed-
ing) and thus constitute an advanced RI set.13 The Complex Trait Consortium is
actively pushing to provide the community with 1,000 ARIs derived from an
advanced intercross between 5 inbred strains and 3 wild derived strains.14
QTL mapping is based on analyzing the pattern of distribution of DNA markers
(SSLPs and SNPs) in order to determine whether any of these markers is linked to
a given phenotypic trait.
The advantages of the RI strains are the same as those of inbred strains. There
are thousands of DNA markers mapped in the RI series for which each strain needs
to be genotyped only once. Therefore, genetic correlations can be established across
measures, across studies and across laboratories thanks to cooperative data banks.
Their main limitation used to be the reduced number of strains in most series
that decreases the power of mapping and allows detecting only QTLs having the
largest effects.
9.5.2 RECOMBINANT CONGENIC STRAINS (RCS)

RCs are formed by crossing two inbred strains, followed by a few (generally 2)
backcrosses of the hybrids to the parental (recipient) strain, with subsequent inbreed-
ing for 20 generations, without selection for any specific marker.15 These RCs consist
of the background of the recipient strain interspersed with homozygous segments

of the donor, the size of which depends upon the number of backcrosses. Two
backcrosses result on average 12.5% of the donor strain genome. For example, CcS1,
CcS2, CcSX are multiple recombinant congenic strains between a BALB/c (C)
recipient and a STS (S) donor.
RCs allow analyzing the individual genes composing a system one by one. A
series of RCs is a powerful tool for genetic and functional dissection of quantitative
traits controlled by up to 5 or 6 nonlinked genes. They can be employed to study
genesgene interactions in various background combinations.
However, as most existing RC strains have been developed for the genetic
analysis of specific traits, they may be of limited use in other applications.
9.5.3 SEGREGATING INBRED STRAINS (SIS)

SIs are inbred strains in which a particular allele or a particular mutation is main-
tained in the heterozygous state. They are developed by brother sister mating but
with heterozygosity of the particular allele selected at each generation. The 129P3/J
mouse strain, for example, segregates for the albino (Tyrc) and chinchilla (Tyrc-ch)
tyrosinase alleles.
9.5.4 MIXED INBRED STRAINS (MIS)

MIs are incipient inbred stocks or mixed strains derived from only two parental
strains and carrying a genetic particularity. B6;129-Acvr2tm/Zuk is a mixed strain
derived from C57BL/6 and a 129 ES cell line, carrying a targeted knockout of the
Acvr2 gene.
9.6 GENETICALLY ENGINEERED MUTANT MICE

This category includes mice carrying a transgene, mice with targeted mutations
(knockouts), and mice with retroviral or chemically induced mutations.
9.6.1 TRANSGENIC MICE

Transgenic mice carry a segment of foreign DNA incorporated into the genome
via pronuclear microinjection, insertion via infection with a retroviral vector
or sometimes via homologous insertion. For instance, Tg(huAPP695.K670N-
M671L)2576 is a mouse carrying the human APP695 gene with the double
mutation Lys670 ASN, MET671Leu, found in a Swedish family with her-
itable Alzheimers disease.16 The APP695s we gene has been inserted in the
cosmid vector of the hamsters prion protein in the hybrid (C57BL/6xSJL)
mouse.
Transgenic mice are generally used as models of genetic human diseases.
One caveat of transgenic lines is that neither the number of transcripts nor the
insertion sites of the foreign gene in the genome are known.
9.6.2 MICE WITH TARGETED MUTATIONS (KNOCKOUTS)

They are created by first introducing gene disruptions, replacements or duplications
into ES cells by homologous recombination between the exogenous (targeting) DNA
and the endogenous (targeted) gene. Genetically modified ES cells are then micro-
injected into host embryos at the 8-cell blastocyst stage. These microinjected
embryos are in turn transferred into pseudopregnant host females that bear chimeric
progeny. Chimeric progeny carrying the targeted mutation in their germ line are then
bred to establish the line.
The knockout can be controlled by adding bits to a gene that act as a switch,
turning the gene on and off (inducible knockout) when a particular hormone or
antibiotic is added or even when the temperature changes (heat shock). The use of
appropriate promoters will allow expression of targeting in selected brain areas.
The knockout technology allows analyzing the function of a gene by performing
a form of experimental blockage (neurogenetic lesioning).
However, this approach encounters important limitations because numerous
genes being also involved in development, the knockout may be nonviable. Com-
pensatory mechanisms may also counterbalance the genetic deficit so that numerous
knockouts display no phenotype.
9.6.3 MICE WITH CHEMICALLY INDUCED MUTATIONS

They are created by exposing male mice to chemical agents, then breeding these
treated males with untreated females. The progeny are screened for phenotypes of
interest.
A variety of chemicals can be used that result in different kinds of mutations:17
Ethylnitrosourea (ENU) is used to induce point mutations.

Chlorembucil, an antibiotic, induces deletion mutations.
Rays exposure results in translocations.
Mutation rate is of 1/1,000 per locus and mutation detection rate of 90%.
These mutant mice are useful tools to elucidate basic biological processes and
to study relationships between gene mutations and disease phenotypes. Although
only recently used in mice, mutagenesis has been successfully employed in Droso-
phila to study learning and memory mutants early in the 1970s. A neuroscience
mutagenesis facility has been created for mice at the Jackson laboratory that is
available online.
9.6.4 CHROMOSOMAL ABERRATION STRAINS

Chromosomal aberration strains include various kinds of chromosomal anomalies
such as rearrangements of DNA segments within chromosomes (inversions),
exchanges of DNA segments between chromosomes (translocations) and deviations
from the normal number of diploid chromosomes in somatic cells (aneuploidy) that
include both Robertsonian chromosomes, i.e., bi-armed (metacentric) chromosomes
formed by the joining of two single-armed (acrocentric or telocentric) chromosomes

at the centromere and whole or partial chromosome trisomies.
Chromosomal aberration strains and Robertsonian F1 hybrids stocks are
available in mice, e.g.:
- In(7)13Rk in C.Cg-In(7)13Rk (In = Inversion)

- Is(7;1)40H in C3H/HeJ-pjIs(7;1)40H (Is = Insertion)
- Rb(6.16)24Lub (Rb = Robertsonian chromosome)
- Tp(Y)1Ct(ySxrb) in B6-Aw-J-EdaTa-6J.Cg-Sxrb Hya- (Tp = Transposition)
- Ts(1716)65Dn in B6EiC3Sn-a/A-Ts(1716)65Dn (Ts = chromosome 16 trisomy)
They constitute useful animal models for studying the effects of chromosomal
aberrations in humans. Moreover, Robertsonian chromosomes in combination can
be used to produce whole chromosome trisomy for specific mouse chromosomes.
For example [Rb(6.16)24Lub Rb(16.17)7Bnr]F1 mice, produce trisomy for chro-
mosome 16.
9.7 OTHER STRAINS
9.7.1 OUTBRED STRAINS OR OUTBRED COLONIES

An outbred strain is obtained by random mating within a closed colony. NMRI,
Swiss Webster, ICR mice, and Wistar, Long-Evans, Zucker, and Sasco-Sprague-
Dawley rats are widely used outbred strains.
They are less expensive than inbred strains, but are genetically unstable across
generations.
9.7.2 WILD- DERIVED INBRED MICE

They are descendants of a pair or a trio of wild-captured mice.
Much more polymorphic than common laboratory mice, they are a valuable tool
for evolution and systematics research. Also, the progeny from interspecific crosses
is especially useful for genetic mapping and they often carry Robertsonian chromo-
somes.
9.8 ORIGINS OF THE OLDEST INBRED

STRAINS OF MICE
As can be seen from Figure 9.4, many of the laboratory inbred strains come from
the same outbred colony, or share common ancestry. Other strains are derived directly
from wild mice. Whereas laboratory rat strains all derive from the Rattus norvegicus
species, there is increasing evidence that laboratory mice have been developed with
contributions from more than one species/subspecies of wild mouse. For example,
some strains carry the Mus musculus domesticus Y-chromosome (e.g., A/J, BALB/cJ,
DBA/1
Littles mice used in DBA
color experiments
DBA/2
C
CBA
CHI
X
CI2I
C3H/SI
C3H/Bi
C3H/An
C3H/He
C3H/He J
C3He B/Fe J
Dealers stock Bogg BALB/c ( OvaC57BL/6J
transferred to
)
in Ohio Albinos A/J
A/St
X
A/Bi
A/He
Cold Spring Harbor A/HeJ
Albinos C5B
C57BL /6
C57BL C57BL /10
Miss Lothrops
C57BR/cd
Stocks
C57BR C57BL /o
C57L
AKR
Furths A & R Stocks RF
SWR
European White Mice
SJL
Webster Swiss
(1909) 12 16 20 24 28 32 36 40 44 48 52 56 60
FIGURE 9.4 The origins and relationships of some of the inbred strains of mice.18
CBA/J, C3H/HeJ, C57BL/6J, DBA/2J, ), while others have the M.m. musculus
type (e.g., AKR/J, SJL/J, SWR/J, etc.).
Most laboratory mice have contributions from both Mus musculus musculus and
Mus musculus domesticus. There is evidence that smaller contributions also may
have come from Mus musculus molossinus and Mus musculus castaneus.3
9.9 STRAINS AND SUBSTRAINS OF LABORATORY

MICE
Established inbred strains may diverge with time into substrains among which there
are discernable genetic differences. This could occur through three different conditions:
If two branches are separated after 20 but before 40 generations of inbreeding,

there is still enough heterogeneity for two genetically different substrains
to result.20
If branches are separated for more than 20 generations from a common
ancestor, genetic variation may have occurred by mutation and genetic drift.
If genetic differences are proven by genetic analysis to have occurred.
For example, IS/Kyo is a substrain of IS rat strain originating at Kyoto University.

Figure 9.5 gives an example of the various substrains that occurred in the C57BL
strain.19
C57BL
GEN. 0 20 40 60 80 100 120
10 Ch
10 Fo
10 Gn
10 Sf
10 Jax
10 Sc
10 Sn
10 Ph
10Y
<?> Ks
6By
6 lcr
6 He
6N
6 Bom
6 Jax
6Y
6Hb
Little 6Mi
How
Mcl
Fr
Rii
A
Rl
He
An
Hf
1Umc
Bcr
Ge
lcrf
Go
Fa
H
LacH
Gr
Ka
St
1920 1930 1940 1950 1960 1970
FIGURE 9.5 Substrains of the C57BL strain. 19
9.10 INTERNATIONAL SUPPLY OF MOUSE AND RAT

RESEARCH MODELS
9.10.1 THE JACKSON LABORATORY AND CHARLES RIVER
LABORATORIES
The Jackson Laboratory (Bar Harbor, Maine) acts as a conservatory of most
inbred, mutant and genetically engineered mice used in biomedical research
around the world. These strains are either in production, or preserved as frozen
embryos.
In July 2001, The Jackson Laboratory and Charles River Laboratories Interna-
tional, Inc., started to cooperate to supply JAX mice to researchers located in
European and Pacific Rim countries.
As detailed in the JAX Bulletin, (N11, Sept. 2001), their distribution agreement
stipulates that the Jackson laboratory ensures that JAX mice raised in Charles
Rivers European and Japanese facilities meet the highest standards for health and
genetic quality. The main points of this agreement are summarized below.
9.10.1.1 Genetic Quality
JAX animals bred in the United States, Europe and Japan originate from pedigreed
identified pairs shipped from The Jackson Laboratory. Authorized breeding colonies
are reinfused with new pedigreed stock from The Jackson Laboratory on a routine
basis to minimize spontaneously occurring genetic drift. Routine monitoring for
genetic quality is performed to ensure both genetic identity and purity.
9.10.1.2 Animal Health Quality
JAX rodents bred in Europe and Japan are monitored for the same pathogenic and
opportunistic agents as are the JAX rodents raised at The Jackson Laboratory, on
the same basis (sampling plan, frequency of testing, diagnostic test methods).
Such genetic and health quality procedures ensure that all mice used in various
countries at different times match the highest standards for genetic uniformity and
makes particularly meaningful comparisons of results.
9.10.2 THE COMPLEX TRAIT CONSORTIUM (CTC)

Another important breeding facility is developing at the University of Memphis,
which will revolutionize QTL mapping research in RI mice. The goal of the CTC
is to promote the development of genetic resources that can be used to understand,
treat and ultimately prevent pervasive human diseases.14 The CTC is actively trying
to generate a collaborative genetic resource called the Collaborative Cross. Starting
with a full diallel cross of eight strains, it should result in a series of 1,000 RIs
available to scientists in the next few years.
9.11 OTHER SPECIES OF INTEREST FOR BEHAVIORAL

NEUROGENETICS RESEARCH
Two more species, the fly Drosophila, and the nematode worm Cenorhabditis
elegans are more and more widely used in behavioral neurogenetics research. They
also gather most of the typical features of a suitable animal resource.
9.11.1 DROSOPHILA MELANOGASTER

D. melanogaster is a very handy tool, highly amenable for behavioral and neuroge-
netic analysis. Its reproductive cycle is 10 to 14 days at 25C. Hundreds of eggs are
laid on the food medium that hatch after 24 h. The development includes the larval
stage (5 days) and the pupa stage (4 days). The adult fly lives about 30 days in the
laboratory. Drosophila has a compact genome (The X/Y sex chromosomes and 3
pairs of chromosomes, the fourth of which being very tiny) that has been entirely
sequenced. The size of the genome is about 165 million bases and contains an
estimated 14,000 genes. Drosophila displays a rather large behavioral repertoire:
learning and memory, courtship, circadian rhythms, food search, olfaction, locomo-
tion, aggression and sleep-like processes can be studied in Drosophila, some of them
at both the larval and adult stages. Most methods used in forward and backward
genetics can be implemented in that species, including polygenic and single-gene
analysis, selective breeding, QTL analysis, mutational analysis (EMS and screening),
namely for memory and biological clocks, visible markers, transposons, genetic
engineering (inducible knockout, reporter genes such as -Galactosidase, GFP).
One limit of Drosophila comes from its resistance to inbreeding so that there
are no truly inbred strains in this species.
9.11.2 CAENORHABDITIS ELEGANS

The terrestrial nematode worm C. elegans also presents most of the characteristics
of a model organism for behavioral neurogenetics. This tiny worm has a very fast
reproductive cycle (3 days) at 20C. Males (XO) are found rarely in laboratory
populations (0.05%). Most individuals are hermaphrodites (XX) that self-fertilize,
so that it is easy to generate homozygous mutant stocks. C. elegans is diploid and
has a small genome with 5 pairs of autosomal chromosomes and a pair of sex
chromosomes. It has been the first animal genome completely sequenced by the end
of 1988. The genome size is 97 Megabases (Mb) and there are about 20,000 protein
coding genes.
The nervous system of C. elegans is very simple with 302 neurons in the
hermaphrodite and 310 in the male that has a more complex phenotype with sexual
behavior. It has also an almost invariant connectivity and neurotransmitters (ACh,
Glu, GABA, DA, 5-HT) that are common with those of mammals. The behavioral
repertoire is reasonably endowed with basic behaviors such as locomotion, feeding,
defecation, egg laying, male mating with the hermaphrodite, sensing its environment
and responding appropriately. Basic learning processes as habituation and sensiti-
zation, some forms of associative memory and internal clocks can also be studied.
Reverse genetics and transgenesis can be used in C. elegans. Hundreds of genetic
markers have been identified. Mutagenesis (EMS and screening) and balancer chro-
mosomes are useful tools in that species.
9.12 USEFUL WEB SITES

Most of the information presented in this chapter can be completed or retrieved from
all purpose and more specialized Web sites, a nonexhaustive list of which is presented
below.
A general purpose virtual library on model organisms can be found at
http://www.ceolas.org. This site is a catalog of Internet resources relating to biolog-
ical model organisms.
9.12.1 WEB SITES FOR THE MOUSE AND THE RAT
The Whole Mouse catalog (this Web site is for mice and rats, formerly
the Mouse and Rat Research Home Page located at CalTech):
http://www.rodentia.com/wmc/.
Mouse Genome Informatics (MGI): http://www.informatics.jax.org.
Neuroscience mutagenesis facility: http://nmf.jax.org.
Complex Trait Consortium (CTC): http://www.complextrait.org.
The Rat Genome Database (Ratmap group, Sweden)
Ratmap: http://ratmap.gen.gu.se/.
Rat Resources: http://ratmap.org/ratres.html.
International Committee on Standardized Genetic Nomenclature for Mice
Chairperson: Dr. Janan T. Eppig ([email protected]).
Rat Genome and Nomenclature Committee, Chairperson: Dr. Eberhard
Guenther ([email protected]).
Current nomenclature rules for naming genes
For mouse: http://www.informatics.jax.org/mgihome/nomen/gene.shtm1#genenom.
For rat: http://rgd.mcw.edu/nomen_rules.html.
Strain names registration
For mouse: http://www.informatics.jax.org/mgihome/submissions/
submissions_menu.shtml.
For rat: http://rgd.mcw.edu.
9.12.2 RODENTS SUPPLIERS
The Jackson Laboratory: http://www.jax.org.

Charles River laboratories: http://www.criver.com.
Taconic: http://www.taconic.com.
Taconic M&B (Europe): http://www.m-b.dk.
RRRC (Rat Resource and Research Center): http://www.nrrrc.mis-
souri.edu.
Harlan: http://www.harlan.com.
9.12.3 WEB SITES FOR DROSOPHILA
A quick and simple introduction to Drosophila melanogaster can be found

at http://www.ceolas.org/fly/introd.html.
The dedicated database to Drosophila is Flybase at http://flybase.net.
9.12.4 WEB SITES FOR CAENORHABDITIS ELEGANS
An introduction to the genetics of C. elegans can be found at

http://www.nematodes.org.
Wormbase is a large-scale database covering information on all worm
genes at http://www.wormbase.org.
The C. elegans www server at UTSMC is the best on line reference for
C. elegans: (http://elegans.swmed.edu/).
elegans Net: (http://members.tripod.com/C.elegans/index.html links on C.
elegans and includes extensive introductory links (ACekit at
http://winw.nbr.wisc.edu/outreach/test/celegans.html ).
REFERENCES
1. Petit, C., Linfluence du mode de croisement sur la structure gntique des popula-
tions; la stabilit des populations exprimentales de faible effectif. Annales de gn-
tique, 1963, 6, 2935.
2. Silver, L.E., Mouse genetics, concepts and applications. 1995 Oxford University
Press. Available online at the Jackson laboratory Mouse Genome Informatics facility:
http://www.informatics.jax.org/silver/frame1.1.shtml and also the CTC Web site:
http://www.complextrait.org.
3. Festing, M.F.W., Inbred strains of mice and their characteristics In: Mouse Genome
Informatics (MGI) 1998 at http://www.informatics.jax.org.
4. Lassalle J.M., Le Pape G., and Mdioni J., A case of behavioral heterosis in mice:
quantitative and qualitative aspects of performance in a water-escape test. J. Comp.
Physiol. Psychol., 1976, 93, 116123.
5. Darvasi, A. and Soller, M., Advanced intercross lines, an experimental population
for fine genetic mapping. Genetics, 1995 141: 11991207.
6. Snell, G.D., Congenic resistant strains of mice. In: Origins of inbred mice, Morse,
H.C., Ed., Academic Press, New York, 1978 pp 131.
7. Flaherty, L., Congenic strains. In: The mouse in biomedical research, Vol. 1, Foster,
H.L., Small, J.D., Fox, J.G. Eds, Academic Press, New York, 1981 pp. 215222.
8. Nadeau, J.H., Singer, J.B., Martin, A., and Lander, E.S., Analysing complex genetic
traits with chromosome substitution strains. Nature Genetics, 2000, 24: 221225.
9. Lewin, B., 1987. Genes, Wiley, New York.
10. Mc Clearn, G.E. and Kakihana, R., Selective breeding for ethanol sensitivity: short
sleep and long sleep mice, in: Development of animal models as pharmacogenetic
tools. G.E. McClearn, R.A. Dietrich and V.G. Ervin, Eds., Research monograph No.
6, Bethesda, MD, US. 1978 Department of Health and Human Services, 147159.
11. Bailey, D.W., 1971. Recombinant inbred strains, an aid to finding identity, linkage,
and function of histocompatibility and other genes. Transplantation, 1971 11:
325327.
12. Taylor, B.A., Recombinant inbred strains: use in gene mapping. In: Origins of inbred
mice, Morse, H.C., Ed., Academic Press, New York, 1978 pp 423438.
13. Peirce, J.L., Lu, L. GU, J., Silver, L.M. and Williams, R., A new set of recombinant
inbred lines from advanced intercross populations in mice. BMC Genetics, 2004 5:7,
117. Article available online from: http://www.biomedcentral.com/14712156/5/7.
14. Williams, R.W. and Churchill, G.A., The collaborative cross: rationale, implemen-
tation and costs. National Institute of General Medical Sciences, Complex trait
workshop report. 1998 www.nigms.nih.gov/newx/reports/genetic_arch.html.
15. Demant, P. and Hart, A.A.M., Recombinant congenic strains a new tool for ana-
lyzing genetic traits determined by more than one gene. Immunogenetics 1986 24:
416422.
16. Hsiao, K.A., Chapman, P., Nilsen, S., Eckman, C., Harigaya, Y., Youkin, S., Yang,
F., and Cole, G., 1996. Correlative memory deficits, An elevation, and amyloid
plaques in transgenic mice. Science, 1996, 274: 99102.
17. Greenspan, R.J., The induction, detection and isolation of mutations. In D. Goldowitz,
D. Wahlsten and R.E. Wimer Eds: Techniques for the Genetic Analysis of Brain and
Behavior. Elsevier Science Publishers BV, 1992, 93110.
18. Staats, J., Origins of the older inbred strains. In: Biology of the Laboratory Mouse.
Green, E.L., Ed., 1966, McGraw Hill.
19. Bailey, D.W., Definitions of inbred strains, substrains, sublines, and F1 Hybrids. In
Handbook of Genetically Standardized Jax Mice. R.R. Fox and B.A. Witham, Eds.
1997 Fifth Edition. The Jackson laboratory, Bar Harbor, Maine.
20. Green, E.L., Genetics and Probability in Animal Breeding Experiments. 1981 Oxford
University Press, New York.
10 Sample Size Requirements

for Experiments on
Laboratory Animals
Douglas Wahlsten
CONTENTS
Introduction............................................................................................................ 149
10.1 Effect size .................................................................................................... 150
10.2 Probability of a False Positive .................................................................... 152
10.3 Probability of Failing to Detect Something Real ....................................... 154
10.4 Estimating Effect Size from Published Data.............................................. 155
10.5 Methods for Estimating Sample Size ......................................................... 156
10.6 Power and Sample Size for Two Independent Groups .............................. 158
10.7 Comparison of Several Inbred Strains........................................................ 159
10.8 Comparing Specific Groups in a One-Way Design ................................... 161
10.9 Strain Treatment Experiments ................................................................. 162
10.10 Bridging the Gap: the 2 2 Design............................................................ 165
10.11 Sample Size for More Specialized Experiments ........................................ 167
References.............................................................................................................. 167
INTRODUCTION
Planning any experiment involves choice of an experimental design and sample size
for each of the groups in the study. The design of a study is dictated by the kind of
question one seeks to answer, and the choice of appropriate control groups is
determined by logic and established genetic principles. Approaches to design are
discussed in numerous articles and chapters, including many in the present volume.
Once the design is chosen, the researcher must decide on the sample size. This will
commonly be done when preparing a grant or thesis proposal. Increasingly, animal
ethics committees also request that the investigator justify the number of animals to
be used. Ethical and financial considerations, alike, demand that the minimum
effective number be employed. At the same time, if too few animals are studied, the
experiment may be unable to detect the very effects it was designed to investigate,
149
thereby wasting the animals as well as the researchers time and the funding agencys
money.
The proper sample size is determined by the experimenters choice of three
values. Two are probabilities of making either of two kinds of errors of statistical
inference, and the third is the size of an effect that one seeks to detect.17
10.1 EFFECT SIZE

Effect size may be expressed in terms of the units of measurement for the phenotype
of interest. Suppose one plans to examine brain weight in milligrams. If control
mice have an average brain weight of 450 mg and closely related mutants have an
average brain weight of 430 mg, then the effect size of the mutant genotype would
be 20 mg. When examining learning of a maze, one might find that controls acquire
a task in 11 trials while mutants require 15 trials, and the effect size would be +4
trials. Whether sizes of the mutations effects on brain weight and learning are similar
would be difficult to say, because there is no way of equating milligrams of brain
tissue with trials in a maze.
Any interesting feature of an animal that we might care to measure is likely to
show individual variation within a group having the same genotype. The extent of
this variation is commonly expressed as the standard deviation. For normally dis-
tributed data, most (about 98%) individuals will score within two standard deviations
of the group mean (see Figure 10.1). For the above example, suppose the standard
deviation of brain weight is 15 mg and of maze learning is 5 trials. The relative sizes
1+
1
Small
2 = .5
Medium
2 = .75
Large
2 = 1.0
-2 -1 0 1 2 3 4
Standard deviations from 1
FIGURE 10.1 Overlap of scores in two normally distributed groups for three effect sizes
ranging from small to large. Note that most scores in a group are within two standard deviations
of its mean.
Sample Size Requirements for Experiments on Laboratory Animals 151
of the effects of the mutation can be assessed by comparing the difference between
group means to the standard deviation within a group. For brain weight, effect size
would be 20 mg/15 mg = 1.33, whereas for learning it would be 4 trials/5 trials
= 0.80. Relative to variation within a group, the mutation would appear to have a
relatively greater effect on brain weight than on maze learning.
The ratio () of group difference in mean (1 2) to standard deviation within
a group () is widely used as an index of effect size when two groups are being
compared, and this is the value that is commonly used for a sample size estimation.
Generally speaking, the smaller the effect that one would like to detect, the larger
must be the sample size.
Because sample size calculations are done before any data are collected, these
involve population values of parameters rather than estimates from sample data.
Symbols used for population parameters and sample statistics are given in Table 10.1.
For a study involving comparisons of two groups, the indicator of effect size is the
hypothesized true value of the population parameter = (1 2)/, whereas its
estimate from sample statistics is d = (M1 M2)/S. The value of d is slightly biased
and can be adjusted when precision is deemed important.14,26 In this presentation,
the true, population parameter is either a Greek letter or boldface, while the estimated
sample statistic is either italicized or the Greek letter with caret ^ to indicate an
estimate from data.
On the basis of my review of many studies of a wide variety of phenotypes,
guidelines are offered for values of that correspond to what are generally regarded
as very small, small, medium, large, and very large effects in neurobehavioral
genetics: = 0.25, 0.50, 0.75, 1.0, and 1.5, respectively (Table 10.2). These are
somewhat larger than values assigned to the same descriptors in psychological
research with humans by Cohen8 and Borenstein et al.5
When more than two groups are involved, as in a study of several inbred strains,
the difference between any two groups will be unique to that pair. Hence, an indicator
that is more broadly representative of the data is needed. Cohen8 has suggested a
convenient approach that compares the population standard deviation among group
means (M) to the standard deviation within groups (): f = M/. When an analysis
of variance (ANOVA) partitions total variance into components among and within
TABLE 10.1
Symbols for True Values of Parameters and Their Estimates From Sample
Statistics
Population Sample
Characteristic Parameter Statistic
Mean M
Standard deviation S
Effect size for 2 groups d
Effect size for J groups 2 est. 2
Cohens effect size for J groups f
Linear contrast of J groups est.
TABLE 10.2
Values Corresponding to Magnitudes of Effect Size for Animal Research in
Neurobehavioral Genetics
2 Groups J Groups
Size of Effect 2 Cohens f 2
Very small 0.25 1.5% 0.1 1%
Small 0.5 6% 0.2 4%
Medium 0.75 12% 0.35 11%
Large 1.0 20% 0.5 20%
Very large 1.5 36% 0.7 33%
groups, a good descriptor of effect size is 2, the proportion of total variance that
is attributable to differences among group means in a fixed effects analysis.13 The
f index has a simple relation to this ratio of variances: 2 = f2/(f2 + 1). Values of
f and 2 for effects of various sizes are compared in Table 10.2. When there are
only two groups in the study, 2 = 2/(2 + 4).
Cohen8 proposed that criteria for small, medium and large effects in a one-way
ANOVA design in psychological research be f = 0.1, 0.25 and 0.4, respectively, which
correspond to 2 = 0.01, 0.06 and 0.14. These values appear to be somewhat lower
than what is commonly observed in neurobehavioral genetics, for example in com-
parisons of several inbred strains. A multigroup difference that accounts for only 1%
of total variance would be seen as not merely small but almost trivially small. Many
inbred-strain studies have found effects exceeding 20% of total varianceand this
would be a reasonable criterion for a large effect in our field. Standards are proposed
for both f and 2 in Table 10.2 for a study with J groups. With these conventions,
the verbal labels of size of the effect correspond to similar values of 2 for a study
with two groups as well as for a large experiment with J groups.
10.2 PROBABILITY OF A FALSE POSITIVE

The size of an effect seen in an experiment is sometimes confused with the signif-
icance of the statistical test. We commonly read reports that describe highly
significant effects of a genetic mutation as though this implies the effect was very
large. It is easy to understand how people new to data analysis might perceive
meaning in this way. The Oxford English Reference Dictionary1 lists the most
common use of the word significance as importance; noteworthiness. It lists the
technical sense employed by statisticians as the extent to which a result deviates
from that expected on the basis of the null hypothesis. Use of the word extent
implies size or magnitude, which is not true for statistical significance. If very large
samples are studied, very small effect sizes may be judged significant by statistical
tests.
A null hypothesis usually contends that a treatment had no effect ( = 0) or
several samples observed in the study were drawn from populations that did not
differ (2 = 0). After collecting the data, the investigator conducts a statistical test,
such as a t test when comparing two groups or the F test in the ANOVA comparing
several groups, in order to assess whether the evidence is sufficient to justify rejection
of the null hypothesis. If the null that there was no real effect can be refuted, then
we may conclude, tentatively, that there was indeed an effect. This arcane logic,
clothed in a most unfortunate choice of words such as significance, often makes
it difficult to keep the characters in the drama of science in their proper roles.
The null may in fact be true or it may be false. If the null is true but we decide
to reject it on the basis of the results our study, then the outcome of our study is a
false positive result and, in statistical parlance, we commit a Type I error of inference.
By convention, we agree to set a criterion for rejecting the null hypothesis based on
the probability of committing a Type I error. Most commonly this criterion, sym-
bolized , is set at one chance in 20. If the data suggest that the probability of
committing a Type I error is less than 5%, the null may be rejected, but this does
not reveal the scientific significance of the result. Instead, it would be better thought
of as the false positive rate (FPR). If we are comparing two independent groups of
animals, perhaps one homozygous for a knockout (/) and the other the wild-type
littermate controls (+/+), and the t test indicates the FPR is .001 or one chance in
1,000, then it is a fairly safe bet that the knockout really did affect the phenotype
we measured.
In those situations where the null is indeed true, the probability of obtaining a
false positive result is not influenced in any way by the sample size. That is, even
an elegant study of large numbers of animals will lead us to commit a Type I error
on about 5% of the tests when the null is true, the same FPR as when only a few
animals are studied.
The choice of the Type I error criterion influences the calculated sample size.
Generally speaking, the smaller the value of , the larger must be the size of the
observed effect in order to warrant rejection of the null, and the larger the sample
size that must be employed. When the results of a simple study depend on just one
statistical test of the null hypothesis, then = .05 may be appropriate. If the study
involves K independent tests, however, or if the study is one of K similar studies
being done independently in different labs, then the Bonferroni corrected Type I
error probability /K is more appropriate. A more sophisticated approach to adjust-
ing for multiple tests has been proposed by Benjamini et al.4
If the researcher has a good idea of the direction of the likely effect of a mutation,
then it would be appropriate to conduct a one-tailed test of the null hypothesis that
two groups do not differ, whereas a two-tailed test is preferred if the direction of
the effect is uncertain. In an ANOVA with multiple groups, this distinction is not
pertinent.
The overbearing emphasis on null hypothesis testing in psychology has been the
subject of debate.7,12 One of the central issues is the credibility of the null in any
particular study. In a genetic linkage study, for example, the hypothesis that alleles
at a marker locus are not related to phenotypic variation is highly credible, and we
expect that most markers will not show evidence of linkage. In such a case, the
proper choice is critically important. Lander and Kruglyak16 proposed the use of
= .0001 for a genome-wide scan in linkage analysis, while Belknap et al.3 argued
for a lower level of as a foil against the demise of a study owing to insensitivity
to real linkage relationships. Given the surprising but common finding that a gene
knockout has little or no effect on many phenotypes, the null may also be credible
in this kind of research. For a multiple inbred strain study, on the other hand, it
would be astounding if no strain main effect were found. Instead, the more interesting
question in strain studies is usually the presence or absence of strain by treatment
interactions or strain correlations among phenotypes.
Whereas the null hypothesis is commonly taken to be no effect at all, methods
are available for evaluating the possibility that the observed effect exceeds some
nonzero effect size.17 This approach has not yet been widely applied in neurobe-
havioral genetics, but it is useful when researchers want to know whether a
mutation has, for example, something more than merely a small effect on a
phenotype.
10.3 PROBABILITY OF FAILING TO DETECT

SOMETHING REAL
A test of significance can be done with just the null ( = 0 or 2 = 0) but no specific
alternative hypothesis in mind. In this situation, the probability of Type I error
can be controlled, but the test will be oblivious to the other kind of error of inference.
Suppose there really is a genetic effect in a study and the null is false. If the statistical
test fails to reject the null and detect the real effect, then a Type II error of inference
has occurred. This kind of error is much more frequent than Type I errors in most
kinds of research, and it is a serious error indeed, given the resources typically
devoted to conducting a study. The probability of a Type II error is symbolized ,
but there is no way to calculate this probability in any general sense, i.e. when 0
or 2 > 0. Instead, we must propose a specific value of effect size that can serve as
the alternative hypothesis. This value should be a plausible size of an effect that we
would like to be able to detect with an experiment.
Suppose that we are seeking to find a medium-sized effect in a comparison of
genotypes +/+ and /. Type I error probability is the probability of rejecting the
null when the hypothesis that = 0 is true. Type II error probability is the
probability of failing to reject = 0 when in fact = 0.75 (see Figure 10.1). If
really does equal 0.75, the possibility of Type I error is entirely moot, and Type II
error is the major peril.
If is the probability of an error, 1 is the probability of doing the right thing
rejecting a false null hypothesis when there truly is a genetic effect. The quantity 1
is termed the power of the statistical test, and we would like our test to have high
power. Type II error probability and power are strongly dependent on sample size; the
larger the sample, the higher the power. Because we choose the values of , and
sample size before conducting the experiment, Type II error probability is under our
control. The choice of power has no bearing at all on the probability of a Type I error
because that mistaken conclusion can occur only when the null is true. The choice of
, on the other hand, influences power; a more stringent Type I error probability, being
a smaller value of , reduces power to detect an effect of a given size.
There is no universally accepted criterion for a proper level of power in research.

Many studies are planned to have power of 80%, meaning that they are willing to
risk a 20% rate of Type II error. It is not clear, however, why an investigator should
be more concerned about Type I than Type II errors. If both are of equal concern,
it would make sense to use = = .05. When working with a novel knockout,
= .05 might be reasonable for the purpose of estimating sample size, but power
of 99% might be appropriate if great effort and expense are involved in creating and
rearing the mice as well as measuring complex phenotypes.
10.4 ESTIMATING EFFECT SIZE FROM PUBLISHED DATA

Statistical analysis is typically done and then reported in an article after data have
been collected, and it is unusual to read in the literature the values of or f that
may have been used to plan the study. If one wishes to use the literature as a guide
to choosing sample size for a future study, then it will be useful to know how to
find or f from published statistics.
In the case of a study with two groups, the task is easy if the authors published
the means, standard deviations and sample sizes of both groups. Presuming the
variances (S2) of the two groups were not exactly the same, a pooled estimate needs
to be obtained by converting each variance to the sum of squared deviations (SS)
from the group mean. For group 1 with n1 subjects, SS1 = (n1 1)S2. The pooled
estimate of variance is (SS1 + SS2)/(n1 + n2 2), and the pooled estimate of the
standard deviation (Spooled) is the square root of this variance. The sample estimate
of effect size is then d = (M1 M2)/Spooled. If a table provides means and standard
errors, then standard error can be converted to standard deviation by dividing by the
square root of sample size for that group.
In some instances the authors do not provide a table of descriptive data and
instead focus attention on the significance test. For example, a report may state that
a group difference was significant (t = 2.17, df = 18, P < .05). Presuming that sample
sizes were similar, it follows that the sample size per group was n = 10. As pointed
out for a study with equal sample sizes,26 the t ratio has a concise relation with effect
size: t2 = d2n/2, so that d2 = 2 t2/n. Thus, for our example, d2 = 2(4.71)/10 = 0.942
and d = 0.97, a large effect size.
Frustration may set in when a report states sparsely that groups differed (e.g.,
P = .007). Provided one can figure out from the methods section that the sample
size was perhaps 15 in each group, all is not lost. Degrees of freedom will be
30 2 = 28, and we can then derive the approximate value of what the obtained t
statistic must have been from the P value and degrees of freedom: t = 2.62. This
leads to d = 0.96, a large effect size.
If the study noted with disappointment that the genetic difference was not
significant at = .05 or stated tersely N.S, there is no way to arrive at their value
of d. At best, one could compute an upper limit to the possible value of d. Suppose
it is clear that n = 25, making degrees of freedom 48 for the t test. The critical value
of t for a two-tailed test at = .05 is t = 2.01, and this ratio corresponds to d =
0.57, a rather small effect if there was any effect at all.
When an ANOVA is done on J groups, it may be possible to estimate M, within,

and thereby f from a table of means and standard deviations or standard errors, as
described for the case of two groups. If the reader finds only the significance test
result (F = 7.3, df = 5/64, P < .05), it follows that there were six groups with about
12 subjects each. The value of the 2 can be found from the relation est.
2 = (F 1)/[F + (dfwithin + 1)/dfbetween], which for the example is est. 2 = 0.31.
From the relation 2 = f2/(f2 + 1), we then estimate that f = 0.67, a very large effect.
In the frustrating happenstance when we read only that P = .034 and that sample
sizes were 20, 25, 22 and 21 in four groups, we know that dfbetween = 3 and dfwithin
= 84. To give P = .034, the value of F must have been F = 3.03, and from this we
find 2 = 0.065 and f = 0.26, a small effect.
How many previous studies should be consulted in order to arrive at a reasonable
guess at the value of or f depends on how large the literature on the subject may
be. An estimate of or f is pertinent to the task at hand only if the methods in a
previous study were similar to those to be used by the researcher. If the literature
offers a wide range of effect sizes, it may well be that there were important meth-
odological differences among the published studies, but this could also be ascribed
to sampling error if samples used in those studies were generally small. Only a formal
meta-analysis can assess these alternatives11,14,20 (see http://www.edres.org/meta). If
a meta-analysis is available, then effect size can be taken from the center of the
confidence interval for in the case of two groups. If the most relevant literature
consists of perhaps fewer than 10 studies, it is not difficult to conduct a quick meta-
analysis of ones own to find the best estimate of effect size.
10.5 METHODS FOR ESTIMATING SAMPLE SIZE

While there is a widespread and well-founded belief that power and sample size
calculations are poorly understood and rarely attempted in our field, there is sufficient
interest to inspire the creation of no fewer than 29 computer programs that do most
of the work for the informed investigator.22 These programs entail a wide range of
capabilities and ease of use, and there currently is no one implementation that
dominates the market. Some are free, while others cost hundreds of dollars. Printed
dissertations on the topic, on the other hand, can be purchased on the used book
market for $20 in the case of Cohen8 and implemented with an electronic calculator.
Several general purpose statistical data analysis programs provide features for esti-
mating power of a test, but in most cases these presume a considerable degree of
expertise from the user.
A mathematically sophisticated approach is available from Kraemer and Thie-
mann,15 where the population value of intraclass correlation is adapted to several
situations and a master table gives degrees of freedom required to achieve various
levels of power under two values of (.05, .01) for both one- and two-tailed tests.
A similar approach based on the F distribution that can be applied to a wide range
of statistical tests is offered by Murphy and Myors17 but provides a master sample
size table only for = .05 and power of 80%. A less accurate but very convenient
set of sample size tables is provided by Cohen8 also for two values of (.05, .01)
for both one- and two-tailed tests. The method of Cohen has been implemented as
the computer programs SamplePower by Borenstein et al.,6 available from SPSS

Inc., and Power and Precision by Borenstein et al.,5 available from Biostat Inc.
Bausell and Li2 rely on a method that is very similar to the Cohen/Borenstein
approach, and they adapt this to analysis of covariance as well as the conventional
ANOVA. A normal approximation to the noncentral t distribution is described by
Wahlsten25,26 for continuous variables and tests of linear contrasts involving one
degree of freedom. Unlike methods that rely on tabled values, it can be applied using
only an electronic calculator with any reasonable choice of , number of tails, desired
power, and effect size.
Four methods are compared in Table 10.3 for the case of two independent groups
when the effect size is large. The values from Bausell and Li2 are the same as for the
Borenstein et al. program. All methods yield reasonably similar values, with Kraemer
and Thiemann15 consistently being a larger number than Cohen.8 Wahlsten26 is gen-
erally between those two but is closer to Kraemer and Thiemann15 for lower levels of
power and to Cohen8 for the higher levels of power. SamplePower by Borenstein et
al.6 yields values between those of Cohen and Wahlsten. Considering that the methods
never yield values differing by more than three or four subjects at the higher levels
of power and two subjects at the lower levels, all four are considered acceptable. The
TABLE 10.3
Four Estimates of Sample Size Needed to Detect a Large Effect Size of = 1.0
Sample Size in Each of Two Groups According to
Kraemer and Borenstein
Tails Power C, Thiemann14 Cohen8 Wahlsten26 et al.6
.05 1 80% 6.18 15 13 15 14
.05 1 90% 8.56 20 18 20 18
.05 1 95% 10.82 25 22 24 23
.05 1 99% 15.77 36 32 34 33
.05 2 80% 7.85 19 17 18 17
.05 2 90% 10.51 24 22 24 23
.05 2 95% 12.99 30 27 28 27
.05 2 99% 18.37 41 38 39 38
.01 1 80% 10.04 23 22 23 22
.01 1 90% 13.02 30 27 29 28
.01 1 95% 15.77 36 33 34 33
.01 1 99% 21.65 48 45 46 45
.01 2 80% 11.68 27 25 26 26
.01 2 90% 14.88 34 31 32 32
.01 2 95% 17.81 40 37 38 38
.01 2 99% 24.03 53 50 51 50
Mean of all 16 values: 31.3 28.7 30.1 29.2
Note: In the method of Wahlsten,25,26 the normal approximation to the noncentral t distribution involves
a constant C, that depends on the standard normal deviates for Type I error (z) and Type II error
(z), such that C, = (z + z)2. The sample size per group is then n = C,,/2 + 2. The resulting value
of n is rounded up to the nearest integer.
accuracy of any estimate of sample size will be undermined by small but inevitable
subject attrition during any large study, and the researcher is well advised to plan on
using a few more animals than the minimum indicated by a sample size calculation
in order to be sure that the final number of good data points is adequate.
10.6 POWER AND SAMPLE SIZE FOR TWO

INDEPENDENT GROUPS
The variable having the greatest impact on power and sample size is effect size. For
an experiment with two groups (e.g., +/+ versus / genotypes), power curves for
several values of effect size are shown in Figure 10.2. A family of curves for the
case with = 0.5 is shown in Figure 10.3. Using SamplePower or similar programs,
it is possible to generate a series of power curves, print them, and then use the figure
to determine sample size simply by drawing a line to the X axis from the point with
the appropriate level of power. For effect sizes between those shown in Figure 10.2,
interpolation between given values will be reasonably close to the desired value,
although a quick calculation would be superior to interpolation.
Perhaps the most difficult thing for someone new to the field to decide is what
effect size should be used to choose sample size. This will depend on two things.
First, we want to know what effect sizes are commonly reported in published articles
in this area of research, for example, in knockout studies. It would be reasonable to
expect an effect size in the range of what other labs have observed; grant review
and ethics committees are not likely to dispute such a figure.
Second, one must consider the stage of the research. If one is about to collect
phenotypes on an entirely new knockout and really has no idea about what the effects
may be or whether there may be noteworthy developmental compensation for the
missing gene products, there would be no firm basis for expecting either a very large
Comparison of 2 Independent Groups

100
90
Statistical Power (%)
80
Large Medium
Very = 1 = 0.75
Large n = 86
60 for 90%
= 1.5 Small Power
= 0.5
40
= .05
20 2-tailed
11 22 38
0
10 20 30 40 50
Number of Animals Per Group
FIGURE 10.2 Power vs. sample size for a two-tailed t test of the difference between two
independent groups when the null hypothesis is that the true group difference is = 0. The
sample size needed to yield a power of 90% is shown as an italicized numeral along the x axis.
Small effect size = 0.5

1.0
0.8 = .05
Power
0.6
0.4
= .01
0.2
52 65 82 97
0.0
0 10 20 30 40 50 60 70 80 90 100
Number of Subjects Per Group
FIGURE 10.3 Power vs. sample size for a t test comparing two groups when the null
hypothesis is = 0 and the alternative is = 0.5. Number of tails (1 or 2) for the test is
indicated in the circle on each line. The sample size needed to yield a power of 80% is shown
as an italicized numeral along the x axis.
or a small effect, and it would seem reasonable to choose sample size in order to
have perhaps 90% power to detect a medium effect of = 0.75. Usually we would
expect a knockout to change a phenotype in a direction that makes the animal less
viable, vigorous or competent, and a one-tailed test would be appropriate. If only
one major phenotype is to be assessed, a choice of = .05 would make sense, but
= .01 would be a better choice if the study will involve measures of several
phenotypes, each to be assessed with a separate test of significance.
In the event that the investigator has only 10 or even fewer mutant animals for
the first study, it is still important to attend to the power curve. For example, if the
test will be done one-tailed with = .05, Figure 10.2 shows that power will be 90%
only for a very large effect size, whereas power will be about 35% if the real effect
size is = 0.75. Thus, by considering the question of power, the investigator will
be forewarned that failure to reject the null that = 0 in no way proves that the
knockout has no effect at all, because the chance of a Type II error occuring would
be quite high for a medium or even large effect. The purpose of doing an initial
study with fewer than 10 mutant animals and a similar number of normal sibs would
be to evaluate the possibility of a severely abnormal phenotype. If such a dramatic
deficiency were not found, it would be wise to use larger samples in future research
with that particular knockout.
10.7 COMPARISON OF SEVERAL INBRED STRAINS

For many phenotypes, the variation among the more common inbred strains is very
large and often exceeds 2 = 0.2 or f = 0.5. The Mouse Phenome Project19
(www.jax.org/phenome) has designated 10 strains on the highest priority A list for
phenotyping, and these strains are chosen largely because of well-documented
differences among them in genotype. Thus, large strain differences are not at all
surprising. The Mouse Phenome Database (MPD) now contains a wide range of
phenotypic data on these strains, and the investigator can rely on these values to
guide an informed choice of f or 2 when estimating sample size to achieve a desired

level of power. It would make good sense to base the hypothesized value on a
phenotype that is in the same neural or behavioral domain as the planned study.
Once the value of 2 and then f have been decided and the number of strains is
also stipulated, the sample size can be estimated from the tables of Cohen8 or the
program of Borenstein et al.6 Figure 10.4 shows the power curves for an ANOVA on
8 inbred strains when their means are uniformly distributed from low to high. For
example, if mice are tested on an elevated plus maze to assess anxiety levels, the
average percent time in the open arms might range from 20 to 55% in steps of 5%.
The population standard deviation of strain means would then be 11.5%, and when
the standard deviation within strains is about 20%, f = 11.5/20 = 0.57, a very large
effect. If the null hypothesis of no strain difference is to be evaluated with = .05,
then only 10 mice per strain would be sufficient to achieve power of 90%. As the
anticipated effect size declines, the necessary sample size becomes much larger.
The sample size is not solely a consequence of , and f. The number of strains
also must be considered, as shown in Figure 10.5 where the effect size is f = 0.35
Comparison of 8 Strains with ANOVA

1.0
0.8
0.6
= .05
= .01
0.4
0.2 Large f = .5
0.0
0 10 20 30 40 50 60 70 80
1.0
0.8
Power
0.6
0.4
0.2 Medium f = .35

0.0
0 10 20 30 40 50 60 70 80
1.0
0.8
0.6
0.4
0.2 Small f = .2
0.0
0 10 20 30 40 50 60 70 80
Sample Size per Group
FIGURE 10.4 Power vs. sample size for an ANOVA comparing means of 8 strains of animals
for three values of Cohens f index of effect size. Black dots indicate the sample size for
power of 90%.
in each case. For 4 strains, 31 mice per strain would suffice to yield power of 90%,
whereas for 8 strains the sample size per group would be considerably less. When
power is at issue, it is apparent that studying a smaller number of strains does not
reduce the total number of animals greatly. For 10, 8, 6, and 4 strains, the total
sample would need to be 170, 160, 144, and 124, respectively. For 2 strains, the
sample would be considerably smaller, but the generality of the results would be
severely restricted. It does not necessarily follow that a study of 10 strains will cost
about the same amount as with 6 strains, because we usually begin the process of
choosing strains with the most common ones that are also least expensive and then
move to less commonly studied strains that cost more per animal.
10.8 COMPARING SPECIFIC GROUPS IN A ONE-WAY

DESIGN
Sometimes the investigator employs a design with multiple groups, but cannot answer
the central questions addressed by the design using the omnibus F test of statistical
significance in the one-way ANOVA. Instead, the intent is to compare specific groups
that differ in biologically interesting ways. Consider the example in Table 10.4
involving four groups of male mice from a backcross of the A B F1 hybrid to strain
A. By convention, the first genotype in a cross is the mother, so that the A B
hybrid is from an A strain female bred with a B strain male. It is of some interest
to know whether the four groups differ significantly, but there is deeper meaning in
specific group comparisons. All four groups have the same complement of autosomal
genes. The backcrosses [(A B) A] and [(B A) A] both have the Y chromo-
some from strain A and a F1 hybrid maternal environment, but one has the mito-
chondrial DNA from strain A and the other from strain B. The backcrosses
[A (A B)] and [A (B A)] differ in the strain contributing the Y chromosome.
Groups [(A B) A] and [A (B A)] both have strain A mtDNA and strain A Y
chromosome, but one has an inbred and the other a hybrid maternal environment.
Table 10.4 proposes plausible results of a study in which all three non-Mendelian
factors are important but to different degrees.
Different # of Strains f = .35 in Each Case

1.0
10
0.8 8
6
Power
0.6 4
2
0.4
0.2 = .05
0.0 17 20 24 31 44
0 10 20 30 40 50
Sample Size per Group
FIGURE 10.5 Power vs. sample size for an ANOVA on different numbers of strains when
f = 0.35. The sample size needed to yield a power of 90% is shown as an italicized numeral
along the x axis.
The method of contrasts13 can be used to compare specific groups in a logical

manner, each comparison being tested with a one degree of freedom t test. For a
contrast among J independent groups with true means j, each group is assigned a
contrast coefficient cj, such that cj = 0. The contrast value is then the sum of
weighted means across all groups: = cjj. For a comparison of any two groups,
the coefficients could be +1 and 1 or +0.5 and 0.5; results will be the same. In
Table 10.4 it is seen that the maternal environment contrast is twice that of the
mtDNA, while the Y effect is midway between them. The method of Wahlsten25,26
can be used to find the required sample sizes. For = .05 one-tailed and power of
95% ( = .05), the sample sizes per each of four groups need to be 51, 24, and 15
to detect the mtDNA, Y, and maternal environment effects, respectively. Presuming
that the investigator is equally interested in the possible presence of all three kinds
of effects, the sample size for the study should be 51 per group.
This logical approach to the dissection of variation among groups can be applied
to large and complex genetic experiments with animals.21,23
10.9 STRAIN TREATMENT EXPERIMENTS

Rarely does a researcher adopt a one-way design simply to compare several inbred
strains. Often the strain or genotype factor is combined with some environmental
treatment to create a two-way factorial design that will be analysed with an ANOVA.
The ANOVA will partition the resulting variance among three sources: main effect
for strain, main effect for environment, and strain environment interaction. Each
of these three effects will have a certain effect size that can be expressed in terms
of Cohens f or 2. In the case of 2, the published value will represent a partial 2
that compares variance for one factor to variance within groups while excluding
from the calculation the variances arising from the other two factors.9,13
TABLE 10.4
Contrasts among Males of Reciprocal Backcrosses to Strain A; Agonistic
Acts in the Home Cage Intruder Test
Contrast Coefficients to Detect Effect
Maternal
Mother Father Mean mtDNA Y Chromosome Environment
AB A 17 1 0 1
BA A 15 1 0 0
A AB 10 0 1 0
A B A 13 0 1 1
2 3 4
n per group 51 24 15
Note: Sample size is found using the method of Wahlsten25,26 with a one-tailed test at = .05 and
power of 95% when within = 3.0. As in the case of two groups (Table 10.3), C, = (z + z)2. For a
one degree of freedom contrast among J groups, n = Ccj2/(/)2 + 2.
In order to estimate sample size for a two-way design when each group is to
have the same sample size, one begins by positing cell means that would constitute
an effect of interest to the investigator. The values should be the kinds of means that
could realistically be observed for the strains, treatment, and phenotype in question,
as suggested by preliminary data or the published literature. If the investigator is
seriously interested in the possibility of a strain treatment interaction, then the
numerical model must express such an interaction. A computation is done to find
the sample size needed to detect the interaction as well as the main effects. In almost
all situations, the sample sizes needed to detect the three effects will be different,
but the study can actually be done with just one sample size, so the best option is
to choose the largest sample size needed to detect the smallest effect of interest.
This will ensure that power to detect the other two effects will be even larger than
the nominal level of power used in the calculation.
Consider the example shown in Table 10.5 where 4 strains reared in either
the usual lab environment or an enriched environment are trained to locate the sub-
merged platform in a water maze. Numbers represent mean latency to escape the
water after 4 days of training. It is expected that the standard deviation of latencies
within a strain will be about = 8.0 sec. In this case, there are major differences
among strains, a substantial effect of environment, and a noteworthy strain envi-
ronment interaction. For strain S1 the environmental treatment has no effect at all, it
improves latencies by 5 sec for strains S2 and S3, whereas it has a large effect of
10 sec on strain S4. This interaction appears to the researchers eye as something
important that ought to be detectable as a significant interaction effect by the ANOVA,
provided enough animals are tested to confer adequate power on the test.
TABLE 10.5
Calculation of Effect Sizes from Hypothetical Results of 4 Strain 2
Environment Design
A. Hypothetical Group Means with Interaction
Deviation
S1 S2 S3 S4 Row Mean
from G
E1 10 15 20 25 17.5 +2.5
E2 10 10 15 15 12.5 2.5
Column mean 10 12.5 17.5 20 G = 15
Dev. from G 5 2.5 +2.5 +5
B. Group Means under Null Hypothesis of Additivity
S1 S2 S3 S4 Row Mean
E1 12.5 15 20 22.5 17.5
E2 7.5 10 15 17.5 12.5
Column mean 10 12.5 17.5 20
C. Interaction Deviations of Cell Means in A from Additive Model in B
S1 S2 S3 S4
E1 2.5 0 0 +2.5
E2 +2.5 0 0 2.5
The next step in the planning process is to determine the three effect sizes. This
is relatively straightforward for the two main effects; find the mean of all values in
a row or column and then find the population standard deviation of all means along
one margin. For the strain effect, this will be the standard deviation of the four
marginal means 10, 12.5, 17.5 and 20, which is M = 3.95 or about 4. For the
environmental treatment we find M = 2.5. Thus, the effect sizes for strain and
treatment main effects are f = 4/8 = .5 and f = 2.5/8 = .31, respectively. Sample
sizes can be found from the tables in Cohen8 or using the SamplePower program by
Borenstein et al.6 The tabular approach requires two steps because the power tables
are set up for a one-way design. For = .05 and power of 90%, Cohens8 table 8.4.4
calls for 15 subjects per group when four groups have means with f = 0.5. The full
factorial study actually has 42 = 8 groups, however. Thus, the next step is to divide
the 15 subjects per strain between the two treatment conditions, resulting in a final
choice of 8 subjects per group. Similarly for the environmental factor, Table 8.4.4
calls for 57 subjects per group, but when these are distributed across the four strains,
the choice is 15 per cell. Full details of the method to find cell sample size from
tabled n are provided in Cohen.8
The SamplePower program uses the same shortcut for finding sample size for
the main effects, while concealing the internal workings of the calculations from the
user. One can enter means for the marginal values of each main effect, but it is not
possible to enter means for all 8 cells separately. Consequently, the program will
not compute the effect size for the interaction. Instead, the user must have a good
understanding of the statistical definition of interaction in order to compute its effect
size using either an electronic calculator, a spreadsheet such as Excel, or a flexible
data analysis program. The procedure is shown in Table 10.5.
The hypothesis of interaction is tested against the null hypothesis that strain and
environment effects are additive. Statistically, interaction is defined as departures
from additivity. For the example in panel A of Table 10.5, the grand mean G of all
eight group means is 15 sec, and the deviations of the two environment means from
this grand mean are +2.5 and 2.5 sec, these being the average effects of each
environment across the 4 strains. For the strains, deviations from 15 range from 5
to +5 sec. To find the deviation of each cell mean that is expected when the main
effects are additive, we simply add the separate deviations. For group S1E1, we add
its environmental deviation +2.5 sec to its strain deviation 5 sec to obtain 2.5 sec
below 15 or 12.5 sec. The results of these additions are shown in panel B of Table
10.5. Note that the marginal means in panel B are identical to those in panel A; the
only difference is that panel A expresses strain environment interaction, whereas
panel B does not. Next, we find the deviation of each cell in panel A from the
additive model in panel B and enter these interaction effects into panel C. Note that
for strains S2 and S3, the means expected under an additive relation (panel B) are
the same as proposed in panel A, while strains S1 and S4 deviate from the additive
model. Interaction effect size is then the standard deviation of these eight interaction
deviations (M = 1.77) divided by = 8 to yield effect size f = 0.22. This value can
then be used to find sample size from either Cohens table 8.4.4 or the SamplePower
program where 0.22 must be entered into the appropriate box, yielding an estimated
sample size of 38 animals in each of the eight groups.
The power curves for the three effects in the ANOVA are portrayed in Figure
10.6. As has been demonstrated for several plausible kinds of interactions,24 at a given
sample size the power to detect the interaction effect is substantially inferior to the
power to detect the main effects. Likewise, to achieve the same level of power as for
the main effects, the sample size to detect the interaction effect must be considerably
larger. The exact ratio of sample sizes to detect interaction and main effects depends
strongly on the specific kind of interaction. To date, no author has provided a satis-
factory rule for judging what constitutes a small, medium and large interaction in
neurobehavioral genetics. The only rule offered here is that the interaction should be
interesting or noteworthy when an experienced investigator examines the pattern of
cell means. Unfortunately, when power to detect an interaction is low, the pattern
may look interesting to the educated eye, yet fail to achieve significance at = .05.
Bausell and Li2 devote two chapters to power and sample size analysis for inter-
actions in several kinds of factorial designs. They note that power to detect interaction
tends to be lower than for main effects and that a formal power analysis is essential
when the investigator is interested in the possibility of interaction effects. Murphy and
Myors17 observe: In general, the study is most likely to have higher power for testing
main effects and lower power for testing complex interactions, and they warn that
very large samples may be needed for detecting complex interaction effects.
10.10 BRIDGING THE GAP: THE 22 DESIGN

One variant of the strain environment factorial experiment, the 22 design, lends
itself to analysis by the method of contrasts, and for this situation a reasonable
standard for what constitutes a noteworthy interaction can be established. The 22
design is often employed in neurobehavioral genetics to assess whether the effects
of a knockout depend on the strain background (epistasis) or whether two genotypes
respond similarly to a treatment (heredityenvironment interaction). It is the design
of choice when comparing the effects of the Y chromosome or the mitochondrial
DNA from two strains backcrossed onto the opposite strain background. Using the
Three Effects in 4 x 2 ANOVA

1.0
0.8
Strain
Strain x Env.
Power
0.6
0.4 Env.
0.2
0.0 8 15 38
0 10 20 30 40
Sample Size in Each of 8 Groups
FIGURE 10.6 Power vs. sample size for main and interaction effects in an ANOVA for an
experiment with 4 strains reared in 2 environments (Table 10.5). The sample size needed to
yield a power of 90% is shown as an italicized numeral along the x axis.
method of Wahlsten,25,26 it is possible to express the relative sample sizes needed to

detect the main effects and the interaction in a most succinct way.
Table 10.6 presents a situation where both genotype and environment have clear
effects on a phenotype but the effect size of environment on genotype 2 is twice as
large as on genotype 1. This is surely an interaction effect that our analysis ought
to be able to detect. As noted in Table 10.3, the sample size per group depends on
a constant C, = (z + z)2 that is determined by our choice of Type I and II error
probabilities. Provided that we plan to evaluate the main effects and interaction effect
in the 22 experiment using t tests on contrasts with the same false positive rate ,
the value of C, will be the same for the genotype, environment and interaction
effects. The sample size per group to detect a contrast is n = C,cj2/(/)2 + 2.
The magnitude of the contrast is considerably greater for the main effects than
the interaction, but the standard deviation within groups is the same in each
contrast. By expressing the contrast coefficients in Table 10.6 as 0.5, the term cj2
= 1 for all three contrasts. Thus, for the situation in Table 10.6 there are only two
quantities that vary in the sample size equation, n and . The ratio of the squared
contrast terms 2 for main effects and interactions is 9.0 when the treatment effect
on one genotype is twice as large as the other in Table 10.6. For this numerical
example, the required sample sizes for the genotype and genotype environment
interaction have a very simple relationship: nGE 2 = 9(nG 2) or, equivalently,
nGE = 9nG 16. The value of nG will depend on the effect size /, being smaller
when the within group variance is smaller, as shown in Figure 10.7. For generally
large effects where the necessary nG is quite small, the necessary sample size nGxE
needed to detect the interaction will be about six times larger than nG, whereas for
generally small effects where nG is itself fairly large, nGE will be more than eight
times larger than nG.
The exact ratio of nGE to nG or nE will depend on the specific model of means
for the four groups in the 22 design, but for most situations where the interaction
is something we might reasonably expect to observe, the need for larger samples to
detect the interaction than the main effects is evident in the algebra.24,25,26
TABLE 10.6
Genotype (G1, G2) by Environment (E1, E2) Experiment That Expresses a
Noteworthy Interaction Where the Environmental Effect on Genotype 2 Is
Twice the Magnitude of the Effect on Genotype 1
Group G1E1 G1E2 G2E1 G2E2
Mean j 10 20 20 40 cj2 = cjj 2
G effect cj 0.5 0.5 +0.5 +0.5 1.0 15 225
E effect cj 0.5 +0.5 0.5 +0.5 1.0 15 225
GE cj for 0.5 +0.5 +0.5 0.5 1.0 5 25

interaction
300 2 x 2 Design with
Sample size per group

Noteworthy Interaction
= = .05
200 n GxE
100
nG
0
0.5 1.0 1.5 2.0
Main Effect Size (/ )
FIGURE 10.7 Sample size needed to yield power of 95% when = .05 for an experiment
in which 2 strains are reared in two environments and the environmental effect on one strain
is twice the size of the effect on the other strain, as shown in Table 10.6. Sample size needed
to detect the interaction is substantially greater than to detect the main effects over a wide
range of effect sizes.
10.11 SAMPLE SIZE FOR MORE SPECIALIZED

EXPERIMENTS
The methods for determining sample size described here are generic and make no
assumptions about the genetics of our subjects. Being generic rather than genetic,
they can be applied in a wide variety of experiments, including many not addressed
in this chapter. They will not be useful for certain kinds of studies such as detection
and mapping of quantitative trait loci, and the reader will want to explore more
specialized treatments for determining sample sizes for that purpose.3,18 Neither do
they inform the question of the number of inbred strains to use or the proper number
of subjects per strain when assessing genetic correlation.10
ACKNOWLEDGMENTS
Preparation of this chapter was supported in part by grant 45825 from Natural Science
and Engineering Research Council of Canada and grant 2 R01 AA012714 from the
National Institutes of Health. The author is grateful to Elizabeth Munn for assistance.
REFERENCES
1. The Oxford English Reference Dictionary, Oxford, Oxford University Press, 1996.
2. Bausell, R.B. and Li, Y.-F., Power analysis for experimental research: a practical
guide for the biological, medical, and social sciences, Cambridge, UK, Cambridge
University Press, 2002.
3. Belknap, J.K., Mitchell, S.R., OToole, L.A., Helms, M.L., and Crabbe, J.C., Type I
and type II error rates for quantitative trait loci (QTL) mapping studies using recom-
binant inbred mouse strains, Behav. Genet., 26: 149160, 1996.
4. Benjamini, Y., Drai, D., Elmer, G., Kafkafi, N., and Golani, I., Controlling the false
discovery rate in behavior genetics research, Behav. Brain Res., 125: 279284, 2001.
5. Borenstein, M., Cohen, J., Rothstein, H., Schoenfeld, D., Berlin, J., and Lakatos, E.,
Power and Precision, Englewood, NJ, Biostat, 2001.
6. Borenstein, M., Rothstein, H., Cohen, J., Schoenfeld, D., and Berlin, J., SamplePower
2.0, Chicago, IL, SPSS, Inc., 2000.
7. Chow, S.L., Statistical Significance: Rationale, Validity, and Utility, London, Sage
Publications, 1996.
8. Cohen, J., Statistical Power Analysis for the Behavioral Sciences, Hillsdale, NJ,
Erlbaum, 1988.
9. Crabbe, J.C., Wahlsten, D., and Dudek, B.C., Genetics of mouse behavior: interactions
with laboratory environment, Science, 284: 16701672, 1999.
10. Crusio, W.E., A note on the effect of within-strain sample sizes on QTL mapping in
recombinant inbred strain studies, Genes Brain Behav., 3: 249251, 2004.
11. Egger, M., Smith, G.D., and Phillips, A.N., Meta-analysis. Principles and procedures,
British Medical Journal, 315: 15331537, 1997.
12. Harlow, L.L., Mulaik, S.A., and Steiger, A.H., Eds., What if There Were No Signifi-
cance Tests? Mahwah, NJ, Lawrence Erlbaum Associates, 1997.
13. Hays, W.L., Statistics, 4th edition, New York, Holt, Rinehart, Winston, 1988.
14. Hedges, L.V. and Olkin, I., Statistical Methods for Meta-Analysis, Orlando, Academic
Press, 1985.
15. Kraemer, H.C. and Thiemann, S., How Many Subjects? Statistical Power Analysis in
Research, Newbury Park, Sage Publications, 1987.
16. Lander, E. and Kruglyak, L., Genetic dissection of complex traits: Guidelines for
interpreting and reporting linkage results, Nat. Genet., 11: 241247, 1995.
17. Murphy, K.R. and Myors, B., Statistical Power Analysis: a Simple and General Model
for Traditional and Modern Hypothesis Tests, Mahwah, NJ, Lawrence Erlbaum Asso-
ciates, 2004.
18. Neumann, P.E., Three-locus linkage analysis using recombinant inbred strains and
Bayes theorem, Genetics, 128: 631638, 1991.
19. Paigen, K. and Eppig, J.T., A mouse phenome project, Mamm. Genome, 11: 715717,
2000.
20. Rosenberg, M.S., Adams, D.C., and Gurevitch, J., MetaWin Version 2.0, Sunderland,
MA, Sinauer Associates, 1997.
21. Sokolowski, M.B., Genetic analysis of behavior in the fruit fly, Drosophila melano-
gaster, in Techniques for the Genetic Analysis of Brain and Behavior. Focus on the
Mouse, D. Goldowitz, D. Wahlsten, and R.E. Wimer, Eds. Amsterdam, Elsevier,
497512, 1992.
22. Thomas, L. and Krebs, C.J., A review of statistical power analysis software, Bulletin
of the Ecological Society of America, 78: 126139, 1997.
23. Wahlsten, D., Behavioral genetics and animal learning, in Psychopharmacology of
Aversively Motivated Behaviors, H. Anisman and G. Bignami, Eds. New York, Ple-
num, 63118, 1978.
24. Wahlsten, D., Insensitivity of the analysis of variance to heredityenvironment inter-
action, Behav. Brain Sci., 13: 109120, 1990.
25. Wahlsten, D., Sample size to detect a planned contrast and a one degree-of-freedom
interaction effect, Psychol. Bull., 110: 587595, 1991.
26. Wahlsten, D., Experimental design and statistical inference, in Molecular-Genetic
Techniques for Behavioral Neuroscience, W.E. Crusio and R.T. Gerlai, Eds. Amster-
dam, Elsevier, 4057, 1999.
11 The Role of Association

Studies in Psychiatric
Disorders
Nicolas Ramoz and Philip Gorwood
CONTENTS
Introduction............................................................................................................ 169
11.1 Association Studies: Definitions and Mechanisms ..................................... 170
11.1.1 Case Control Association Study ...................................................... 170
11.1.2 Family-Based Association Studies .................................................. 171
11.2 Advantage of Association Studies............................................................... 172
11.2.1 Power................................................................................................ 172
11.2.2 Linkage Disequilibrium ................................................................... 173
11.3 Limitations of Association Studies.............................................................. 173
11.3.1 False Positives: Linkage Disequilibrium
and Control Groups.......................................................................... 174
11.3.2 Statistical Power: Accept or Reject an Association........................ 175
11.3.3 Consequences in Neurobehavioral Genetics ................................... 175
11.4 Progress in Neurobehavioral Genetics ........................................................ 176
11.4.1 One Gene Many Phenotypes (the Case
of Phenotypical Heterogeneity) ....................................................... 177
11.4.2 One Phenotype Many Genes (the Case
of Genetic Heterogeneity)................................................................ 177
11.4.3 Interaction between the Environment and Genetic
Vulnerability..................................................................................... 178
11.5 Conclusions .................................................................................................. 180
References.............................................................................................................. 180
INTRODUCTION
In the beginning of the twenty-first century, the fiftieth anniversary of the discovery
of DNA was celebrated, and despite the rapid progress in both human genetics and
molecular biological technologies since, neurobehavioral genetics has provided us
with only its first promising successes. The human genome has a physical size of
3109 base pairs that encode and regulate approximately 32,000 genes. Some of
169
these genes play a role in the aetiology of psychiatric disorders. However, the model
of segregation for these disorders does not follow a simple Mendelian trait but
follows complex traits where both polygenetic and environmental factors influence
the susceptibility to neurobehavioral diseases. The use of association studies to
complement linkage analyses provides a powerful tool to identify the genetic com-
ponent in psychiatric disorders.
11.1 ASSOCIATION STUDIES: DEFINITIONS AND

MECHANISMS
Association studies compare the occurrence in a population (case and control) or
kindred (family based) of two or more features (usually phenotype vs. genotype)
with a frequency greater than that anticipated on the basis of chance alone. Associ-
ation studies are widely used in genetics to search for the involvement and/or the
localization of a gene in the risk of certain disorders.
11.1.1 CASE CONTROL ASSOCIATION STUDY

The case control association study analyzes the distributions of allele frequencies or
genotypes in different groups of unrelated patients, with and without a certain phe-
notype, to those from matched controls represented by unaffected individuals. The
case/control study can be performed with one marker or a combination of markers
(haplotypes). The test determines the existence of a significant association, usually
with a fixed 5% risk of type I error. These parameters are explained in Table 11.1.
The measure of association can be both qualitative, with a presence or an absence
of a significant association, and quantitative, with a strength of association evaluated
by risk difference (RD), relative risk (RR), odds ratio (OR), and attributable risk
(AR). The RD computation evaluates the presence of the genetic marker in the
affected group, while taking into account the existence of subjects without the phe-
notype who also carry the marker. The RR calculation is a ratio of these probabilities,
TABLE 11.1
Measures of Association
Phenotype Genetic Marker
Present Absent Total
Affected a b a+b
Controls c d c+d
Total a+c b+d N
Notes: 2 (qualitative approach) = (obscalc)/calc. Where a, b, c, d are the observed (obs) numbers,
and (a+b)*(a+c)/N is the calculated (calc) value for a, and so on for b, c, d.
RD (risk difference) = [a/(a+b)][c/(c+d)]; RR (relative risk) = [a/(a+c)] / [b/(b+d)]; OR (odds ratio) =

ad/bc; AR (attributable risk) = [a/(a+c)]*[(OR1)/OR].
The Role of Association Studies in Psychiatric Disorders 171
instead of a difference, whereas, the OR value is the ratio of odds for each population.
The significance of the association can be computed from the OR.1 Finally, AR
provides the strength of the association and reflects the impact of the allele in
explaining the phenotype, since it is dependent on the OR and also the allele frequency
in the general population.
The case/control association study is a useful tool that provides a powerful
statistical test. However, one major limitation of association analysis can be caused
by population stratification, i.e., when cases and controls are not ethnically matched.
11.1.2 FAMILY-BASED ASSOCIATION STUDIES

There are several other approaches that can be used to address the stratification bias,
and all use family-based association studies.
Thus, the haplotype relative risk (HRR) test provides OR by comparing allele
frequencies of a marker or a haplotype in affected offspring with frequencies in a
virtual control builds with parental alleles not transmitted to affected offspring
(Figure 11.1).2,3
The transmission disequilibrium test (TDT) is another approach that considers
parents who are heterozygous for an allele associated with the disease, and evaluates
the frequency of transmission vs. nontransmission of marker alleles to affected off-
spring.4 TDT is also considered as being a test for linkage in presence of association,
especially for fine-mapping.5 Several tests were developed to include multi-allelic
markers and haplotypes, such as the extended TDT (ETDT).6 The major limitation
of TDT is the use of complete triads because families with one missing parent could
introduce a bias. To counter this problem, the sib-transmission disequilibrium test (S-
TDT) was designed by Spielman and Ewens.7 This test compares the allele frequen-
cies between affected and unaffected siblings in order to measure difference. Since
complex disorders may not be characterized by dichotomous traits, without covariates,
test of association for quantitative traits (QTDT) was designed for nuclear families
of any size, with or without the parental phenotype.8 One again, statistical power of
QTDT is greater than that with the genotyping of parents. The information in pedi-
grees of any size, in particular the large or extended pedigrees, could be taken into
account by using the pedigree disequilibrium test (PDT).9 The PDT provides the
statistical significance of linkage disequilibrium for the entire pedigree. Of the tests
discussed so far, the PDT appears to be the most powerful.9 Furthermore, additional
statistical tests based on likelihood approaches, and not on test, are also available.
A/B C/D
A/C B/D
Affected offspring Virtual control

(Observed genotype) (Built genotype with non-transmitted alleles)
FIGURE 11.1 The haplotype relative risk (HRR) method.
These tests include TRANSMIT program, which can compute association for
marker(s) or haplotype(s) to affected subjects or to one randomly selected from
families, either in the to absence of parental genotypes or to unknown haplotype
phase.10 These family-based association tests are now available in packages, e.g.,
FBAT.11
11.2 ADVANTAGE OF ASSOCIATION STUDIES

To identify the impact of markers, haplotypes, regions, or candidate genes in
the risk for a disorder, the association studies are based on strategy of comparing
correlations between phenotype and genotype in case/control or family-based
populations.
11.2.1 POWER
Whole genome scans followed by genetic linkage analyses are considered to be the
most powerful strategies to identify genes of heritable diseases. Linkage analyses
are successfully used for genetic disorders harboring Mendelian inheritance patterns,
but may not be applicable for common diseases that have a more complex pattern
of inheritance. Linkage analyses present several methodological limitations for deter-
mining polygenic disorders when the assumption of a single major gene is incorrect,
or genetic heterogeneity and sporadic cases are present, or penetrance of suscepti-
bility genotypes is reduced, or when the genetic aetiology is not well defined.12
These features all characterize the genetics of animal and human neurobehavior, and
therefore association studies may be more relevant.
In fact, linkage analyses are comparisons of the likelihood to observe a link
between markers and loci (transmitted together with greater frequency than 50%),
while association studies compare discrete traits and alleles (found together with
greater frequency than by chance alone). Linkage analyses attempt to detect physical
mapping on the chromosome while association studies look for a causal relation-
ship. Linkage analyses also need the screening of multigenerational families, while
association studies require only the studying of unrelated individuals within a
population, or triad families. According to these characteristics, association studies
should be the preferred initial choice rather than linkage studies. Association studies
are more appropriate for polygenic (many additive genes, each with a small effect),
multiloci (two or more genes involved), multifactorial (epigenetic factors, such as
environment, play a significant role in the individual vulnerability), heterogeneous
(phenotype contains different entities that are dependent on several factors) disor-
ders or traits.
Furthermore, the power of association studies vs. linkage studies to detect a
significant role of a gene involved in complex human disease has been compared
recently.13 When the genotypic RR is low (<2.0), then the number of families needed
to detect a significant linkage is so large that it becomes impractical to achieve
(12,000 to 300,000 for allele frequencies between 0.01 and 0.50). In comparison,
the number of triplets to be investigated with the TDT method is much lower (340
to 5800), for allele frequencies between 0.01 and 0.50. Simulations of the power
of association studies can be performed with the recent genetic power calculator
program (GPC; http://statgen.iop.kcl.ac.uk/gpc/).14 These results have shown that
association studies are relevant for genetic susceptibility or liability, but not for
genetic determinism in complex genetic diseases, such as neurobehavior disorders.
11.2.2 LINKAGE DISEQUILIBRIUM

Association studies are also statistical tools that can be used to test the allele
frequencies of some genetic polymorphisms that are potentially close to and in
linkage disequilibrium (LD) with the gene involved in the feature analyzed. This
analysis is therefore dedicated to localization studies, whether for systematic screen-
ing of the whole genome or for a fine-mapping in specific candidate regions of
interest.
A certain phenotype can be influenced by a specific allele, and the polymorphisms
of a closed genetic marker should be associated with this phenotype if they are in LD.
Association studies could be used in conjunction with LD to perform to human genetic
analysis. In the absence of selection, population heterogeneity, or other confounding
phenomena, LD should by theory be linearly related to distance. The utilization of
linkage disequilibrium concepts in association studies has some limitations that could
be partially raised, such as the gap between physical and linkage distance.
Single nucleotide polymorphism (SNP) is a recently developed tool in molecular
biology and genetics. SNPs are a useful and powerful tool. Genotyping of SNPs is
simple, efficient, quick, easily automatizable, and cheap. It is powerful because the
resolution with SNPs is 10 times higher than microsatellites, short tandem repeats
(STRs), or variable number of tandem repeats (VNTRs). SNPs are present in the
human genome once every 600 base pairs. Typically, SNPs segregate as specific
combinations or haplotypes, or blocks that harbour a high linkage disequilibrium
and a limited diversity.15 In other words, few SNPs are request and sufficient to
screen for association between a disorder trait and a gene. The choice of the SNPs
and their information regarding their allelic frequencies and linkage disequilibrium
in populations are available in databases, in particular with the international consor-
tium HAPMAP (www.hapmap.org). Looking for a strong and specific excess of an
allele, genotype or haplotype in a group of subjects compared to unaffected matched
controls according to the presence of linkage disequilibrium allows to the identifi-
cation of disease associations between genes and human traits. However, it is exposed
to certain limitations.
11.3 LIMITATIONS OF ASSOCIATION STUDIES

A major problem in association studies is false positives. They occur when the
populations of case controls are not matched ethnically, and when there is a lack of
parent data in family-based or with a low frequency of the minor allele of the
genotyped marker. The other limitation is the statistical power of the association
study.
11.3.1 FALSE POSITIVES: LINKAGE DISEQUILIBRIUM

AND CONTROL GROUPS
The genetic heterogeneity of a population is directly linked to the linkage

disequilibrium between variants or markers in the genome. Thus, the relation
between physical distance and linkage disequilibrium depends on so-called hot
or cold spots of recombination that can be detected with frequent, or low,
recombinations on small distance of chromatides.16 Telomeric regions present a
higher frequency of recombinations, particularly to centromeric regions, than
the rest of the genome. Furthermore, the decay of linkage disequilibrium in
succeeding generations is governed not only by the recombination frequency
(i.e., the frequency of crossovers between the relevant loci during meiosis), but
also the number of generations since the introduction of the mutation into the
population.17 For example, the half life of an association would be 69 gener-
ations or about 2,000 years at a recombination fraction of 0.01.18 Nevertheless,
the shared segments extended over large genetic distances (1 megabases) in a
Finnish sample established over 2,000 years ago.19 Thus, linkage disequilibrium
may not be used as a constant and uniform concept. In conclusion, linkage
disequilibrium studies should be more productive among genetically isolated
populations with a well-documented history.20
The second limitation of association studies is based on the required comparison
of cohorts (affected vs. unaffected). First of all, association studies have to control
the stratification effect, i.e., one of the samples, affected or controls, derived from
an ethnic group sharing other DNA variants for reasons unrelated to illness. Control
and affected cohorts have thus to be matched for ethnic origins. This kind of
limitation is now well controlled and measured by genotyping specific markers
across the genome and is likely independent of the goal of the study. The control
group based on volunteers in human association studies is also subject to bias as
volunteer subjects have rates of psychopathology that far exceed the population
expectations.12,2123 The use of unrelated relatives, spouses, is also a potential artefact
because they are more frequently affected (assortative mating).24 The other way is
to perform family-based association study when parental data are available, as we
described in Section 11.1.2.
The limitation may come from the proband to study. This question is still
conflicting. For some authors, few severely affected probands from numerous famil-
ial cases are required; for others, much more representative affected subjects are
sufficient. The question is also unresolved for controls cohort: (1) extracted from
the general population, regardless of their phenotype (if the mutation analyzed is
involved, affected probands should be different from the whole population),
(2) extracted from the general population, but excluding affected subjects, and
(3) without any potentially linked disease, in the proband interview and also in his
first (and sometimes second) degree family. This last cohort of controls (labeled
super-control) has been criticized as favoring the observation of coaggregation of
different diseases in the affected group.25
Finally, in multifactorial disorders and complex traits, the existence of incom-
plete penetrance (i.e., subjects with the vulnerability gene that do not express the
TABLE 11.2
Estimation of Each Sample Size Needed to Show a Significant Association
According to Allele Frequency in Controls and the Expected Difference
between Patients and Controls
Allele Frequency in Control (N)
Expected Difference in Allele Frequency between Patients and Controls

N 1% 2% 5% 10% 20% 30% 40% 50%
1% 2578 822 207 80 32 19* 13* 10*
2% 4332 1270 284 100 38 22 15* 11*
5% 9312 2515 485 223 68 34 21 14*
10% 16862 4386 778 223 68 34 21 14*
20% 29254 7442 1249 334 92 44 26 17*
30% 39052 9598 1574 407 106 48 27 17*
40% 43259 10854 1754 443 111 48 26 14*
50% 44875 11214 1790 443 106 44 21 8*
*= Sample size lower than 20 subjects requires special test corrections.
phenotype) renders the choice of the control group even more critical. For example,
for research on a common disorder such as alcoholism, it may be more relevant to
recruit controls that drink alcohol regularly without developing alcohol abuse or
dependence, instead of abstainers, who may quickly develop alcoholism if they were
exposed to alcohol consumption.
11.3.2 STATISTICAL POWER: ACCEPT OR REJECT AN ASSOCIATION

The statistical power and specificity of the association detected or rejected are other
problems. Small sample size and low heterogeneity of allele frequencies are impor-
tant limitations in the opportunity to show linkage disequilibrium between two
closely linked markers.26 The chances of detecting the association are reduced if the
allele conferring susceptibility is common among unaffected individuals.27 The sam-
ple size required to detect a significant association can be evaluated (Table 11.2).
These estimations are relevant for testing a single hypothesis (relationship between
a marker and a phenotype), without prior notion of the type of association (associ-
ation or exclusion), and with low type I and type II errors ( = 5% and = 10%).
The number of patients that have to be included are given according to the frequen-
cies of the studied allele in the control population (from 0.01 to 0.50) and the
expected difference in allele frequency between patients and controls. These power
computations of association studies can now be simulated by investigators prior to
promoting any cohort or performing any genetic screening with the genetic power
calculator program (GPC; http://statgen.iop.kcl.ac.uk/gpc/).14
11.3.3 CONSEQUENCES IN NEUROBEHAVIORAL GENETICS

Statistical corrections may be applied to assess the question of false-positive and
the power of association studies to accept or reject the hypothesis.
Thus, the importance of the likelihood of false-positive results in the genetics

of behavior and psychiatric disorders, on the basis of the number of candidate genes
(namely all genes expressed in the brain) could be raised.28 Bonferronis correction
could be performed by dividing the 0.05 first risk error by the number of markers
screened, and a statistical significance of 0.00001 should be achieved to avoid a
false-positive rate of 5%. Using a significant level of 0.0001, 80% of positive findings
would be false, and the traditional level of 0.05 would yield 99.5% false-positive
results. According to such a false-positive rate and considering that half of the genes
are expressed in the brain, there are potentially 100,000 association studies which
could be carried out for each neurobehavioral trait studied. Regarding such a nearly
impossible-to-reach level of significance, alternative approaches have been proposed.
However, consistent replications, on independent cohorts, would be the best evidence
for a true association.29
Considering all the advantages and limitations of the methodology previously
described for association studies, the role and the impact of these approaches need
to be discussed further and compared with the different tests and statistical correc-
tions. One of the frequent criticisms of positive association studies is the lack of
reproducibility. Some journals propose to refuse any more positive case-control
genetic association studies of complex traits.30 Association studies have a lower
chance to throw light on the biological mechanism underlying the association. A
major disease gene at another locus in linkage disequilibrium with the associated
marker locus or a minor gene possibly located at the marker locus itself cannot be
distinguished. Association studies are nevertheless extremely powerful in detecting
small, partial or complex genetic effects, although their high sensibility is associated
with low specificity. Often, association studies are in fact specifically required, and
different approaches and tests may easily increase their validity. There are family-
based association studies described, in Section 11.1.2, that do not request control
group but needed parental data. Instead of preferring sensibility to specificity, a step-
by-step strategy could be chosen, especially as it is in accordance with the recruit-
ment of affected subjects. Beginning studies with a case-control analysis is strong,
but the recruitment of parents may be done later. DNA from parents can be used to
see if the allele found in excess in the affected sample is more frequently transmitted
to the affected proband than not transmitted from the parents. Affected sibs can be
later recruited (sibpair analysis), and once multiplex families are detected, linkage
studies are easier to perform (LOD score linkage analysis). Combined with high-
throughput genotyping of markers, including SNPs, case-control or family-based
association studies allow identification of the genetic component in neurobehavioral
disorders.3133
11.4 PROGRESS IN NEUROBEHAVIORAL GENETICS

Apart from the apparent absence of monogenic inheritance, even in some families,
there are three main reasons explaining how no definite results were found in
psychiatric genetics, although tremendous efforts have been made to depict which
genes are involved. The case-control association study approach may be partially
helpful in each case.
11.4.1 ONE GENE MANY PHENOTYPES (THE CASE

OF PHENOTYPICAL HETEROGENEITY)
International and validated classifications (such as ICD or DSM) of psychiatric disor-

ders are mainly based on clinical intuition, are partly influenced by epidemiological
studies, and depend, to a moderate degree, on treatment response. But none of the
decisive factors are relying on biological aspects, for an obvious lack of clear results.
But classifications based on biological specificities are much easier to use when
looking for genes. A good example of this problem in psychiatry is the distinction
between mood depressive disorder (MDD) and generalized anxiety disorder (GAD).
Indeed, the vulnerability factors involved, and even more specifically for the genes
involved, are much more shared than specific. The proposition has even been made
that GAD and MDD are two different aspects of the same disorder, an example of
genetic pleiotropy.34
At the molecular genetic level, the gene which has been the most frequently
analyzed is, without a doubt, the gene coding for the serotonin transporter (5-HTT).
From the year 2000, more than 500 studies analyzed this gene in more than 20
different phenotypes, mainly psychiatric disorders, traits or temperaments. Some of
these findings are now convincing, for example regarding the role of the short allele
in the risk of suicidal behavior.3536 Yet, it is difficult to disentangle at which level
the short allele of the 5-HTT gene is increasing the risk for psychiatric morbidity,
as this allele has been found associated at least once in each step of mood disorders,
including the temperament: harm-avoidance,37 which is a well-known risk factor
for depression; general anxiety disorder,38 which has a high comorbidity with mood
disorders; major depressive disorder and bipolar disorder per se;3940 suicidal behav-
ior,3536 which is a relatively frequent complication of the latter, and antidepressant
treatment response.41
An alternative approach is to focus on intermediate phenotype.42 Such phenotype
might be closer to vulnerability allele (as relying on simpler biological and physi-
ological mechanisms) and are directly related to the concept of genetic vulnerability
to the disorder (as, for example, being found in excess in nonaffected relatives of
the proband). The well-known role of the amygdala in anxiety has been assessed
regarding the role of the 5-HTT gene. An experimental paradigm was thus used, in
showing the face of a very anxious person, and looking at which part of the cortex
was activated. This study showed that subjects with the short allele are activating
the amygdala, contrary to those who do not have this vulnerability allele.43 As
serotonin has a core role in balancing internal vs. external stimuli, and as the
amygdala hyperactivity is probably a core mechanism in anxious disorder, such case-
control approach might be particularly more powerful to depict the genes involved
in such a complex disorder with relatively low heritability.
11.4.2 ONE PHENOTYPE MANY GENES (THE CASE

OF GENETIC HETEROGENEITY)
There are not many psychiatric disorders for which some evidence exists about the
clear involvement of some vulnerability genes. Autism might be one of them. Jamain
et al.44 reported mutations in the X-linked neuroligin 3 and 4 genes in Swedish

sibpairs with autism. A frameshift mutation in NLGN4 appeared de novo in the
mother, cosegregated with an affected brother with Asperger syndrome and was
absent in a normal brother. This frameshift mutation was not present in 600 unrelated
control X-chromosomes. A mis-sense mutation in NLGN3, was found in the mother
and two sibs, one with autism and another with Asperger syndrome, but no other
relatives were studied. It was not found in 300 unrelated control X-chromosomes.
Laumonnier et al.45 reported later on a large French family in which 10 males had
nonspecific X-linked mental retardation, two had autism, and one had pervasive
developmental disorder. All affected patients were found to have the same frameshift
mutation (1253delAG) in the NLGN4 gene. A third study scanned and sequenced
2.5 Mb of the same gene in 150 patients with autism, and revealed an association
of NLGN4 structural variants at highly conserved amino acids with an estimated
attributable risk for autism of about 3% in these cohorts.46
Thus, important evidence was in favor of the conclusion that the first vulnera-
bility genes in autism were found. On the other hand, screening a large set of
individuals affected with autism (N = 196) did not reveal any mutation for these
two genes associated with the disorder,47 showing that many mutations are probably
involved in autism. Using case-control or within-family association studies is an
important tool to cope with the difficulty of finding rare mutations.
11.4.3 INTERACTION BETWEEN THE ENVIRONMENT AND GENETIC

VULNERABILITY
Case-control association studies also offer a unique opportunity to test the interaction
between genes and the environment. Archibold Garrod48 was probably one of the
first to raise the importance of the interaction between environment and genetics
when studying complex disorders, response to treatment or even pure genetic dis-
orders. Assessing such potential interaction may be particularly important in psy-
chiatric disorders for which various stressors are nearly systematically found before
the onset of the disease.
First of all, the estimate of the populations attributable risk for genetic, as well
as for environmental, factors might be biased when such interaction is not taken into
account.49 If a gene has an important but specific impact on the risk for a disorder
when, and only when, a specific environmental condition is present, then we could
expect a large variability in findings, according to the prevalence of the environ-
mental conditions. A good example to highlight this problem is the risk for aggressive
behavior at adult age. As antisocial, aggressive and impulsive behaviors belong to
the antisocial personality disorder spectrum, a personality disorder that has substan-
tial heritability (around 50%), a cohort study analyzed the role of the gene coding
for the MAO-A (the monoamine oxydase-A degrades all monoamines, including
serotonin) in the risk of aggressive behavior at adulthood.50 The rare allele (which
codes for a hypoactive enzyme) was indeed associated with an increased risk of
aggression at adulthood, but specifically for subjects who were exposed to violence
during childhood. In this example, it is the assessment of both these two factors that
allowed the possibility to show how genes (in this case the MAO-A gene) and
environment (here aggression from the parents during childhood) are significantly
involved in the risk of aggression in adults.
A second interesting aspect is the greater chance to discover which environmen-
tal factors are interacting with genes, in order to propose a prevention strategy. For
example, the role of cannabis in schizophrenia has been the topic of numerous
discussions, the difficulty being to disentangle the role of cannabis as a marker (i.e.,
those who will ultimately develop schizophrenia are more prone to use cannabis)
or a risk factor (i.e., those who use cannabis increase their risk for schizophrenia).
One study, focusing on geneenvironment interactions, showed that cannabis is a
risk factor, but mainly for a subgroup of at-risk patients as having familial history
of psychosis.51
The adequate way to assess the existence of a geneenvironment interaction
depends on the selected sample. The golden standard would be based on the prospec-
tive collection of both environmental and lifestyle data and DNA, if (1) the disease,
(2) the vulnerability allele(s), and (3) major environmental risk factors are frequent
enough. Collecting DNA at the end of the cohort would expose to stratification bias
when DNA is not obtained from all patients and controls. For rare disorders, such as
autism, a case-control approach assessing geneenvironment interaction retrospec-
tively might be more appropriate, although more exposed to different risks. Such
biases could include unnatural selection (for more severe or treated patients), popu-
lation stratification, survivor bias (suicide being a major problem for bipolar disorder
for example) and imperfect recall (age at onset and number of lifetime major depres-
sive episodes are sometimes difficult to remember for old depressed patients).
Statistically, there are two ways to demonstrate a significant geneenvironment
interaction. Table 11.3 could be used, defining a, b, c and d as the frequency of affected
patients, taking into account presence vs. absence of the vulnerability factor and/or
the vulnerability allele. Testing the departure from the multiplicative model of inter-
action is the most frequently used way, comparing the relative risk of the population
b (exposed to both factors) from the relative risk of a (only environmental risk)
multiplied by the relative risk of d (only genetic risk). Relative risks are created with
the reference category c for which the relative risk is, by definition 1 (no exposition),
and using the 95% confidence interval of each relative risk to assess the significance
of the difference between these two relative risks. The alternative (the additive model)
is to use rate differences, the effect of genes (frequency d) plus the effect of environment
(frequency a) being smaller than the effect of both factors (b).
TABLE 11.3
Assessing GeneEnvironmental Interaction on the Basis of the Percentage
of Affected Patients Exposed (vs. Unexposed) to Each Factor
Environmental Exposure Genotype
Wild Type Variant
Exposed a b
Unexposed c d
11.5 CONCLUSIONS
Association studies have one of the highest risk of false-positive findings, are
exposed to numerous bias because of the selection of probands (frequently variable
in different studies, even when having the same diagnosis) and controls [from too
healthy (super-normal controls) to to close with the analysed phenotype (general
population)], are extremely sensitive to the problem of mismatching controls with
cases for ethnic background (stratification bias), and are particularly exposed to the
problem of the nonlinearity of linkage disequilibrium through the genome. Yet, the
association study approach is one of the most powerful, i.e., being able to discover
the role which has a small attributable risk, and has many advantages that are summed
up in this chapter. Thus, this approach should be used cautiously, in conjunction
with more specific techniques, and be more appropriate for complex traits, such as
polygenic (many additive genes each with a small effect), multiloci (two or more
genes involved), multifactorial (epigenetic factors such as environment play a sig-
nificant role in the individual vulnerability), heterogeneous (phenotype contains
different entities which are dependent on several factors) disorders or traits. As all
these characteristics may be particularly true for neuropsychiatric disorders, associ-
ation studies still have a role to play.
REFERENCES
1. Mantel, N. and Haenzel, W., Statistical aspects of the analysis of data from retro-
spective studies of disease, J. Natl. Cancer Inst., 22, 719, 1959.
2. Falk, C.T. and Rubinstein, P., Haplotype relative risk: an easy reliable way to construct
a proper control sample for risk calculation, Ann. Hum. Gen., 51, 227, 1987.
3. Ott, J., Statistical properties of the haplotype relative risk, Genet. Epidemiol., 6, 127,
1989.
4. Spielman, R.S., McInnis, R.E., and Ewens, W.J., Transmission Test for Linkage
Disequilibrium: The insulin gene region and insulin-dependent diabetus mellitus
(IDDM), Am. J. Hum. Genet., 52, 506, 1993.
5. Spielman, R. and Ewens, W., The TDT and other family-based tests for linkage
disequilibrium and association, Am. J. Hum. Genet., 59, 983, 1996.
6. Sham, P.C. and Curtis, D., An extended transmission/disequilibrium test (TDT) for
multi-allele marker loci, Ann. Hum. Genet., 59, 323, 1995.
7. Spielman, R.S. and Ewens, W.J., A sibship test for linkage in the presence of asso-
ciation: the sib transmission/disequilibrium test, Am. J. Hum. Genet., 62, 450, 1998.
8. Abecasis, G.R., Cardon, L.R., and Cookson, W.O., A general test of association for
quantitative traits in nuclear families, Am. J. Hum. Genet., 66, 279, 2000.
9. Martin, E.R. et al., A test for linkage and association in general pedigrees: the pedigree
disequilibrium test, Am. J. Hum. Genet., 67, 146, 2000.
10. Clayton, D., A generalization of the transmission/disequilibrium test for uncertain-
haplotype transmission, Am. J. Hum. Genet., 65, 1170, 1999.
11. Laird, N., Horvath, S., and Xu, X., Implementing an untried approach to family based
tests of association, Genet. Epidemiol., 19, S36, 2000.
12. Risch, S. et al., Ensuring the normalcy of normal volunteers, Am. J. Psychiatry,
147, 682, 1990.
13. Risch, N. and Merikangas, K., The future of genetic studies of complex human
diseases, Science, 273, 1516, 1996.
14. Purcell, S., Sham, P., and Daly, N.J., Parental phenotypes in family-based association
analysis, Am. J. Hum. Genet., 76, 249, 2005.
15. Patil, N. et al., Blocks of limited haplotype diversity revealed by high-resolution
scanning of human chromosome 21, Science, 294, 1719, 2001.
16. Chakravarti, A. and Nei, M., Utility and efficiency of linked marker genes for genetic
counseling. II. Identification of linkage phase by offspring phenotypes, Am. J. Hum.
Genet., 34, 531, 1982.
17. Jorde, L.B., Linkage disequilibrium as a gene-mapping tool, Am. J. Hum. Genet., 56,
11, 1995.
18. Owen, M. and McGuffin, P., Association and linkage: complementary strategies for
complex disorders, J. Med. Genet., 30, 638, 1993.
19. Peterson, A.C., et al., The distribution of linkage disequilibrium over anonymous
genome regions, Hum. Mol. Genet., 4, 887, 1995.
20. Nimgaonkar, V., In defense of association studies, Mol. Psychiatry, 2, 275, 1997.
21. Gibbons, R., Davis, J., and Hedeker, D., A comment on the selection of healthy
controls for psychiatric experiments, Arch. Gen. Psychiatry, 47, 785, 1990.
22. Shtasel, D. et al., Volunteers for biomedical research. Recruitment and screening of
normal controls, Arch. Gen. Psychiatry, 48, 1022, 1991.
23. Buckley, P. et al., Schizophrenia research: the problem of controls, Biol. Psychiatry,
32, 215, 1992.
24. Suarez, B.K. and Hampe, C.L., Linkage and association, Am. J. Hum. Genet., 54,
554, 1994.
25. Kendler, K., The super-normal control group in psychiatric genetics: possible evi-
dence for coaggregation, Psychiatric Genetics, 1, 45, 1990.
26. Greenberg, D.A., Linkage analysis of necessary disease loci versus susceptibility
loci, Am. J. Hum. Genet., 52, 135, 1993.
27. Cox, N.J. and Bell, G.I., Disease associations. Chance, artifact, or susceptibility
genes?, Diabetes, 38, 947, 1989.
28. Crowe, R.R., Candidate genes in psychiatry: an epidemiological perspective, Am. J.
Med. Genet., 48, 74, 1993.
29. Kidd, K.K., Association of disease with genetic markers: deja vu all over again, Am.
J. Med. Genet., 48, 71, 1993.
30. Paterson, D., Case control association studies in complex traits-the end of an era?,
Mol. Psychiatry, 2, 277, 1997.
31. Hodge, S.E., What association analysis can and cannot tell us about the genetics of
complex disease, Am. J. Med. Genet., 54, 318, 1994.
32. Baron, M., Association studies in psychiatry: a season of discontent, Mol. Psychiatry,
2, 278, 1997.
33. Inoue, K. and Lupski, J.R., Genetics and genomics of behavioural and psychiatric
disorders, Cur. Op. Gen Dev., 13, 303, 2003.
34. Gorwood, P., The generalized anxiety disorder and major depressive disorder comor-
bidity: an example of genetic pleitropy? Eur. Psychiatry, 19, 27, 2004.
35. Anguelova, M., Benkelfat, C., and Turecki, G., A systematic review of association
studies investigating genes coding for serotonin receptors and the serotonin trans-
porter: II. Suicidal behavior, Mol. Psychiatry, 8, 646, 2003.
36. Lin, P.Y., and Tsai, G., Association between serotonin transporter gene promoter
polymorphism and suicide: results of a meta-analysis. Biol. Psychiatry, 15, 1023,
2004.
37. Munafo, M.R., Clark, T., and Flint, J., Does measurement instrument moderate the
association between the serotonin transporter gene and anxiety-related personality
traits? A meta-analysis. Mol. Psychiatry, 10, 415, 2005.
38. You, J.S., Hu.,S., Chen, B., and Zhang, H.,G., Serotonin transporter and tryptophan
hydroxylase gene polymorphisms in Chinese patients with generalized anxiety dis-
order. Psychiatr. Genet., 15, 7, 2005.
39. Hoefgen, B., Schulze, T.G., Ohlraun, S., von Widdern, O., Hofels, S., Gross, M. et al.
The power of sample size and homogeneous sampling: association between the 5-
HTTLPR serotonin transporter polymorphism and major depressive disorder. Biol.
Psychiatry, 57, 247, 2005.
40. Cho, H.J., Meira-Lima, I., Cordeiro, Q., Michelon, L., Sham, P., Vallada, H., and
Collier, D.A., Population-based and family-based studies on the serotonin transporter
gene polymorphisms and bipolar disorder: a systematic review and meta-analysis.
Mol. Psychiatry. 2005 Apr. 12 (online).
41. Malhotra, A.K., Murphy, G.M., Jr, and Kennedy, J.L., Pharmacogenetics of psycho-
tropic drug response. Am. J. Psychiatry, 161, 780, 2004.
42. Leboyer, M., Bellivier, F., Nosten-Bertrand, M., Jouvent, R., Pauls, D., and Mallet,
J., Psychiatric genetics: search for phenotypes. Trends Neurosci., 21, 102, 1998.
43. Hariri, A.R., Mattay, V.S., Tessitore, A., Kolachana, B., Fera, F., Goldman, D., Egan,
M.F., and Weinberger, D.R., Serotonin transporter genetic variation and the response
of the human amygdala. Science, 297, 400, 2002.
44. Jamain, S. et al., Mutations of the X-linked genes encoding neuroligins NLGN3 and
NLGN4 are associated with autism. Nat. Genet., 34, 27, 2003.
45. Laumonnier, F. et al., X-linked mental retardation and autism are associated with a
mutation in the NLGN4 gene, a member of the neuroligin family. Am. J. Hum. Genet.,
74, 552, 2004.
46. Yan, J. et al., Analysis of the neuroligin 3 and 4 genes in autism and other neurop-
sychiatric patients. Mol. Psychiatry, 10, 329, 2005.
47. Gauthier, J., Bonnel, A., St-Onge, J., Karemera, L., Laurent, S., Mottron, L., Fom-
bonne, E., Joober, R., and Rouleau, G.A., NLGN3/NLGN4 gene mutations are not
responsible for autism in the Quebec population. Am. J. Med. Genet. B Neuropsychi-
atr. Genet., 132, 74, 2005.
48. Garrod, A., The incidence of alkaptonuria: a study in chemical individuality. Lancet,
2, 1616, 1902.
49. Hunter, D., Gene-environment interactions in human diseases. Nature reviews genet-
ics, 6, 87, 2005.
50. Caspi, A., McClay, J., Moffitt, T.E., Mill, J., Martin, J., Craig, I.W., Taylor, A., and
Poulton, R., Role of genotype in the cycle of violence in maltreated children. Science,
297, 851, 2002.
51. Henquet, C., Krabbendam, L., Spauwen, J., Kaplan, C., Lieb, R., Wittchen, H.U., and
van Os, J., Prospective cohort study of cannabis use, predisposition for psychosis,
and psychotic symptoms in young people. BMJ, 330, 11, 2005.
12 Family and Twin Methods

Keith E. Whitfield and Tracy L. Nelson
CONTENTS
12.1 Introduction .................................................................................................. 183

12.2 Behavioral Genetics and Health .................................................................. 183
12.3 General Behavioral Genetic Methods.......................................................... 184
12.4 Genetics, Environment, and Health: Integration of Approaches ................ 186
References.............................................................................................................. 188
12.1 INTRODUCTION
Traditionally, behavioral scientists and epidemiologists have tended to attribute
health and behavioral variation to environmental sources. The recent completion of
mapping the human genome opens a new era of discovering genetic influences on
behavioral and neuroscience phenotypes. Behavioral genetics offers a theoretical
and statistical approach for assessing both the genetic and environmental contribu-
tions to individual variation in health. This paper provides an introduction to the
concepts and statistical techniques used in behavioral genetic research.
12.2 BEHAVIORAL GENETICS AND HEALTH

Behavioral genetics (also known as quantitative genetics) became a field of inquiry
in part in opposition to the developmental, or environmentalist, view that only
factors from the environment are involved in interindividual variability in the devel-
opment of children.1 Just as erroneous a view is that genes are the most important
factors involved in interindividual differences. Recent advances in molecular-
genetics techniques have greatly increased our ability to understand the contribution
of genes to health and illness. The National Human Genome Research Institute
(NHGRI) of the National Institutes of Health announced in June 2000 that it had
developed a working draft of the human genome. However, the inclination to believe
that genes account for all or none of the differences between people is genetic
determinism that does not accurately represent the assumptions of behavior genetic
methodologies.
Some suggest that disease processes arise from early environment.2 If the pre-
natal environment does impact diseases like type II diabetes or hypertension and
identical twins are highly concordant for these conditions, genetic explanations are
183
not the only viable interpretations. These conditions may be due to adverse envi-
ronments in the womb.
One of the potential dangers in emphasizing the presence of genetic influences
and ignoring environmental factors is that society and policy makers will invest less
in changing environments. One must keep in mind that behavior genetics methods
operate under assumptions of the collective contribution of genes and environment,
not nature vs. nurture. So the interventions to improve health will be most effective
when combinations of environmental manipulations and gene therapies (when appro-
priate) are developed. In this article, a brief introduction to behavior genetic meth-
odology is presented, followed by a brief review of some recent research on health
using this approach from the past 15 years.
12.3 GENERAL BEHAVIORAL GENETIC METHODS

Traditionally, behavioral scientists and epidemiologists have tended to attribute
behavioral variation to only environmental sources. During the past decade, however,
most researchers have come to accept the notion that genetic factors are, in part,
responsible for individual variation on many measures of behaviors. Using a behav-
ioral genetic approach, origins of individual differences are conceptualized to derive
from additive genetic and environmental sources of variance1 and thus, in theory,
account for the total behavioral or phenotypic variation.
Additive genetic variation is the sum of the effects from genes influencing a
trait. In some disease processes such as phenylketonuria or Huntingtons chorea,
single genes have been found to be responsible in the formation of the disease states.
These examples of one-gene one-disease (OGOD) scenarios are typically not the
case for most behavioral phenomena nor the premise of quantitative genetic analyses.
The underlying premise is that multiple genes, or pleiotropy,1 are involved in genetic
effects that are observed in a trait.
Although the name behavioral genetics (sometimes used in place of quantita-
tive genetics) implies an emphasis on heritable aspects of behavior, the identification
of environmental influences on a phenotype is equally important in this type of
analysis. Environmental effects are partitioned into those that are common (shared)
or unique (nonshared).1 Common/shared environmental variation is the phenotypic
variation due to the subjects living in the same family, thus sharing the same
environment.1
Unique/nonshared environmental variation is the component of phenotypic vari-
ance that can be attributed to the environmental factors not shared by family members
and thus the factor that makes members of the same family different from one
another. Plomin and Daniels3 illustrate how unique environmental variance tracks
how individuals from the same family differ from one another on numerous dimen-
sions of behavior.
Behavior genetic designs frequently take the form of twin studies. The classic
twin study involves comparisons between identical or monozygotic (MZ) and same
sex fraternal or dizygotic (DZ) twin pairs. There are several variations of this design
(for examples see Segal9) but the classic twin study tends to be the most frequently
used design. An assumption of behavioral genetics is that MZ twin pairs share 100%
Family and Twin Methods 185
of their segregating genes and DZ twin pairs share on average 50%. Using data from
twin pairs and operating under this assumption, the heritability of a trait can be
analyzed using several techniques. Heritability is considered the proportion of phe-
notypic variance that is due to genetic variation. A popular simple method for cal-
culating heritability is to calculate twice the difference between the intraclass corre-
lations of MZ and DZ twin pairs (for the calculation of intraclass correlations see
Plomin et al.1). Shared environmental variance can be calculated by doubling the
intraclass correlation for the DZ twin pairs and subtracting the MZ intraclass corre-
lation. By subtracting the sum of the heritability and shared environmental estimates
from 1.00, an estimate of nonshared environmental variance is obtained.
While twin designs are a useful and interesting method for evaluating genetic
and environmental contributions, family studies are a popular alternative. Family
studies use the same basic analysis techniques as those in twin studies, but the
expected covariances for pairs of relatives are estimated. The following table high-
lights some of the common associations used in family studies.
Coefficient for Shared

Additive Genetic Dominant Genetic Environmental
Types of Relatives Variance Variance Variance
Spousespouse 0 0 1
Parentoffspring (living
0.5 0 1
together)
Full sibs (living together) 0.5 0.25 1
Full sibs (living apart) 0.5 0.25 0

Half sibs (living
0.25 0 1
together)
Aunt/uncleniece/nephew 0.25 0 0
First cousins (living 0.125
0 0
apart)
Source: From Khoury et al.4
As in other social sciences fields, structural equation modeling (SEM) or bio-

metrical modeling, the fitting of observed data to models of genetic and environ-
mental effects5 is the preferred analytical method of behavioral geneticists. These
SEM techniques provide parameter estimates that represent the relative contribution
of genetic and environmental influences. Variances and covariances are used in the
calculation of the components of variance (genetic, shared environmental, and unique
environmental). The benefit of these techniques over the intraclass correlation
method described earlier is that multivariate extensions can be performed to simul-
taneously assess shared and unique genetic and environmental relationships among
two or more variables.
An example of a path diagram of a classic univariate twin model is shown in
Figure 12.1 (for examples of other models).6 Path diagrams are useful in that they
provide a visual display of correlational and causal relationships between variables.7
By employing path analysis, specific hypotheses about relationships between the
1.00
1.00 MZ
0.50 DZ
E1 C1 A1 A2 C2 E2
e c h h c e
P1 P2
FIGURE 12.1 General structural equation model.
variables are quantified by parameter estimates or path coefficients.6 In the model

shown in Figure 12.1, the overall phenotypic variance is explained using three
components: A-additive genetic variation, C-common or shared family environmen-
tal variation, and E-unique environmental variation. In Figure 12.1, A, C, and E are
latent variables, and h, c, and e are parameter estimates. In this figure, P1 represents
the observed score for twin 1 and P2 represents the score for twin 2. It should be
noted that the example provided here is the general example and that P could
represent any phenotype, trait, or variable.
In model fitting, observed data are compared to expected values and the result
is a familiar 2 statistic. Using these techniques, parameters are dropped from the
model to see if a more parsimonious model (one with fewer parameters or latent
variables, yet still fitting the data) is available to explain the data. Statistical
significance of these parameters is assessed from maximum likelihood-ratio 2
comparisons of the models after the parameters have been dropped. Significance
of parameters is evaluated by taking the difference between the 2s of the full and
reduced model and using that difference as a 2. The degrees of freedom are
calculated by taking the difference between the degrees of freedom for the full and
reduced. If there is a significant difference between two models, the parameter that
was dropped is significant.6
Using this basic model then, if MZ twins are more alike than DZ twins, pheno-
typic variance can be attributed, in part, to genetic sources. If the genetic variation
estimated by the parameter h represents a significant source of variation, this param-
eter cannot be dropped from the model without causing a significant change in P2
between the models. Squaring that parameter, h, provides an estimate of heritability.
The contributions of shared and unique environmental factors to the phenotypic
variation are also tested in this manner. For ease of comprehension, parameter
estimates are typically standardized (procedure outlined in the LISREL manual).8
12.4 GENETICS, ENVIRONMENT, AND HEALTH:

INTEGRATION OF APPROACHES
The National Human Genome Research Institute (NHGRI) of the National Institutes
of Health announced in June of 2000 that they had developed a working draft of
the human genome. This historic event places science on the doorstep of limitless
Family and Twin Methods 187
possibilities including new insights about diseases and how to treat and prevent
them. The trepidation about genetic approaches in the study of health by some
scientists may arise from the inherent power in identifying biological mechanisms
of disease and illness. Knowing the sequence of the genome, however, is only the
beginning. Equally important will be our knowledge of how the environment influ-
ences health, disease, and complex behaviors associated with health. This integrative
geneticenvironmental perspective places behavioral genetic approaches at the fore-
front of useful research designs and theory that will advance the search for answers
about health. Twin studies and other behavioral genetic designs have not been used
to their fullest potential to this end.
One of the areas of most interest in the search for answers about health and
disease is why disparities exist across economic and ethnic groups. It is a correct
assumption that basing investigations of health differentials across ethnic groups
solely on the basis of genetic differences will not yield accurate identification of the
mechanisms responsible for health disparities. Preconceived notions about geneti-
cally based racial inferiority have hindered advances in understanding and reducing
health disparities. Attempting to explain the differential health burden ethnic minor-
ities experience by genetic differences goes against probability given there are
considerably small genetic differences across racial groups and more variability
within each group. The role of genetic influences, however, cannot be completely
dismissed. The manner in which genes have the potential of playing a role in creating
health differentials requires further explanation. It is not genes defining individuals
from different ethnic groups that is key to the elucidation of health differentials per
se. Instead, describing health differentials as arising from insults to a complex system
represented by the interaction between genes and environments, which creates excess
burden of chronic illness and disease within some groups, is a more accurate
perspective.
In contrast to simply focusing on genetic explanations, there is ample information
that differences in environmental factors between ethnic groups account for dispar-
ities in health status. Previous research on the significant impact combinations of
socio-demographic and psychosocial factors in disease processes and complex
behaviors is perhaps our best indicator that science must avoid a reductionistic view.
Genetic reductionism assumes knowing and manipulating the genome will cure all
our ills. Rather, we must understand how genetic and environmental influences work
in concert to account for health conditions and the psychosocial variables that impact
health. Much of previous research has focused on the behaviors and social structures
that produce differences in health and disease across ethnic groups. One of the future
and formidable challenges to using the information ascertained from adding genetic
information to examinations of health differentials is to understand the underlying
effect genes have on health and aging within complex environments or contexts. We
may find that the polymorphisms that occur in genotypes are deleterious or protective
factors related to disease and health that are created, modified, or triggered by cultural
and contextual factors.
Complementary, interdisciplinary approaches are desperately needed to harness
the important findings that can from the Human Genome Project and continued
epidemiological research in the exploration of the underlying causes of health and
illness and the related psychosocial behaviors. Continued use of behavior genetic
designs (and modified designs such as those mentioned here) will significantly
advance our knowledge. Both conceptual and statistical advances in these methods
are still required. These methods have the potential to provide the backdrop for
exciting new revelations about how genes and environment work in concert to create
health and illness.
ACKNOWLEDGMENTS
Keith E. Whitfield is supported by the National Institute on Aging (5-RO1-AG13662-
04). Tracy L. Nelson is supported by USPHS Grant K01-DK-64647-01.
REFERENCES
1. Plomin R., DeFries, J. C. and McClearn, G. E. Behavioral Genetics: A Primer. W.
A. Freeman and Co. New York. 1990, P53 ff.
2. Barker, D. J. A new model for the origins of chronic disease. Medical Health Care
Philosophy, 4 (1), 3135, 2001.
3. Plomin, R. and Daniels, D. Why are children in the same family so different from
one another? Behavioral and Brain Sciences, 10 (1), 116, 1987.
4. Khoury, M. J., Beaty, T. H., and Cohen, B. H. Fundamentals of Genetic Epidemiology:
chapter 7, Oxford University Press, Chap 7, 1993.
5. Jinks, J. L. and Fulker, D. W. Comparison of the biomertical genetical, MAVA, and
classical approaches to the analysis of human behavior. Psychological Bulletin, 73,
311349, 1970.
6. Neale, M.C. and Cardon, L. R. (Eds.). Methodology for Genetic Studies of Twins and
Families. Dordrecht, the Netherlands: Kluwer Academic Press, 1992.
7. Loehlin, J. C. Latent Variable Models. Baltimore: Lawrence Erlbaum, 1987.
8. Jreskog, K. G. and Srbom, D. LISREL 7: A Guide to the Program and Applications
(2nd ed.). Chicago: SPSS, Inc., 1989.
9. Segal, N.L. The importance of twin studies for individual differences research. J.
Counsel. Dvlp., 68(6), 612622, 1990.
13 GeneEnvironment
Interactions
Byron C. Jones and Leslie C. Jones
CONTENTS
13.1 Introduction: What Are GeneEnvironment Interactions

and Why Are They Important?.................................................................... 189
13.2 How Do Genes and Environments Cooperate?........................................... 190
13.2.1 GeneEnvironment Correlation ....................................................... 190
13.2.1.1 Passive GenotypeEnvironment Correlation.................... 190
13.2.1.2 Reactive GenotypeEnvironment Correlation.................. 190
13.2.1.3 Active GenotypeEnvironment Correlation ..................... 190
13.2.2 GeneEnvironment Interaction ........................................................ 190
13.2.2.1 Definition Vp = Vg + Ve + (Vgxe) + V .......................... 191
13.2.2.2 Evidence for GeneEnvironment ..................................... 191
13.2.2.2.1 GenotypeEnvironment
Interaction in Humans .................................... 191
13.2.2.2.2 GeneEnvironment
Interaction in Animals.................................... 195
13.3 General Considerations ................................................................................ 197
References.............................................................................................................. 198
13.1 INTRODUCTION: WHAT ARE

GENEENVIRONMENT INTERACTIONS
AND WHY ARE THEY IMPORTANT?
Some of the most important work on neurobehavioral genetics is in the study of
psychiatric disorders in twins. As you will recall, monozygotic twins share identical
genotypes and dizygotic twins share, on average, 50% of genes (alleles) in common.
While we know that there is a genetic component to schizophrenia, most estimated
concordance rates among monozygotic twins range from 40 to 50%, and for dizy-
gotic twins, most estimates are below 20%.1 This tells us that genes are involved in
schizophrenia, but the environment also plays a major role. So, those carrying one
or more allelic configurations are susceptible to environmental events that produce
the disease. Just what the genetic configurations and environmental events are
constitutes a subject of intensive study. Environment can be broadly construed and
189
can include internal as well as external events. Thus, exposure to drugs, hormones,
toxins, or teratogens in utero, nutritional status, viral infections, early rearing con-
ditions, and social situations in adulthood may all be considered as environment.
13.2 HOW DO GENES AND ENVIRONMENTS

COOPERATE?
As discussed in earlier chapters, genes do not work in isolation. They are influ-
enced by the environment and in turn may help select individuals into specific
environments.
13.2.1 GENEENVIRONMENT CORRELATION

Geneenvironment interaction sometimes is confused with geneenvironment
correlation. Geneenvironment correlation may be considered as the degree to which
genetically influenced propensities in the individual cause that individual to be
associated with a characteristic environment. Plomin, Loehlin and DeFries2 describe
three types of geneotypeenvironment correlations as described below.
13.2.1.1 Passive GenotypeEnvironment Correlation
Under this condition, individuals inherit the same relevant alleles and are exposed
to the same environment as other family members. For example, some forms of
antisocial behaviors show significant heritability. It is also nearly certain that those
at genetic risk are exposed to family members who exhibit many of the same
behavioral characteristics. The importance of this correlation has obvious implica-
tions for developmental and intervention concerns.2
13.2.1.2 Reactive GenotypeEnvironment Correlation
In this situation, genetically influenced behavioral tendencies in the individual elicit

characteristic responses from others. Thus children who are impulsive and show an
inability to suppress inappropriate behavior are more likely to receive harsh parenting
than children with little tendency to behave inappropriately.
13.2.1.3 Active GenotypeEnvironment Correlation
Children who are verbally and socially gifted tend to seek out social situations that
match their tendencies. This is termed niche-picking. Scarr and McCartney3 have
exploited genotypeenvironmental correlation to help explain some of the phenom-
ena in child development, parental styles, treatment of behavior disorders, and
childfamily relations.
13.2.2 GENEENVIRONMENT INTERACTION

Geneenvironment interaction is a quite different matter, although there may be
co-extant geneenvironment correlations. In this situation, individuals with specific
GeneEnvironment Interactions 191
genotypes respond differently to the environment than do others of differing

genotypes. There are obvious implications for such interactions. For example, given
an individuals genotype, we could assess susceptibility to adverse environmental
conditions.
13.2.2.1 Definition Vp = Vg + Ve + (Vgxe) + V
Phenotypic variability (or more properly, variance) can be partitioned into a number
of components. Thus Vg is the genetic part and consists of additive and dominant
genetic effects. Ve is the environmental source and V is measurement error. Vgxe
is the component left over when Vg, Ve, and V are accounted for (in the long run,
we assume that V sums to zero).
Some graphical representations are shown below.
13.2.2.2 Evidence for GeneEnvironment
In Cloningers 1987 seminal paper on alcoholism typologies,4 two kinds of alcohol-

related problems were described. In the type II alcoholism, the afflicted individuals
are mostly male, there is an early onset of drinking, and there is a strong association
with sensation-seeking and conduct disorder. The estimated heritability is quite high
and there seems to be little influence from the environment. Type I alcoholism is
also genetically influenced, but its expression is dependent upon the environment
difficult relationships, job-related troubles, etc. So, the individual carries a comple-
ment of alleles that interact with environmental situations that shape the phenotype.
13.2.2.2.1 GenotypeEnvironment Interaction in Humans

The malfunctioning of the raphe serotonin (5HT) system has long been implicated
in depression, and polymorphisms in genes involved in the regulation of that system
have been considered candidate risk factors. In 1996, a report by Lesch and
colleagues5 described an association between anxiety-related traits and a polymor-
phism at the promoter region of the gene (SLC6A4) for the serotonin transporter
(5HTT). This polymorphism (5HTTLPR) is a 44 base-pair insertion or deletion that
results in a long (l) or short (s) allele. The l allele exhibits higher transcriptional
G1, G2
Response
E1 E2
FIGURE 13.1 In this representation, we can see that the environment has an effect on the
phenotype, but both genotypes respond equally to the environmental change.
G1
Response
G2
E1 E2
FIGURE 13.2 In this figure, we observe that there is no effect of the environment, but an
effect of genotype.
G1
Response
G2
E1 E2
FIGURE 13.3 This figure shows, not only a significant effect of genotype, a possible effect
of environment, not only but also a significant genotypeenvironment interaction.
G1
Response
G2
E1 E2
FIGURE 13.4 In this figure, we see a significant interaction between genotype and environ-
ment but no main effect of either genotype or environment.
activity on the order of 1.4- to 1.7-fold and cells transfected with the l allele evince
increased serotonin uptake compared with cells transfected with the s allele. Lesch
then performed an association study between allelic configuration and affective
disorders. Despite the rather dramatic differences in 5HT biology, the allelic differ-
ences accounted for rather modest variance in anxiety-related traits.
In other studies concerning the relationship between 5HTTLPR polymorphism

and affective disorders, either no association was found or, like the Lesch study, the
strength of association was modest at best.6,7
One reason for conflicting results among studies, suggested by Hoefgen et al.,7
could be insufficient sample sizes and low power. These researchers reported an
association between the 5HTTLPR and depression, in which the s allele mildly
increases risk, and they attributed this to the size of their sample (466 patients, 836
controls) and careful adjustment for various stratification factors. Still, the effect
that they found was modest. The lack of effect found in many studies and the modest
effect found by Hoefgen et al. could also be explained by how these studies isolated
the allele as the putative sole factor. In fact, the allele may exert its effects by
interacting with environmental factors, and possibly other genetic polymorphisms.
In 2003 Caspi and colleagues8 studied a cohort of 847 New Zealand native
Caucasians who had been studied every 2 years from age 3 to 21. When these
individuals were 26 years of age, the subjects were classified based on their
5HTTLPR genotype, l/l, l/s and s/s. They were able to then classify individuals
based on history of depression and history of stressful life events (SLEs). Thirty
percent of the subjects reported no SLEs in the past 5 years. Twenty-five percent of
subjects had experienced one SLE, while 20, 11, and 15% had experienced 2, 3, or
more than 4 SLEs, respectfully. Distribution of SLEs showed that the 5HTTLPR
gene had no apparent effect on stress exposure, as there were no significant differ-
ences in exposure among allele groups. Next, subjects were categorized by symptoms
of major depression. Seventeen percent of the subjects met the DSM-IV criteria for
having experienced a major depressive episode within the past year. Among subjects
who reported SLEs, those with at least one s allele were significantly more likely
than subjects homozygous for the l allele to report depression. In fact, subjects who
had reported four or more SLEs and carried two s alleles had a 43% likelihood of
evincing depression.
Since the results of this study were reported, at least two groups of researchers
have replicated their findings.9,10 One recent study supported the findings of Caspi
et al. by showing that individuals who had been subjected to childhood mistreatment
were more likely to develop a depressive episode if they contained an s allele of the
5HTTLPR gene. Kaufman and colleagues11 studied children who had been removed
from their homes, for various reasons, by the State of Connecticut Department of
Children and Families (DCF) within the past 6 months. They also included a sample
of community controls (CCs). More than 80% of the children in the DCF group had
experienced more than two or more forms of maltreatment. Most DCF children
(81%) had experienced neglect, but other forms of maltreatment included physical
abuse (63%), sexual abuse (19%), emotional maltreatment (79%), and exposure to
domestic violence (60%). Their results supported a geneenvironment interaction,
because the s allele of the 5HTTLPR gene increased the likelihood of depression
only in the children who had experienced childhood maltreatment. Analysis of the
effect of social supports showed that their presence moderates in a compensatory
manner the interaction between the 5HTTLPR genotype and childhood mistreatment.
Children were grouped into high-risk, moderate-risk, and low-risk categories (s/s
without social supports, s/l or low social supports, and l/l plus high social supports,
respectively), and it was determined that the depression scores of high-risk children
doubled those of children who had the s/s genotype but positive social supports.10
The authors concluded that social supports further moderated the interaction between
the 5HTTLPR gene and childhood maltreatment, reducing the (increased) risk for
depression among children with the s allele.10 This study both supports the
geneenvironment interaction model and emphasizes further the relative importance
of environmental factors.
In summary of the findings of the four studies in agreement, it can be concluded
that the s allele of the 5HTTLPR significantly increases susceptibility to depression
only in the presence of SLEs, and this increased vulnerability can be at least partially
mitigated for by other environmental factors, including social supports.
Hariri and colleagues12 used an anxiety- and fear-provoking task, involving
photos of fearful and angry facial expressions, in conjunction with fMRI, to provide
evidence of (right) amygdala hyper-reactivity among subjects carrying the s allele
of 5HTTLPR. Indeed, the association between 5HTTLPR and amygdala hyper-
reactivity fit nicely with the current neurobiological hypothesis of depression.13
Of particular interest, however, was that the subjects tested in this study were healthy
volunteers with no history of psychiatric illness, including depression. It seems
likely, therefore, that the amygdala hyper-reactivity among subjects carrying the s
allele indicates that the 5HTTLPR may in fact mediate the intensity of an individuals
reaction to negative stimuli, and, hence, increased susceptibility to SLE-induced
depression.
Pezawas and colleagues14 used voxel-based morphometry (VBM) to test the
hypothesis that allelic configuration of the 5HTTLPR involves neural circuits that
function in emotional regulation. VBM is a highly sensitive method that measures
differences in gray matter volume between subjects. Using VBM, Pezawas and
colleagues detected reduced gray matter in carriers of the s allele in the right
amygdala and the perigenual anterior cingulate cortex (pACC). The volume of pACC
and amygdala gray matter were shown to be positively correlated and functionally
connected, as shown by fMRI. Two subregions of the pACC are functionally con-
nected to the amygdala, one showing positive and the other showing negative cor-
related activity. The authors proposed that the three brain regions may form a
feedback loop, modulating amygdala reactivity. In carriers of the s allele, the cou-
pling of the amygdala to these regions was impaired, especially the connection
between the amygdala and the positively associated subregion of the pACC.
Thus, a body of research findings is accumulating that suggests that the s allele of
the 5HTTLPR causes neurobiological differences evident even in normal subjects
with no history of psychiatric illness. These neurobiological differences do predict
personality traits related to depression in healthy s-carriers, but do not necessarily
cause depression. When life stressors are not reported, s-carriers do not have an
increased risk of developing depression; however, when a sufficient amount of life
stress is experienced, s-carriers exhibit an increased risk toward major episodes
of depression.
Because the subjects in both the fMRI and VBM studies were healthy individ-
uals, the authors also concluded that the phenotypic effects were developmental.11
This conclusion is in agreement with recent evidence that the impact of reduced
transcription of the 5HTT early in development causes irreversible alterations in the

regulation of serotonin in relevant brain regions.15
13.2.2.2.2 GeneEnvironment Interaction in Animals

For more than half a century, we have known that environmental perturbations during
infancy can have long lasting effects on neurobehavioral development in rodents
and primates. Thus, brief interruption of maternal contact in rats16 has dramatic
effects on adult stress response, and social isolation in infancy causes severe social
behavioral problems in primates.17 In the case of rodents, one of the observations
has been that the brief maternal separation or brief disruptions in preweanling rats
tends to alter the hormonal response to stress in adulthood.18 Moreover, early envi-
ronmental enrichment has been shown to increase cerebral cortical thickness and
increase the functioning of synaptic processes.19 The story that grew out of such
studies was that infantile stimulation had nearly universally beneficial effects, some-
how tempering the rodent to better face lifes challenges in adulthood. In 1970,
Henderson20 published an important paper in which he showed that rearing practices
in the laboratory interact with genetically influenced behavioral differences in mice
so as to reduce those behavioral differences. The latter findings thus raise the question
as to whether early experience might have different effects, depending on genetic
background. For providing an answer to this question, mice are better study subjects
than rats because there are more genetically defined strains and lines than found in
rats (although the gap is narrowing rapidly). The notion that early handling (i.e.,
brief separation from the nest environment) has universally beneficial effects is called
into question by a study by Raymond et al.21 who showed that early handling caused
compromised immune response in adult C57BL/10J but not BALB/cJ mice.
A more comprehensive study designed to address the effect of early handling
in genetically defined mice was reported recently by Garipy and colleagues.22 The
subjects for this study were male mice from two lines selected by Cairns23 for
differential isolation-induced aggression, based on latency to attack and frequency
of attacking an intruder. The low aggressive line is NC100 and the high aggression
line is NC900. Two experimental treatments were administered. The first treatment
was infantile handling. On every other day, the pups were removed from the nest
and the entire litter was placed into a plastic beaker for 60 seconds and subsequently
gently returned to the nest. At 21 days of age, the animals were weaned into isolate
housing or group housing with four animals of same handling condition per group.
When the animals were 51 to 60 days of age, they were subjected to a mild stressor,
exposure to an open field for 10 minutes. At 20 minutes after this exposure, the
animals were sacrificed and blood and brains harvested for measurement of corti-
costeronethe primary glucocorticoid stress hormone in rodentsand density of
dopamine D1 receptors in mesolimbic and nigro-striatal areas.
Subsequently, a separate group of animals was subjected to infantile handling
vs. nonhandled controls to examine the effect of this treatment on aggression in
adulthood. Because group-housed animals rarely show aggression, all of the animals
were housed in isolation from weaning until testing.
200 Baseline Isolated Group

Corticosterone (ng/ml)
160
120
80
40
0
NC100 NC900 NC100 NC900 NC100 NC900
Non-Handled
Handled
FIGURE 13.5 Handling reduced serum corticosterone response to stress (right two panels) under
isolate and group housing conditions in the low aggressive animals (NC100). Moreover, group
housing in this line also reduced the response and both treatments had additive effects on the
corticosterone response. The high aggressive line presents a different story, as there was no effect
of handling under either housing condition. (Reprinted from Garipy, J. -L. et al., Pharmacology
Biochemistry and Behavior, 73(1), 717. Copyright 2002, with permission from Elsevier.)
Isolate Housing Group Housing
16 16 Non-Handled
Non-Handled
Handled
14 Handled 14
fmol/mg protein-1
12 12
10 10
8 8
6 6
4 4
2 2
0 0
NC100 NC900 NC100 NC900
FIGURE 13.6 The effect of infantile handling on the density of dopamine D1 receptors. Note
here that handling dramatically increased the density of dopamine receptors in the nucleus
accumbens. There was no effect of differential housing on this measure and there was no
effect of either handling or housing on D1 receptor densities in the caudate-putamen.
(Reprinted from Garipy, J. -L. et al., Pharmacology Biochemistry and Behavior, 73(1), 717.
Copyright 2002, with permission from Elsevier.)
The results from this study showed the hypothalamo-pituitary-adrenal axis to

be a good candidate for showing genotypeenvironment effects (Figure 13.5); how-
ever the mesolimbic, dopamine-D1-related functions are sensitive to handling,
regardless of the genetic architecture (Figure 13.6). Figure 13.7 illustrates that
600 Non-Handled 600 Non-Handled

Handled Handled
500 500
Latency to Freeze (sec)

Latency to Attack (sec)
400 400
300 300
200 200
100 100
0 0
NC100 NC900 NC100 NC900
25 25
Non-Handled Non-Handled
Handled
Handled
20 20
15 15
Freeze
Attack
10 10
5 5
0 0
NC100 NC900 NC100 NC900
FIGURE 13.7 Handling had no effect on latency to attack in the low aggressive line but
significantly shortened this measure in the high aggressive mice. The same was true for a
number of attacks, although in the low aggressive line, there was a nonsignificant trend toward
an increase in attack frequency in the low aggressive mice. Latency to freeze was also affected
by handling, more so in the low line than in the high line, and the number of times that the
animal was seen motionless (freezing) was reduced in the low line and abolished in the high
line, although there was a clear floor effect in these animals.
aggressive and fear-like behaviors are also sensitive to geneenvironment interac-

tions. Thus, it is clear that not all neurobehavioral systems are equally susceptible
to geneenvironment. Is it the case then that systems that are stress-related are more
susceptible to geneenvironment than other systems? Or, is it the genetic makeup
of the individual that determines which system will be more responsive?
13.3 GENERAL CONSIDERATIONS

It is clear that geneenvironment-based influences on brain and behavior are common
and play a major role in development and function. Learning about the short allele
of the 5HTTLPR is important from the standpoint that major depression is the most
common psychiatric disorder. Now that we know at least part of the genetic basis
for increased risk, there is the possibility that this knowledge can help equip us for
more effective means of prevention, prophylaxis, and treatment. In a recent, provoc-
ative article, Tienari and colleagues24 related childhood rearing environments to risk
for developing schizophrenia-spectrum disorder (SPD) in young adulthood. The
subjects consisted of adoptees whose mothers had been diagnosed with (SPD) and
a group of adoptees whose mothers had not been so diagnosed. The rearing envi-
ronment was evaluated in terms of factors indicating adversity. When the adoptees
reached young adulthood, all were assessed for SPD. Among those individuals whose
biological mothers were free of SPD, the environment, adverse or benign, had no
effect on the frequency diagnosed with SPD. In the group whose biological mothers
had been diagnosed with SPD, those who were reared in an adverse environment
had a significant increase in SPD diagnoses. The value of interventiongiven
biologically derived risk in this exampleis clear.
The research in animals gives us the prospect to understand what systems seem
to be more affected by geneenvironment and what kinds of environments are
important. The kinds of genetic material that are involved are also important. For
example, does geneenvironment involve specific structural genes or promoters, for
example 5HTTPLR? Are housekeeping genes involved? Do genes involved in
geneenvironment involve specific networks?
REFERENCES
1. Cardno, A. G. and Gottesman, I. I. Twin studies of schizophrenia: from bow-and-
arrow concordances to Star Wars Mx and functional genomics. Am J Med Genet 97:
1217, 2000.
2. Plomin, R., Loehlin, J. D. and DeFries, J. C. Genetic and environmental components
of environmental influences. Developmental Psychology 21: 391402, 1985.
3. Scarr, S. and McCartney, K. How people make their own environments: a theory of
genotypeenvironment effects. Child Development 54: 424435, 1983.
4. Cloninger, C. R. Neurogenetic adaptive mechanisms in alcoholism. Science 236:
410416, 1987.
5. Lesch, K. P. Geneenvironment interaction and the genetics of depression. J. Psychi-
atry Neurosci 29: 174184, 2004.
6. Anguelova, M., Benkelfat, C., Turecki, G. A systematic review of association studies
investigating genes coding for serotonin receptors and the serotonin transporter: I.
Affective disorders. Mol Psychiat 8: 574591, 2003.
7. Hoefgen, B., Schulze, T. G., Ohlraun, S., Widdern, O. V., Hofels, S., Gross, M.,
Heidmann, V., et al. The power of sample size and homogeneous sampling: associ-
ation between the 5-HTTLPR serotonin transporter polymorphism and major depres-
sive disorder. Biol Psychiat 57: 247251, 2005.
8. Caspi, A., Sugden, K., Moffitt, T. E., Taylor, A., Craig, I. W., Harrington, H. L.,
McClay, J. M., et al. Influence of Life Stress on Depression: Moderation by a
polymorphism in the 5-HTT gene. Science 301: 386389, 2003.
9. Kendler, K. S., Kuhn, J. W., Vittum, J., Prescott, C. A., and Riley, B. The interaction
of stressful life events and a serotonin transporter polymorphism in the prediction of
episodes of major depression. Arch Gen Psych 62: 529535, 2005.
10. Gillespie, N. A., Whitfield, J. B., Williams, B., Heath, A. C., and Martin, N. G. The
relationship between stressful life events, the serotonin transporter (5-HTTLPR) gen-
otype, and major depression. Psychol Med 35: 101111, 2005.
11. Kaufman, J., Yang, B-Z., Douglas-Palumberi, H., Houshyar, S., Lipschitz, D., Krystal,
J. H., and Gelernter, J. Social supports and serotonin transporter gene moderate
depression in maltreated children. PNAS 101(49):1731617321, 2004.
12. Hariri, A. R., Drabant, E. M., Munoz, K. E, Kolachana, B. S., Mattay, B. S., Egan,
M. F., and Weinberger, D. R. A susceptibility gene for affective disorders and the
response of the human amygdala. Arch Gen Psych 62: 146152, 2005.
13. Hamann, S. Blue Genes: wiring the brain for depression. Nature neurosc 8: 701703,
2005.
14. Pezawas, L., Meyer-Lindenberg, A., Drabant, E. M., Verchinski, B. A., Munoz,
K. E., Kolachana, B. S., Egan, M.F., Mattay, V.S., Hariri, A.R., and Weinberger, D.R.
5-HTTLPR polymorphism impacts human cingulate-amygdala interaction: a genetic
susceptibility mechanism for depression. Nature Neurosc 8(6): 828834, 2005.
15. Ansorge, M. S., Zhou, M., Lira, A., Hen, R., Gingrich, J. A. Early-life blockade of
the 5-HT transporter alters emotional behavior in adult mice. Science 306(5697):
879881, 2004.
16. Treiman, D. M. and Levine, S. Plasma corticosteroid response to stress in four species
of wild mice. Endocrinology 84: 676680, 1969.
17. Harlow, H. F., and Soumi, S. Social recovery by isolation-reared monkeys. Proc Natl
Acad Sci USA 68: 15341538, 1971.
18. Levine, S. Developmental determinants of sensitivity and resistance to stress. Psych-
oendocrinology 30: 939946, 2005.
19. Rosenzweig, M. R. and Bennett, E. L. Psychobiology of plasticity: effects of training
and experience on brain and behavior. Behavioural Brain Research 78: 5765, 1996.
20. Henderson, N. D. Genetic influences on the behavior of mice can be obscured by
laboratory rearing. J Comp Physiol Psychol 72: 505511, 1970.
21. Raymond, L. N., Tokuda, S., Reyes, E. and Jones, B.C. Differential immune response
in two handled inbred strains of mice. Physiology and Behavior, 37: 295297, 1986.
22. Garipy, J-L., Rodriguiz, R. M. and Jones, B. C. Handling and genetic effects on the
stress system, social behavior and dopamine function. Pharmacol Biochem Behav,
73(1): 717, 2002.
23. Cairns, R. B. Aggression from a developmental perspective: genes, environments and
interactions. Ciba Found Symp 194: 4556, 1996.
24. Tienari, P., Wynne, L. C., Sorri, A., Lahti, I., Laksy, K., Moring, J., Naarala, M.,
Nieminen, P. and Wahlberg, K.-E. Genotypeenvironment interaction in schizophre-
nia-spectrum disorder. Br J Psychiat 184: 216222, 2004.
14 And Now It Starts to Get

Interesting: GeneGene
Interactions
Yamima Osher
CONTENTS
14.1 Introduction .................................................................................................. 201

14.2 Epistasis........................................................................................................ 202
14.2.1 Two Genes, Interacting .................................................................... 202
14.2.2 Three Can Play, Too ........................................................................ 203
14.3 Not Only in Humans.................................................................................... 203
14.4 GeneGene Interactions in Common Human Diseases .............................. 204
14.5 Statistical Methods....................................................................................... 204
14.6 Summary ...................................................................................................... 205
References.............................................................................................................. 205
14.1 INTRODUCTION
Genetics was discovered in the nineteenth century by an Austrian monk named
Mendel, who spent years observing the reproduction of pea plants (in those
days there was no HBO). Mendel noticed that baby pea plants often inherited
certain characteristics of the mommy and daddy pea plants, such as height, eye
color, and personality. Mendel found that, by mating a certain pea plant with a
certain other pea plant, he could cause a third pea plant to go into a violent
jealous rage. What can we learn from these experiments? I have no idea.
Dave Barry
In the beginning, there was the OGOD model: one-gene one-disease.1 Even before
the age of molecular genetics, the study of patterns of inheritance made it clear that
this model was not applicable to the study of personality traits. When the first
replicated reports were published of positive associations between a candidate gene
and a personality traitdopamine D4 receptor gene (D4DR) and novelty seeking
201
(NS)2,3the authors speculated that that there might be 10 or so different genes,

each of small effect size, responsible for the trait.3 The assumption was that the
effects of these genes would be additive, each one responsible for about 10% of the
observed genetic variance. However, as Richard Ebstein notes in Chapter 18 of this
book, finding these genes has proven frustratingly difficult and results are often
inconsistent. More recently, evidence has begun to accumulate, which suggests that
part of the answer to this conundrum may be found in genegene interactions.
14.2 EPISTASIS
Plomin1 provides an excellent explanation of genegene interactions, commonly
referred to as epistasis. He explains that while dominance is an intralocus interaction
with a given allele interacting in a nonadditive way with the allele at the same locus
on the homologous chromosomeepistasis refers to the interlocus interaction of
alleles at other loci. In Plomins words, [C]onsider two loci (A and B) that affect
a phenotypic character. Both the additive genetic values and the dominance devia-
tions are summed across the two loci. However, a particular combination of a certain
allele at locus A and another at locus B may influence the phenotype in ways not
explainable by the additive and dominance effects. Epistasis refers to this sort of
effect (p. 292). It should also be pointed out that epistatic effects are not limited
to the interaction of only two genes, but may include any number of players.
14.2.1 TWO GENES, INTERACTING

Let us begin with a relatively simple example of an interaction between two candidate
genes involved in neurotransmitter systems, and its effect on a personality (temper-
ament) trait. Harm avoidance4 is an anxiety-related trait which has sometimes5,6 but
not always (for review see Lesch7) been found to be associated with the short allele
of the 5-HTTLPR (serotonin transporter) gene. Might the impact of this allele be
mitigated, or influenced, by another allele which could be related to curiosity,
adventurousness, and the desire for stimulationspecifically, the long allele of the
D4DR? One could speculate that individuals prone to anxiety might express this
trait more if, at the same time, they felt the urge or the need to expose themselves
to exciting (but also anxiety-provoking) experiences and stimuli. Interestingly, this
seems to be the pattern that emerges when the interaction of the 5-HTTLPR and the
D4DR genotypes is examined: whether the dependent variable is anxious behavior
in infants8 or harm avoidance in adults as measured by Cloningers temperament
and character inventory (TCI)9, the group with BOTH the s/s 5-HTTLPR genotype
AND the long allele of the D4DR exon III polymorphism show the highest scores.
That this is a true interaction and not an additive effect can be clearly seen in the
Szekely9 data, as the D4DR genotype has no influence on harm avoidance scores
in the absence of the s/s 5-HTTLPR genotype.
Findings regarding genegene interactions in personality genetics have been
noticeably robust (in contrast to the inconsistent findings regarding most single
candidate genes). An illustrative example, from the early years of molecular person-
ality genetics: Ebstein et al.10 looked at the interaction between a serotonin-receptor
And Now It Starts to Get Interesting 203
polymorphism (5-HT2c) and the D4DR polymorphism. The 5-HT2c polymorphism

involves a cystein to serine substitution. The relatively rare 5-HT2Cser allele was
associated with significantly reduced reward dependence (RD) and persistence
scores (persistence was originally one of the four subscales of Cloningers reward
dependence, but was later extracted and considered a fourth temperament dimen-
sion). This effect was dramatically increased in those subjects who also had the long
allele of the D4DR exon III polymorphism, such that having the 5-HT2Cser allele and
the long D4DR allele together reduced mean RD scores by more than two standard
deviations. Originally, however, this finding was regarded with some skepticism, as
the group of subjects with the rarer forms of both alleles was quite small (n = 6).
Within 2 years, however, an independent group from Germany failed to replicate
the main effects (long D4DR/higher NS, 5-HT2Cser /lower RD [found only a trend])
but did find a highly significant effect of the interaction of the two alleles on levels
of RD.11 The lowest mean RD score was again seen in the 5-HT2Cser allele/long D4DR
allele group.
14.2.2 THREE CAN PLAY, TOO

In addition to the serotonin and dopamine neurotransmitter systems, a third system,
catechol O-methyltransferase (COMT), has also been found to play a role in per-
sonality and temperament. COMT and monoamine oxidase (MAO) are the two
enzymes that account for the first steps in the elimination of catecholamines.12 The
physiological substrates of COMT include dopamine, norepinephrine and adrenaline
but in contrast to MAO, serotonin is not a substrate for this enzyme.13 A common
COMT variant coding for a thermolabile low activity enzyme has been identified.14,15
A single amino acid substitution (val108 5 met) in exon IV accounts for differences
in thermolability: val/val homozygotes display high levels, val/met intermediate
levels and met/met 4 to 5 times lower COMT activity.16 When the effects on
personality of three polymorphismsD4DR exon III, 5-HTTLPR, and COMT
were examined simultaneously, a complex set of interactions was revealed.17 The
long allele of D4DR was associated with higher NS scores except in the presence
of the short allele of 5-HTTLPR, and this effect was most powerful in the presence
of the high activity (val/val) version of COMT; the effect of D4DR on NS virtually
disappeared, however, in those subjects with the val/met version of COMTregard-
less of 5-HTTPLR genotype. In this case, as with the 5-HT2c and D4DR interaction
noted above, this finding was subsequently replicated by a different group in a
separate population.18
14.3 NOT ONLY IN HUMANS

Association studies in human populations are not the only possible way to study
genegene interactions. Research with knockout and other types of genetically
manipulated animals has strongly supported the case for complex genegene
interaction effects, as the effect of many specific manipulations has been found to
be profoundly influenced by the genetic background of the animal.19 Single-gene
knockout mice have now been interbred in order to study interactions between
alterations in two or more genes. A recent review by Murphy et al.20 summarizes

studies that looked at behavioral, physiological and/or neurochemical phenotypes in
crosses between mice with single allele or single gene targeted disruptions in the
serotonin transporter (SERT) and five other mouse genes, all of which are known
to be polymorphic in humans and all of which are considered candidate genes for
neuropsychiatric and other disorders: dopamine transporter (DAT), brain-derived
neurotropic factor (BDNF), monoamine oxidase type A (MAOA), serotonin 1B
receptor (5-HT1B), and norepinephrine transporter (NET). An impressive variety of
interaction effects was found. As summarized in the review, some interactions were
additive, such as effects on brain serotonin levels of the SERT BDNF cross.
An epistatic effect on striatal serotonin concentrations was found in the SERT
DAT cross, as was an epistatic effect on behavior (reduced cocain-conditioned place
preference). Other interactions were synergistic, antagonistic, or even more
complex.20
14.4 GENEGENE INTERACTIONS IN COMMON

HUMAN DISEASES
As the importance of genegene interactions in personality is becoming increasingly
clear, it may be said that the role of such interactions is already considered a fact
in the realm of many illnesses and disorders. It has even been proposed that epistasis
is a ubiquitous component of the genetic architecture of common human diseases,
and that complex interactions are more important than the independent main effects
of any one susceptibility gene (Moore,21 emphasis mine). Understanding some of
these interactions could have important implications for somatic therapies of neu-
ropsychiatric disorders, as illustrated by the work of Anttila et al.22 That group has
found that those schizophrenia patients who had specific variants of two polymor-
phismsthe met/met form of COMT, and the c/c version of the gene NOTCH4
(involved in the regulation of generating neurons and glial cells)had more than
ten times the risk of not responding to conventional antipsychotic drugs, as opposed
to patients with other genotypes.
14.5 STATISTICAL METHODS

Many of the genegene interactions discussed above have been pinpointed using
relatively simple mathematical toolsANOVA or MANCOVA (multiple analysis
of covariance) analyses. This is the analysis of choice in association studies when
the dependent variable is continuous (as are personality or temperament trait
measures) and the genetic variables are nominal (specific allele/s present or absent).
Sex is most appropriately entered as an independent variable (main effect), while
other factors known to influence the dependent variable, such as age, are entered as
covariates. Note that significant interaction effects can be found even in the absence
of main effects for the individual genotype or allele. When study designs include
measures of relatedness (twins, sib pairs, linkage studies using parentoffspring
triads, and so on), the analysis of choice will be some form of multiple regression.
Increasingly sophisticated analyses are being proposed, which should help carry this
work forward. For a good discussion of sample size requirements for finding
genegene and geneenvironment interactions in various models of association
studies (case-control, affected sib, and case-parent/ case only designs), see Gauder-
man.23 Holmans24 discusses the conditions under which the use of a modification of
logistic regression should be used to examine possible genegene interactions in
linkage studies. An expanded method for model fitting which uses a conditional
logistic approach and allows for the analysis of genegene interactions at unlinked
loci is presented by Cordell et al.25
14.6 SUMMARY
What, as asked above, can we learn from these experiments? One important lesson
might be that when we begin to think about genetic engineering for human person-
ality, caveat emptorlet the buyer be very very beware. We already know that many
human genes are highly pleiotropic, and that changing a gene could have far-ranging
but unintended effects. Now things have become even more complex. Not only the
multiple effects of any given gene must be taken into account, but also all possible
interactions with any number of other genes. A second lesson, perhaps more imme-
diate than the first, is that meta-analysis may not be the most appropriate way to
understand the matrix of replications and non-replications of the effects of individual
candidate genes on personality or temperament. It seems increasingly likely that
summing studies across different populations will yield false negatives, as differing
allele frequencies for other genes interacting with the candidate gene will be impact-
ing upon the dependent variable. Systematic estimation of allele frequencies in
different populations, and wide-ranging examination of possible interactions
between known candidate genes will be required in order to elucidate the true nature
of genetic factors in personality, and no less, in the search for the genetic components
of major mental illnesses.
REFERENCES
1. Plomin, R., DeFries, J.C., McClearn, G.E., and Rutter, M. (1997). Behavioral Genet-
ics. NY: W.H. Freeman and Company.
2. Ebstein, R.P., Novick, O., Umansky, R., Priel, B., Osher, Y., Blaine, D., Bennett, E.R.,
Nemanov, L., Katz, M., and Belmaker, R.H., (1996). Dopamine D4 receptor (D4DR)
exon III polymorphism associated with the human personality trait of Novelty Seek-
ing. Nat Genet 12, 7880.
3. Benjamin, J., Li, L., Patterson, C., Greenberg, B.D., Murphy, D.L., and Hamer, D.H.
(1996). Population and familial association between the D4 dopamine receptor gene
and measures of Novelty Seeking. Nat Genet 12, 8184.
4. Cloninger, C.R., Przybeck, T., Svrakic, D., and Wetzel, R. (1994). The Temperament
and Character Inventory: A guide to its development and use. St. Louis: Center for
Psychobiology of Personality, Washington University.
5. Lesch, K.P., Bengel, D., Heils, A., Sabol, S.Z., Greenberg, B.D., Petri, S., Benjamin,
J., Muller, C.R., Hamer, D.H., and Murphy, D.L. (1996). Association of anxiety-
related traits with a polymorphism in the serotonin transporter gene regulatory region.
Science 274:15271531.
6. Osher, Y., Hamer, D., and Benjamin, J. (2000). Association and linkage of anxiety-
related traits with a functional polymorphism of the serotonin transporter gene reg-
ulatory region in Israeli sibling pairs. Mol Psychiat 5:216219.
7. Lesch, K.P. (2001). Molecular foundation of anxiety disorders. J Neural Transm
108:717746.
8. Lakatos, K., Nemoda, Z., Birkas, E., Ronai, Z., Kovacs, E., Ney, K., Toth, I., Sasvari-
Szekely, M., and Gervai, J. (2003). Association of dopamine receptor gene and
serotonin transporter promoter polymorphisms with infants response to novelty. Mol
Psychiatry 8:9097.
9. Szekely, A., Ronai, Z., Nemoda, Z., Kolmann, G., Gervai, J., and Sasvari-Szekely,
M. (2004). Human personality dimensions of persistence and harm avoidance asso-
ciated with D4DR and 5-HTTLPR polymorphisms. Am J Med Genet 126B:106110.
10. Ebstein, R.P., Segman, R., Benjamin, J., Osher, Y., Nemanov, L., and Belmaker, R.H.
(1997). 5-HT2C Serotonin receptor gene polymorphism associated with human per-
sonality trait of reward dependence: interactins with D4DR and D3DR polymor-
phisms. Am J Med Genet 74:6572.
11. Kuhn, K.U., Meyer, K., Nothen, M.M., Gansicke, M., Papassotiropoulos, A., and
Maier, W. Allelic variants of dopamine receptor D4 (DRD4) and serotonin receptor
5HT2c (HTR2c) and temperament factors: replication tests. Am J Med Genet
88(2):16872.
12. Boulton, A.A. and Eisenhofer, G. (1998). Catecholamoine metabolism: From molec-
ular understanding to clinical diagnosis and treatment. Overview. Adv Pharmacol
42:273292.
13. Mannisto, P.T. (1998). Catechol O-methyltransferase: characterization of the protein,
its gene, and the preclinical pharmacology of COMT inhibitors. Adv Pharmacol
42:324328.
14. Weinshilboum, R. and Dunnette, J. (1981). Thermal stability and the biochemical
genetics of erythrocyte catechol O-methyl-transferase and plasma dopamine-beta-
hydroxylase. Clin Genet 19:426437.
15. Lotta T., Vidgren J., Tilgmann C., Ulmanen I., Melen K., Julkunen I., and Taskinen
J. (1995). Kinetics of human soluble and membrane-bound COMT: a revised mech-
anism and description of the thermolabile variant of the enzyme. Biochemistry
34:42024210.
16. Lachman, H.M., Papolos, D.F., Saito, T., Yu, Y.M., Szumlanski, C.L., and Weinshil-
boum, R.M. (1996). Human COMT pharmacogenetics: description of a functional
polymorphism and its potential application to neuropsychiatric disorders. Pharma-
cogenetics 6:243250.
17. Benjamin, J., Osher, Y., Kotler, M., Gritsenko, I., Nemanov, L., Belmaker, R.H., and
Ebstein, R.P. (2000). Association between TPQ traits and three functional polymor-
phisms: D4DR, 5-HTTLPR, and COMT. Mol Psychiat 5:96100.
18. Strobel, A., Lesch, K.P., Jatzke, S., Paetzold, F., and Brocke, B. (2003). Further
evidence for a modulation of NS by DR exon III, 5-HTTLPR, and COMT val/met
variants. Mol Psychiatry 8:371372.
19. McClearn, G. E. (2004). Nature and nurture: Interaction and coaction. Am J Med Gen
124B: 124130.
20. Murphy, D.L., Uhl, G.R., Holmes, A., Ren-Patterson, R., Hall, F.S., Sora, I., Detera-
Wadleigh, S., and Lesch, K.P. (2003). Experimental gene interaction studies with
SERT mutant mice as models for human polygenetic and epistatic traits and disorders.
Genes Brain Behav 2:350364.
21. Moore, J.H. (2003). The ubiquitous nature of epistasis in determining susceptibility
to common human diseases. Hum Hered 56:7382.
22. Anttila, S., Illi, A., Kampman, O., Mattila, K.M., Lehtimaki, T., and Leinonen, E.
(2004). Interaction between NOTCH4 and COMT genotypes in schizophrenia patients
with poor response to typical neuroleptics. Pharmacogenetics 14: 303307.
23. Gauderman, W.J. (2002). Sample size requirements for association studies of
genegene interaction. Am J Epidemiol 155: 478484.
24. Holmans, P. (2002). Detecting genegene interactions using affected sib-pair analysis
with covariates. Hum Hered 53:92102.
25. Cordell, H.J., Barratt, B.J., and Clayton, D.G. (2004). Case/pseudocontrol analysis
in genetic association studies: A unified framework for detection of genotype and
haplotype associations, genegene and geneenvironment interactions, and parent-
of-origin effects. Genet Epidemiol 26:167185.
15 Schizophrenia: Study
of a Genetically Complex
Phenotype
Michael F. Pogue-Geile and Irving I. Gottesman
CONTENTS
Abstract .................................................................................................................. 209

15.1 Introduction .................................................................................................. 210
15.2 Defining the Phenotype ............................................................................... 210
15.3 Does Genetic Variation Contribute to the Etiology
of Schizophrenia?......................................................................................... 211
15.3.1 Family Studies.................................................................................. 211
15.3.2 Twin Studies..................................................................................... 213
15.3.3 Adoption Studies.............................................................................. 214
15.4 What Is the Nature of Genetic Effects on Schizophrenia? ......................... 216
15.4.1 Only Genetic Effects? ...................................................................... 216
15.4.2 How Many Genes?........................................................................... 217
15.5 Where on the Chromosomes Are Schizophrenia
Liability Genes Located? ............................................................................. 219
15.6 Which Genes Contribute to Schizophrenia Liability? ................................ 220
15.7 Conclusions .................................................................................................. 222
References.............................................................................................................. 222
ABSTRACT
Genetic research on the phenotype of schizophrenia is reviewed to illustrate the strat-
egies, problems, and findings encountered in the study of a genetically complex,
relatively common, dysfunctional behavioral phenotype. Schizophrenia provides an
excellent case history for these purposes because it has one of the longest histories
and most socially costly presence of any psychopathological phenotype. Family, twin,
and adoption studies are consistent in indicating that genetic effects on schizophrenia
are both important and complex. Genetic model-fitting to risk data for different
relative classes also suggests that: (1) environmental experiences that are uncorrelated
among siblings probably interact with the relevant genotypes to contribute to (or protect
from) schizophrenia, (2) action of abnormal alleles in multiple genes seems likely
209
for most schizophrenia cases, and (3) phenotypically normal (non-schizophrenic) indi-
viduals with unexpressed genetic liability exist, confusing molecular genetic studies
as false negatives. Genetic linkage and association studies attempting to locate and
identify individual genes of probably modest effect have to-date produced numerous
positive and negative findings with frequent failures to replicate. However, a second
generation of more sophisticated studies suggests increased reasons for cautious opti-
mism that the nature of the genetic effects on schizophrenia will be revealed in the
foreseeable future.
15.1 INTRODUCTION
Schizophrenia, which was first defined and initially named dementia praecox by
the German psychiatrist Emil Kraepelin in 1896,31 is a diagnosis based upon abnor-
malities of experience and behavior that include hallucinations, delusions, disorga-
nized speech, affective flattening, and bizarre behavior (for current criteria see DSM
IV1). Individuals with the diagnosis suffer severe personal and family distress as
well as frequent chronic impairments in social and occupational functioning. Not
only is the illness usually clinically severe, it is also relatively common and wide-
spread, with approximately a 1% lifetime risk in the general population, world-
wide.16,19 It is therefore a major public health problem warranting concerted effort
to understand its etiology as a guide to effective prevention and treatment.
The aim of the present chapter is a circumscribed one. For pedagogical purposes,
we will consider schizophrenia as an example of a genetically complex behavioral
phenotype and will selectively survey research on genetic effects in order to demon-
strate the major questions, strategies, problems, and findings that have been con-
fronted in its study and that may therefore be relevant to other behavioral phenotypes.
This is a daunting task, as schizophrenia has both one of the longest and most active
histories of genetic research of any psychopathological phenotype, beginning with
Rdin57 in 191612,22 and continuing until today. With these caveats, it should be clear
that the following is of necessity not a comprehensive and critical review of all genetic
findings on schizophrenia.8,10,21,29,43
15.2 DEFINING THE PHENOTYPE

A common issue in the study of many behavioral phenotypes is that they are usually
developed and defined based on nongenetic criteria. In the case of schizophrenia, the
definition was initially developed based upon clinicians observations and hypotheses
that a number of signs and symptoms clustered together cross-sectionally and longi-
tudinally in a manner that might suggest the existence of some common, albeit
unknown, pathophysiology and perhaps etiology. Sometimes the mapping of such
phenotypically defined behavioral syndromes on to genetic causes works well and
relatively simply, as was the case for the clinically developed definition of Hunting-
tons disease, which is primarily caused by necessary and sufficient mutations of a
single gene.46 However, it is certainly not necessarily the case that one disorder is the
result of one gene mutation.50 If the relationship between a clinical phenotype and
Schizophrenia 211
genetic causes is a complex one, then it can be argued that the particular phenotype
may not be the most useful one for genetic analysis. In which case, it makes sense
to begin to define new phenotypes and endophenotypes using genetic validating
criteria that may reflect gene effects more simply.13,61 This boot strapping strategy
in which the results of genetic studies serve to change the definition of the phenotype
studied is an important approach in general and one that appears to have particular
relevance to schizophrenia. However, for reasons of space we will focus here primarily
on the clinically defined phenotype of schizophrenia itself and not on potentially
related phenotypes, such as schizoaffective disorder, schizotypal personality disorder,
or various cognitive and biological abnormalities in the long causal chain from gene
to endophenotype to a named clinical phenotype, despite their potential interest and
importance.13,52,53,66 For similar reasons, we will also not consider in detail any poten-
tial genetic differences among the various operational definitions of clinical schizo-
phrenia itself.
15.3 DOES GENETIC VARIATION CONTRIBUTE

TO THE ETIOLOGY OF SCHIZOPHRENIA?
One of the first questions to be asked concerning the etiology of any phenotype is
whether variation in genotypes in the population contributes at all to individual
differences in the phenotype (it usually does). In this section we will briefly consider
several designs that have addressed this question for schizophrenia. Those aspects
of the designs most relevant to detecting genetic rather than environmental influences
will be stressed.
15.3.1 FAMILY STUDIES

Traditional family studies of first-degree relatives (parents, offspring, and siblings)
are popular and often a first step in studying etiology because of the relative ease
of subject recruitment. However, the inferences that can be drawn from them are
somewhat limited, with the greatest limitation being that genetic and shared envi-
ronmental causes for any observed phenotypic resemblance among relatives cannot
be distinguished because family members share half their genes and many similar
environments and experiences. Nevertheless, the presence of some family resem-
blance for the phenotype under study is a sine qua non for the presence of genetic
influence, although phenotypes that are primarily caused by interactions among a
large number of gene loci (epistasis) or among a large number of shared environ-
mental experiences may show little or no family resemblance, unless identical (MZ)
twins are studied. Detailed comments on the methodology of family studies are
provided in Whitfield and Nelson (see Chapter 12 of this volume).
Beginning with Rdin in 1916,58 who worked with Kraepelin in Munich, numer-
ous family studies of schizophrenia have been performed and the findings prior to
1991 have been summarized by Gottesman10 (see Table 15.1). More recent and
methodologically stronger individual studies24,37 echo these earlier aggregated obser-
vations. Evidence of family resemblance for a categorical characteristic such as
schizophrenia in studies using selected ascertainment via patient probands comes
TABLE 15.1
Risk of Definite or Probable Schizophrenia among Relatives of
Schizophrenic Probands Aggregated across Studies
% Genes Shared
Relationship to Schizophrenic with Observed
Proband Schizophrenic Proband Morbid Risk (%)1
Monozygotic cotwins 100 48
Dizygotic cotwins 50 17
Siblings 50 9
Offspring (1 parent affected) 50 13
Parents 50 6
Second degree2 25 34
General population 0 1
1 Estimate of the probability of receiving a diagnosis of schizophrenia at some time during ones life
(all risks age-corrected, except twins).
2 Includes half-siblings, uncles/aunts, nephews/nieces, and grandchildren.
Source: After Gottesman I.I., Schizophrenia Genesis: The Origins of Madness. WH Freeman & Co.,
New York, 1991. With permission.
primarily from observing an increased age adjusted lifetime risk for schizophrenia
among the different relative classes (e.g., parents, offspring, siblings, etc.) compared
with the lifetime risk among control samples or the general population, which is
usually estimated to be about 1%.19 As can be seen in Table 15.1, risks for schizo-
phrenia are increased among all classes of relatives, indicating familiality for schizo-
phrenia, which by itself is consistent either with the effects of shared genes and/or
shared environmental experiences.
If, for the sake of argument at this stage, we were to presume that this observed
resemblance were due only to shared genes and not shared environment, then the data
suggest several tentative interpretations. First, there is a relative risk (risk among
family members/risk in the general population) among individuals sharing 50% of
genes with a schizophrenic proband (i.e., offspring and siblings) of approximately 10
times the general population rate. Although substantial, a risk of 13% among offspring
of a schizophrenic parent is considerably less than the 50% predicted from a rare,
single gene with dominance and complete penetrance (i.e., all individuals with a
disease genotype [cf. Huntingtons disease] become schizophrenic). The similarity
between risks in offspring and siblings argues against genetic recessive effects. Some-
what of an anomaly is the risk among parents, who also share 50% of their genes
with a schizophrenic proband, but only have about half the risk of offspring and
siblings. This discrepancy is usually interpreted as being a result of the common-sense
screening for mental health that occurs in the mating process and the resulting reduced
reproductive fitness (number of offspring) among schizophrenic patients. The risk
among second-degree relatives (e.g., uncles/aunts, nephews/nieces, and grandchil-
dren), who share 25% of their genes with the proband, is also less than half of the
risk among first-degree relatives, the prediction if only a single-gene were involved
Schizophrenia 213
(see below). Therefore, if these family data were due solely to genetic effects, their
pattern is inconsistent with a simple, single-gene transmission, but instead tentatively
suggests a complex mode of transmission involving perhaps multiple genes and/or
incomplete penetrance (i.e., not all individuals with a schizophrenia genotype develop
schizophrenia). Of course, based only on family study data, these findings could also
all be explained by experiences shared within families. Thus, although the family
study data are consistent with genetic influences on schizophrenia, they cannot prove
their existence (or rather, fail to rule them out).
15.3.2 TWIN STUDIES

The classical twin study attempts to surmount this constraint of the family study by
taking advantage of an experiment of nature (the existence of two kinds of twinning)
and making an important assumption about environmental effects. Inferences from
the classical twin study are based on comparisons of phenotypic resemblance
between two twin groups (monozygotic, MZ vs. dizygotic, DZ) that differ in genetic
similarity (100% vs. 50% shared genes), but are assumed to be equivalent on
phenotype-relevant environmental similarity. Thus, the extent to which MZ twins
are observed to resemble each other phenotypically more than DZ twins can be
attributed to genetic effects, if the equal environmental similarity assumption is
correct. Detailed discussion of twin study methodology is provided in Whitfield and
Nelson (see Chapter 12 of this volume).
Beginning with Luxenburger36 in Munich in 1928, a number of twin studies of
schizophrenia have been reported and have again been helpfully aggregated10 (see
Table 15.1). Once again, results from more recent and methodologically sophisti-
cated studies reprise these summaries of earlier work.3,30,47 Resemblance for a cat-
egorical trait such as a diagnosis of schizophrenia from twin studies using selected
ascertainment through probands is usually indexed by the probandwise concordance,
or rate of schizophrenia among MZ and DZ cotwins of schizophrenic probands.38
As can be seen in Table 15.1, the MZ concordance rate is considerably greater than
that for DZ twins (48% vs. 17%), which suggests the importance of genetic influ-
ences on the etiology of schizophrenia, although this interpretation relies on the
validity of the assumption that experiences relevant to schizophrenia are shared
equally between MZ and DZ twins. For other phenotypes, such as personality traits,
evidence generally supports this, but for schizophrenia, it remains an untested, albeit
plausible assumption. Again, like the family data discussed above, the twin data
point toward genetic complexity. First, the MZ concordance rate is not 100%, which
may suggest an important role (in addition to genes) for environmental experiences
that are not always shared between twins and therefore the existence of individuals
with a genetic liability for schizophrenia who are not clinically schizophrenic (i.e.,
incomplete penetrance). Furthermore, the MZ concordance rate is considerably more
than twice the DZ rate, which would not be predicted if a single gene without
recessivity were operating. Thus, twin studies of schizophrenia provide evidence for
genetic influences of some sort on schizophrenia and hint at genetic complexities
and environmental effects.
A hybrid twin-family design has also been employed with schizophrenia to

investigate further these results from classical twin studies. Schizophrenic twins and
their adult offspring were initially studied by Fischer7 in Denmark and subsequently
followed up by Gottesman and Bertelsen.11 One of the questions addressed by these
studies was the presence of genetic liability in discordant MZ twin pairs.32 One
interpretation of the MZ twin concordance/discordances described above in the clas-
sical twin studies involves extreme etiological heterogeneity and posits that there are
basically two distinct kinds of schizophreniathose caused solely by genetic factors
and those caused solely by environmental insults (environmental phenocopies). In
this view, genetic liability would not be necessary for schizophrenia, although it might
be sufficient and concordant MZ twin pairs would have genetic schizophrenia and
discordant pairs environmental schizophrenia. An alternative interpretation of the
classical twin study results emphasizes relative etiological homogeneity and instead
hypothesizes that concordant and discordant MZ twin pairs all have at least some
genetic liability. In the discordant pairs, however, the diagnosed twin has been exposed
to environmental experiences that interact or co-act with their genetic liability to
produce clinical schizophrenia, but the undiagnosed twin has not (or vice versa for
protective environmental experiences). Presumably in concordant pairs, both mem-
bers were exposed to the relevant experiences. Under this model, some genetic
liability would probably be necessary, but not sufficient for schizophrenia. In order
to address the questions of whether genetic liability may be necessary and/or sufficient
to cause schizophrenia, Fischer,7 Gottesman and Bertelsen identified schizophrenic
twins who had become parents by crossing Denmarks National Twin Register with
other national registers. The twins adult offspring were then identified and their rate
of schizophrenia was determined using the national psychiatric registry, hospital
charts, and interviews. The risk for schizophrenia among the offspring of the non-
schizophrenic MZ cotwins of schizophrenic twins was 17%, which was significantly
greater than the 2% found among the offspring of non-schizophrenic DZ cotwins of
schizophrenic twins. This result suggests that discordant MZ twins have significant
genetic liability that can be transmitted to their offspring even though they do not
manifest it themselves. Data from this design reinforce the importance of genetic
influences, demonstrate the existence of unexpressed genotypes, and indicate the
importance of (unspecified) environmental experiences that are uncorrelated between
twins that may interact with genetic liability. These data suggest that genetic effects
may be necessary but not sufficient to cause schizophrenia.
15.3.3 ADOPTION STUDIES

Adoption studies are the second general design that attempts to estimate genetic
effects and to overcome the limitations of the traditional family study.51 Adoption
studies aim to uncorrelate the effects of shared genes and shared experiences within
families by identifying individuals (adoptees) who share only half their genes with
one set of relatives (biological parents and siblings) and only their family environment
with another set (adoptive parents and siblings). If genes and environments are truly
uncorrelated in the adoption study, then any observed phenotypic resemblance
between adoptees and their biological relatives can be attributed to shared genetic
Schizophrenia 215
effects and any observed phenotypic resemblance between adoptees and adoptive
relatives can be attributed to shared environmental effects. The two inferences, how-
ever, depend on the absence of selective placement of adoptees in adoptive homes
that induces a correlation between their genotypes and rearing environments that are
relevant to the phenotype under study. Selective placement is often indexed by an
observed resemblance for relevant phenotypes between biological parents (as an
estimate of the adoptees genotype) and adoptive parents (as an estimate of the
adoptees rearing environment). For example, it could be that adoptees were selec-
tively placed in adoptive homes based on matching the religion or language spoken
(e.g., Spanish or English) of the adoptive parents with that of the adoptees biological
parents as part of an adoption agency policy attempting to increase similarity between
adoptees and their adoptive families for ethnic/cultural characteristics. In such a
situation (selective placement for and environmental transmission of language spo-
ken) it would be obviously incorrect to attribute any observed resemblance for lan-
guage spoken between adoptees and biological relatives to genetic effects. Instead,
adoptees and biological relatives spuriously resemble each other for language spoken
due to the effects of selective placement and cultural transmission from adoptive
parents to adoptees. Similar examples (e.g., selective placement for hair color) could
be developed in which selective placement spuriously inflates estimates of environ-
mental effects based on resemblance between adoptees and their adoptive relatives.
Unfortunately, this important assumption of the adoption design is difficult to evaluate
for studies of schizophrenia because so little is known about which phenotypes are
relevant and should be correlated between biological parents and adoptive parents.
More optimistically, if researchers do not know what to measure to detect selective
placement, then it seems unlikely that adoption agencies do either.
Adoption studies of relatively rare, categorical phenotypes such as schizophrenia
rarely use unselected random samples of adoptees, but typically employ some
selective ascertainment strategy to increase statistical power and efficiency. Two
complimentary techniques have been used for estimating genetic effects based on
the resemblance between adoptees and their biological relatives. The adoptees
method ascertains index families through a schizophrenic biological parent who
has adopted away an offspring and control families through a demographically
matched non-schizophrenic biological parent who has adopted away an offspring.
The rate of schizophrenia in the adopted away offspring (adoptees) of the index
and control biological parents is then assessed and compared. If genetic influences
contribute to schizophrenia, then schizophrenia should be more frequent in the index
adoptees compared with the control adoptees. In contrast, the adoptees families
method ascertains index probands who are schizophrenic adoptees and control
probands who are demographically matched non-schizophrenic adoptees. The rate
of schizophrenia in the biological relatives (adoptees families) of the index
adoptees is then assessed and compared with that among the biological relatives of
the control adoptees.
The adoption strategy was pioneered in schizophrenia research by Heston in
Oregon in 196617 and Rosenthal56 and Kety26 from the NIMH who studied adoptees
and their families in Denmark. The results across these and other adoption studies68
have been consistent in indicating significant resemblance for schizophrenia between
biological relatives and adoptees using both adoptees and adoptees families methods,
thus suggesting the presence of genetic influences (see Gottesman16 for a detailed
review). For example, a final report from Kety and colleagues of their Danish adoptees
families study based on personal interviews found a morbid risk for schizophrenia
(using DSM-III criteria) among the first degree biological relatives of schizophrenic
adoptees of 8% compared to about 1% in the first degree biological relatives of control
adoptees.23,27 Inclusion of diagnoses of schizoaffective disorder, schizotypal, and
paranoid personality disorder as part of a schizophrenia spectrum increases the
prevalences to 24% vs. 5%, respectively. The convergence of evidence indicating
genetic influences from both adoption and twin studies is impressive. Although each
kind of study makes assumptions that could be argued separately, they are different
assumptions, and thus strengthen the case for an important role for genetic effects
among the distal causes of schizophrenia.
15.4 WHAT IS THE NATURE OF GENETIC EFFECTS

ON SCHIZOPHRENIA?
Decades of work around the world have established indirectly that genes affect risk
for schizophrenia. As important as this fact is, however, it is just the beginning of
the important questions concerning the nature of these genetic effects. One of the
first such questions is: Is schizophrenia liability due only to genes or is environmental
variation also important?
15.4.1 ONLY GENETIC EFFECTS?

As discussed above, the discordance rate of about 50% among MZ twins argues
strongly for the importance of nongenetic factors that are uncorrelated among
family members. Typically, such factors are termed nonshared environmental
effects, which might include such experiences as accidents, peer relations, and
other environmental experiences that are uncorrelated in siblings and MZ twins.
However, it may also be that differences between individuals with identical gen-
otypes arise for purely stochastic, chance reasons, as suggested by asymmetries
in paired bilateral body structures,71 and epigenetic factors.70 Although these pos-
sibilities cannot be resolved currently, it nevertheless seems clear that some such
nonshared factors contribute in an important manner along with genotypes to the
liability to schizophrenia. The specific nature of these nonshared nongenetic factors
remains one of the many major questions concerning the etiology of schizophrenia.
In contrast to the apparent importance of nonshared effects, other results, for
example from adoption studies, find little evidence for important effects of expe-
riences that are shared among relatives (often termed shared environmental effects).
Thus, to be plausible, models of genetic effects on schizophrenia must include a
role for nonshared environmental effects, although apparently shared experiences
are not needed. Incorporation of either sort of environmental effect is also often
considered as one way of allowing for incomplete penetrance of the genotype,
where penetrance is the probability of developing schizophrenia given the relevant
genotype.
Schizophrenia 217
15.4.2 HOW MANY GENES?

Assuming that any genetic model of schizophrenia needs to include nonshared envi-
ronmental effects, the next important question concerns how many genes contribute
to schizophrenia liability. The most parsimonious genetic model for the transmission
of schizophrenia, the generalized single locus (GSL) model,20 involves only one gene,
although for reasons discussed above, incomplete penetrance due to nonshared envi-
ronmental effects also needs to be included. More complex alternatives incorporate
multiple genes in various ways (in addition to the usual nonshared environmental
effects).
Oligogenic models posit a small, defined number of genes. One popular
oligogenic model specifies that all risk-increasing genes are required to produce
schizophrenia. For example, if such an oligogenic model hypothesizes the existence
of three genes that can contribute to risk for schizophrenia, then only individuals
with all three liability alleles are at risk for schizophrenia. Individuals with one or
any two alleles will not become schizophrenic. This sort of model involves genetic
epistasis, or interaction among genetic loci, in that the effect of each individual
allele depends on the presence of the other alleles.
A more complex multiple gene model is the related multifactorial threshold
(MFT) model,6 which was first proposed for schizophrenia by Gottesman and
Shields in 1967.15 This model posits: (1) a large, unspecified number of genes each
of relatively small effect (polygenes) and nonshared environmental experiences
that are interchangeable and additive in their effect on liability and (2) a categorical
threshold on the liability dimension beyond which an individual becomes schizo-
phrenic. For example, if a total of eight different genetic loci and three nonshared
experiences are hypothesized to increase risk for schizophrenia and the threshold
to produce schizophrenia is four factors, then an individual with any four or more
of the eight alleles or three experiences will develop schizophrenia. Importantly,
although the MFT model assumes additive effects of genetic and environmental
factors on quantitative liability, the imposition of a threshold makes for an epistatic
or interactive effect on the categorical diagnosis itself as the effect of any one gene
depends upon the effects of others in determining whether an individual crosses
the threshold and receives a diagnosis. The primary difference therefore between
typical MFT and oligogenicepistatic models is in the number of risk-increasing
genes hypothesized. There are also usually differences between the two in where
the threshold is placed. The typical MFT posits a greater number of potential risk-
increasing genes than is necessary to produce schizophrenia, whereas the oligo-
genicepistatic model usually hypothesizes that the number required to produce
schizophrenia equals the number of potential risk-increasing genes. This difference
has the effect that the MFT (but not the oligogenicepistatic model) predicts
etiological variation among schizophrenic patients because one individual may
have risk-increasing alleles for gene numbers 1, 2, 3, and 4, whereas another patient
may have alleles for genes 1, 2, 3, and 5.
How is one to decide among these several theoretical possibilities for the number
of genes affecting liability to schizophrenia? Currently, the primary data available to
evaluate these models are the risks for schizophrenia observed among the different
classes of relatives of schizophrenic probands. In addition, segregation analyses have

been applied to raw data from individual studies.67 The former strategy uses one of
the genetic models described above to generate predicted risks for schizophrenia
among different relative classes that can then be compared with the observed risks
(see Table 15.1) to evaluate whether the hypothesized model fails to fit the observed
data within sampling variation.39,54 The consistent finding from such analyses, as well
as segregation analyses, is that a single gene GSL model even with incomplete
penetrance does not fit the observed risks, whereas the multiple gene MFT or oligo-
genic models cannot be statistically rejected. The essential problem is that the GSL
model predicts a linear decrease in risk across the different relative classes because
the probability of sharing alleles at one genetic locus with a schizophrenic proband
decreases by a constant 50% between adjacent relative classes (e.g., from MZ twin
to first-degree relatives (including DZ twins, siblings, offspring) 50%/100% = 50%;
from first-degree relatives to second-degree relatives 25%/50% = 50%). It is important
to note that this linear prediction holds only if there is no recessivity, as seems to be
the case, because sibling and offspring risks are generally equal. If penetrance is set
at approximately 50% due to nonshared experiences (so that the observed risk of
about 50% in MZ cotwins is correctly predicted), then the simple GSL model
predicts a risk of 25% for first-degree relatives (50% probability of sharing allele
with proband 50% penetrance of allele = 25%) and a risk of 12.5% for second
degree relatives (25% probability of sharing allele with proband 50% penetrance =
12.5%); a linear decrease. Inspection of Table 15.1 shows that the observed risks do
not follow such a linear pattern and that this GSL model considerably over-predicts
the risks in first- (25% predicted vs. about 12% observed) and second-degree relatives
(12.5% predicted vs. 3 to 4% observed). Alternatively, the penetrance may be reduced
enough (to about 25%) to fit more closely the observed first (50% 25% penetrance =
12.5% predicted vs. about 12% observed) and second-degree risks (25% 25%
penetrance = 6% predicted vs. 3% observed), but in this case the GSL considerably
under-predicts the MZ twin risk (100% 25% penetrance = 25% predicted vs. 50%
observed). Importantly, these prediction errors of the GSL model hold even in the
face of a kind of genetic heterogeneity in which schizophrenia is caused by several
different single major genes in different families because a linear pattern is predicted
for each single major locus and the average across individual families remains linear.
An important assumption of all these analyses (in addition to no recessivity) is that
there are either minimal effects of environmental experiences shared within families
(which seems largely justified from the literature), or that their effects also decrease
linearly at a slope of .50 across relative classes. However, if environmental experi-
ences relevant to schizophrenia were shared only among MZ twins, then a more
complex GSL model that incorporated this feature would probably be able to fit the
observed data and not under-predict the MZ twin risk. Overall, therefore, after making
the reasonable assumptions of no special MZ twin shared environment and no reces-
sivity, it is highly improbable that a single gene causes schizophrenia. In contrast to
the problems of the GSL models, the more complex multiple-gene hypotheses fit the
observed data much more easily because they predict a curvilinear decrease in risk
across relatives rather than a linear relationship. As a simplified exercise (ignoring
issues of allele frequency effects and phenocopies), let us step through a two-gene
Schizophrenia 219
epistatic oligogenic model with 50% penetrance of the joint two locus genotype due
to nonshared experiences (to predict MZ twin risk satisfactorily). The probability
that a MZ twin of a schizophrenic proband shares alleles at two loci with the proband
is 100% (because MZ twins share all their genes) and given a penetrance of 50% the
model predicts a risk for MZ twins of 50% (100% 50% penetrance = 50%
predicted vs. about 50% observed). The probability that a first-degree relative shares
alleles at two loci with a proband is 25% (probability of sharing at one locus, 50%,
probability of sharing at second locus, 50%, = 25%) and thus the model predicts a
risk of 12.5% for first-degree relatives (25% 50% penetrance = 12.5% predicted
vs. about 12% observed). Similarly, the probability that second-degree relatives share
two loci with a proband is 6.25% (25% 25% = 6.25%) and their predicted risk
would be 3% (6.25% 50% penetrance = 3.125% predicted vs. 3% observed). This
rough example shows how well even a simple multigene model can fit the observed
data and MFT models fit similarly well with the heritability of liability estimated at
about 80%.39
Given such results, it is very likely that schizophrenias genetic contributions
involve two or more gene loci acting together. However, it is difficult to distinguish
among epistatic oligogenic models with a few necessary genes, MFT models with
many possible genes of quite small effect, or MFT models with some genes of
moderate effect along with many of small effect. Attempts at distinguishing some
of the multigene alternatives have investigated the properties of integrative mixed
models, in which there is both one gene of major effect and a polygene background
both contributing to schizophrenia risk.45 Although fitting mixed models to individual
data sets has usually resulted in a failure to detect a single gene of major effect,67
analyses of aggregated data sets suggest some interesting possibilities. Gottesman
and McGue14 found the following three general scenarios that were statistically
consistent with observed familial risks: (1) presence of a highly penetrant, but very
rare, single major locus and high polygene heritability, (2) presence of a low pene-
trance, but relatively common, single locus and high polygene heritability, and, (3)
absence of a single locus and high polygene heritability. Risch55 used the same data
and found that several oligogenic models also fit the observed data well. In summary,
it appears quite likely that most cases of schizophrenia are caused by two or more
gene loci acting along with nonshared environmental influences, although the further
details of the number and importance of the loci remain speculative at this time.9 It
is also important to note that the individual genes in such multigene systems need
not be specific to schizophrenia liability.
15.5 WHERE ON THE CHROMOSOMES ARE

SCHIZOPHRENIA LIABILITY GENES LOCATED?
Studies that attempt to discover the chromosomal location of liability genes for
schizophrenia usually employ some variant of linkage analysis. Although such studies
have a long history in research on schizophrenia,40 considerably antedating the recent
molecular revolution, the first positive findings using DNA markers were reported
by Gurling and colleagues in 1988 for a region on chromosome 5.60 Foreshadowing
future developments, a failure to replicate was also published as a companion piece
in the same journal issue.25 Since these initial reports, the history of linkage studies
in schizophrenia has been an active one, but until recently with few positive findings
and fewer replications. This brief history can be divided into two phases. Early linkage
studies often seemed to be predicated on a search for the gene for schizophrenia,
despite the evidence discussed above suggesting the involvement of multiple genes.
Designs often employed large multiplex families, with many affected members in an
attempt to identify families segregating a single major gene. Candidate chromosomal
regions were usually investigated based on suggestions from other observations.
Sample sizes were usually relatively small based on the optimistic assumption of
finding a gene of major effect. Methods of analysis usually emphasized parametric
linkage approaches that depend on specifying a model of transmission that includes
values for: dominance/recessivity, frequency of phenocopies, penetrance of geno-
types, and frequency of alleles. Incorrect specification of these parameters should
decrease power to detect linkage, however, testing of multiple models tends to
inflate alpha error rates. Phenotype definitions were multiple, including diagnoses
of schizophrenia, schizophrenia-spectrum, and often even more inclusive diagnostic
groupings. Failures to replicate positive findings during this early phase were often
optimistically interpreted in terms of genetic heterogeneity across studies.
The frustrations of these initial studies and methodological developments have
encouraged a more sophisticated and realistic wave of studies and reports.29,42,44,49
Increasingly the view is that the aim of linkage studies is to identify multiple genes,
probably each of small to modest effect. Studies now typically utilize complete
genome scan approaches and, given the multitude of statistical tests applied, failures
to replicate are more often interpreted from the perspective of statistical false positives,
leading to proposals for more stringent statistical criteria.33 Designs have been increas-
ingly based on affected sibling pairs and analyzed using nonparametric affected
pedigree member techniques.34,69 In contrast to the parametric methods, these analytic
approaches do not require specification of a transmission model and include only
schizophrenic relatives, thus avoiding the potential problems of incorrect model spec-
ification and diagnostic false negatives among non-schizophrenic relatives, although
at a cost of statistical power. Sample sizes have also become much larger. Meta-
analyses and pooling of data across studies have also become more common (some-
times mandated by grant funders) and have been very useful. Although still very much
of a moving target, results from some of these newer studies have generated consid-
erable enthusiasm recently about potential replicated linkages on chromosomes 1, 2,
6, 8, 13, and 22 and others that have been reported by several groups.43,49,59,63 However,
skeptics remain28 and it will be for future work to resolve these controversies.2,29,35,41
15.6 WHICH GENES CONTRIBUTE

TO SCHIZOPHRENIA LIABILITY?
Answers to questions about specific contributing genes must rely on genetic asso-
ciation studies in which associations between particular alleles and the schizophrenia
phenotype are identified. Valid phenotypeallele associations may arise either
because the allele actually contributes causally to the phenotype or because it is in
Schizophrenia 221
linkage disequilibrium with a locus that does. Linkage disequilibrium occurs when
alleles at different loci become correlated. Association studies have the advantage
of being much more powerful than linkage studies to detect loci of small effect;
however, they are much less sensitive than linkage studies to loci beyond a narrow
distance surrounding the marker. Association studies employ either case-control
designs or various sorts of within-family designs to avoid spurious findings due to
population stratification. Increasingly, association studies utilize multiple polymor-
phisms (i.e., single nucleotide polymorphisms, SNPs) within a gene, in order to
measure haplotypes, which increases power. A haplotype is the sequence of alleles
across loci on a particular chromosome. Given their precision, genome-wide asso-
ciation studies have been impractical to date, perhaps requiring well over 100,000
markers to screen the entire genome.56 New technology using microchips that allows
rapid genotyping of individuals for such dense maps of polymorphisms, however,
is becoming less expensive and will make this strategy increasingly useful. Tech-
niques, such as DNA pooling, in which genotyping is performed on pools of
groups of individuals (e.g., all cases vs. all controls), simultaneously, are also
reducing genotyping costs, although the statistical issues surrounding performing
thousands of tests remain complicating factors.
For these reasons, association studies to date have employed some sort of
candidate strategy to narrow the search to a smaller number of markers. Candidate
strategies have either been based on hypotheses drawn from models of pathology
or on positional information provided by linkage studies. For schizophrenia, there
have been numerous such association studies using candidate genes most frequently
chosen because they might be relevant to schizophrenias hypothesized pathophys-
iologies (e.g., genes coding for characteristics affecting dopamine or glutamate
neurotransmission). To date, most such association results have been negative and
positive findings have been rarely replicated. One particular variation of this strategy
that should hold special promise is exemplified by the history of the RGS4 (Regulator
of G protein signaling 4, chr 1) gene. RGS4 was identified as a candidate through
a gene expression study of post-mortem brain in schizophrenia, which found it to
be under-expressed.42 Based on this gene-expression finding, gene association stud-
ies using RGS4 as a candidate have been performed, with promising results to date.4
As knowledge of pathology increases and more genes are identified and sequenced,
these approaches should become more fruitful.48,49
In contrast, positional candidate strategies have generated considerable excite-
ment recently. Based on promising results from linkage studies mentioned above,
markers in several candidate chromosomal regions have been investigated using
association techniques. Although still controversial, several genes have been pro-
posed as potential contributors to schizophrenia liability based on this strategy. The
following genes have had at least some positive replications (along with some
negative): Dystrobrevin-binding protein 1 (DTNBP1 dysbindin, chr 6),64 Neuregulin
1 (NRG1, chr 8),62 and D-amino acid oxidase (DAO) and D-amino acid oxidase
activator (DAOA) (chr 13)5. Future study will show whether the current enthusiasm
for these particular genes is well placed; denser SNP-otyping may reveal
unexpected candidate genes very close to the genes named here.
15.7 CONCLUSIONS
The aim of this chapter is to provide a case history of the primary strategies used
to investigate genetic influences on the complex phenotype of schizophrenia, as well
as to provide a sketch of current findings. The Internet is your best friend for keeping
up with insertions and deletions in the corpus of received knowledge. From the long
history of research on schizophrenia, the following points seem certain enough to
conclude: (1) genes play a major causal role in schizophrenia, (2) effects of multiple
genes, combined with environmental and stressful experiential factors, epigenetic,
and stochastic factors are safe bets for explaining most schizophrenia cases, (3) overall
effects at any single gene locus are small with unhelpful positive predictive power,
(4) some sort of genetic heterogeneity across schizophrenia patients is very likely,
(5) environmental experiences that are uncorrelated among siblings probably interact
with the relevant genotypes to contribute to (or protect from) schizophrenia liability,
and (6) non-schizophrenic individuals with unexpressed genetic liability exist and
may become transmitters of liability. It is very likely that most other psychological
and psychopathological phenotypes share many of these features.
Although we have covered many aspects of genetic research on schizophrenia, we
have also omitted many important areas due to space constraints.18 Among those that
may prove most relevant to future genetic studies are work on environmental influences,
geneenvironment interaction, etiological heterogeneity, endophenotype and phenotype
definition, developmental neurobiology,65 and evolutionary issues. Despite these omis-
sions, we have tried to communicate at least a hint of the clinical importance, intellectual
challenge, and excitement to be found in the etiological study of schizophrenia.
REFERENCES
1. American Psychiatric Association: Diagnostic and Statistical Manual of Mental Dis-
orders (4th ed.), American Psychiatric Association Press, Washington, DC, 1994.
2. Badner, J.A. and Gershon, E.S., Meta-analysis of whole-genome linkage scans of
bipolar disorder and schizophrenia. Mol. Psychiatry, 2002, 7: 405411.
3. Cardno, A.G. and Gottesman, I.I., Twin studies of schizophrenia: From bow-and-
arrow concordances to star wars Mx and functional genomics. American Journal of
Medical Genetics, 2000, 97, 1217.
4. Chowdari, K.V., Mirnics K., Semwal, P. et al., Association and linkage analyses of
RGS4 polymorphisms in schizophrenia. Hum. Mol. Genet., 2002, 11: 13731380.
5. Chumakov, I., Blumenfeld M., Guerassimenkom, O., et al., Genetic and physiological
data implicating the new human gene for D-amino acid oxidase in schizophrenia.
Proc. Natl. Acad. Sci. USA, 2002, 99: 1367513680.
6. Falconer, D.S., The inheritance of liability to certain diseases estimated from inci-
dence among relatives. Annals of Human Genetics, 1965, 29: 5176.
7. Fischer, M., Psychoses in the offspring of schizophrenic monozygotic twins and their
normal co-twins. British Journal of Psychiatry, 1971, 115: 981990.
8. Gershon, E.S. and Cloninger, C.R., Genetic Approaches to Mental Disorders. Amer-
ican Psychiatric Association Press, Washington, DC, 1994.
9. Gottesman, I.I., Complications to the complex inheritance of schizophrenia. Clinical
Genetics, 1994, 46, 116123.
Schizophrenia 223
10. Gottesman, I.I., Schizophrenia Genesis: The Origins of Madness. WH Freeman &
Co., New York, 1991.
11. Gottesman, I.I. and Bertelsen A., Confirming unexpressed genotypes for schizophre-
nia: Risks in the offspring of Fischers Danish identical and fraternal discordant twins.
Archives of General Psychiatry, 1989, 46: 867872.
12. Gottesman, I.I. and Bertelsen, A., Editorial: Legacy of German psychiatric genetics:
Hindsight is always 20/20. American Journal of Medical Genetics, 1996, 67: 317322.
13. Gottesman, I.I. and Gould, T.D., The endophenotype concept in psychiatry:
Etymology and strategic intentions. Am. J. Psychiatry. 2003, 160: 636645.
14. Gottesman, I.I., and McGue, M., Mixed and mixed-up models for the transmission
of schizophrenia. In D. Cichetti, W. Grove (Eds.), Thinking Clearly about Psychology:
Essays in Honor of Paul E. Meehl. University of Minnesota Press, Minneapolis, 1991,
pp. 295312.
15. Gottesman, I.I. and Shields J., A polygenic theory of schizophrenia. Proceedings of
the National Academy of Sciences of the United States of America, 1967, 58: 199205.
16. Gottesman, I.I. and Shields, J., Schizophrenia: The Epigenetic Puzzle. Cambridge
University Press, New York, 1982.
17. Heston, L.L., Psychiatric disorders in foster home reared children of schizophrenic
mothers. British Journal of Psychiatry, 1966, 112: 819825.
18. Hirsch, S.R. and Weinberger, D.R., Schizophrenia. Blackwell, Oxford, UK, 1995.
19. Jablensky, A., Schizophrenia: Recent epidemiologic issues. Epidemiologic Reviews,
1995, 17, 1020.
20. James, J.W., Frequency in relatives for an all-or-none trait. Annals of Human Genetics,
1971, 35: 4749.
21. Kendler, K.S. and Diehl, S.R., The genetics of schizophrenia: A current, geneticepi-
demiologic perspective. Schizophrenia Bulletin, 1993, 19: 261285.
22. Kendler, K.S. and Zerbin-Rdin, E., Abstract and review of Studien ber Vererbung
und Intstehung geistiger Strungen. I. Zur Vererbung und Neuentstehung der Demen-
tia Praecox, American Journal of Medical Genetics, 1996, 67: 338342.
23. Kendler, K.S., Gruenberg, A.M., and Kinney, D.K., Independent diagnoses of adop-
tees and relatives as defined by DSM-III in the provincial and national samples of
the Danish adoption study of schizophrenia. Archives of General Psychiatry, 1994,
51: 456468.
24. Kendler, K.S., McGuire, M., Gruenberg, A.M., OHare, A., Spellman, M., and Walsh,
D., The Roscommon family study: I. Methods, diagnosis of probands and risk of
schizophrenia in relatives. Archives of General Psychiatry, 1993, 50: 527540.
25. Kennedy, J.L., Giuffra, L.A., Moises, H.W., Cavalli-Sforza, L.L., Pakstis, A.J., Kidd,
J.R., Castiglione, C.M., Sjogren, B., Wetterberg, L., Kidd, K.K., Evidence against
linkage of schizophrenia to markers on chromosome 5 in a northern Swedish pedigree.
Nature, 1988, 336: 167170.
26. Kety, S.S., Rosenthal, D., Wender, P.H., Schulsinger, F., The types and prevalence of
mental illness in the biological and adoptive families of adopted schizophrenics. In
D. Rosenthal, S.S. Kety (Eds.): The Transmission of Schizophrenia. Pergamon Press,
Oxford, 1968, pp. 345362.
27. Kety, S.S., Wender, P.H., Jacobsen, B., Ingraham, L.J., Jansson, L., Faber, B., Kinney,
D.K., Mental illness in the biological and adoptive relatives of schizophrenic adoptees:
Replication of the Copenhagen study in the rest of Denmark. Archives of General
Psychiatry, 1994, 51: 442455.
28. Kidd, K.K., Can we find genes for schizophrenia? American Journal of Medical
Genetics (Neuropsychiatric Genetics), 1997, 74: 104111.
29. Kirov G., ODonovan, M.C., and Owen, M.J., Finding schizophrenia genes. The
Journal of Clinical Investigation, 2005, 115: 14401448.
30. Klning, U., Schizophrenia in twins: Incidence and risk factors. Department of Psy-
chiatric Demography: University of Aarhus, Aarhus, Denmark. Unpublished doctoral
dissertation, 1996.
31. Kraepelin, E., Psychiatrie, Barth, Leipzig, 1896, pp. 426441.
32. Kringlen, E. and Cramer, G., Offspring of monozygotic twins discordant for schizo-
phrenia. Archives of General Psychiatry, 1989, 46: 873877.
33. Lander, E. and Kruglyak, L., Genetic dissection of complex traits: Guidelines for
interpreting and reporting linkage results. Nature Genetics, 1995, 11: 241247.
34. Lander, E. and Schork, N.J., Genetic dissection of complex traits. Science, 1994, 265,
20372048.
35. Lewis, C.M. et al. Genome scan meta-analysis of schizophrenia and bipolar disorder,
part II: Schizophrenia. Am. J. Hum. Genet. 2003, 73: 3448.
36. Luxenburger, H., Vorlufiger Bericht ber psychiatrische Serienuntersuchungen an
Zwillingen. Zeitscrift fr die gesamte Neurologie und Psychiatrie, 1928, 116:
297326.
37. Maier, W., Lichtermann, D., Minges, J., Hallmeyer, J., Heun, R., Benkert, O., and
Levinson, D.F., Continuity and discontinuity of affective disorders and schizophrenia.
Archives of General Psychiatry, 1993, 50, 871883.
38. McGue, M., When assessing twin concordance, use the probandwise not the pairwise
rate. Schizophrenia Bulletin, 1992, 18: 171176.
39. McGue, M. and Gottesman, I.I., Genetic linkage in schizophrenia: perspectives from
genetic epidemiology. Schizophrenia Bulletin, 1989, 15: 453464.
40. McGuffin, P. and Sturt, E., Genetic markers in schizophrenia. Human Heredity, 1986,
36: 6588.
41. McGuffin, P., Owen, M.J., and Gottesman, I.I., Psychiatric Genetics and Genomics.
London and New York: Oxford University Press, 2004.
42. Mirnics, K., Middleton, F.A., Stanwood, G.D., Lewis, D.A., and Levitt, P., Disease-
specific changes in regulator of G-protein signaling 4 (RGS4) expression in schizo-
phrenia. Mol. Psychiatry, 2001, 6: 293301.
43. Moises, H.W., Yang, L., Kristbjarnarson, H., Wiese, C. et al., An international two-
stage genome-wide search for schizophrenia susceptibility genes. Nature Genetics,
1995, 11: 321324.
44. Moldin, S.O. and Gottesman, I.I., Genes, experience, and chance in schizophrenia:
Positioning for the 21st century. Schizophrenia Bulletin, 1997, 23: 547561.
45. Morton, N.E. and MacLean, C.J., Analysis of family resemblance: III. Complex
segregation analysis of quantitative traits. American Journal of Human Genetics,
1974, 26: 489503.
46. Nance, M.A., Invited editorialHuntingtons disease: Another chapter rewritten.
American Journal of Human Genetics, 1996, 59, 16.
47. Onstad, S., Skre, I., Torgersen, S., and Kringlen, E., Twin concordance for DSM-III-
R schizophrenia. Acta Psychiatrica Scandinavica, 1991, 83: 395401.
48. Owen, M.J., ODonovan, M.C., and Harrison, P.J., Schizophrenia: Disorder of the
synapse? British Medical Journal, 330: 158159, 2005.
49. Owen, M.J., Williams, N.M., and ODonovan, M.C., The molecular genetics of
schizophrenia: New findings promise new insights. Mol. Psychiatry, 2004, 9: 1427.
50. Plomin, R., Owen, M.J., and McGuffin P., The genetic basis of complex human
behaviors. Science, 1994, 264: 17331739.
Schizophrenia 225
51. Pogue-Geile, M.F. and Rose R.J. Psychopathology: A behavior genetic perspective.
In T. Jacob (Ed.), Family Interaction and Psychopathology. Plenum Press, New York,
1987, pp. 629650.
52. Pogue-Geile, M.F. (2003). Schizophrenia spectrum disorders. In Encyclopedia of the
Human Genome, D. Cooper (Ed.), Vol. 5, pp. 185189. John Wiley & Sons, Ltd.,
Chichester, UK.
53. Prescott C.A. and Gottesman, I.I., Genetically mediated vulnerability to schizophre-
nia. Psychiatric Clinics of North America, 1993, 16, 245267.
54. Risch, N., Genetic linkage and complex diseases, with special reference to psychiatric
disorders. Genetic Epidemiology, 1990, 7: 316.
55. Risch, N., Linkage strategies for genetically complex traits: I Multilocus models.
American Journal of Human Genetics, 1990, 46: 222228.
56. Risch, N. and Merikangas, K., The future of genetic studies of complex human
diseases. Science, 1996, 273, 15161517.
57. Rosenthal, D., Wender, P.H., Kety, S.S., Schulsinger, F., Welner, J., stergaard, L.,
Schizophrenicsoffspring reared in adoptive homes. In D. Rosenthal, S.S. Kety (Eds.),
The Transmission of Schizophrenia. Pergamon Press, Oxford, 1968, pp. 377391.
58. Rdin, E., Zur Vererbung und Neuentstehung der Dementia Praecox. Springer, Berlin,
1916.
59. Schizophrenia Linkage Collaborative Group for chromosomes 3,6, and 8: Additional
support for schizophrenia linkage on chromosomes 6 and 8: A multicenter study.
American Journal of Medical Genetics (Neuropsychiatric Genetics), 1996, 67,
580594.
60. Sherrington, R., Brynjolfsson, J., Petursson, H., Potter, M., Dudleston, K., Barra-
clough, B., Wasmuth, J., Dobbs, M., and Gurling, H., Localization of a susceptibility
locus for schizophrenia on chromosome 5. Nature, 1988, 336, 164167.
61. Shields J. and Gottesman I.I., Cross-national diagnosis of schizophrenia in twins.
Archives of General Psychiatry, 1972, 27, 725730.
62. Stefansson, H. et al., Neuregulin 1 and susceptibility to schizophrenia. Am. J. Hum.
Genet. 71: 877892, 2002.
63. Straub, R.E., MacLean, C.J., ONeill, F.A., Burke, J., Murphy, B., Duke, F., Shinkwin,
R., Webb, B.T., Zhang, J., Walsh, D., Kendler, K.S., A potential vulnerability locus
for schizophrenia on chromosome 6p2422: Evidence for genetic heterogeneity.
Nature Genetics, 1995, 11, 287293.
64. Straub, R.E. et al. Genetic variation in the 6p22.3 gene DTNBP1, the human ortholog
of the mouse dysbindin gene, is associated with schizophrenia. Am. J. Hum. Genet,
2002, 71: 337348.
65. Thompson, J.L., Pogue-Geile, M.F., and Grace, A.A. Developmental pathology,
dopamine, and stress: A model for the age of onset of schizophrenia symptoms.
Schizophrenia Bulletin, 2004, 30, 875900.
66. Thompson, J.L., Watson, J.R., Steinhauer, S.R., Goldstein, G., and Pogue-Geile, M.F.
Indicators of genetic liability to schizophrenia: A sibling study of neuropsychological
performance. Schizophrenia Bulletin, 2005, 31, 8596.
67. Vogler, G., Gottesman, I.I., McGue, M.K., Rao, D.C., Mixed-model segregation
analysis of schizophrenia in the Lindelius Swedish pedigrees. Behavior Genetics,
1990, 20: 461472.
68. Wahlberg, K-E., Wynne, L.C., Oja, H., Keskitalo, P., Pyklinen, L., Lahti, I., Moring,
J., Naarala, M., Sorri, A., Seitamaa, M., Lksy, K., Kolassa, J., Tienari, P., Geneenvi-
ronment interaction in vulnerability to schizophrenia: Findings from the Finnish
adoptive family study of schizophrenia. American Journal of Psychiatry, 1997, 154:
355362.
69. Weeks, D.E. and Lange, K., The affected-pedigree-member method of linkage anal-
ysis. American Journal of Human Genetics, 1988, 42: 315326.
70. Wong, A.H.C., Gottesman, I.I., and Petronis, A., Phenotypic differences in genetically
identical organisms: the epigenetic perspective. Human Molecular Genetics, 2005,
14, 1118.
71. Woolf, C.M., Does the genotype for schizophrenia often remain unexpressed because
of canalization and stochastic events during development? Psychological Medicine,
1997, 27, 659668.
16 Genetics of Major
Affective Disorders
Wade Berrettini
CONTENTS
Introduction............................................................................................................ 227
16.1 Bipolar Disorders ......................................................................................... 228
16.1.1 Genetic Epidemiology: Family Studies........................................... 228
16.1.2 Genetic Epidemiology: Twin Studies .............................................. 228
16.1.3 Molecular Linkage Studies .............................................................. 230
16.1.4 Linkage Disequilibrium Studies ...................................................... 231
16.2 Recurrent Unipolar Disorders...................................................................... 236
16.2.1 Genetic Epidemiology: Family Studies........................................... 236
16.2.2 Genetic Epidemiology: Twin Studies .............................................. 237
16.2.3 Molecular Linkage Studies .............................................................. 237
Summary ................................................................................................................ 237
References.............................................................................................................. 238
INTRODUCTION
This chapter reviews some aspects of the genetic epidemiology and molecular genetic
research on bipolar disorders (BDs) and recurrent unipolar disorders (RUPs). Genetic
concepts of linkage and linkage disequilibrium (LD) are reviewed. Given that the
inherited susceptibilities for BD and RUP are explained by multiple genes of small
effect, simulations indicate that universal confirmation of vulnerability genes cannot
be expected due to power issues, sampling variation, and genetic heterogeneity. With
this background, several valid linkages of BD to genomic regions are reviewed,
including some that may be shared with schizophrenia. These results suggest that
nosology must be changed to reflect the genetic origins of the multiple disorders
that are collectively described by the term BD. The briefer history of RUP molecular
linkage and LD studies is also reviewed.
227
16.1 BIPOLAR DISORDERS

16.1.1 GENETIC EPIDEMIOLOGY: FAMILY STUDIES
The optimal design for a BD family study is one in which relatives of controls and
relatives of BD probands are directly interviewed, and diagnoses are made by
individuals who are blind to the identity and family origin of the individual under
scrutiny. Family studies of bipolar illness show that a spectrum of mood disorders
is found among the first-degree relatives of BD probands: DDBDI, BDII with major
depression (hypomania and RUP illness in the same person), schizoaffective disor-
ders and RUP illness.112 The family studies suggest shared liability for BD and RUP
disorders.
No BD family study conducted in an optimal manner reports increased risk for
schizophrenia (SZ) among relatives of BD probands. Similarly, no SZ family study
reports increased risk for BD disorders among relatives of SZ probands; however,
several SZ family studies report increased risk for RUP and schizoaffective disor-
ders among relatives of SZ probands.1114 These family studies are consistent with
some degree of overlap in susceptibility to RUP and schizoaffective disorders for
relatives of BP probands and relatives of SZ probands. Kendler et al.14 specifically
note an increase in risk for psychotic affective disorders among the relatives of SZ
probands.
Potash et al.15,16 reported that psychotic affective disorders cluster in families.
Risk for psychotic affective disorders was significantly higher among the relatives
of psychotic BD probands compared to the risk for relatives of nonpsychotic BD
probands. This raises the possibility that the partial overlap in risk BD and SZ
nosological categories is due to a subset of BD characterized by psychotic symptoms.
This subset of BD is probably quite common, as the majority of bipolar (BP)
probands from the Potash et al.16 study were psychotic.
16.1.2 GENETIC EPIDEMIOLOGY: TWIN STUDIES

Twin studies of BD have been conducted over the past 70 years. In nearly all of
these reports, RUP illness in the cotwin of a BP index case was grounds for
categorizing the twin pair as concordant. Bertelsen et al.17 and Allan et al.18 report
that approximately 20 percent of concordant monozygotic twin pairs were comprised
of a BP index twin and a RUP cotwin. These older studies (Table 16.1) were
conducted prior to the introduction of operationalized diagnostic criteria and semi-
structured interviews. However, the older results are quite consistent with the more
recent studies,14,19 reporting significantly higher monozygotic twins concordance
rates compared with those for dizygotic twins (Figure 16.1). Recent estimates of
heritability are approximately 80%, and approximately 30% of this is shared liability
with RUP illness.19
Mendlewicz and Rainer25 reported a controlled adoption study of BD probands,
including a control group of probands with poliomyelitis. The biological relatives
of the BP probands had a 31% risk for BD or unipolar (UP) disorders, as opposed
to 2% in the relatives of the control probands. The risk for affective disorder
Genetics of Major Affective Disorders 229
TABLE 16.1
Concordance Rates for Affective Illness in Monozygotic and
Dizygotic Twins*
Monozygotic Twins Dizygotic Twins
Concordant Pairs/Total Pairs (%) Concordant Pairs/Total Pairs (%)
Study
Luxenberger20 3/4 75.0 0/13 0.0
Rosanoff et al.21 16/23 69.6 11/67 16.4
Slater22 4/7 57.1 4/17 23.5
Kallman23 25/27 92.6 13/55 23.6
Harvald and Hauge24 10/15 66.7 2/40 5.0
Allen et al.18 5/15 33.3 0/34 0.0
Bertelsen et al.17 32/55 58.3 9/52 17.3
Totals 95/146 65.0 39/278 14.0
*Data not corrected for age. Diagnoses include both bipolar and unipolar illness.
Heritability Estimates: BPD Twin Studies

N=34
80
70
CONCORDANCE
N=30
60
50
40 1977
1977
30 2003
N=37
20 N=37
10
0
MZ DZ
FIGURE 16.1 Two BD twin studies, conducted 25 years apart, are compared for concor-
dance rates in monozygotic (MZ) and dizygotic (DZ) twins. These studies yield very similar
results, including estimates of heritability of ~80%. Twins were clinically ascertained
through a BPD proband. Co-twins were concordant if BPD or RUP was present. From
McGuffin P., et al. Arch Gen Psychiatry. 2003;60:497502, and Bertelsen A, et al. Br J
Psychiatry. 1977;130:330351.
in biological relatives of adopted BD patients was similar to the risk in relatives of

BD patients who were not adopted (26%). Adoptive relatives do not show increased
risk compared with relatives of control probands.
Wender et al.26 and Cadoret et al.27 studied UP and BD probands. Although
evidence for genetic susceptibility was found, adoptive relatives of affective probands
had a tendency to excess affective illness themselves, compared with the adoptive
relatives of controls. Von Knorring et al.28 did not find concordance in psychopa-
thology between adoptees and biological relatives when examining the records of
56 adoptees with UP disorders.
16.1.3 MOLECULAR LINKAGE STUDIES

Linkage refers to the observation that two DNA sequences found near each other
on the same chromosome tend to be inherited together more often than expected by
chance within families. Such DNA sequences are said to be linked. The log of odds
(LOD) score refers to the probability that observed cosegregation of alleles at distinct
DNA sequences within a family has occurred because the two DNA sequences are
linked. An LOD score greater than 3 is evidence, but not proof, that two DNA
sequences are linked. The numerical value of the LOD score is dependent on the
proposed mode of inheritance (dominant, recessive, sex-linked) and penetrance.
Because the LOD score is dependent on these parameters (mode of inheritance and
penetrance), it is sometimes termed a parametric statistic. This dependence on
inheritance mode and penetrance distinguishes the LOD score from nonparametric
statistics (including affected sibling pair and affected pedigree member methods),
because such statistics are not dependent on mode of inheritance or penetrance.
These nonparametric statistics use kinship coefficients to estimate the randomly
expected degree of DNA sharing among affected members of the same family. For
example, siblings share 50% of DNA sequences randomly because they have the
same parents. If one has DNA samples from 1,000 pairs of ill siblings, one could
search through the genomes of those 2,000 persons to find regions where DNA
sequences that are shared are significantly greater than the baseline 50% rate.
Regions of increased DNA sharing may harbor genes that may explain in part why
both members of each sibling pair are affected.
What level of statistical significance should be required for declaring linkage?
A recommended level of statistical significance for an initial report (p < ~0.00002)
is a stringent criterion, based on simulations that indicate this level of significance
would occur less than five times randomly in 100 genome scans for linkage.29 This
statistical criterion assumes that all the genetic information within the pedigrees
studied would be extracted, an assumption that is not true in practice. Typically, no
more than approximately 80% of the genetic information in a pedigree series is
extracted through genotyping; however, as in any area of science, no single report
of linkage should be accepted as valid without independent confirmation. The
requirement for independent confirmations (at p < 0.01) is not waived, no matter
the level of statistical significance achieved in a single report. This confirmation
requirement should be seen within the context that valid linkages will not be con-
firmed in some studies. Indeed, nonconfirmations should be expected, intuitively,
because of population (ethnic) differences, sampling procedures and genetic heter-
ogeneity.
The probability of confirmation in simulations has been examined by Suarez
et al.,30 who simulated a disorder caused by any one of six loci and determined that
a large sample size and substantial time will be required for an initial linkage to be
confirmed in a second sample. It is clear from Suarezs simulations that consistent
detection of a locus of moderate effect cannot be expected. Nonconfirmatory studies
will always occur when an initially detected linkage is valid.
One of the most critical issues in confirmation of reported linkages is power.
Attempts at confirmation of a reported susceptibility locus should state what power
has been achieved to detect the locus initially described. For example, if a locus
increases risk for BP illness by a factor of two, it may be necessary to study
approximately 200 affected sibling pairs in order to have adequate (90%) power to
detect such a locus.31
Unfortunately, few studies address this key issue. If 200 affected sibling pairs
are required to achieve adequate (90%) power to detect a previously described locus,
then a publication with less than 150 sibling pairs does not address the central issue
of confirmation. However, such power-limited publications may have an important
role in meta-analyses, in that they identify invaluable sources of additional data.
Comprehensive scans of the human genome have been completed with sufficient
numbers (e.g., >100) of BP individuals.3240 If a major locus (explaining >50% of
the risk in >50% of BPD persons) existed, it would have been detected in many of
these studies. Thus, no such major locus exists for BPD. There are several confirmed
reports of loci of smaller effect, which can be termed susceptibility loci. These loci
are neither necessary nor sufficient for disease, but increase risk for the disorder in
a non-Mendelian manner.
From these genome scans and from additional, smaller studies, a picture has
emerged in which there are approximately 10 confirmed DBD linkage regions across
the genome. It is highly probable that additional confirmed BD linkages will be
identified through future linkage studies. These BP linkage regions are confirmed
by virtue of at least one study with strong statistical significance (p < 0.0001) and
at least two confirmatory studies (p < 0.01). As noted elsewhere,41 in some cases,
these confirmed BP linkage regions overlap with schizophrenia linkage reports,
suggesting that the same loci may be involved in some aspects of both disorders. In
Figure 16.2, these are mapped onto an ideogram of the human genome.
The studies which support these findings are listed in Table 16.2, with the iden-
tified primary report cited as the first study with genome-wide statistical significance.
Two methodological approaches have been used to conduct meta-analyses of
BP linkage studies.73,74 Badner and Gershon73 analyzed linkage results using a
multiple scan probability approach, in which p values are combined across studies,
after adjusting for the size of the linkage region. These authors concluded that
two genomic regions, 13q32 and 22q,1113 were the most promising loci for DBD
disorder. Segurado et al.75 used the method of Levinson et al.74 which ranks the
p values across the genome of each study, then sums the rankings for each genomic
bin. In this approach, no genomic region reached genome-wide significance,
although the region that seemed most promising was the pericentromeric region
of 18.75
16.1.4 LINKAGE DISEQUILIBRIUM STUDIES

The human genome consists of approximately three billion base pairs of DNA.76
The recent completion of draft genomic sequences of the human genome76 is
consistent with approximately 35,000 to 40,000 genes. Physical distance along the
linear sequence of DNA can be expressed in terms of base pairs of DNA. The
most common sequence variation in the human genome is a single nucleotide
polymorphism (SNP), where at the same position on different chromosomes there
FIGURE 16.2 (SEE COLOR INSERT FOLLOWING PAGE 236) Confirmed linkage loci
for bipolar disorder (B), schizophrenia (S), and both disorders (*) on a diagram of the human
genome.
are two different nucleotides (from the possible four found among homo sapiens:
adenine [A], guanine [G], thymidine [T], and cytosine [C]). SNPs with a common
minor allele (frequency of approximately 20% or more) occur approximately every
1,000 base pairs of DNA.76 Analysis of closely spaced SNPs in outbred populations
suggests a complex pattern of inheritance in which recombination is inhibited in
a small region of DNA, such that blocks of DNA (containing multiple SNPs) tend
to be inherited intact over many generations.77 Thus, blocks of DNA are shared
among present-day individuals who may have had a common ancestor 10,000
generations ago. These blocks are variable in length and often contain multiple
SNPs, but among outbred human populations the block length rarely exceeds
approximately 100,000 base pairs. Alleles of SNPs within a block form a haplotype
(a set of alleles) that is usually inherited together across many generations. Such
SNPs are said to be in strong LD with one another. LD refers to the fact that
two (or more) alleles can be found together in unrelated individuals more often
than predicted by chance. The interested reader is referred to primary reports
concerning LD.77
LD is a useful tool to investigate the relatively small genomic regions that
have been implicated in the genetic origination of DBD through linkage studies
(Table 16.2). In this process, SNPs spaced across genes in the linkage region are
assessed in large groups (ideally at least several hundreds) of ethnically matched
TABLE 16.2
Confirmed Linkage Regions in Bipolar Disorder
Location Primary Report Independent Confirmations Comments
18p11 Berrettini et al., Stine et al., 1995; Nothen et al., 1999; Paternal parent-of-
1994 and 1997 Turecki et al., 1999; Bennett et al., 2002 origin effect; see
Schwab et al., 1998;
Lin and Bale, 1997
21q22 Straub et al., 1994 Detera-Wadleigh et al., 1996; Smyth
et al., 1996; Kwok et al., 1999;
Morissette et al., 1999; Aita et al., 1999
22q11 Kelsoe et al., 2001 Detera-Wadleigh et al., 1999; Lachman Velocardiofacial
et al., 1997 syndrome region;
also a SZ locus: Gill
et al., 1996
18q22 Stine et al., 1995 McInnes et al., 1996; McMahon et al., See Freimer et al.,
1997; De Bruyn et al., 1996; McInnis 1996
et al., 2003
12q23 Morissette et al., Ewald et al., 2002; Maziade et al., 2002; Primary report in a
1999 Ekholm et al., 2003; Curtis et al., 2003; Canadian isolate;
Dawson et al., 1995 Abkevich et al., 2003
8q24 Cichon et al., 2001 McInnis et al., 2003; Dick et al.,
2003
13q32 Detera-Wadleigh Kelsoe et al., 2001; Potash et al., 2003; See Brzustowicz et al.,
et al., 1999 Liu et al., 2004; Badenohop et al., 2001 1999; Blouin et al.,
1998; Chumakov
et al., 2002
16p12 Ewald et al., 2002 Ekholm et al., 2003; Dick et al., 2003
4q32 Ekholm et al., 2003 Adams et al., 1998; McInnis et al., 2003;
Liu et al., 2004
4p15 Blackwood et al., Ewald et al., 2002; Cichon et al., 2001; See Ginns et al., 1998
1996 Morissette et al., 1999; Detera-
Wadleigh et al., 1999
cases and controls. Investigators compare allele and genotype frequencies among
groups of cases and controls.
There have been a multitude of LD studies in BD over the past decade. In a
typical report, allele and genotype frequencies in the BD cases and controls are
examined at a single candidate gene variant. If nominally significant differences in
allele or genotype frequencies are found between groups, the authors might conclude
that the variant influences the risk for BD disorder.
Most often, these studies have assembled a small group of BD patients and
unaffected controls from a population. These studies have typically employed one
variant in a single candidate gene which is selected based on presumed central

nervous system (CNS) function, in relation to BD pathophysiologic theories. Unfor-
tunately, the nearly complete absence of pathophysiologic data in BDBD makes the
process of rational candidate gene selection difficult.
Additionally, these studies have typically involved smaller numbers of BD
patients than is optimal, given that the effect size of individual alleles on risk must
be small. Lastly, these studies have often used gene variants which are not known
to confer functional differences in the gene. Despite these difficulties, there are
several candidate genes which deserve mention.
One promising candidate gene is the G72 locus on 13q32, the site of a confirmed
linkage in BD and SZ (Table 16.2). G72 is a primate-specific brain-expressed gene
that activates D-amino acid oxidase.69 D-amino acid oxidase may control levels of
D-serine, which regulates glutamatergic receptors.78 Chumakov et al.69 identified a
haplotype from G72 SNPs (without obvious functional significance), which were in
LD with SZ in a French-Canadian sample. This has been confirmed in distinct SZ
populations, including Russian69 and German79 populations, although different hap-
lotypes have been associated in different ethnic populations. Similarly, in BD there
have been several positive findings with distinct haplotypes in different populations,
including American80,81 and German79 BD samples. Although the data is promising,
no clear functional variants have been defined at this locus.
A second promising candidate gene is brain-derived neurotropic factor (BDNF),
for which there are several positive reports at a functional mis-sense variant.82,83 In
the several European-origin populations studied at a G/A SNP (Val/Met), the G (Val)
allele was over-transmitted. Egan et al.84 demonstrated that this SNP confers func-
tional difference.
These reports are promising because they consistently identify a functional
variant contributing to genetic risk for DBD. However, there are small studies of
Japanese and Chinese BD patients that do not show any evidence of this variant
influencing risk for BD disorder.85,86 The effect may be limited to populations
of European origins. Alternatively, the negative studies may have been under-
powered. It must be remembered that some variants will be relatively specific to
particular ethnic groups. Consider that the protective effect of aldehyde dehydroge-
nase deficiency on risk for alcoholism is easily demonstrated in Chinese, Korean,
and Japanese populations because the deficiency allele has a frequency of approx-
imately 30%.87,88 Much larger sample sizes are required to detect this influence in
European populations because the deficiency allele frequency is lower by an order
of magnitude.
There have been numerous independent-association studies of BD and RUP
and an MAOA (CA) repeat polymorphism in European89,94 and Asian95,96 popula-
tions. Those studies reporting a positive association90,9395 generally detect an over-
representation of allele 5 or 6 of the MAOA (CA) n repeat among BD patients
compared with controls, which is an observation that may be particularly evident
with respect to women. The effect size is small, the odds ratio being 1.49,94 and the
sample size required for adequate power to detect is larger than most of the negative
studies.89,91,92,96,97 There is also an MAOA promoter polymorphism.98 These studies
involve multiple ethnic groups, case-control methods, and family-based designs,
with some studies having limited power to detect a small effect size. Thus, it is
understandable that conflicting studies are reported.
Another intensively studied candidate gene is the serotonin transporter (5HTT),
a functional candidate gene for which multiple BDLD studies have been published.
The 5HTT represents a logical candidate gene, as many antidepressants act through
binding to the 5HTT protein.99 There are two variants of the 5HTT that have been
studied in BD, and both have functional significance, based on in vitro analysis of
these noncoding polymorphisms. The first variant is an insertion/deletion polymor-
phism in the 5HTT promoter region. The shorter allele has much less transcriptional
activity than the longer allele.100 Moreover, the shorter allele has been associated
with anxiety-related personality traits in humans.101 The second variant is a variable
number of tandem repeats (VNTRs) polymorphism in intron 2. The two most
common alleles are the 10 and 12 repeats, which confer differential transcriptional
activity in an embryonic stem cell line.102 Collier et al.100 first reported that the 5HTT
intron 2 VNTR allele 12 was in LD with BD among patients from the U.K. Collier
et al. also reported that the short allele of the 5HTT promoter variant was more
common among 454 European BD and RUP patients, compared with 570 European
controls, although the statistical significance was marginal (p = 0.03), emphasizing
the small effect size involved. Analysis by genotype suggested that homozygosity
for the short allele was associated with BD (p < 0.05) and RUP ( p < 0.01).
Rees et al.104 confirmed the observation of Collier et al.100 that allele 12 of the
intron 2 VNTR was in LD with BD among 171 BD probands, compared with 121
controls (p = 0.031). Similarly, Rees et al. studied BD patients and controls, reporting
an excess of BD patients among individuals homozygous for the shorter promoter
allele, implying a recessive mode on inheritance. Note that the sample sizes for Rees
et al.104 and for Collier et al.103 were in the hundreds.
Vincent et al.105 studied an initial sample of approximately 100 BD probands
from Canada, confirming the observation that the promoter short allele was in LD
with BD, compared with approximately 100 controls; however, he then failed to
confirm this observation in a second set of approximately 100 BD probands. Sam-
pling variation and the small effect size, coupled with limited power of this sample
size, are probable explanations for these results.
Gutierrez et al.106 studied BD probands and controls from Spanish origin. They
reported no evidence for LD with 5HTT alleles. This may be secondary to the ethnic
background of patients or to small sample size. Bocchetta et al.107 studied approxi-
mately 55 Sardinian parentchild BD trios, finding no evidence for transmission
disequilibrium in the 5HTT gene, although sample size was a limiting factor in their
conclusions. Studying 123 BD parentchild trios of European origin, Mundo et al.108
reported no evidence for LD with the 5HTT promoter alleles. Mynett-Johnson
et al.109 studied approximately 100 Irish BD parentchild trios from multiplex fam-
ilies, reporting that a haplotype including the shorter promoter allele and a 3'UTR
SNP conferred risk for BD.
Kirov et al.110 studied 122 parentchild trios of British ethnic background, with
no nominally significant results at either polymorphism. In a study of 50 Indian BD
patients and controls using the VNTR variant, no evidence for LD was reported,111
this result being limited by the small sample size. From another ethnic perspective,
Mendes de Oliveira et al.112 studied a small number of Brazilian BD patients, finding

no evidence for LD with the 5HTT promoter polymorphism. Kunugi et al.113 studied
these two polymorphisms in a Japanese sample of 191 patients with affective dis-
orders (142 bipolar and 49 unipolar) and 212 controls. They report nominally
significant LD between the VNTR and BD, with no evidence for LD with the
promoter variant.
Furlong et al.114 reported results of a meta-analysis for approximately 1,400
individuals of European origin, including 772 controls, 375 bipolar patients and 299
unipolar patients. Although there was no evidence for LD with affective disorders
for the VNTR, a marginally significant result was found for the short allele of the
5HTT promoter polymorphism. This result is important because it suggests that
samples in the thousands will be necessary to draw firm conclusions, due to the
small effect sizes involved.
What clinical characteristics might describe those BD probands with illness
secondary to 5HTT alleles? One preliminary answer is derived from two studies
on independent populations. Coyle et al.115 studied BD women with post-partum
psychotic episodes, demonstrating a large effect size for the 5HTT promoter short
allele, with an attributable risk of 69%. Ospina-Duque et al.116 studied 100 BD
patients from a Columbian population isolate and approximately 100 controls, show-
ing that those patients with psychosis as part of the phenotype showed LD with the
shorter 5HTT promoter allele. Their conclusions may be similar to those of Coyle
et al.115 and suggest that this deserves further study.
In another large European study, Mendlewicz et al.117 examined the genetic
contribution of the 5-HTT promoter polymorphism in a case-control sample, includ-
ing 539 RUP patients, 572 BD patients, and 821 controls. No evidence of LD was
found for RUP or BD, and subdividing the sample according to family history,
suicidal attempts, or psychotic features did not reveal any role of the promoter variant
in the genetic susceptibilities to these disorders.
16.2 RECURRENT UNIPOLAR DISORDERS

16.2.1 GENETIC EPIDEMIOLOGY: FAMILY STUDIES
In considering optimally designed family studies of RUP (with blinded interviews
and simultaneous examination of control relatives), there are five reports in the
literature.1,2,10,12,118 These five reports yield remarkably similar results, each study
concluding that the first-degree relatives of RUP probands were at increased risk for
RUP disorders, compared to first-degree relatives of control probands. Across the
five studies, there was a two- to fourfold increased risk for RUP among the first-
degree relatives of RUP probands.
Characteristics of RUP disorders that yield a more heritable phenotype include
early onset (e.g., before age 30)27,118122 and a high degree of recurrence.13,121,124126
A third characteristic that may identify a separate group of disorders is the presence
of psychosis.14 Additional genetic subtypes of RUP may be identified through exam-
ination of comorbidities with panic disorder and other anxiety disorders, as well as
with alcoholism.127129
16.2.2 GENETIC EPIDEMIOLOGY: TWIN STUDIES

A recent review of twin studies in RUP disorder estimated heritability at 37%, with
a substantial component of unique individual environmental risk but little shared
environmental risk.130 These twin studies included four community-ascertained
samples131134 and two clinically ascertained samples, one from the United
Kingdom135 and one from Sweden.132 The results are quite consistent in concluding
that genetic influence is a significant factor in risk for RUP, independent of ascer-
tainment and country or origin.
16.2.3 MOLECULAR LINKAGE STUDIES

There have been just several RUP genome scans with greater than approximately
100 affected individuals, in contrast to BD disorder. Holmans et al.136 reported on
the first phase of a multisite collaborative effort. The sample consisted of 297
informative multiplex families (containing 685 informative affected relative pairs,
555 sibling pairs, and 130 other pair types). Affected cases had RUP with onset
before age 31 for probands or 41 for other affected relatives; the mean age at onset
was 18.5 and the mean number of depressive episodes was 7.3, indicating a highly
recurrent form of illness. Families were excluded if there was a BD first-degree or
second-degree relative. Linkage was observed on chromosome 15q25.326.2 (empir-
ical genome-wide p = 0.023). The linkage was not sex-specific. This was the sole
significant linkage peak observed by this group.
Abkevich et al.65 reported a genome scan on 110 Utah pedigrees (each with at
least four affected individuals), in which there were 784 RUP individuals, 161
persons with single-episode major depressive disorder and 162 BD individuals, who
were also considered affected. They observed a highly significant linkage signal at
12q23 (p = 0.0000007), confirming a previously identified DBD locus (Table 16.2).
There were no other linkage peaks approaching statistical significance. It is probable
that this study has detected the same BD 12q23 locuseven though the families
were ascertained from a RUP proband because most kindreds probably had at least
one BD individual. These results confirm family and twin studies, suggesting genetic
overlap between BD and RUP disorders, and this study identifies the 12q23 region
as a locus which increases risk for both BD and RUP disorders.
Zubenko et al.137 reported on a genome scan of 81 families ascertained through
a proband with early-onset, nonpsychotic RUP disorder. They describe a highly
significant linkage (p < 0.0001) of this phenotype to 2q35 near marker D2321, which
is near a candidate gene, CREB1 (cyclic AMP response element binding protein 1).
Sequence variants in the CREB gene were found to segregate with RUP disorder
among women in 2 of these 81 extended kindreds,138 thus nominating CREB as a
RUP susceptibility gene. These intriguing results await independent confirmation.
SUMMARY
Family, twin, and adoption studies of BD and RUP disorders are reviewed. They
are, in general, consistent with substantial heritable components to risk, with the
BD disorders having higher heritability than the RUP disorders. Multiple regions of
the genome, including 18p11, 18q22, 12q24, 21q21, 13q32, 4p15, 4q32, 16p12,
8q24, and 22q11, have been implicated by several independent groups in the genetic
origins of BD. It is likely that most of these regions will yield susceptibility genes
within the next 5 years through the application of LD mapping methods to large
sample sizes. LD approaches to candidate genes have yielded several promising
prospects, including G72 and BDNF for DBD. For each of these candidate genes,
there are several independent BD populations yielding data consistent with the
existence of one or more haplotypes as susceptibility sequences. Only several
genome scans for RUP disorders have been published, and confirmations are
required.
REFERENCES
1. Gershon, ES, Hamovit, J, Guroff, JJ, et al. A family study of schizoaffective, bipolar
I, bipolar II, unipolar, and normal control probands. Arch Gen Psychiatry
1982;39:115767.
2. Weissman, MM, Gershon, ES, Kidd, KK, et al. Psychiatric disorders in the relatives
of probands with affective disorder. Arch Gen Psychiatry 1984;41:1321.
3. Baron, M, Gruen, R, Anis, L, Kane, J. Schizoaffective illness, schizophrenia and
affective disorders: morbidity risk and genetic transmission. Acta Psychiatrica Scand
1983;65:25362.
4. Winokur, G, Tsuang, MT, Crowe, RR. The Iowa 500: Affective disorder in relatives
of manic and depressed patients. Am J Psychiatry 1982;139:20912.
5. Winokur G, Coryell, Keller, M, et al. A family study of manic-depressive (bipolar I)
disease. Is it a distinct illness separable from primary unipolar depression? Arch Gen
Psychiatry 1995;52:36773.
6. Helzer, JE and Winokur, G. A family interview study of male manic-depressives.
Arch Gen Psychiatry 1974;31,737.
7. James, NM and Chapman, CJ. A genetic study of bipolar affective disorder. Br J
Psychiatry 1975;126:44956.
8. Johnson, GFS and Leeman, MM. Analysis of familial factors in bipolar affective
illness. Arch Gen Psychiatry 1977;34:107483.
9. Angst, J, Frey, R, Lohmeyer, R, and Zerben-Rubin, E. Bipolar manic depressive
psychoses: results of a genetic investigation. Hum Genet 1980;55:23754.
10. Tsuang, MT, Winokur, G and Crowe, RR. Morbidity risks of schizophrenia and
affective disorders among first-degree relatives of patients with schizophrenia, mania,
depression, and surgical conditions. Br J Psychiatry 1980;137:497504.
11. Taylor, MA, Berenbaum, SA, Jampala, VC and Cloninger, CR. Are schizophrenia
and affective disorder related? Preliminary data from a family study. Am J Psychiatry
1993;150:27885.
12. Maier, W, Lichtermann, D, Minges, J, et al. Continuity and discontinuity of affective
disorders and schizophrenia. Results of a controlled family study. Arch Gen Psychiatry
1993;50:87183.
13. Gershon, ES, DeLisi, LE, Hamovit, J, et al. A controlled family study of chronic
psychoses. Arch Gen Psychiatry 1988;45:32836.
14. Kendler, KS, McGuire, M, Gruenberg, AM, et al. The Roscommon family study.
Arch Gen Psychiatry 1993;50:52740.
15. Potash, JB, Willour, VL, Chiu, YF, et al. The familial aggregation of psychotic
symptoms in bipolar disorder pedigrees. Am J Psychiatry 2001;158:125864.
16. Potash, JB, Chiu, Y-F, MacKinnon, DF, et al. Familial aggregation of psychosis in a
replication set of 69 bipolar pedigrees: Am J Med Genet 2003;116B:907.
17. Bertelsen, A, Harvald, B, and Hauge, M. A Danish twin study of manic-depressive
disorders. Br J Psychiatry 1977;130:33051.
18. Allen, MG, Cohen, S, Pollin, W, Greenspan, SI. Affective illness in veteran twins.
A diagnostic review. Am J Psychiatry 1974;131:12349.
19. McGuffin, P, Rijsdijk, S, Andrew, M, et al. The heritability of bipolar affective
disorder and the genetic relationship to unipolar depression. Arch Gen Psychiatry
2003;60: 497502.
20. Luxenberger, H. Psychiatrisch-neurologische Zwillings pathologie. Zentralblatt fur
diagesamte Neurologie and Psychiatrie 1930;14, 567,14580.
21. Rosanoff, AJ, Handy, L, and Plesset, IR. The etiology of manic-depressive syndromes
with special reference to their occurrence in twins. Am J Psychiatry 1935;91:72562.
22. Slater, E. The inheritance of manic-depressive insanity. Proceedings of the Royal
Society of Medicine 1953;29:98190.
23. Kallman, F. In: Hoch, PH and Zubin, J(eds). Depression. New York:Grune & Strat-
ton;1954:124.
24. Harvald, B, Hauge, M. Biopolar disorders, in, Genetics and the Epidemiology of
Chronic Diseases. (PHS Publ No 1163). Edited by Neal, JV, Shaw, MW, and Shull,
WJ. Washington, DC, U.S. Department of Health, Education, and Welfare, 1965 pp
6176.
25. Mendlewicz, J and Rainer, J.D. Adoption study supporting genetic transmission in
manic-depressive illness. Nature 1977;268:3279.
26. Wender, H, Kety, SS, Rosenthal, D, et al. Psychiatric disorders in the biological and
adoptive families of adopted individuals with affective disorders. Arch Gen Psychiatry
1986;43:9239.
27. Cadoret, RJ, Woolson, R, and Winokur, G. The relationship of age of onset in unipolar
affective illness to risk of alcoholism and depression in parents. J Psychiatr Res
1977;13:13742.
28. Von Knorring, AL, Cloninger, CR, Bohman, M, and Sigvardsson, A. An adoption study
of depressive disorders and substance abuse. Arch Gen Psychiatry 1983;40:94350.
29. Lander, E, and Kruglyak, L. Genetic dissection of complex traits: Guidelines for
interpreting and reporting linkage results. Nature Genet 1995;11:2417.
30. Suarez, B, Harpe, CL, and Van Eerdewegh, P. Problems of replicating linkage claims
in psychiatry. In: Gershon ES, Cloninger CR (eds). Genetic Approaches to Mental
Disorders. Proceedings of the 82nd Annual Meeting of the American Psychopatho-
logical Assn., Washington (DC): American Psychiatric Press;1994:2346.
31. Hauser, ER, Boehnke, M, Guo, SW, and Risch, N. Affected sib pair interval mapping
and exclusion for complex genetic traits. Genetic Epidemiology 1996;13:11737.
32. Detera-Wadleigh, SD, Badner JA, Berrettini, WH, et al. A highdensity genome scan
detects evidence for a bipolar-disorder susceptibility locus on 13q32 and other poten-
tial loci on 9q32 and 18p11.2. Proc Natl Acad Sci USA 1999;96:560456099.
33. Dick, DM, Foroud, T, Flury, L, et al. Genome-wide linkage analyses of bipolar
disorder: a new sample of 250 NIMH Genetics Initiative pedigrees. Am J Hum Genet
2003;73:10714.
34. McInnis, MG, Lan, T-H, Willour, VL, et al. Genome-wide scan of bipolar disorder
in 65 pedigrees: supportive evidence for linkage on 8q24, 18q22, 4q32, 2p12, and
13q12. Mol Psychiatry 2003;8:28898.
35. Bennett, P, Segurado, R, Jones, I, et al. The Wellcome trust UK-Irish bipolar affective
disorder sibling pair genome screen: First stage report. Mol Psychiatry 2002;7: 189200.
36. Rice, JP, Goate, A, Williams, JT, et al: Initial genome scan of the NIMH Genetics
Initiative bipolar pedigrees: Chromosomes 1, 6, 8, 10, and 12. Am J Med Genet
1997;74:24753.
37. Ekholm, JM, Kieseppa, T, Hiekkalinna, T, et al. Evidence of susceptibility loci on
4q32 and 16p12 for bipolar disorder. Hum Mol Genet 2003;12:190715.
38. Cichon, S, Schumacher, J, Muller, DJ, et al. A genome screen for genes predisposing
to bipolar affective disorder detects a new susceptibility locus on 8q. Hum Mol Genet
2001;10:293344.
39. Kelsoe, JR, Spence, MA, Loetscher, E, et al. A genome survey indicates a suscepti-
bility locus for bipolar disorder on chromosome 22. Proc Natl Acad Sci
2001;98:58590.
40. Liu, J, Juo, SH, Dewan, A, et al. Evidence for a putative bipolar disorder locus on
2p13-16 and other potential loci on 4q31, 7q34, 8q13, 9q31, 10q21 24, 13q32, 14q21
and 17q11-12. Molecular Psychiatry 2003;8:33342.
41. Berrettini, WH: Evidence for shared susceptibility in bipolar disorder and schizo-
phrenia. Am J Med Genet 2003;123C:5964.
42. Berrettini, W, Ferraro, T, Choi, H, et al. Linkage studies of bipolar illness. Arch Gen
Psych 1997;54:329.
43. Stine, OC, Xu, J, Koskela, R, et al. Evidence for linkage of bipolar disorder to
chromosome 18 with a parent-of-origin effect. Am J Hum Genet 1995;57:138494.
44. Nothen, MM, Cichon, S, Rohleder, H, et al. Evaluation of linkage of bipolar affective
disorder to chromosome 18 in a sample of 57 German families. Mol Psychiatry
1999;4:76.
45. Turecki, G, Grof, P, Cavazzoni, P, et al. Lithium responsive bipolar disorder, unilin-
eality and chromosome 18: A linkage study. Am J Med Genet 1999;88:41115.
46. Schwab, SG, Hallmayer, J, Lerer, B, et al. Support for a chromosome 18p locus
conferring susceptibility to functional psychoses in families with schizophrenia, by
association and linkage analysis. Am J Hum Genet 1998;63:1139.
47. Lin, JP, and Bale, SJ. Parental transmission and D18S37 allele sharing in bipolar
affective disorder. Genetic Epidemiology 1997;14:6658.
48. Straub, RE, Lehner, T, Luo, Y, et al. A possible vulnerability locus for bipolar affective
disorder on chromosome 21q22.3. Nature Genet 1994;8:2916.
49. Detera-Wadleigh, SD, Badner, JA, Goldin, LR, et al. Affected sib-pair analyses reveal
support of prior evidence for a susceptibility locus for bipolar disorder on 21q. Am
J Hum Genet 1996;58:127985.
50. Smyth, C, Kalsi, G, Brynjolfsson, J, et al. Further tests for linkage of bipolar affective
disorder to the tyrosine hydroxylase gene locus on chromosome 11p15 in a new series
of multiplex British affective disorder pedigrees. Am J Psychiatry 1996;153:2714.
51. Kwok, JB, Adams, LJ, Salmon, JA, et al. Non-parametric simulation-based statistical
analyses for bipolar affective disorder locus on chromosome 21q22.3. Am J Med
Genet 1996;88:99102.
52. Morissette, J, Villeneuve, A, Bordeleau, L, et al. Genome-wide search for linkage of
bipolar affective disorders in a very large pedigree derived from a homogeneous
population in Quebec points to a locus of major effect on chromosome 12q23-q24.
Am J Med Genet 1999;88:56787.
53. Aita, VM, Liu, J, Knowles, JA, et al. A comprehensive linkage analysis of chromo-
some 21q22 supports prior evidence for a putative bipolar affective disorder locus.
Am J Hum Genet 1999;64:21017.
54. Detera-Wadleigh, SD, Badner, JA, Berrettini, WH, et al. A high density genome scan
detects evidence for a bipolar-disorder susceptibility locus on 13q32 and other poten-
tial loci on 1q32 and 18p11.2. Proc Natl Acad Sci USA 1999;96:56049.
55. Lachman, HM, Kelsoe, JR, Remick, RA, et al. Linkage studies support a possible
locus for bipolar disorder in the velocardiofacial syndrome region on chromosome
22. Am J Med Genet 1997;74:1218.
56. Gill, M, Vallada, H, Collier, D, et al. A combined analysis of D22S278 marker alleles
in affected sib-pairs: support for a susceptibility locus for schizophrenia at chromo-
some 22q12. Am J Med Genet 1996;67:405.
57. McInnes, LA, Escamilla, MA, Service, SK, et al. A complete genome screen for
genes predisposing to severe bipolar disorder in two Costa Rican pedigrees. PNAS
1996;93:130605.
58. McMahon, FJ, Hopkins, PJ, Xu, J, et al. Linkage of bipolar affective disorder to
chromosome 18 markers in a new pedigree series. Am J of Hum Genet 1997;61:
1397404.
59. De Bruyn, A, Souery, D, Mendelbaum, K, et al. Linkage analysis of families with
bipolar illness and chromosome 18 markers. Biol Psychiat 1996;39:67988.
60. Freimer, NB, Reus, VI, Escamilla, MA, et al. Genetic mapping using haplotype,
association and linkage methods suggests a locus for severe bipolar disorder (BPI)
at 18q22-q23. Nature Genet 1996;12:43641.
61. Ewald, H, Flint, T, Kruse, TA, et al. A genome-wide scan shows significant linkage
between bipolar disorder and chromosome 12q24.3 and suggestive linkage to chro-
mosomes 1p22-21, 4p16, 6q14-22, 10q26 and 16p13.3. Mol Psychiatry 2002;7:
73444.
62. Maziade, M, Roy, MA, Rouillard, E, et al. A search for specific and common
susceptibility loci for schizophrenia and bipolar disorder: a linkage study in 13 target
chromosomes. Mol Psychiatry 2002;6:68493.
63. Curtis, D, Kalsi, G, Brynjolfsson, J, et al. Genome scan of pedigrees multiply affected
with bipolar disorder provides further support for the presence of a susceptibility
locus on chromosome 12q23-q24, and suggests the presence of additional loci on 1p
and 1q. Psychiatr Genet 2003;13:7784.
64. Dawson, E, Parfitt, E, Roberts, Q, et al. Linkage studies of bipolar disorder in the
region of the DarierAEs disease gene on chromosome 12q23-24.1. Am J Med Genet
1995;60:94102.
65. Abkevich, V, Camp, NJ, Hensel, CH, et al. Predisposition locus for major depression
at chromosome 12q22-12q23.2. Am J Hum Genet 2003;73:127181.
66. Badenohop, RF, Moses, MJ, Scimone, A, et al. A genome screen of a large bipolar
affective disorder pedigree supports evidence for a susceptibility locus on chromo-
some 13q. Mol Psychiatry 2001;6:396.
67. Brzustowicz, LM, Honer, WG, Chow, EWC, et al. Linkage of familial schizophrenia
to chromosome 13q32. Am J Hum Genet 1999;65:1096103.
68. Blouin, JL, Dombroski, B, Nath, SK, et al. Schizophrenia susceptibility loci on
chromosomes 13q32 and 8p21. Nat Genet 1998;20:70.
69. Chumakov, I, Blumenfeld, M, Guerassimenko, O, et al. Genetic and physiological
data implicating the new human gene G72 and the gene for D-amino acid oxidase in
schizophrenia. PNAS 2002;99:133657.
70. Adams, LJ, Mitchell, PB, Fielder, SL, et al. A susceptibility locus for bipolar affective
disorder on chromosome 4q35. Am J Hum Genet 1998;62:108491.
71. Blackwood, DHR, He, L, Morris, SW, et al. A locus for bipolar affective disorder on
chromosome 4p. Nature Genet 1996;12:42730.
72. Ginns, EI, St Jean, P, Philibert, RA, et al. A genome search for chromosomal loci
linked to mental health wellness in relatives at high risk for bipolar affective disorder
among the Old Order Amish. PNAS 1998;95:15531.
73. Badner, JA, and Gershon, ES. Meta-analysis of whole-genome linkage scans of
bipolar disorder and schizophrenia. Mol Psychiatry 2002;7:40511.
74. Levinson, DF, Levinson, MD, Sequardo, R, and Lewis, CM. Genome scan meta-
analysis of schizophrenia and bipolar disorder, part I: Methods and power analysis.
Am J Hum Genet 2003;73:1733.
75. Segurado, R, Detera-Wadleigh, SD, Levinson, DF, et al. Genome scan meta-analysis
of schizophrenia and bipolar disorder part III: bipolar disorder. Am J Hum Genet
2003;3:4962.
76. Venter, JC, Adams, MD, Myers, EW, et al. The sequence of the human genome.
Science 2001;291: 130451.
77. Gabriel, SB, Schaffner, SF, Nguyen, H, et al. The structure of haplotype blocks in
the human genome. Science 2002;296:22257.
78. Johnson, JW, and Ascher, P. Glycine potentiates the NMDA response in cultured
mouse brain neurons. Nature 2003;325:52931.
79. Schumacher, J, Jamra, RA, Freudenberg, J, et al. Examination of G72 and D-amino
acid oxidase as genetic risk factors for schizophrenia and bipolar affective disorder.
Mol Psychiatry 2004;9:2037.
80. Hattori, E, Liu, C, Badner, JA, et al. Polymorphisms at the G72/G30 gene locus on
13q33 are associated with bipolar disorder in two independent pedigree series. Am J
Hum Genet 2003;72:113140.
81. Chen, YS, Akula, N, Detera-Wadleigh, SD, et al. Findings in an independent sample
support an association between bipolar affective disorder and the G72/G30 locus on
chromosome 13q33. Mol Psychiatry 2004;9:8792.
82. Sklar, P, Gabriel, SP, McInnis, MG, et al. Family-based association study of 76
candidate genes in bipolar disorder: BDNF is a susceptibility locus. Mol Psychiatry
2002;7:57993.
83. Neves-Pereira, M, Mundo, E, Muglia, P, et al. The brain-derived neurotrophic factor
gene confers susceptibility to bipolar disorder: evidence from a family-based associ-
ation study. Am J Hum Genet 2002;71:6515.
84. Egan, MF, Goldberg, TE, Kolachana, BS, et al. Effect of COMT Val108/158 Met
genotype on frontal lobe function and risk for schizophrenia. PNAS 2001;98: 691722.
85. Nakata, K, Ujike, H, Sakai, A, et al. Association study of the BDNF gene with bipolar
disorder. Neurosci Lett 2003;337:1720.
86. Hong, CJ, Huo, SJ, Yen, FC, et al. Association study of a brain-derived neurotrophic
factor genetic polymorphism and mood disorders, age of onset and suicidal behavior.
Neuropsychobiology 2003;48:1869.
87. Harada, S, Agarwal, DP, and Goedde, HW. Aldehyde dehydrogenase deficiency as
cause of the flushing reaction to alcohol in Japanese. Lancet ii 1982;8253:982.
88. Thomasson, HR, Edenberg, HJ, Crabb, DW, et al. Alcohol and aldehyde dehydroge-
nase genotypes and alcoholism in Chinese men. Am J Hum Genet 1991;48:67781.
89. Craddock, N, Daniels, J, Roberts, E, et al. No evidence for allelic association between
bipolar disorder and MAOA gene polymorphisms. Am J Med Genet 1995;60:3224.
90. Lim, LC, Powell, J, Sham, P, et al. Evidence for a genetic association between alleles
of MAOA gene and bipolar affective disorder. Am J Med Genet 1995;60:32531.
91. Nothen, MM, Eggermann, K, Albus, M, et al. Association analyses of the MAOA
gene in bipolar affective disorder by using family based controls. Am J Human Genet
1995;57:9757.
92. Parsian, A, and Todd, RD. Genetic association between monoamine oxidase and
manic-depressive illness: comparison of relative risk and haplotype relative risk data.
Am J Med Genet 1997;74:4759.
93. Furlong, RA, Ho, L, Rubinszstein, JS, et al. Analysis of the MAOA gene in bipolar
affective disorder by association studies, meta-analyses and sequencing of the pro-
moter. Am J Med Genet 1999;88:398406.
94. Preisig, M, Bellivier, F, Fenton, BT, et al. Association between bipolar disorder and
MAOA gene polymorphisms: results of a multicenter study. Am J Psychiatry
2000;157:94859.
95. Kawada, Y, Hattori, M, and Dai, XY. Possible association between MAOA gene and
bipolar affective disorder. Am J Hum Genet 1995;56:3356.
96. Muramatsu, T, Matsushita, S, Kanba, S, et al. MAO genes polymorphisms and mood
disorder. Am J Med Genet 1997;19:4946.
97. Turecki, G, Grof, P, Cavazzoni, P, et al. MAOA: Association and linkage studies with
lithium responsive bipolar disorder. Psychiatric Genet 1999;9:136.
98. Kunugi, H, Ishida, S, Kato, T, et al. A functional polymorphism in the promoter
region of MAOA gene and mood disorders. Mol Psychiatry 1999;4:3935.
99. Ramammorthy, S, Bauman, AL, Moore, KR, et al. Antidepressant and cocaine sen-
sitive human transporter: Molecular cloning, expression and chromosomal localiza-
tion. Proc Natl Acad Sci 1993;90:25426.
100. Collier, DA, Arranz, MJ, Sham, P, et al. The serotonin transporter is a potential
susceptibility factor for bipolar affective disorder. Neuroreport 1996;7:16759.
101. Lesch, KP, Bengel, D, Heils, A, et al. Association of anxiety related traits with a poly-
morphism in the serotonin transporter gene regulatory region. Science 1996; 274: 152731.
102. Fickerstrand, CE, Lovejoy, EA, and Quinn, JP. An intronic polymorphic domain often
associated with susceptibility to affective disorders has allele dependent differential
enhancer activity in embryonic stem cells. FEBS Lett 1999;458:1714.
103. Collier, DA, Stober, G, Li, T, et al. A novel functional polymorphism within the
promoter of the serotonin transporter gene: possible role in susceptibility to affective
disorders. Mol Psychiatry 1996a;1:45360.
104. Rees, M, Norton, N, Jones, T, et al. Association studies of bipolar disorder at the
human serotonin transporter gene(hSERT;5HTT). Mol Psychiatry 1997;2:398402.
105. Vincent, JB, Masellis, M, Lawrence, J, et al. Genetic association analysis of serotonin
system genes in bipolar affective disorder. Am J Psychiatry 1999;156:1368.
106. Gutierrez, B, Arranz, MJ, Collier, DA, et al. Serotonin transporter gene and risk for
bipolar affective disorder: An association study in Spanish population. Biol Psychiatry
1998;43:8437.
107. Bocchetta, A, Piccardi, MP, Palmas, MA, et al. Family-based association study
between bipolar disorder and DRD2, DRD4, DAT, and SERT in Sardinia. Am J Med
Genet 1999;88:5226.
108. Mundo, E, Walker, M, Tims, H, et al. Lack of linkage disequilibrium between
serotonin transporter protein gene (SLC6A4) and bipolar disorder. Am J Med Genet
2000;96:37983.
109. Mynett-Johnson, L, Kealey, C, Claffey, E, et al. Multimarker haplotypes within the
serotonin transporter gene suggest evidence of an association with bipolar disorder.
Am J Med Genet 2000;96:8459.
110. Kirov, G, Rees, M, Jones, I, et al. Bipolar disorder and the serotonin transporter gene:
A family-based association study. Psychol Med 1999;29:124954.
111. Saleem, Q, Ganesh, S, Vijaykumar, M, et al. Association analysis of 5HT transporter
gene in bipolar disorder in the Indian population. Am J Med Genet 2000;96:1702.
112. Mendes de Oliveira, JR, Otto, PA, Vallada, H, et al. Analysis of a novel functional
polymorphism within the promoter region of the serotonin transporter gene (5-HTT)
in Brazilian patients affected by bipolar disorder and schizophrenia. Am J Med Genet
1998;81:2257.
113. Kunugi, H, Hattori, M, Kato, T, et al. Serotonin transporter gene polymorphisms:
ethnic difference and possible association with bipolar affective disorder. Mol Psy-
chiatry 1997;2:45762.
114. Furlong, RA, Ho, L, Walsh, C, et al. Analysis and meta-analysis of two serotonin
transporter gene polymorphisms in bipolar and unipolar affective disorders. Am J
Med Genet 1998;81:5863.
115. Coyle, N, Jones, I, Robertson, E, et al. Variation at the serotonin transporter gene
influences susceptibility to bipolar affective puerperal psychosis. Lancet
2000;356:14901.
116. Ospina-Duque, J, Duque, C, Carvajal-Carmona, L, et al. An association study of bipolar
mood disorder (type I) with the 5-HTTLPR serotonin transporter polymorphism in a
human population isolate from Colombia. Neurosci Lett 2000;292:199202.
117. Mendlewicz, J, Massat, I, Souery, D, et al. Serotonin transporter 5HTTLPR polymor-
phism and affective disorders: No evidence of association in a large European mul-
ticenter study. Eur J Hum Genet 2004;12:37782.
118. Weissman, MM, Wickramaratne, P, Adams, PB, et al: The relationship between panic
disorder and major depression: a new family study. Arch Gen Psychiatry 1993;
50:76780.
119. Mendlewicz, J and Baron, M. Morbidity risks in subtypes of unipolar depressive
illness: Differences between early and late onset forms. Br J Psychiatry
1981;139:4636.
120. Weissman, MM, Merikangas, KR, Wickramaratne, P, et al. Understanding the clinical
heterogeneity of major depression using family data. Arch Gen Psychiatry 1986;
43:4304.
121. Bland, RC, Newman, SC, and Orn, H. Recurrent and nonrecurrent depression. Arch
Gen Psychiatry 1986;43:10859.
122. Stancer, HC, Persad, E, Wagener, DK, and Jorna, T. Evidence for homogeneity of
major depression and bipolar affective disorder. J Psychiatr Res 1987;21:3753.
123. Ekholm, JM, Kieseppa, T, Hiekkalinna, T, et al. Evidence of susceptibility loci on
4q32 and 16p12 for bipolar disorder. Hum Mol Genet 2000;12:190715.
124. Reich, T, Van Eerdewegh, P, Rice, J, et al. The familial transmission of primary major
depressive disorder. J Psychiatr Res 1987;21:61324.
125. Kendler, KS, Neale, MC, Kessler, RC, et al. The clinical characteristics of major
depression as indices of the familial risk to illness. Br J Psychiatry 1994;165:6672.
126. Kendler, KS, Gardner, CO, and Prescott, CA. Clinical characteristics of major depres-
sion that predict risk of depression in relatives. Arch Gen Psychiatry 1999;56:322.
127. Merikangas, KR, Risch, NJ, and Weissman, MM. Co-morbidity and co-transmission
of alcoholism, anxiety and depression. Psychol Med 1994;24:6980.
128. Winokur, G, Cadoret, R, Dorsab, J, et al. Depressive disease: a genetic study. Arch
129. Nurnberger, JI, Foroud, T, Flury, L, et al. Evidence for a locus on chromosome 1 that
influences vulnerability to alcoholism and affective disorder. Am J Psychiatry
2001;158:71824.
130. Sullivan, P, and Kendler, K. Genetic case-control studies in neuropsychiatry. Arch
131. Bierut, LJ, Heath, AC, Bucholz, KK, et al. Major depressive disorder in a community-
based twin sample: Are there different genetic and environmental contributions for
men and women? Arch Gen Psychiatry 1999;56:55763.
132. Kendler, KS, Pedersen, N, Johnson, L, et al. A pilot Swedish twin study of affective
illness, including hospital and population-ascertained subsamples. Behav Genet
1995;25:21732.
133. Lyons, MJ, Eisen, SA, Goldberg, J, et al. A registry-based twin study of depression
in men. Arch Gen Psychiatry 1998;55:46872.
134. Kendler, KS, and Prescott, CA. A population-based twin study of lifetime major
depression in men and women. Arch Gen Psychiatry 1999;56:3944.
135. McGuffin, P, Katz, R, Watkins, S, and Rutherford, J: A hospital-based twin register of
the heritability of DSM-IV unipolar depression. Arch Gen Psychiatry 1996;53:12936.
136. Holmans, P, Zubenko, GS, Crowe, RR, et al. Genome-wide significant linkage to
recurrent early-onset major depressive disorder on chromosome 15q. Am J Hum Genet
2004;74:115467.
137. Zubenko, GS, Maher, B, Hughes, HB, et al. Genome-wide linkage survey for genetic
loci that influence the development of depressive disorders in families with recurrent
early-onset major depression. Am J Med Genet 2003;123B:118.
138. Zubenko, GS, Hughes, HG, Stiffler, JS, et al. Sequence variations in CREB1 co-
segregate with depressive disorders in women. Mol Psychiatry 2003;8:61118.
139. McInnis, MG, Lan, TH, Willour, VL, et al. Genome-wide scan of bipolar disorder
in 65 pedigrees: supportive evidence for linkage at 8q24, 18q22, 4q32, 2p12, and
13q12. Mol Psychiatry 2003;8:28898.
140. Cichon, S, Schumacher, J, Muller, DJ, et al. A genome screen for genes predisposing
to bipolar affective disorder detects a new susceptibility locus on 8q. Hum Mol Genet
2001;10:293344.
141. Liu J, Juo SH, Dewan A, et al. Evidence for a putative bipolar disorder locus on
2p13-16 and other potential loci on 4q31, 7q34, 8q13, 9q31, 10q21-24, 13q32, 14q21
and 17q11-12. Mol Psychiatry 2004;8:33342.
142. Kelsoe, JR, Spence, MA, Loetscher, E, et al. A genome survey indicates a possible
susceptibility locus for bipolar disorder on chromosome 22. Proc Natl Acad Sci
USA 2001;98:58590.
143. Dick, DM, Foroud, T, Flury, L, et al. Genome-wide linkage analyses of bipolar
disorder: a new sample of 250 NIMH Genetics Initiative pedigrees. Am J Hum Genet
2003;73:10714.
144. Potash, JB, Chiu, Y-F, MacKinnon, DF, et al. Familial aggregation of psychosis in a
replication set of 69 bipolar pedigrees: Am J Med Genet 2003;116B:907.
145. Berrettini, W, Ferraro, T, Goldin, L, et al. Chromosome 18 DNA markers and manic
depressive illness: evidence for a susceptibility gene. Proc Natl Acad Sci USA
1994;91:591821.
17 Pedigree Analyses and the

Study of Chimpanzee
(Pan troglodytes)
Personality and Subjective
Well-Being
Alexander Weiss and James E. King
CONTENTS
17.1 Overview ...................................................................................................... 248

17.1.1 Chimpanzee Natural History and Behavior..................................... 249
17.1.2 Human and Chimpanzee Personality............................................... 249
17.1.3 Human and Chimpanzee Subjective Well-Being ............................ 250
17.1.4 Study Goals ...................................................................................... 250
17.1.4.1 Reasons for Using Pedigree Analysis Methods............... 250
17.1.4.2 Common Features of Pedigree Analysis Methods .......... 250
17.2 Symmetric Differences Squared .................................................................. 251
17.2.1 Squared Phenotypic Differences and the Variance ......................... 251
17.2.1.1 A Latent Variable Model for Estimating Variances
and Covariances................................................................ 251
17.2.1.2 Using Multiple Regression to Solve for
Variance Estimates............................................................ 252
17.2.1.3 Computing Proportions of Variance................................. 252
17.2.2 Results When Using Symmetric Differences Squared.................... 253
17.2.3 Additional Considerations............................................................... 253
17.2.3.1 Advantages of Symmetric Differences Squared .............. 253
17.2.3.2 Least Squares Estimation and Negative
Variance Components ....................................................... 253
17.2.3.3 Using Monte Carlo Procedures to Determine
Statistical Significance...................................................... 254
17.2.4 Concluding Thoughts ...................................................................... 254
17.3 Multitrait Derivative Free Restricted Maximum Likelihood Analysis ....... 254
17.3.1 Foundations of the Univariate Case ................................................ 254
247
17.3.1.1 Hendersons Mixed Model Equation and the Estimation

of Variance and Covariance Components ........................ 254
17.3.2 Study Goals ...................................................................................... 255
17.3.2.1 Choosing a Model Using Akaikes
Information Criterion........................................................ 256
17.3.2.2 Comparing Nested Models Using Chi-Square Tests ....... 256
17.3.3 Considerations.................................................................................. 256
17.3.3.1 Advantages of Using Multitrait Derivative Free
Restricted Maximum Likelihood Analysis ..................... 257
17.3.3.2 Drawbacks of Using Multitrait Derivative Free
Restricted Maximum Likelihood Analysis ..................... 257
17.4 Concluding Thoughts and Other Uses For Multitrait Derivative
Free Restricted Maximum Likelihood Analysis ......................................... 257
17.4.1 Other Applications of Multitrait Derivative Free Restricted
Maximum Likelihood Analysis ....................................................... 257
17.4.1.1 Genetic Association Studies ............................................. 257
17.4.1.2 Environmental Association Studies.................................. 258
17.4.1.3 Assessing Gene Environment Interactions.................... 258
17.4.2 Concluding Thoughts....................................................................... 258
Authors Note ........................................................................................................ 258
References.............................................................................................................. 259
Appendix A: Using Animal Breeders Toolkit to Compute Relatedness
Among Individuals within a Pedigree ......................................................... 261
17.1 OVERVIEW
In this chapter we describe our research on the behavior genetics of personality and
subjective well-being (SWB) in our chimpanzee cousins, a research program that
we plan to extend to the other great ape speciesorangutans, gorillas, and bonobos.
Although we had been familiar with classic behavior genetics designs before initi-
ating this research, we quickly realized that these approaches were not suitable for
studying great apes in naturalistic and semi-naturalistic settings. Our subjects could
not be selectively bred, they could not be conveniently grouped into sets of monozy-
gotic and dizygotic twins, and their pedigrees did not resemble trees but, instead,
kudzu. Fortunately, we stumbled upon two types of pedigree analysis that allowed
us to hack through the familial wilderness and estimate the relative contributions of
genetic and environmental effects on individual differences in chimpanzee person-
ality and SWB. The first part of this chapter will describe reasons why the behav-
iorgenetic study of chimpanzees is interesting and compelling. The second part
will describe how multiple regression analysis can be used to estimate relevant
variance components and will highlight the advantages and disadvantages of this
method. The third part will describe the use of mixed-model equations and a
restricted maximum likelihood analysis to estimate variance and covariance com-
ponents. Finally, we will offer concluding thoughts and suggest other potential uses
for these analyses.
Pedigree Analyses 249
17.1.1 CHIMPANZEE NATURAL HISTORY AND BEHAVIOR

An important reason for studying the behavior genetics of chimpanzee personality
and SWB is that chimpanzees are our closest living nonhuman relatives. Indeed, we
shared a common ancestor only 5 to 6 million years ago, and consequently we now
share a little more than 99% of our genes.13 Genetic analysis of chimpanzee person-
ality thus enables us to do more than engage in armchair speculation of the hominid
condition; we can investigate that condition in a chimpanzee model of early humans.
That chimpanzees show evidence of some of the more complex behavioral and
cognitive characteristics possessed by humans is an outgrowth of this genetic sim-
ilarity. For example, several years of research by the Rumbaughs4 demonstrated that,
in contrast with the claims of some linguists,5 chimpanzees and bonobos have
language abilities previously thought to exist only in humans. Theory of mind6,7 and
moral behavior8 have also been observed in chimpanzees, both in the field and
laboratory. Culture, once touted as the highest human accomplishment, appears a
feat that also is not above the abilities of the lowly chimpanzee.9,10 These
humanchimpanzee similarities in behavior and cognition appear to be evolutionary
homologies, not independent convergence of similar traits in evolutionarily indepen-
dent species.
17.1.2 HUMAN AND CHIMPANZEE PERSONALITY

There is now a consensus that in humans five broad dimensionsneuroticism,
extraversion, openness to experience, agreeableness, and conscientiousnessunder-
lie human personality variation.11 This model of human personality has come to be
known as the five-factor model (FFM). The domains of the FFM have biological
origins since they are heritable,12 mostly stable in adulthood,11 and are found in
multiple Western and non-Western cultures.13
Human SWB also appears to have biological origins because it is substantially
heritable14 and stable over time.15 Additionally, the best predictors of human SWB
are personality dimensions, especially neuroticism and extraversion.16
Research on chimpanzee personality indicates that chimpanzee personality and
SWB are similar to their human counterparts. First, factor analysis of chimpanzee
personality descriptors yields a six-factor solution; the first and largest of these
factors was a broad, chimpanzee-specific factor that was primarily a blend of low
neuroticism and agreeableness and high extraversion.17 Since the item with the
highest loading on this factor was dominant, the factor was designated as dominance.
The remaining five factorsextraversion, dependability, agreeableness, emotional
stability, and opennessare analogous to the five human domains. Also, similar to
human personality findings,11 chimpanzee personality factors were mostly stable
over time.41 In addition, recent human personality studies showing that the factor
structure as well as age and sex differences generalize across cultures13 led us to
ask the same questions about chimpanzee personality. We found that, as in humans,
the chimpanzee personality factor structure as well as age and sex differences
generalized from zoo environments to an African chimpanzee sanctuary whose
habitat closely resembled that of wild chimpanzees.19
17.1.3 HUMAN AND CHIMPANZEE SUBJECTIVE WELL-BEING

Similarities also exist between measures of human and chimpanzee SWB. One study
that examined SWB in chimpanzees used a measure based on the major components
of human SWB: the balance of positive and negative moods, pleasure derived from
social interactions, success in achieving goals, and a global measure of happiness.18
As with human measures of SWB, about half of the variance of chimpanzee SWB
was predicted by personality, namely, dominance and dependability.*,18 In addition
to these findings, there was evidence that chimpanzee SWB is as stable over time
as human SWB.18
17.1.4 STUDY GOALS

Because of the similarity of factor structures of humans and chimpanzees and, of
course, the close genetic correspondence between the two species, we sought to
answer a series of questions regarding chimpanzee personality and SWB. First, to
what extent are individual differences in chimpanzee personality factors and SWB
a reflection of genetic differences among individuals, i.e., heritable, and to what
extent are these individual differences a manifestation of differences in the shared
zoo environment and nonshared environmental effects? Additionally, with respect
to SWB, we asked whether heritable or nonheritable characteristics of individuals
mothers contributed to individual differences in SWB. Finally, we wished to deter-
mine whether the correlation between personality and SWB in chimpanzees could
be explained by common genetic or environmental factors.
17.1.4.1 Reasons for Using Pedigree Analysis Methods
Our primary obstacle was the low number of discrete family units in the sample. The
relative absence of discrete family units dramatically reduced the number of geneti-
cally independent pairs, thereby precluding use of most conventional techniques for
genetic analysis. Fortunately, approaches that consider the interrelationships between
all possible pairs of individuals in a sample enabled us to surmount this obstacle.
17.1.4.2 Common Features of Pedigree Analysis Methods
These approaches incorporate pedigrees based on the identity of each subjects

sire (father) and dam (mother), the environment, and phenotype scores for as many
individuals and generations as possible. Pedigree analyses have been used for many
years in the agricultural industry with samples that include hundreds or thousands
of animals. Use of these pedigree analyses has been focused on enhancing agricul-
turally and economically important traits through selective breeding. Detailed
pedigree records incorporated into official studbooks are maintained for several
species that are well represented in zoo populations, including endangered species
such as chimpanzees. For the studies described in this chapter, the chimpanzee
* The relationship between dependability and subjective well-being in chimpanzees may be accounted
for by the fact that the chimpanzee dependability factor is comprised of several items defining human
low-neuroticism.
studbook was the essential component to unraveling the behavior genetics of chim-
panzee personality and SWB.20 Since our studies, a new edition has been released.21
17.2 SYMMETRIC DIFFERENCES SQUARED

Symmetric differences squared (SDS) is an approach based on multiple regression
analysis.22 As with other pedigree analyses, all possible pairs of individuals in the
sample are included in the analysis. Studies using simulated23 and real24 data sets
indicate that the parameter estimates provided by SDS analysis tend to be more
accurate than those provided by methods based on analysis of variance.
17.2.1 SQUARED PHENOTYPIC DIFFERENCES AND THE VARIANCE

The foundation of SDS is that the expected value of the squared phenotypic differ-
ences of all possible pairs of individuals in a sample of independent scores is equal
to two times the variance of that phenotype in the sample. However, if phenotype
scores of related individuals are correlated, the expected value of the squared phe-
notypic differences will be lower than that in the population. This is illustrated by
the following equation:22
E(Yi Yj)2 = 2[var(Y) cov(Yi, Yj)]
where i < j and Yi and Yj are the phenotype scores for individuals i and j.
17.2.1.1 A Latent Variable Model for Estimating Variances

and Covariances
Variance and covariance estimates can be estimated with a simple latent variable model
(see Figure 17.1). In this model, phenotype scores are the result of additive genetic
(A), zoo (Z), and nonshared environment effects plus error (E). Additive genetic effects
will be correlated between individuals to the extent that two individuals are genetically
related (Rij, Wrights coefficient of relatedness). One thing to note is that, while
computing Rij is a relatively simple task in samples containing mostly nuclear family
units, it is considerably more difficult to do with complex pedigrees, especially those
where inbreeding has occurred. Fortunately, free software packages, e.g., Animal
Breeders Toolkit,25 can compute the degree of relatedness between all pairs of indi-
viduals in a sample and provide inbreeding coefficients by which the degree of
relatedness must be adjusted (see Appendix A in this chapter for more details).
The correlation of zoo effects is denoted rzij and equal to 1 if the pair lives in the
same zoo enclosure and 0 if they do not. Finally, the nonshared environment effects
plus error are assumed to be uncorrelated.22 If we let a2, z2, and e2 be the variances
attributable to additive genetic, zoo, and error, respectively, it then follows that
var(Y) = a2 + z2 + e2
cov(Yi, Yj) = Rija2 + rzij z2.

Rij
Ai Aj
a a
Ei e Yi Yj e Ej
z z
Zi Zj
rzij
FIGURE 17.1 The correlation in trait Y for any given pair of animals, i and j, is described
as a function of six latent variables with variances equal to 1. Latent variables include additive
genetic (A), shared zoo (Z ), and nonshared environment plus error (E) effects. Paths a, z and
e are the effects of each variable on the trait of interest. Rij is Wrights coefficient of relationship
and rzij is a dummy-coded variable equal to 1 if a pair lives in the same zoo and 0 if it does
not. (From Behavior Genetics, Weiss, A., King, J.E. and Figueredo, A.J. The heritability of
personality factors in chimpanzees (Pan troglodytes), 2000, vol. 30, 213221. With permission
from Kluwer Academic Publishers.)
17.2.1.2 Using Multiple Regression to Solve

for Variance Estimates
These values can then be substituted in the first equation22 to yield a formula that
can be solved as the linear regression equation
.5E(Yi Yj)2 = e2 + a2(1 Rij) + z2(1 rzij).
The squared differences among all possible pairs are regressed onto (1 Rij)
and (1 rzij). The unstandardized beta weight for the variable reflecting (1 Rij) is
equal to the variance due to genetic effects (a2) and the unstandardized beta weight
for the variable reflecting (1 rzij) is equal to the variance due to shared zoo effects
(z2). Finally, the intercept term is equal to the variance due to the nonshared envi-
ronment plus error (e2).22
17.2.1.3 Computing Proportions of Variance
With these estimates, it is a simple matter to compute heritability, the proportion of

variance due to additive genetic effects.22
h2 = a2 / (a2 + z2 + e2)
Substituting the a2 in the numerator with z2 or e2 would yield the proportion of

variance resulting from shared zoo and nonshared environment plus error effects,
respectively.
17.2.2 RESULTS WHEN USING SYMMETRIC DIFFERENCES SQUARED

We used SDS to estimate the heritability of personality dimensions in 145 zoo-
housed chimpanzees.26 Approximately 63% of the dominance factor variance was
heritable. In addition, 21% of the dependability variance was heritable, but was not
statistically significant. Finally, virtually all of the remaining variance for all factors
was accounted for by nonshared environment effects plus error; shared zoo effects
were negligible for all six factors.26 There are several possible reasons for the
discrepancy between these findings and those of the behavior genetics of the FFM
in humans, which indicate that all five domains are heritable.12 First, the sample size
was relatively small and it did not include twins. This problem may be especially
important as a recent review of the heritability of personality literature indicates that
some of the genetic variance in human personality is the result of nonadditive genetic
effects, i.e., genetic effects reflecting interactions among genes, which are not easily
assessed in non-twin designs.27 A second potential reason for our inability to replicate
the heritability of the five human-like personality traits may have been that domi-
nance accounted for the lions (or chimpanzees) share of common factor variance
among adjectival descriptors.
On the other hand, the dearth of shared or common environmental effects is
relatively consistent with the results from behavior genetic analyses of human per-
sonality dimensions.12 Specifically, shared zoo effects on chimpanzee personality
are almost as nonexistent as shared family effects on human personality.28
17.2.3 ADDITIONAL CONSIDERATIONS

No one technique is perfect, so, when choosing any analysis technique, one should
be aware of practical and methodological advantages and disadvantages. This is as
true for quantitative genetic analysis as for conventional statistics.
17.2.3.1 Advantages of Symmetric Differences Squared
The major advantage of SDS is that, if the genetic relatedness is known for all possible
pairs of individuals, there is no need for a specialized software package. We conducted
our analysis using the PROC REG command in SAS,29 but it could just as easily
have been computed using the regression modules of any other statistics package.
17.2.3.2 Least Squares Estimation and Negative Variance

Components
A possibly disquieting characteristic of SDS is that the least squares estimation

procedure may produce negative variance estimates. The authors of this technique
recommend that any negative variance estimates be treated as zeros.22
In unpublished analyses, we noticed that other statistical approaches that are based
on restricted maximum likelihood analyses (described later) will estimate as zero the
variance components that were negative in the SDS analysis.26 Modern statistical
packages allow one to solve regression equations via restricted maximum likelihood
as opposed to least squares regression, thereby avoiding negative variance estimates.
17.2.3.3 Using Monte Carlo Procedures to Determine

Statistical Significance
The use of all possible pairs of subjects in the SDS procedure means that each
subject will be present in n 1 pairs. This lack of independence means that standard
procedures for assessing the statistical significance of these estimates will be
incorrect. The statistical significance of these estimates can, however, be addressed
by Monte Carlo simulation, in which the variable representing degree of relatedness
is randomly assigned to subject pairs. Repeated iterations of the analysis can then
be used to estimate the sampling distribution of these estimates. We used 1,000
iterations to estimate these sampling distributions.30 The proportion of heritability
estimates from the simulation that were greater than or equal to the estimates
derived when the degrees of relatedness were properly assigned was used as an
estimate of p.
17.2.4 CONCLUDING THOUGHTS

While SDS can be used to estimate genetic and environmental effects, it is probably
best for single trait analyses, especially when sample sizes are small. The relative
ease of implementation also makes it potentially useful for initial data exploration
before progressing to other analyses.
17.3 MULTITRAIT DERIVATIVE FREE RESTRICTED

MAXIMUM LIKELIHOOD ANALYSIS
Multitrait derivative free restricted maximum likelihood (MTDFREML31) analysis is
another means by which variance and covariance components can be estimated in
samples with complex pedigrees. Like SDS, MTDFREML is based on all pairs of
individuals. MTDFREML is based on a specific form of the mixed-model equation.32
17.3.1 FOUNDATIONS OF THE UNIVARIATE CASE

In the univariate case,** this approach is described in the following formula. It
assumes that the phenotype of each individual, i, is composed of fixed () and
random effects including, but not limited to, additive genetic (a); heritable maternal
(m); nonheritable maternal (c); and nonshared environment effects plus error (e).32
yi = i + ai + mi + ci + ei.
17.3.1.1 Hendersons Mixed Model Equation and the

Estimation of Variance and Covariance Components
A more general form of this equation can be used to describe phenotypic deviations
of each individual from a sample or population mean. This general equation is based
** This approach can be used to estimate variances and covariance of and among multiple traits.29,30
on a series of vectors and matrices. The first of these vectors, i, is a vector of deviations
from a grand mean attributable to fixed effects (zoos in our case). Other vectors include
a, m, and c, which indicate the degree to which subjects deviations from the zoo mean
are the result of additive genetic, heritable maternal, and nonheritable maternal effects,
respectively. The matrices include X and Za, which indicate each individuals zoo
membership and identifying number, respectively. The final matrices are Zm, and Zc,
which indicate the identifying number of each chimpanzees motherused to assess
heritable maternal and nonheritable maternal effects, respectively.***
The products of these vectors and matrices and the residual term, e, which
represents deviations resulting from nonshared environmental effects plus error, can
be used to predict, y, a vector of phenotypes, in a sample or population.
y = X() + Za(a) + Zm(m) + Zc(c) + e
These summed and squared deviations are variances. If a given phenotype is

heritable, the similarity of deviations of two individuals from the zoo mean will be
dependent, in part, on how related they are. If some heritable characteristic of
mothers contributes to a phenotype score of an offspring over and above the genetic
effects based on the relatedness of mother and offspring, heritable maternal effects
are said to be present. In this case, the similarity of deviations of two individuals
from the zoo mean will be dependent, in part, on how related their mothers are. If
some nonheritable characteristic of mothers has an effect on the phenotype of their
offspring, i.e., there are nonheritable maternal effects, the similarity of deviations
of two individuals will be more similar if they share the same mother than if they
have different mothers, regardless of the relatedness of the two mothers.
With two n n matrices, A, indicating how related all possible pairs are, and I,
with 1s in the diagonal and 0s elsewhere, we can decompose the phenotypic variance
by weighting additive genetic and heritable maternal variance by A and nonheritable
maternal and nonshared environmental variance by I.
2y = 2 + A2a + A2m + I2c + I2e
These variance components can then be used to compute the proportions of

variance and covariance that arise from genetic or different environmental effects.
17.3.2 STUDY GOALS

We used MTDFREML31 to solve for these variance and covariance components in a
sample of 128 zoo chimpanzees that were rated on personality and SWB.33 MTD-
FREML uses an iterative procedure that tries different values for variance components
until it achieves a set that minimizes the difference between the expected and predicted
phenotype scores.34,35 Our previous study indicated that dominance was the only
*** If the source of these effects is the same, as when two types of maternal effects are investigated, the
matrices will be identical.
personality trait that was heritable,26 so we focused exclusively on the genetic and
environmental sources of variances and covariance for dominance and SWB.33
17.3.2.1 Choosing a Model Using Akaikes

Information Criterion
Along with the parameter estimates, MTDFREML provides a statistic (2log) indi-
cating how well the model replicated the actual data. The lower this statistic, the better
the model fit. We compared seven different models using two different approaches.
The first approach was based on Akaikes information criterion (AIC).36 AIC is based
on the notion that model fit must be balanced with model parsimony and, hence,
penalizes the degree of model fit by the number of parameters estimated (k).
AIC = 2log + 2k
The model with the lowest AIC, then, is the one that best balances model fit
and parsimony. In our analysis this model was one in which 66% of the dominance
and 40% of the SWB variance were heritable.33 The heritability estimate for domi-
nance was nearly identical to what we obtained in univariate analyses with SDS26
and the heritability estimate for SWB was nearly identical to what we obtained using
SDS and another estimation procedure, Method R.37,38 The covariance between
dominance and SWB was almost entirely due to shared genes. Furthermore, heritable
maternal effects accounted for approximately 22% of the SWB, but none of the
variance in dominance.33 Finally, largely independent nonshared environmental
effects plus error accounted for 27% of the remaining dominance and 35% of the
remaining SWB variance.33
17.3.2.2 Comparing Nested Models Using Chi-Square Tests
The second approach, nested models comparisons, is based on the fact that the 2log
statistic is distributed as a chi-square. Hence, one can test whether parameters are
significant by comparing the 2log of models with and without those parameters.39
If the 2log difference between models is not significant, this indicates that the
additional parameters were not contributing to the model beyond what would be
expected by chance and, hence, the model with fewer parameters should be adopted.
In our analysis, comparing models using nested models comparison indicated
that there was no significant difference between the model that we chose based on
the comparison of the AIC statistics and a model that was identical except for not
including heritable maternal effects. We therefore accepted this more parsimonious
model that did not include heritable maternal effects as a plausible alternative.33
17.3.3 CONSIDERATIONS
As with SDS, using MTDFREML to solve for variance and covariance components
has advantages and disadvantages. These should be considered carefully when plan-
ning a behavior genetic study in samples with complex pedigrees.
17.3.3.1 Advantages of Using Multitrait Derivative Free

Restricted Maximum Likelihood Analysis
One of the main advantages is that the MTDFREML approach does not produce
any negative variance components. Additionally, it estimates sources of covariance
between two or more traits and even between different effects. Another advantage
is that these analyses provide model fit parameters that can easily be used to compare
the fit and parsimony of different models. Finally, there are several free programs
that can be used to conduct these analyses. We used MTDFREML,31 but there are
similar programs such as DFREML.40
17.3.3.2 Drawbacks of Using Multitrait Derivative Free

Restricted Maximum Likelihood Analysis
However, these programs are not as flexible as standard statistical packages. Thus,
if one wanted to use the program to investigate gene environment interactions, it
would require considerable knowledge in FORTRAN programming. Another disad-
vantage is that even with small samples, a thorough analysis can take a long time.
The author spent from 10:00 PM to 8:00 AM computing variances and covariances
with the universitys mainframe computer for the seven models we compared.33
17.4 CONCLUDING THOUGHTS AND OTHER USES

FOR MULTITRAIT DERIVATIVE FREE RESTRICTED
MAXIMUM LIKELIHOOD ANALYSIS
Pedigree analyses are a powerful means of addressing questions about genetic and
environmental effects in samples with complex pedigrees. Certainly, these methods
could also be applied to studies of human populations.
17.4.1 OTHER APPLICATIONS OF MULTITRAIT DERIVATIVE FREE

RESTRICTED MAXIMUM LIKELIHOOD ANALYSIS
Another potential use of these analyses is that, if one knows the heritability of a
trait and the pedigree, one can obtain breeding values for individuals. Breeding
values reflect the amount by which individuals phenotypic values deviate from the
sample mean because of additive genetic effects. Calculation of an individuals
breeding value is dependent upon having phenotypic data from genetic relatives and
knowing the heritability of the trait. Breeding values have been used extensively in
the agricultural industry to increase the efficiency of selective breeding. Selectively
breeding animals with the highest breeding values has been used, for example, to
produce more milk with fewer cows in the United States.
17.4.1.1 Genetic Association Studies
Because breeding values reflect the genetic contribution to a given phenotype, they
could be used to improve the power of genetic association studies. Genetic association
studies typically compare individuals who are high in some phenotype with those
who are low or average in the phenotype. Because the phenotype is a blend of additive
and nonadditive genetic variance as well as environmental effects, the phenotype
score is only an approximation of an individuals genetic predisposition for that trait.
The use of breeding values in these studies would increase the likelihood of detecting
associations.
17.4.1.2 Environmental Association Studies
Similarly, pedigree analyses can also output the degree to which individuals differ
from their expected mean based on measured environmental effects. This potentially
useful measure is the obverse of breeding values and would show individual differ-
ences in susceptibility to environmental modification of the trait being studied. An
interesting question is whether this environmental susceptibility is itself heritable.
Those studying the environmental determinants of behavior could attempt to associate
different environmental events and conditions to these environmental susceptibility
values and reduce the likelihood that genetic effects confound their findings.
17.4.1.3 Assessing Gene Environment Interactions
Environmental susceptibility scores could be used to identify individuals who are

particularly susceptible to environmental and social sources of stress that may lead
to depression and other psychopathological disorders. The reasoning would be that
those individuals who have high phenotypic psychopathology scores and high envi-
ronmental susceptibility scores are particularly vulnerable to stressful social and
environmental conditions. Two individuals with identical phenotypic psychopathol-
ogy scores might differ markedly in their environmental susceptibility scores, thereby
implying different underlying causes of psychopathology in the two individuals.
Likewise, environmental susceptibility scores would be a basis for a new approach
to individual differences in personality.
17.4.2 CONCLUDING THOUGHTS

Although the classic twin method is powerful and invaluable, especially as it provides
broad-sense heritability estimates, it is our hope that pedigree analyses will be more
widely adopted. Pedigree analyses offer the ability to analyze data on a wider range
of samples and can provide more accurate estimates of genetic variances and cova-
riances because they use all possible relationships.
AUTHORS NOTE
We wish to thank Virginia Landau and the staff of the ChimpanZoo program and
all the raters at the different zoos for making our research possible. We also wish
to thank our past collaborators A. J. Figueredo and R. Mark Enns. Finally, we would
like to thank R. R. McCrae and Antonio Terracciano for their comments on an earlier
manuscript.
REFERENCES
1. Purvis, A. (1995). A composite estimate of primate phylogeny. Philosophical Trans-
action of the Royal Society of London, Series B: Biological Sciences 348, 405421.
2. Sarich, V.M. and Wilson, A.C. (1967). Immunological time scale for hominid evo-
lution. Science 158, 12001203.
3. Wilson, A.C. and Sarich, V.M. (1969). A molecular time scale for human evolution.
Proceedings of the National Academy of Sciences U.S.A. 63.
4. Rumbaugh, D.M. and Savage-Rumbaugh, E.S. (1994). Language in comparative
perspective. In Animal Learning and Cognition, N.J. Mackintosh, ed. (San Diego,
CA: Academic Press), pp. 307333.
5. Chomsky, A.N. (1972). Language and the Mind (New York: Harcourt Brace
Jovanovich).
6. Hare, B., Call, J., and Tomasello, M. (2001). Do chimpanzees know what conspecifics
know? Animal Behaviour 61, 139151.
7. Whiten, A. (1997). The Machiavellian mindreader. In Machiavellian Intelligence II:
Extensions and Evaluations, A. Whiten and R.W. Byrne, eds. (New York: Cambridge
University Press), pp. 144173.
8. de Waal, F.B.M. (1996). Good Natured: The Origins of Right and Wrong in Humans
and Other Animals. (Cambridge, MA: Harvard University Press).
9. de Waal, F.B.M. (2001). The Ape and the Sushi Master Cultural Reflections by a
Primatologist (New York: Basic Books).
10. Whiten, A., Goodall, J., McGrew, W.C., Nishida, T., Reynolds, V., Sugiyama, Y.,
Tutin, C.E.G., Wrangham, R.W., and Boesch, C. (1999). Cultures in chimpanzees.
Nature 399, 682685.
11. McCrae, R.R. and Costa, P.T., Jr. (2003). Personality in Adulthood: A Five-Factor
Theory Perspective (New York: Guilford).
12. Bouchard, T.J., Jr. and Loehlin, J.C. (2001). Genes, evolution, and personality. Behav-
ior Genetics 31, 243271.
13. McCrae, R.R. (2001). Trait psychology and culture: Exploring intercultural compar-
isons. Journal of Personality 69, 819846.
14. Lykken, D.T. and Tellegen, A. (1996). Happiness is a stochastic phenomenon. Psy-
chological Science 7, 186189.
15. Eid, M. and Diener, E. (2004). Global judgements of subjective well-being: Situa-
tional variability and long-term stability. Social Indicators Research 65, 245277.
16. Diener, E. (1996). Traits can be powerful, but are not enough: Lessons from subjective
well-being. Journal of Research in Personality 30, 389399.
17. King, J.E. and Figueredo, A.J. (1997). The Five-Factor Model plus Dominance in
chimpanzee personality. Journal of Research in Personality 31, 257271.
18. King, J.E. and Landau, V.I. (2003). Can chimpanzee (Pan troglodytes) happiness be
estimated by human raters? Journal of Research in Personality 37, 115.
19. King, J.E., Weiss, A., and Farmer, K.H. (2005). A chimpanzee (Pan troglodytes)
analogue of cross-national generalization of personality structure: zoological Parks
and an African Sanctuary. Journal of Personality 73, 389410.
20. Fulk, R. (1999). Chimpanzee Species Survival Plan: Master Plan 19981999, North
Carolina Zoological Park: Asheboro, NC.
21. Ross, S. (2003). North American Regional Studbook for the Chimpanzee (Pan trogl-
odytes), Lincoln Park Zoo: Chicago.
22. Grimes, L.W. and Harvey, W.R. (1980). Estimation of genetic variances and covari-
ances using Symmetric Differences Squared. Journal of Animal Science 50, 632644.
23. Bruckner, C.M. and Slanger, W.D. (1986). Symmetrical differences squared and
analysis of variance procedures for estimating genetic and environmental variances
and covariances for beef-cattle weaning weight: I. Comparison via simulation. Journal
of Animal Science 63, 17791793.
24. Bruckner, C.M. and Slanger, W.D. (1986). Symmetrical differences squared and
analysis of variance procedures for estimating genetic and environmental variances
and covariances for beef-cattle weaning weight: II. Estimates from a data set. Journal
of Animal Science 63, 17941803.
25. Golden, B.L., Snelling, W.M., and Mallinckrodt, C.H. (1992). Animal Breeders
Toolkit: Users Guide and Reference Manual, Colorado State University Agricultural
Experiment Station: Fort Collins.
26. Weiss, A., King, J.E., and Figueredo, A.J. (2000). The heritability of personality
factors in chimpanzees (Pan troglodytes). Behavior Genetics 30, 213221.
27. Bouchard, T.J., Jr. (2004). Genetic influence on human personality traits: A survey.
Current Directions in Psychological Science 13, 148151.
28. Rowe, D.C. (1994). The Limits of Family Influence (New York: Guilford).
29. SAS Institute, I. (1999). SAS/STAT Users Guide, Version 8 (Cary, NC: SAS Institute).
30. Cheverud, J.M., Falk, D., Vannier, M., Konigsberg, L., Helmkamp, R.C., and Hilde-
bolt, C. (1990). Heritability of brain size and surface features in rhesus monkeys
(Macaca mulatta). Journal of Heredity 81, 5157.
31. Boldman, K.G., Kriese, L.A., Van Vleck, L.D., Van Tassel, C.P., and Kachman, S.D.
(1995). A Manual for Use of MTDFREML [draft], Animal Research Station, United
States Department of Agriculture: Clay Center, NE.
32. Henderson, C.R. (1975). Best linear unbiased estimation and prediction under a
selection model. Biometrics 32, 423447.
33. Weiss, A., King, J.E., and Enns, R.M. (2002). Subjective well-being is heritable and
genetically correlated with Dominance in chimpanzees (Pan troglodytes). Journal of
Personality and Social Psychology 83, 11411149.
34. Meyer, K. (1989). Restricted maximum likelihood to estimate variance components
for animal models with several random effects using a derivativefree algorithm.
Genetics, Selection, Evolution 21, 317340.
35. Meyer, K. (1991). Estimating variances and covariances for multivariate animal
models by restricted maximum likelihood. Genetics, Selection, Evolution 23.
36. Akaike, H. (1987). Factor analysis and AIC. Psychometrika 52, 317332.
37. Weiss, A., King, J.E., and Enns, R.M. (1999). Can we breed a happier ape? The
heritability of subjective well-being in zoo chimpanzees (Pan troglodytes). Behavior
Genetics 29, 374.
38. Reverter, A., Golden, B.L., Bourdon, R.M., and Brinks, J.S. (1994). Method R
variance-components procedure: Application on the simple breeding value model.
Journal of Animal Science 72, 22472253.
39. Loehlin, J.C. (1998). Latent Variable Models: An Introduction to Factor, Path, and
Structural Analysis (Mahwah, NJ: Lawrence Erlbaum Associates).
40. Meyer, K. (1998). DFREML, 3.1 ed., Self Published: Armidale, NSW, Austra-
lia.Meyer, K. (1998). DFREML, 3.1 ed., Self Published: Armidale, NSW, Australia.
41. King, J.E., Weiss, A., and Sisco, M.M. Age differences in chimpanzee personality,
under review.
APPENDIX A: USING ANIMAL BREEDERS TOOLKIT

TO COMPUTE RELATEDNESS AMONG INDIVIDUALS
WITHIN A PEDIGREE
The Animal Breeders Toolkit (ABTK)25 is one of several software packages that
compute degrees of relatedness (Rij) and inbreeding coefficients (f) in a sample. It
is comprised of a package of commands that perform the necessary matrix algebra
needed to compute these as well as other parameters. Recently, SAS29 has incorpo-
rated a new procedure, INBREED, which performs the same function. There are
other free and commercial products as well.
To compute the degree of relatedness with ABTK one must first have a space-
delimited pedigree file in which the first column represents the stud or identity
number of the individual, the second the identity number of his or her father, and
the third the stud number of his or her mother. If either parents identity is unknown,
it should be represented by a .. We have included a small section of the pedigree
from a sample of chimpanzees at the Yerkes Regional Primate Center below:
.
.
.
86 16 21
87 37 24
88 8 40
89 8 28
90 37 44
91 . .
92 53 36
93 . 41
94 16 23
95 37 45
96 8 11
.
.
.
The first step is to reorder and check the pedigree file:
chkord ped.old > ped.chk 2 > ped.bad
This command reorders the pedigree so that parents precede offspring and checks
whether there are any problems, e.g., individuals that appear as both fathers and
mothers. This procedure is required for computing the degrees of relatedness and it
is generally a good idea to apply this to any pedigree file including those that will
be used in other pedigree analyses such as MTDFREML. The ped.old file is the
original space-delimited pedigree file. The ped.chk file is the properly ordered
version of the pedigree file, i.e., where parents precede their offspring, which is
required for the next steps. The ped.bad file will contain a listing of any errors,
e.g., individuals that appear as mothers and fathers in the file.
The second step is to create a file containing the inverse numerator relationship
matrix (A1) and a file containing inbreeding coefficients (f) for the subjects in the
analysis from the reordered pedigree file that we created in the previous step.
ainv -i ped.chk -o ped.inbreed -v ped.nrm
where ped.inbreed and ped.nrm are the files containing the inbreeding coef-
ficients for all subjects and the inverse numerator relationship matrix for all possible
pairs of subjects, respectively.
The third step is to invert the inverse numerator relationship matrix:
invert -i ped.nrm -o ped.output
Here, ped.nrm is the file derived from the previous command and ped.out-
put contains a matrix, A, indicating the degree of relatedness between all pairs of
individuals.
Unfortunately, the format of the matrix might be difficult to use in a conventional
statistics packages. Hence, our next two steps use ABTK to change the format of
the ped.output file so that each row includes the stud numbers of the individuals
within a pair and their degree of relatedness. The commands to do this are as follows:
d2s ped.output ped.newoutput
s2t ped.newoutput
Here, ped.output is the file we created in the prior step and ped.newout-
put is the tree version of this file where each row represents a single pair of
individuals.
If there is no inbreeding, i.e., the ped.inbreed file is empty; the previous
step was the last. However, if there was some inbreeding, then the degree of relat-
edness for every pair of individuals in which an inbred animal was present needs to
be adjusted by dividing the degree of relatedness by the following:
( fi + 1)( f j + 1)
where fi is equal to the inbreeding coefficient of one individual in the pair and fj is
equal to the inbreeding coefficient of the other individual of the pair. If an individual
is not inbred, the inbreeding coefficient is equal to 0.
These file names are only examples; any file name can be used.
18 The Elusive World of

Personality Genes:
Cherchez le Phenotype
Richard P. Ebstein, Rachel Bachner-Melman,
Jonathan Benjamin, and Robert H. Belmaker
CONTENTS
18.1 Personality.................................................................................................... 263
18.1.1 What Is Personality and How Is It Measured? ............................... 263
18.2 Heritability ................................................................................................... 266
18.3 Molecular Genetics ...................................................................................... 266
18.4 Related Phenotypes ...................................................................................... 267
18.4.1 Fibromyalgia .................................................................................... 267
18.4.2 Altruism............................................................................................ 268
18.5 The Social Personality ................................................................................. 270
18.6 Linkage Studies............................................................................................ 271
18.7 Importance.................................................................................................... 273
References.............................................................................................................. 273
18.1 PERSONALITY
18.1.1 WHAT IS PERSONALITY AND HOW IS IT MEASURED?

The term personality typically is used in two different ways. One refers to the
totality of a persons psychological nature. Personality in this sense is to be distin-
guished from physique (an individual is comprised of body and soulif we separate
out the explicitly spiritual we are left with body and personality), and, in a different
context, from mental illness (we distinguish between someones behavior during
illness and her normal personality). For most authors, personality does not encom-
pass all mental phenomena. Few would include simple perception, memory, and
concentration as aspects of personality, and intelligence is also frequently not con-
sidered part of personality, although a recent book1 on personality genetics did
include a chapter on cognitive ability. All agree that the emotional aspects of the
psyche are included in personality. This approach to personality considers the
263
universal aspects of people, not how they differ from each other. Traditional theories
of personality, of which Freuds remains the outstanding example, purport to explain
what is man, what makes him tick, to suggest a physiology of personality. If
genomics aims to elucidate the entire genome of a species, and biochemistry and
physiology aim to account for the functions of that genome and its protein products,
and if genetics is limited to the study of the consequences of genetic differences
between members of the species, then by analogy we can say that this first sense of
personality is a genomic approach. Its subjects include the unconscious, moti-
vation, instincts, defense mechanisms, the life cycle, interpersonal relations, condi-
tioning and self-actualization. While this branch of psychology has had a major
influence on twentieth-century culture, much of it is psychoanalytic, metaphoric,
and untestable, it has hardly been studied by geneticists.
A second sense of the term personality, and the one we will use throughout
the remainder of this chapter, is the individual psychological aspects of people that
make them recognizable, which is to say different from one another. In this sense
we say that Jones is miserly, Smith is gregarious, and, even, that Brown is a
character. The personalities of fictitious characters like Falstaff, Othello, and Scarlet
OHara seem familiar to millions. However, this intuitive grasping of personality is
based on one or two outstanding traits of each individual and there are no obvious
or systematic relationships among the various examples of traits mentioned above.
A general, empirical theory of personality differences must include a battery of
dimensional variables, and the number of variables must be manageable. The
attributes chosen should account for most of the variance in personality, and hope-
fully also make clinical sense to the psychologist, that is, somehow relate to what
seem to be fundamental and universal aspects of personality in the first sense of the
term as described above. It is moreover a part of the definition of a personality trait
or a personality type that the description is relatively enduring and stable for the
individual concerned, not a temporary condition.
Virtually all approaches to personality assessment used in genetic research have
used either direct laboratory evaluation of dimensional behaviors or pencil-and-paper
questionnaires that inventory dimensional personality traits. Laboratory assessments
are objective but labor-intensive both for collecting data and for analyzing them.
Furthermore, they sample a single moment in time. Self- or other report question-
naires may be more subjective, but have the advantage that they typically sample
subjects characteristics over extended periods.
The most widely used general pencil-and-paper personality inventories assume
that rating an individual accurately on the important dimensions of personality gives
a reasonable and economical description of that individuals personality. The number
and makeup of dimensions chosen are typically derived by one of the following
three approaches.
Approaches based on psychopathology derive the personality traits from psychi-
atric illnesses. Psychiatric illnesses are assumed to be extreme variants, and therefore
the clearest expressions, of normal personality. Thus the Minnesota Multiphasic
Personality Inventory2 grades people on the degree to which they resemble patients
diagnosed with hysteria, obsessivecompulsive disorder, schizophrenia, and so on.
This approach is often reflected in popular discourse; we say that someone is
The Elusive World of Personality Genes 265
compulsive, hysterical, and so on, but the approach has not been widely used in
genetic research on human personality.
Theoretical approaches derive from a theory of personality in the first sense
described above. Cloninger developed the Tridimensional Personality Questionnaire
(TPQ)3 based on a biological model employing the monoamines dopamine, norad-
renaline, and serotonin, which function as neurotransmitters in subcortical brain
systems. Animal and human findings from lesioning, imaging and pharmacological
studies involving these pathways implicate serotonin (5-HT) pathways in behaviors
devoted to avoiding harm or escaping punishment, and to the associated emotion of
anxiety (called harm avoidance (HA) in the TPQ). Human subjects with high HA
would be called neurotic on many other instruments; those with low HA are
stable or healthy or well adjusted. Similar studies implicate noradrenergic
(NE) pathways in approach behaviors, which the TPQ calls reward dependence
(RD). Human beings high on RD are sentimental and affectionate; those with low
RD are tough and pragmatic. Finally dopaminergic (DA) pathways are implicated
in drug use, sensation seeking and explorative behaviors, and emotions like curiosity
and recklessness, which the TPQ calls novelty seeking (NS). Individuals with low
NS are deliberate and frugal. The eight possible combinations of the two extremes
of the distributions of these three dimensions yield personality constellations held
to reflect clinically recognized personality disorders such as antisocial psychopathy
and obsessivecompulsive personality disorder. A later version of the TPQ is the
Temperament and Character Inventory (TCI).
The third approach derives from everyday speech. Natural language is considered
to have proved itself during evolution to be a reasonably accurate and adaptive way of
describing people. Personality inventories have been constructed by listing thousands of
personality descriptors culled from dictionaries of everyday language and reducing them
to sizable numbers by factor analysis. The most popular such inventory in clinical and
academic use is the neopersonality inventory (NEO-PI)4 which assesses five main per-
sonality factors. Neuroticism resembles TPQ HA. Extraversion is the degree of interest
in other people and in social dominance. Openness to new ideas and experiences is the
opposite of traditionalism. Agreeableness is a measure of how likeable an individual is.
Conscientiousness assesses conscience, thoughtfulness, planning and order, and their
opposites: spontaneity and lack of prudence. Extraversion and conscientiousness (with
opposite signs) correlate with TPQ NS. While details of dimensions differ between
instruments, a number of recent factor analyses concur that five main factors adequately
describe human personality differences, and these have been termed the big five.
Given the uncertainty inherent in measuring personality traits, it seems advisable,
where possible, to study putative personality traits simultaneously with more than
one instrument and/or laboratory assessment. Good correlations between apparently
similar traits in two or more assessments, and converging findings of associations
between a given polymorphism and those traits in two or more assessments, enhance
our confidence in those findings. This is all the more necessary, given the difficulty
in replicating molecular genetic findings in personality genetics, similar to the
problems encountered in other complex phenotypes.
Additionally, personality geneticists study traits of particular interest, which may
not be included in general personality inventories (see below).
18.2 HERITABILITY
Personality dimensions measured by self-report questionnaires such as the NEO and
TPQ are moderately heritable (4060%).58
18.3 MOLECULAR GENETICS

An important step in genetic studies of human personality consisted of several initial
investigations in 1996 that suggested association between specific genes and
personality traits assessed by self-report questionnaires. The two common polymor-
phisms first studied, the dopamine D4 receptor (DRD4)9,10 and the serotonin
transporter promoter region (5-HTTLPR),11 continue to be the focus of many inves-
tigations relevant not only to narrow definitions of personality (e.g., extraver-
sion/novelty-seeking, harm avoidance/neuroticism), but also to an ever widening
circle of related phenotypes including substance abuse,1221 attention deficit,2226
shyness,27 fear responses,28 and depression.29
The associations between the DRD4 exon III repeat region and novelty
seeking30,31 and the 5-HTTLPR with neuroticism32,33 have also been the subjects
of recent meta-analyses. Although the DRD4 exon III repeat region is not asso-
ciated overall with novelty seeking, a functional promoter region polymorphism
upstream (-C521T) first studied in a Japanese population34 shows more promising
results following meta-analysis,31 In an overall analysis of results with 5-HTTLPR
(n = 5,629 subjects), Sen et al.32 found suggestive evidence for an association
between the 5-HTTLPR short allele (s) and increased anxiety-related personality
trait scores (p = 0.087). However, they also found strong evidence for heteroge-
neity. This heterogeneity is largely explained by substantial variation between
the studies in the inventory used. When the analysis was stratified by inventory
type, there was a significant association between 5-HTTLPR and NEO neuroti-
cism (p = 0.000016), a nonsignificant association between 5-HTTLPR and
TCI/TPQ harm avoidance (p = 0.166), and no association between 5-HTTLPR
and other anxiety-related personality traits (p = 0.944). Similar meta-analysis
findings for 5-HTTLPR were reported by Schinka et al.33 who suggest that the
success of future personality genetics research will be maximized by the use of
personality measures from both the psychobiological and five-factor models. It
is perplexing and frustrating that NEO neuroticism and TPQ harm avoidance,
which at face value appear to measure very similar traits, produce such divergent
results.
A parallel strategy is to look at these common polymorphisms across a range
of related phenotypes, some of which might capture associations that are often
hard to pin down when a single self-report questionnaire is used to inventory
subjects. In the following section we discuss some results using the DRD4 and 5-
HTTLPR polymorphisms as examples of how looking at extended, or broader,
phenotypes may shed light on complex phenotypegenotype relationships in per-
sonality genetics.
18.4 RELATED PHENOTYPES

18.4.1 FIBROMYALGIA
Fibromyalgia syndrome (FM) is an under-diagnosed musculoskeletal condition of
unknown etiology affecting more than 10% of patients attending general medical
clinics and 15 to 20% attending rheumatology clinics.35 It is seen predominantly in
women. Patients complain that they ache all over. A number of other symptoms are
often present, particularly fatigue, morning stiffness, sleep disturbance, paresthesias,
and headaches. In 1999, a German group reported an association between this
syndrome and the short 5-HTTLPR allele36 consistent with the observed psychoso-
cial profile of these women who score high on anxiety-related personality traits.
Toward confirming this first report, we genotyped a group of 99 female fibromyalgia
patients from two Israeli ethnic groups.37 Additionally, we assessed each patient with
the Tridimensional Personality Questionnaire (TPQ). The percentage of patients
showing the short/short genotype is twice that observed in the control population
(35% vs. 17%). The difference in 5-HTTLPR genotype frequencies between
the fibromyalgia and control group was highly significant (chi-square = 25.31,
p = 0.00019, df = 2). Additionally, between-subjects testing showed a highly signi-
ficant effect of diagnosis on three of the TPQ personality traits. Novelty seeking is
lower in FMS (F = 14.46, p < 0.0001), whereas harm avoidance (F = 31.89,
p < 0.0001) and persistence (F = 41.38, p < 0.0001) are higher. No effect on reward-
directed behavior was observed. Intriguingly, there was also a significant association
between 5-HTTLPR genotype and the TPQ harm avoidance trait (F = 6.21,
p = 0.013), an association we have consistently failed to observe38 in our large
nonclinical population using this personality inventory.
We were also prompted in these women to examine the DRD4 exon III repeat
since as a group they score low on TPQ novelty seeking traits.39 As we predicted,
a significant decrease in the frequency of the 7 repeat genotype was observed in the
fibromyalgia patients (chi-square = 29.91, p < 0.008, df = 14), consistent with low
novelty seeking in these women. Although in the control population no association
was detected between TPQ novelty seeking and the DRD4 7 repeat, in the pheno-
typically extreme, albeit smaller, FM group a significant association was observed
between the 7 repeat and novelty-seeking (one-way ANOVA: F = 4.47, p = 0.038).
Patients with at least one 7 repeat had higher novelty-seeking scores as is observed
in some but not all nonclinical populations (subjects with at least one 7-repeat NS =
14.57 3.26 SD, vs. all others NS = 12.72 3.49 SD).
Consequently, by looking at a clinical group of women with an extreme
personality profile, high harm avoidance and low novelty seeking we were able
to verify the sometimes elusive associations between the short 5-HTTLPR allele
and the DRD4 7 repeat, respectively. We consider these results proof of principle
that reliance on a single personality measure in one population category may be
inadequate to demonstrate small effect sizes of common polymorphisms modestly
contributing a small fraction of the genetic variance in complex traits such as
personality.
18.4.2 ALTRUISM
We have also examined an intriguing dimension of human personality, altruism, in
a large nonclinical population for association with several common polymorphisms
germane to a broader view of human personality than that examined with the usual
inventories (TPQ or NEO).40
The paradox of human altruism, helping others and thereby reducing ones own
fitness, has confounded evolutionary biologists since the days of Darwin. Not only
is altruism a puzzle for evolutionary biologists but this trait has also perplexed
psychologists who have questioned whether there is such a thing as an altruistic
personality.41,42 However, altruistic behavior is commonplace, and a unique feature
of human altruism is that it extends beyond Hamiltons concept of inclusive fit-
ness,43 which explains altruistic acts by including helping genetically related indi-
viduals, and even beyond reciprocal altruism and reputation-based altruism.44 Out
of all the animals, only humans practice wholesale mutual aid among genetically
unrelated individuals.
Although theoretical understanding of the evolutionary45 and psychological
mechanisms46 that underlie altruism has greatly increased in the past several decades,
almost nothing is known regarding specific genes contributing to this behavior.
However, several twin studies4752 reported significant heritability of prosocial atti-
tudes. As expected of a trait that is partially influenced by genes, prosocial or
altruistic dispositions show individual differences with origins in early childhood
and stability over developmental time.53,54
The mechanisms facilitating prosocial behavior and human altruism are of inter-
est not only to evolutionary biology, but also to students of psychopathology. Psy-
chiatric research has for obvious reasons focused on negative factors that adversely
affect mental well-being. However, there is growing interest in positive or protective
factors that play a role in promoting normal development. Prosocial or altruistic
attitudes, such as cooperativeness, helpfulness, sharing, and being empathetic, are
important elements that facilitate social networks thought to promote mental well-
being. For example, prosocial behavior correlates with childrens academic excel-
lence, allows them to resist social pressures for antisocial activities, and to engage
themselves with empathy in others emotional experiences.55 Additionally, positive
social interactions undoubtedly eased by altruistic behaviors have been shown to
exert powerful beneficial effects on health outcomes and longevity.56 Importantly,
self-report measures of altruism correlate with peers and teachers evaluation.53,57
Any account of prosocial behavior and altruism in humans should include the
identification of particular genetic polymorphisms partially contributing to this trait.
Toward the goal of identifying specific genes associated with altruism and prosocial
attitudes, 354 nonclinical families with multiple siblings were inventoried for scores
on the selflessness scale.58 This questionnaire measures the propensity to ignore
ones own needs and serve the needs of others, or in other words, altruism. Subjects
were also inventoried on Cloningers TPQ59 since the reward subscale of this ques-
tionnaire taps into elements of human altruism, such as empathy. We examined two
dopaminergic genes in these subjects that we hypothesized might contribute to
prosocial or altruistic traits, based on the role a single variant of these genes plays
in novelty seeking, which includes antisocial aspects, and in a childhood behavioral

disorder, attention-deficit hyperactivity disorder (ADHD). This disorder is often
comorbid with antisocial behavior such as conduct disorder and oppositional defiant
disorder. The dopamine D4 receptor (DRD4) exon III 7 repeat (D4.7) and the DRD5
148 bp microsatellite variant both have been robustly associated with ADHD.22,60
We reasoned that if one variant contributes to antisocial traits, then, conversely, the
absence of this variant or the presence of other variants might contribute to altruistic
behavior. Additionally, we examined three DRD4 promoter region SNPs.34 We also
genotyped 3 SNPs61 in the insulin-like growth factor 2 gene (IGF2), an imprinted
gene on chromosome 11p15.5 that is an attractive candidate because some studies
connect growth factors of this class with survival of dopamine neurons specifi-
cally,6269 and with neural development overall.7075
Significant association with scores on the selflessness scale was observed
with the most common D4.4 allele (FBAT multivariate test, bi-allelic mode:
chi-square = 14.06, p = 0.0008, df = 2). No association was observed with the
second most common allele in this population, the D4.7 repeat, the risk allele in
ADHD. Evidence was also observed showing a negative correlation between the
DRD5 148 bp microsatellite repeat, the risk allele in ADHD (p = 0.027). Signif-
icant association was also observed with the IGF2 ApaI G allele (p = 0.006).
Association was also observed between TPQ reward, a temperament that taps
into some elements of altruism such as empathy, and the D4.4 repeat (p = 0.0016).
The demonstration that common polymorphisms pertinent to dopamine path-
ways are associated with altruistic or prosocial attitudes in humankind is consistent
with a central role of the brain reward center in mediating this phenotype. This is
one of the first reports to characterize specific genes acting as positive or protective
factors hypothesized to promote normal behavioral development and mental well-
being.
The provisional finding that the DRD4 exon III 4 repeat is associated with a
positive personality trait, altruism, is intriguing though not altogether surprising.
The DRD4 7 repeat has been the focus of extensive investigation both regarding
ADHD and novelty seeking/impulsivity, behavioral phenotypes with considerable
negative features. It makes sense therefore that the most common 4 repeat would
be associated with a more positive phenotype such as prosocial or altruistic behavior.
Evidence from haplotype data indicates that the relatively new D4.7 repeat allele
originated as a rare mutational event (or events) that nevertheless increased to high
frequency in human populations by positive selection.76,77 The estimated origin of
this event is approximately ~40,000 to 50,000 years ago, a crucial period in the
history of our species when, in one key wave of migration, there was a spread of
modern humans out of Africa. Ding et al.76 hypothesize that the currently observed
frequencies of the D4.4 and D4.7 alleles of this two-allele system are an example
of balanced selection. We conjecture that the balanced maintenance of both the D4.4
and D4.7 repeats is related to the need for diverse behavioral phenotypes in human
populations partially determined by this gene, altruistic and prosocial (D4.4) versus
a more aggressive, novelty seeking or perhaps even antisocial type (D4.7). Indeed,
models suggest that populations composed of extremely altruistic individuals would
be unstable.44
18.5 THE SOCIAL PERSONALITY

Although progress has been made in unraveling the molecular genetic architecture
of individual personality traits,30,31,78,79 little is known regarding the genetic basis of
social behaviors (measures of the interaction between at least two individuals) in
humans. Much more has been elucidated on the genetic control of social behavior
in animals, including insects and lower vertebrates.8082 In particular, research over
the past two decades has revealed the molecular mechanisms by which two peptide
hormones, vasopressin and oxytocin, shape social behavior in species from fish to
rodents.83 However, despite speculation,84 little evidence has been forthcoming link-
ing these hormones to corresponding human behaviors.
Arginine-vasopressin (AVP) in rodents has been associated with the modulation
of a broad range of behavioral phenotypes including social recognition and learning,
affiliative behaviors, aggression, dominantsubordinate relationships, parental
behavior, grooming, and categories of pair bonding such as monogamy.85 Of partic-
ular interest for the present study, however, are affiliative behaviors that interact with
but are distinct from reproductive pair bonding. For example, in squirrel monkeys
vasopressin has been shown to modulate malemale interactions,86 and in humans
a facet of aggressive behavior.87 Extrapolating from these studies, we suggest that
vasopressin might modulate a range of human behaviors distinct from romantic pair
bonding as observed in the prairie vole. Strengthening this notion is a recent study
showing linkage between the AVPR1A receptor and autism, a behavior disorder of
which dysfunctional social interaction is a core symptom.88
In our study,89 we examined linkage between microsatellites in the vicinity of
the AVPR1A receptor, which in animal studies regulates brain regional expression
of this gene,90 and self-report questionnaires designed to measure two aspects of
complex social behavior, self-presentation91 and sibling relationships.92 We hypoth-
esized that only these two social constructs, among those inventoried in this group
of subjects, represent human equivalents of behaviors observed in animals and likely
modulated by vasopressin. None of the other measures of behavior we inventoried
were significantly associated with the vasopressin receptor except for abnormal
eating behavior.93
We evaluated the perceived quality of the relationship between siblings using
the Sibling Relationship Questionnaire (SRQ)94 and assessed three dimensions
including: (a) relative status/power (b) warmth/closeness, and (c) conflict. In a
factor analysis of the original questionnaire,94 the authors found the following non-
orthogonal factor pattern: conflict is manifested as quarreling, competition and
antagonism. Warmth/closeness is manifested as intimacy, prosocial behavior, com-
panionship, admiration, perceived similarity and affection. Relative status/power
refers to the degree and direction of asymmetry in the sibling relationship, one
extreme indicating greater power by one sibling, and the other greater power by
the other sibling.
Snyder95 first proposed that people differ in the extent to which they monitor
(observe and regulate) their social self and balance their inner directives as opposed
to their perceptions of how others expect them to behave. Lennox and Wolfe91 revised
Snyders original self-monitoring scale into the revised self-monitoring scale
(RSMS) and the concern for appropriateness scale (CFA), which together comprise
the social psychological construct of self-presentation. The two have generally been
found to be orthogonal96,97 and more psychometrically sound than Snyders scale.98
The RSMS measures acquisitive self-presentation, or getting ahead, an active and
flexible social approach aimed at gaining power and status. The CFA measures a
defensive and fearful social approach associated with conformity, aimed at gaining
acceptance and approval and avoiding social threats. Whereas both self-presenta-
tional styles reflect social orientations with a high degree of concern for social cues
and social approval, the RSMS correlates positively with measures of social adap-
tation such as self-esteem and extraversion,97 whereas the CFA correlates negatively
with such measures.
As discussed above, AVPR1A molds social behavior across the vertebrates, and
it makes sense that self-presentation, a construct that characterizes the style (dom-
inating, fitting in) an individual adopts in participating in group activities should
also be linked to this receptor. We further hypothesized that the SRQ construct of
conflict, and possibly that of power, would tap into the aggressive style of behav-
iors that are also molded by vasopressin.
We also thought it of some interest to examine sibling relationships because
such relationships are often the first dyadic social relationship that individuals
experience, and the style people adopt with their siblings may have ramifications
for their subsequent relationships with nonrelated persons. Indeed, we observe in
this cohort a correlation between the CFA and two of the SRQ scales (conflict and
warmth).
Suggestive linkage was observed between both microsatellites (RS1 and RS3)
and the SRQ Conflict Scale (RS1: chi-square = 13.65, LOD = 2.96, p = 0.0001;
RS3: chi-square = 14.54, LOD = 3.16, p = 0.00007) and the CFA self-presentational
style (RS1: chi-square = 8.25, LOD = 1.79 p = 0.002; RS3: chi-square = 8.81,
LOD = 1.91, p = 0.002).
The provisional linkage observed between microsatellites in the AVPR1A pro-
moter region and scores on the CFA and the SRQ conflict scale constitutes the first
substantial evidence that vasopressin, which mediates species-specific social behav-
ior across the vertebrates, may play a similar role in humans. Although social
behavior and the acquisition of social skills in humans constitute the cornerstone of
society and culture, few if any genetic studies have attempted to relate individual
differences in social skills to specific genes. Moreover, social skills are relevant to
a variety of psychiatric disorders including autism, schizophrenia, and externalizing
behavior problems in children. From an evolutionary perspective, it is not surprising
that a hormone that almost universally affects a spectrum of social and affiliative
behaviors in lower animals85 has a parallel role in the human, a social animal highly
dependent on group interactions for individual and species survival.
18.6 LINKAGE STUDIES

Four studies99102 have now identified a broad region on chromosome 8p that harbors
a locus that contributes to individual differences in a personality trait that is a measure
of emotional lability. Two studies found linkage to harm avoidance99,101 and used
identical self-report questionnaires, the TPQ, whereas the Fullerton et al.100 report
used Eysencks psychoticism scale and found linkage to neuroticism.103 Subjects who
score high on harm avoidance can be described as worrying and pessimistic, fearful
and doubtful, and shy and fatigable. High scorers on Eysencks neuroticism dimension
can be similarly described. The Cloninger et al.99 genome scan used subjects recruited
from families with an alcoholic proband as part of the COGA study.104 The highest
LOD score (3.2) using 291 microsatellite markers for a site to harm avoidance was
near marker D8S1106 (8p 21-23) at 27 cM. The Fullerton British study genotyped
561 extremely discordant and concordant sibling pairs selected from a population of
34,580 families. The observed linkage to neuroticism near marker D8S277 (8 cM)
gave a LOD score of 2.9.
In our own study102 we have now genotyped altogether 24 microsatellite markers
in 377 families. Using three methods (maximum likelihood binomial or MLB, MER-
LIN and an associated one parameter model) we observed significant results (p-values
from 0.002 to 0.0004) for linkage to harm avoidance in this region. A peak multipoint
LOD score of 2.76 (p-value 0.0002) was obtained with the MLB method. The region-
wide empirical p-value was 0.002 [0.001 0.0046]. Although the peak position varied
somewhat according to the method (D8S1048 for MLB, D8S1463 for the two other
methods), for three methods D8S1810 (~60 cM) is within 1 to 2 cM of the peak for
harm avoidance. This marker is of particular interest since it is proximate (<0.5 cM)
to the core haplotype that in several recent studies shows significant association with
schizophrenia near neuroregulin 1.105,106 Although association studies with microsat-
ellite markers need to be interpreted cautiously, using the Haplotype Trend Regression
test one marker, D8S499 (~60 cM), showed an empirical p-value of 2 105 for allele
3, which confers a decreased harm avoidance score.
Altogether, our current linkage and association results suggest the possibility
that the same locus near the neuroregulin 1 gene on chromosome 8p confers risk
for both an anxiety-related personality trait as well as schizophrenia. We hypothesize
that this common genetic factor may contribute to emotional lability during early
development which constitutes a predisposing factor for major psychosis.
Curtis et al.108 recently reanalyzed the original dataset first reported by Cloninger et
99
al. using a new method of linkage analysis for quantitative traits that deals with extended
pedigrees. As well as supporting linkage of HA to D8S549 as originally reported,99 this
method also produces an MALOD of 2.4 (p = 0.002) near DBH and several positive
LOD scores for novelty seeking, the largest being MALODs of 3.1 (p = 0.0004) near
D12S391 and 3.4 (p = 0.0003) near D17S1299. There is no support for linkage of novelty
seeking or HA to the regions around DRD4 and 5-HTTLPR, respectively.
Neale et al.107 analyzed a genome scan for neuroticism on a sample of 129 sib-
pair families (113 with a single sibling pair, 18 with multiple sibling pairs) containing
a total of 201 possible sibling pairs, ascertained for concordance on nicotine depen-
dence. Using Merlin-Regress, and replicated peaks for neuroticism described by
prior studies, on chromosomes 1 and 11 with LOD scores of 2.52 and 1.97 (p-value
of 0.003 and 0.0108), respectively, and have some evidence for a novel finding on
chromosome 12 with a LOD of 2.85 (p-value of 0.0014).
18.7 IMPORTANCE
Advances in the molecular genetics of human personality could have major
implications for both psychology and psychiatry. In the field of psychology,
alleles or alleleallele interactions affecting personality traits with large effect
sizes would invite renewed explorations of the neurochemical or neuroanatomical
effects involved. Discoveries of such effects could potentially revolutionize under-
standing of the physical bases of the complex and hitherto intractable phenotypes
of personality traits. Such discoveries might also raise the specter of a pill for,
or against, every emotion or trait, a sort of plastic surgery of the psyche. History
provides ample evidence of our readiness to use and abuse psychoactive sub-
stances. Thus far, however, findings have been limited to modest effect sizes on
monoamine systems that were already predicted by knowledge outside the field
of genetics.
In psychiatry, chromosomal regions or interactions between regions, or alleles
or alleleallele interactions, affecting personality traits with large effect sizes would
invite new approaches to genetics for mental illnesses. We have already alluded to
the idea that mental illnesses may be extreme manifestations of normal personality
traits, or of maladaptive combinations of these traits. In analogous fashion, idio-
pathic hypertension and adult-onset diabetes may represent harmful expressions of
genes, or harmful interactions between genes, that originally evolved to code for
proteins that enable normal blood pressure and carbohydrate metabolism. Modern
environments may differ from those in which those genes were adaptive. Genes
affecting personality traits may also have been selected when both the external
environment and intrapersonal and interpersonal existence differed from modern
experience. Discoveries of chromosomal regions or alleles affecting, for example,
harm avoidance, invite investigations of those same regions or alleles in clinical
anxiety disorders. In practice, findings so far have been so few that any promising
leads probably will be followed up with initial investigations in all or almost all
psychiatric disorders.
The region on chromosome 8 implicated in harm avoidance will undoubtedly
be further scrutinized for linkage to psychiatric disorders. To the extent that future
genetic findings in personality are unexpected, they may suggest novel approaches
to psychiatric illnesses.
ACKNOWLEDGMENTS
This research was partially supported by the Israel Science Foundation founded by
the Israel Academy of Sciences and Humanities (RPE).
REFERENCES
1. Benjamin, J., Ebstein, R.P. & Belmaker, R., H. Molecular Genetics and the Human
Personality, 355 (American Psychiatric Publishing, Inc., Washington, D.C., 2002).
2. Butcher, J.N. MMPI-2: Minnesota Multiphasic Personality Inventory-2: manual for
administration, scoring, and interpretation, ix, 212 p. (University of Minnesota Press,
Minneapolis, 2001).
3. Cloninger, C.R. A unified biosocial theory of personality and its role in the develop-
ment of anxiety states. Psychiatr Dev 4, 167226 (1986).
4. Costa, P.T., Jr. & McCrae, R.R. Stability and change in personality assessment: the
revised NEO Personality Inventory in the year 2000. J Pers Assess 68, 8694 (1997).
5. Jang, K.L., Livesley, W.J. & Vernon, P.A. Heritability of the big five personality
dimensions and their facets: a twin study. J Pers 64, 57791 (1996).
6. Heiman, N., Stallings, M.C., Hofer, S.M. & Hewitt, J.K. Investigating age differences
in the genetic and environmental structure of the tridimensional personality question-
naire in later adulthood. Behav Genet 33, 17180 (2003).
7. Heath, A.C., Cloninger, C.R. & Martin, N.G. Testing a model for the genetic structure
of personality: a comparison of the personality systems of Cloninger and Eysenck.
J Pers Soc Psychol 66, 76275 (1994).
8. Bergeman, C.S. et al. Genetic and environmental effects on openness to experience,
agreeableness, and conscientiousness: an adoption/twin study. J Pers 61, 15979
(1993).
9. Ebstein, R.P. et al. Dopamine D4 receptor (D4DR) exon III polymorphism associated
with the human personality trait of Novelty Seeking. Nat Genet 12, 7880 (1996).
10. Benjamin, J. et al. Population and familial association between the D4 dopamine
receptor gene and measures of Novelty Seeking. Nat Genet 12, 814 (1996).
11. Lesch, K.P. et al. Association of anxiety-related traits with a polymorphism in the
serotonin transporter gene regulatory region [see comments]. Science 274, 152731
(1996).
12. Lusher, J.M., Chandler, C. & Ball, D. Dopamine D4 receptor gene (DRD4) is asso-
ciated with Novelty Seeking (NS) and substance abuse: the saga continues. Mol
Psychiatry 6, 4979 (2001).
13. Comings, D.E. et al. Studies of the 48 bp repeat polymorphism of the DRD4 gene
in impulsive, compulsive, addictive behaviors: Tourette syndrome, ADHD, patholog-
ical gambling, and substance abuse. Am J Med Genet 88, 35868 (1999).
14. Mel, H. et al. Additional evidence for an association between the dopamine D4
receptor (D4DR) exon III seven-repeat allele and substance abuse in opioid dependent
subjects: Relationship of treatment retention to genotype and personality. Addiction
Biology 3, 473481 (1998).
15. Kotler, M. et al. Excess dopamine D4 receptor (D4DR) exon III seven repeat allele
in opioid-dependent subjects. Mol Psychiatry 2, 2514 (1997).
16. Li, T. et al. Association analysis of the dopamine D4 gene exon III VNTR and heroin
abuse in Chinese subjects. Mol Psychiatry 2, 4136 (1997).
17. Muramatsu, T., Higuchi, S., Murayama, M., Matsushita, S. & Hayashida, M. Asso-
ciation between alcoholism and the dopamine D4 receptor gene. J Med Genet 33,
1135 (1996).
18. Lerman, C. et al. Depression and self-medication with nicotine: the modifying influ-
ence of the dopamine D4 receptor gene [see comments]. Health Psychol 17, 5662
(1998).
19. Shields, P.G. et al. Dopamine D4 receptors and the risk of cigarette smoking in
African-Americans and Caucasians. Cancer Epidemiol Biomarkers Prev 7, 4538
(1998).
20. Lerman, C. et al. Interacting effects of the serotonin transporter gene and neuroticism
in smoking practices and nicotine dependence. Mol Psychiatry 5, 18992 (2000).
21. Hu, S., Brody, C.L., Fisher, C. et al. Interaction between the serotonin transporter
gene and neuroticism in cigarette smoking behavior. Mol Psychiatry 5, 1818 (2000).
22. Faraone, S.V., Doyle, A.E., Mick, E. & Biederman, J. Meta-analysis of the association
between the 7-repeat allele of the dopamine D(4) receptor gene and attention deficit
hyperactivity disorder. Am J Psychiatry 158, 10527 (2001).
23. Cadoret, R.J. et al. Associations of the serotonin transporter promoter polymorphism
with aggressivity, attention deficit, and conduct disorder in an adoptee population.
Compr Psychiatry 44, 88101 (2003).
24. Kent, L. et al. Evidence that variation at the serotonin transporter gene influences
susceptibility to attention deficit hyperactivity disorder (ADHD): analysis and pooled
analysis. Mol Psychiatry 7, 90812 (2002).
25. Retz, W., Thome, J., Blocher, D., Baader, M. and Rosler, M. Association of attention
deficit hyperactivity disorder-related psychopathology and personality traits with the
serotonin transporter promoter region polymorphism. Neurosci Lett 319, 1336 (2002).
26. Manor, I. et al. Family-based association study of the serotonin transporter promoter
region polymorphism (5-HTTLPR) in attention deficit hyperactivity disorder. Am J
Med Genet 105, 915. (2001).
27. Arbelle, S. et al. Relation of shyness in grade school children to the genotype for the
long form of the serotonin transporter promoter region polymorphism. Am J Psychi-
atry 160, 6716 (2003).
28. Hariri, A.R. et al. Serotonin transporter genetic variation and the response of the
human amygdala. Science 297, 4003 (2002).
29. Caspi, A. et al. Influence of life stress on depression: moderation by a polymorphism
in the 5-HTT gene. Science 301, 3869 (2003).
30. Kluger, A.N., Siegfried, Z. & Ebstein, R.P. A meta-analysis of the association between
DRD4 polymorphism and novelty seeking. Mol Psychiatry 7, 7127 (2002).
31. Schinka, J.A., Letsch, E.A. & Crawford, F.C. DRD4 and novelty seeking: results of
meta-analyses. Am J Med Genet 114, 6438 (2002).
32. Sen, S., Burmeister, M. & Ghosh, D. Meta-analysis of the association between a
serotonin transporter promoter polymorphism (5-HTTLPR) and anxiety-related per-
sonality traits. Am J Med Genet 127B, 859 (2004).
33. Schinka, J.A., Busch, R.M. & Robichaux-Keene, N. A meta-analysis of the associa-
tion between the serotonin transporter gene polymorphism (5-HTTLPR) and trait
anxiety. Mol Psychiatry 9, 197202 (2004).
34. Okuyama, Y. et al. Identification of a polymorphism in the promoter region of DRD4
associated with the human novelty seeking personality trait. Mol Psychiatry 5, 649
(2000).
35. Buskila, D. Fibromyalgia, chronic fatigue syndrome, and myofascial pain syndrome.
Curr Opin Rheumatol 13, 11727 (2001).
36. Offenbaecher, M. et al. Possible association of fibromyalgia with a polymorphism in
the serotonin transporter gene regulatory region. Arthritis Rheum 42, 24828 (1999).
37. Cohen, H., Buskila, D., Neumann, L. & Ebstein, R.P. Confirmation of an association
between fibromyalgia and serotonin transporter promoter region (5- HTTLPR) poly-
morphism, and relationship to anxiety-related personality traits. Arthritis Rheum 46,
8457 (2002).
38. Ebstein, R.P. et al. No association between the serotonin transporter gene regulatory
region polymorphism and the Tridimensional Personality Questionnaire (TPQ) tem-
perament of harm avoidance. Mol Psychiatry 2, 2246 (1997).
39. Buskila, D., Cohen, h., Neuman, L. & Ebstein, R.P. An association between fibro-
myalgia and the dopamine D4 receptor exon III repeat polymorphism and relationship
to novelty seeking personality traits. Mol Psychiatry 9, 7301 (2004).
40. Bachner-Melman, R. et al. Dopaminergic polymorphisms associated with self-report
measures of human altruism: A fresh phenotype for the dopamine D4 receptor. Mol
Psychiatry 10, 3335 (2005).
41. Gergen, K.-J., Gergen, M.-M. & Meter, K. Individual orientations to prosocial behav-
ior. Journal of Social Issues 28, 105130 (1972).
42. Batson, C.D. The Altruism Question: Toward a Social Psychological Answer, ix, 257
p. (L. Erlbaum, Associates, Hillsdale, N.J., 1991).
43. Hamilton, W.D. The genetical evolution of social behaviour. I. J Theor Biol 7, 116
(1964).
44. Sigmund, K. & Hauert, C. Primer: altruism. Curr Biol 12, R2702 (2002).
45. Fehr, E. & Fischbacher, U. The nature of human altruism. Nature 425, 78591 (2003).
46. Eisenberg, N. Prosocial behavior, empathy, and sympathy. In Well Being: Positive
Development across the Life Course. Crosscurrents in Contemporary Psychology
(eds. Bornstein, M.H. & Davidson, L.) 253265 (Lawrence Erlbaum Associates,
Publishers, Mahwah, NJ, 2003).
47. Emde, R.N. et al. Temperament, emotion, and cognition at fourteen months: the
MacArthur Longitudinal Twin Study. Child Dev 63, 143755 (1992).
48. Loat, C.S., Asbury, K., Galsworthy, M.J., Plomin, R. & Craig, I.W. X inactivation as
a source of behavioural differences in monozygotic female twins. Twin Res 7, 5461
(2004).
49. Rushton, J.P., Fulker, D.W., Neale, M.C., Nias, D.K. & Eysenck, H.J. Altruism and
aggression: the heritability of individual differences. J Pers Soc Psychol 50, 11928
(1986).
50. Loehlin, J.C. & Nichols, R.C. Heredity, Environment, & Personality: A Study of 850
Sets of Twins, xii, 202 p. (University of Texas Press, Austin, 1976).
51. Matthews, K.-A., Batson, C.D., Horn, J. & Rosenman, R.-H. Principles in his nature
which interest him in the fortune of others: The heritability of empathic concern for
others. Journal of Personality 49, 237247 (1981).
52. Zahn-Waxler, C., Robinson, J.-L. & Emde, R.-N. The development of empathy in
twins. Developmental Psychology 28, 10381047 (1992).
53. Eisenberg, N. et al. Prosocial development in early adulthood: a longitudinal study.
J Pers Soc Psychol 82, 9931006 (2002).
54. Eisenberg, N. et al. Consistency and development of prosocial dispositions: a longi-
tudinal study. Child Dev 70, 136072 (1999).
55. Bandura, A., Caprara, G.V., Barbaranelli, C., Gerbino, M. & Pastorelli, C. Role of
affective self-regulatory efficacy in diverse spheres of psychosocial functioning. Child
Dev 74, 76982 (2003).
56. Seeman, T.E. Health promoting effects of friends and family on health outcomes in
older adults. Am J Health Promot 14, 36270 (2000).
57. Caprara, G.V., Barbaranelli, C., Pastorelli, C., Bandura, A. & Zimbardo, P.G. Proso-
cial foundations of children's academic achievement. Psychol Sci 11, 3026 (2000).
58. Bachar, E. et al. Rejection of life in anorexic and bulimic patients. International
Journal of Eating Disorders 31, 4348 (2001).
59. Cloninger, C.R. A systematic method for clinical description and classification of
personality variants. A proposal. Arch Gen Psychiatry 44, 57388 (1987).
60. Lowe, N. et al. Joint Analysis of the DRD5 Marker Concludes Association with
Attention-Deficit/Hyperactivity Disorder Confined to the Predominantly Inattentive
and Combined Subtypes. Am J Hum Genet 74, 34856 (2004).
61. Gaunt, T.R., Cooper, J.A., Miller, G.J., Day, I.N. & O'Dell, S.D. Positive associations
between single nucleotide polymorphisms in the IGF2 gene region and body mass
index in adult males. Hum Mol Genet 10, 1491501 (2001).
62. Quesada, A. & Micevych, P.E. Estrogen interacts with the IGF-1 system to protect
nigrostriatal dopamine and maintain motoric behavior after 6-hydroxdopamine
lesions. J Neurosci Res 75, 10716 (2004).
63. Shavali, S., Ren, J. & Ebadi, M. Insulin-like growth factor-1 protects human dopam-
inergic SH-SY5Y cells from salsolinol-induced toxicity. Neurosci Lett 340, 7982
(2003).
64. Stefansson, H. et al. Neuregulin 1 and susceptibility to schizophrenia. Am J Hum
Genet 71, 87792 (2002).
65. Bach, L.A. & Leeding, K.S. Insulin-like growth factors decrease catecholamine
content in PC12 rat pheochromocytoma cells. Horm Metab Res 34, 48791 (2002).
66. Offen, D. et al. Protective effect of insulin-like-growth-factor-1 against dopamine-
induced neurotoxicity in human and rodent neuronal cultures: possible implications
for Parkinson's disease. Neurosci Lett 316, 12932 (2001).
67. Clarkson, E.D. et al. IGF-I and bFGF improve dopamine neuron survival and behav-
ioral outcome in parkinsonian rats receiving cultured human fetal tissue strands. Exp
Neurol 168, 18391 (2001).
68. Guan, J. et al. N-terminal tripeptide of IGF-1 (GPE) prevents the loss of TH positive
neurons after 6-OHDA induced nigral lesion in rats. Brain Res 859, 28692 (2000).
69. Zawada, W.M. et al. Growth factors improve immediate survival of embryonic dopam-
ine neurons after transplantation into rats. Brain Res 786, 96103 (1998).
70. Hemberger, M. et al. H19 and Igf2 are expressed and differentially imprinted in
neuroectoderm-derived cells in the mouse brain. Dev Genes Evol 208, 393402.
71. Pham, N.V., Nguyen, M.T., Hu, J.F., Vu, T.H. & Hoffman, A.R. Dissociation of IGF2
and H19 imprinting in human brain. Brain Res 810, 18 (1998).
72. Mori, H. et al. Expression of mouse igf2 mRNA-binding protein 3 and its implications
for the developing central nervous system. J Neurosci Res 64, 13243 (2001).
73. Albrecht, S. et al. Variable imprinting of H19 and IGF2 in fetal cerebellum and
medulloblastoma. J Neuropathol Exp Neurol 55, 12706 (1996).
74. Pera, E.M., Wessely, O., Li, S.Y. & De Robertis, E.M. Neural and head induction by
insulin-like growth factor signals. Dev Cell 1, 65565 (2001).
75. Pera, E.M., Ikeda, A., Eivers, E. & De Robertis, E.M. Integration of IGF, FGF, and
anti-BMP signals via Smad1 phosphorylation in neural induction. Genes Dev 17,
30238 (2003).
76. Ding, Y.C. et al. Evidence of positive selection acting at the human dopamine receptor
D4 gene locus. Proc Natl Acad Sci U S A 99, 309314 (2002).
77. Wang, E. et al. The Genetic Architecture of Selection at the Human Dopamine
Receptor D4 (DRD4) Gene Locus. Am J Hum Genet 74, 931944 (2004).
78. Bouchard, T.J., Jr. & Loehlin, J.C. Genes, evolution, and personality. Behav Genet
31, 24373 (2001).
79. Ebstein, R.P., Benjamin, J. & Belmaker, R.H. Personality and polymorphisms of
genes involved in aminergic neurotransmission. Eur J Pharmacol 410, 205214
(2000).
80. Giraud, T., Pedersen, J.S. & Keller, L. Evolution of supercolonies: the Argentine ants
of southern Europe. Proc Natl Acad Sci U S A 99, 60759 (2002).
81. Krieger, M.J. & Ross, K.G. Identification of a major gene regulating complex social
behavior. Science 295, 32832 (2002).
82. Keller, L. & Parker, J.D. Behavioral genetics: a gene for supersociality. Curr Biol 12,
R1801 (2002).
83. Young, L.J., Lim, M.M., Gingrich, B. & Insel, T.R. Cellular mechanisms of social
attachment. Horm Behav 40, 1338 (2001).
84. Taylor, S.E. et al. Biobehavioral responses to stress in females: tend-and-befriend,
not fight-or-flight. Psychol Rev 107, 41129 (2000).
85. Ferguson, J.N., Young, L.J. & Insel, T.R. The neuroendocrine basis of social recog-
nition. Front Neuroendocrinol 23, 20024 (2002).
86. Winslow, J. & Insel, T.R. Vasopressin modulates male squirrel monkeys' behavior
during social separation. Eur J Pharmacol 200, 95101 (1991).
87. Thompson, R., Gupta, S., Miller, K., Mills, S. & Orr, S. The effects of vasopressin
on human facial responses related to social communication. Psychoneuroendocrinol-
ogy 29, 3548 (2004).
88. Kim, S.J. et al. Transmission disequilibrium testing of arginine vasopressin receptor
1A (AVPR1A) polymorphisms in autism. Mol Psychiatry 7, 5037 (2002).
89. Bachner-Melman, R. et al. Linkage between vasopressin receptor AVPR1A promoter
region microsatellites and measures of social behavior in humans. Journal of Indi-
vidual Differences 36, 45160 (2005).
90. Hammock, E.A. & Young, L.J. Variation in the vasopressin V1a receptor promoter
and expression: implications for inter- and intraspecific variation in social behaviour.
Eur J Neurosci 16, 399402 (2002).
91. Lennox, R.-D. & Wolfe, R.-N. Revision of the Self-Monitoring Scale. Journal of
Personality and Social Psychology 46, 13491364 (1984).
92. Buhrmester, D. & Furman, W. Perceptions of sibling relationships during middle
childhood and adolescence. Child Dev 61, 138798 (1990).
93. Bachner-Melman, R. et al. Association between a vasopressin receptor AVPR1A
promoter region microsatellite and eating behavior measured by a self-report ques-
tionnaire (Eating Attitudes Test) in a family-based study of a non-clinical population.
Int J Eat Disord 36, 45160 (2004).
94. Furman, W. & Buhrmester, D. Children's perceptions of the qualities of sibling
relationships. Child Dev 56, 44861 (1985).
95. Snyder, M. Self-monitoring of expressive behavior. Journal of Personality and Social
Psychology 30, 526537 (1974).
96. Wolfe, R.N., Lennox, R.D. & Cutler, B.L. "Getting along" and "getting ahead":
Empirical support for a theory of protective and acquisitive self-presentation. Journal
of Personality and Social Psychology 50, 356361 (1986).
97. Miller, M.L., Omens, R.S. & Delvadia, R. Dimensions of social competence: Per-
sonality and coping style correlates. Personality and Individual Differences 12,
955964 (1991).
98. Day, D., Schleicher, D.J., Unckless, A.L. & Hiller, N.J. Self-Monitoring personality
at work: A meta-analytic investigation of construct validity. Journal of Applied Psy-
chology 87, 390401 (2002).
99. Cloninger, C.R. et al. Anxiety proneness linked to epistatic loci in genome scan of
human personality traits. Am J Med Genet 81, 3137 (1998).
100. Fullerton, J. et al. Linkage analysis of extremely discordant and concordant sibling
pairs identifies quantitative-trait Loci that influence variation in the human personality
trait neuroticism. Am J Hum Genet 72, 87990 (2003).
101. Zohar, A.H. et al. Tridimensional personality questionnaire trait of harm avoidance
(anxiety proneness) is linked to a locus on chromosome 8p21. Am J Med Genet 117B,
669 (2003).
102. Dina, C. et al. Fine mapping of a region on chromosome 8p gives evidence for a
QTL contributing to individual differences in an anxiety-related personality trait, TPQ
harm avoidance. Am J Med Genet B 132, 1048 (2004).
103. Eysenck, H.J. The Scientific Study of Personality, (Routledge & Kegan Paul, London,
1952).
104. Schuckit, M.A. et al. A genome-wide search for genes that relate to a low level of
response to alcohol. Alcohol Clin Exp Res 25, 3239 (2001).
105. Stefansson, H. et al. Neuregulin 1 and susceptibility to schizophrenia. Am J Hum
Genet 71, 87792 (2002).
106. Stefansson, H. et al. Association of neuregulin 1 with schizophrenia confirmed in a
Scottish population. Am J Hum Genet 72, 837 (2003).
107. Neale, B.M., Sullivan, P.F. & Kendler, K.S. A genome scan of neuroticism in nicotine
dependent smokers. Am J Med Genet B 132, 659 (2005).
108. Curtis, D. Re-analysis of collaborative study on the genetics of alcoholism pedigrees
suggests the presence of loci influencing noveltyseeking near D12S391 and
D17S1299. Psych Gen 14, 1515 (2004).
19 Aggression: Concepts and

Methods Relevant to
Genetic Analyses in Mice
and Humans
Stephen C. Maxson and Andrew Canastar
CONTENTS
Summary ................................................................................................................ 281
19.1 Introduction .................................................................................................. 282
19.2 Mice.............................................................................................................. 282
19.3 Humans......................................................................................................... 286
19.4 Mice and Humans ........................................................................................ 287
References.............................................................................................................. 288
SUMMARY
Aggression is a complex, social behavior involving at least two individuals. Some
conceptual and methodological issues relevant to genetic analyses of aggression in
mice and humans are considered.
In animals, there are several types of aggression. In male and female mice, these
are offense, defense, infanticide, and predation. In females, there are also types of
aggression associated with pregnancy and lactation. There are strain differences for
each type, and genes have been identified for offense and defense. At present, 38
genes have been reported to affect some aspect of offense. There are also environ-
mental effects that include prenatal maternal environment, postnatal maternal envi-
ronment, postweaning environment, and test situation. All of these must be considered
in critically evaluating experimental data and experimental design for the genetics of
a type of aggression. The behaviors observed and measured are also a factor. Some
of these issues are discussed in detail for offense in males.
Although there may be different types of aggression in humans, it is difficult to
identify these. Consequently, aggression in humans is often treated as a single,
unitary trait. This lumping together of different types of aggression may render
difficult the interpretation and evaluation of genetic analyses for human aggression.
Regardless, there is some evidence that individual differences in personality traits
associated with aggressiveness are due to genetic variants. Also, three genes may
281
have been shown to have effects on impulsive aggression, and there is a genotype
interaction for one of these. However, there is no proof at present that criminal
behaviors associated with physical assault are substantially heritable, other than as
interactions of specific genes with specific environments.
19.1 INTRODUCTION
There are many definitions of aggression and of aggressive behavior. One of these
is overt behavior involving intent to inflict noxious stimulation to or behave destruc-
tively toward another organism.28 Another of these is any form of behavior directed
toward the goal of harming or injuring another living being who is motivated to
avoid such treatment.3 These definitions specify broadly the behavioral domain
considered in this chapter.
It has long been recognized that, at least in animals, there are many types of
aggression or aggressive behavior. These differ in mechanism, eliciting stimuli,
development, function, and phylogeny. Consequently, they may also differ in their
genetics. That is to say, the same genes may not be causes of differences or act in
development of each type of aggression.
Regardless, aggression has long been of interest as a behavioral trait for genetic
analyses.16,17 Although most of the research on genetics of aggression has been done
with mice, much of the scientific and public interest in the genetics of aggression is
sparked by concerns about the role of nature and nurture in human aggression. This
includes the causes of not only adaptive types of aggression, but also aggression in
psychopathologies, crime, and war.5,28 Here, some conceptual and methodological
issues involved in research on the genetics of aggression are considered, first for mice
and then for humans.
19.2 MICE
In male and female mice, the four types of aggression are offense, defense, infan-
ticide, and predation.16 There are also female specific types of aggression which
occur during pregnancy or lactation.4 Between strain differences exist for each type
of aggression,10 and genes have been identified for offense and defense.16,19,21,23
Offense and defense by males and females (not pregnant or lactating) are ago-
nistic behaviors. Infanticide and predation by males and females are not agonistic
behaviors. The agonistic behaviors have the common function of adaptation to
situations involving physical conflict between members of the same species. Each
type of aggressive behavior can be characterized in mice by function and bite targets.
These are listed in Table 19.1. For each type, there are also differences in mechanism
and development. Thus, it may be expected that although some genes may affect all
types of aggression, others will affect only one or a few kinds. Also, conceptual and
methodological issues in identifying genes for each type of aggression may be
different. Because most of the current research on the genetics of aggression in mice
focuses on offense in males, this part of the chapter will consider this behavior in
mice. Information on the other types of aggressive behavior in mice can be found
in references 2, 6, 14, 16.
Aggression 283
TABLE 19.1
Types of Mouse Aggression
Type Bite Target Function
Offense Back, rump, tail Obtaining and restraining resources
Defense Face, shoulders Self-protection from injury by
others
Pregnancy Anywhere Prevent reproductive termination
Lactation Flanks, neck Prevent reproductive termination
Infanticide Anywhere Reproductive termination
Predation (insects) Head, thorax Food
Elsewhere, we have reviewed in detail the many aspects of the environment that
can have an effect on offensive aggression in mice17,18,21 and that must be considered
in critically evaluating research results and in designing experimental studies on the
genetics of offense. These include prenatal maternal environment, postnatal parental
environment, postweaning environment, and test conditions. Also, the behavioral
measures used are a factor. Effects of test conditions and behavioral measure will
be considered in more detail.
Offense is obviously a social behavior involving at least two mice, and most
genetic analyses of aggression in mice use dyadic tests. For genetic analyses, there
are three types of dyadic tests. These are homogeneous set test, the panel of testers,
and the standard opponent test. In the homogeneous set test, all encounters are
between mice of the same genotype (strain or F1 hybrids). In the panel of testers,
encounters are between mice of the same or different genotype (strains or F1
hybrids), and each experimental group of genotypes (strains or F1 hybrids) is tested
against a panel of genotypes (strains or F1 hybrids). The panel of testees (experi-
mental group) and panel of testers (opponent group) may be of the same or different
array of genotypes (strains or F1 hybrids). Because in both the homogeneous set
test and panel of testers, the genotypes of both individuals in the dyadic encounter
must be known, only isogenic populations may be used in the two paradigms.
Isogenic populations include inbred strains, recombinant inbred strains, consomic
strains, recombinant congenic strains, congenic strains, coisogenic strains, and F1
hybrids of two such strains. With these two paradigms, only recombinant inbred
strains, recombinant congenic, congenic (such as knockout mutants), and coisogenic
strains (such as transgenes and insertional mutants) can be used to identify genes
with effects on offense. In the standard opponent test, all encounters are between
mice of one or another genotype and mice of a single, standard genotype (stock,
strain, or F1 hybrid). Since only the genotype of the standard opponent must be
known, the standard opponent test can be used with mice from isogenic or hetero-
genic populations. The latter include F2, F3, and other Fn hybrids, backcrosses,
heterogeneous stocks, and selected lines. These can be used to map quantitative trait
loci (QTLs) with effects on offense.
Strain or genetic differences in male offense can depend on the genotype not
only of the testee (experimental group), but also of the tester (opponent group). For
example, in a homogeneous set test, the mean number of fights/animal/week was

higher for males of the BALB/cBy strain than for males of the C57BL/6By strain,
whereas in a standard opponent test, the mean number of fights/animal/week was
higher for males of the C57BL/6By than for BALB/cBy males. The standard oppo-
nent was of the CXBH recombinant inbred strain which fought rarely in homoge-
neous set tests. Also, effects of the male-specific part of the Y chromosome on
percent of attacking males depended on the genotype of the opponent. An effect of
the male specific part of the Y chromosome on offense was detected in homogeneous
set tests but not with an inbred mouse strain (A/J) males as the standard opponent.
However, the same genetic effects are sometimes obtained with opponents of
more than one type. For example, males mutant for Nos1 (nitric oxide synthase) are
more aggressive than nonmutant males with opponents of two different genotypes.
Similarly, males mutant for Esr1 (-estrogen receptor) are less aggressive than non-
mutant males with either opponents of the same genotype or with olfactory bulbec-
tomized Swiss-Webster males.
These opponent effects may be mediated by the chemosignals or behavior of
the stimulus mouse. Testosterone-dependent chemosignals of one male are known
to affect the aggressive behavior of another male. There are strain differences in the
potency of these signals and in the receptivity to these signals. For example, different
levels of aggression are elicited from a testee mouse in response to urine of three
inbred strains daubed onto gonadectomized opponents. Strain differences in behavior
of the tester or opponent mouse may also be involved in the differential elicitation
of aggression by the tester or opponent. This has been shown for strains of opponents
that are aggressive or passive.13 Aggressive males elicit different qualities and quan-
tities of fighting than passive males. There is also evidence that standard opponents
of the same genotype but with different treatment to eliminate their attack and
interactive behaviors elicit different patterns of offensive behaviors from testee mice.
These treatments include olfactory bulbectomy, gonadectomy, and experience of
repeated defeats.
Offense is a complex behavior consisting of four motor patterns. These are chase,
sideways offensive posture, upright offensive posture, and attack. The latency, fre-
quency, and duration of each behavior can be measured. The development of video
recording and computer analysis of animal behavior, including offense, has made it
possible to describe, record, and measure all aspects of the motor patterns of offense,
to examine the temporal patterns of behavioral elements within and between episodes
of offensive behaviors, and to assess total time allocated by a mouse to broad
behavioral categories such as nonsocial activity, social investigation, defense, attack,
threats, and displacement activity (digging and self-grooming). Such detailed anal-
ysis of offense is common in studies of drug but not of gene effects.
Many genetic studies of offense have either used a single composite score or a
few measures, usually for the attack component, of this complex social behavior.
This is perhaps necessary when mapping QTLs with segregating populations or for
screening chemically induced mutants. Having many measures in QTL analysis can
be a source of two problems. These are the possibility of more false positives and
of selecting population extremes for DNA genotyping. The selection of single index
measures should be based on phenotypic or genotypic correlations as estimated from
Aggression 285
inbred strains. There are several reports of phenotypic correlations among latency
to attack, number of attacks, and accumulated attack time. Also, it has been reported
that in tests with A/J males as standard opponents, the percentage of males that
attack is correlated with many other measures of offense across several strains.
However, once the QTL is identified and congenic strains for it are established, there
should be complete analyses of this genes effects on offense. This can and should
always be done for congenic or coisogenic strains with knockout mutants, other
mutants, or transgenes.
Just as drugs do not always affect all aspects of offense, genes may not always
do so. For example, it has been reported that the proportion of mice attacking has
a mode of inheritance across several strains and their hybrids different from that for
number of attacks and accumulated attack time. Also, effects of genes on the Y
chromosome on offense are dependent on the behavioral measure. In a standard
opponent test, the CBA/H and NZB Y chromosomes differ in effect on proportion
of males that attack at least once but not in effect on latency to attack. Similarly,
the Nos1 mutant affects the frequency but not latency of attacks in a standard
opponent test. Hence, in both types of research, complete ethological or behavioral
analysis is essential.
These and other methodological issues must be considered in evaluating reports
of gene effects on offense and in designing research to test gene effects on offense.
To date, there is some evidence for effects of 38 genes on male offense.17,21 These
are listed in Table 19.2. Several Slc6a327 and Gnai225 genes have been studied
intensively, leading to proposals for a role of genes expressed in the hippocampus,11
TABLE 19.2
Genes and Male Offense
Gene Name Chromosome
Adora1a Adenosine1A receptor 1
Adora2a Adenosine2A receptor 10
Adra2C Adrenergic alpha 2C receptor 5
Ar Androgen receptor X
Avpr1b Argenine vasopressin 1B receptor 1
B2m Beta2-microglobulin 12
Bcr Breakpoint cluster region 10
CamK2a -Calcium/calmodulin kinase II 18
Ckb Brain creatinine kinase 12
Comt Catecho-o-methyl transferase 16
Cyp19 Aromatase 9
Esr1 -Estrogen receptor 6
Esr2 -Estrogen receptor 12
Fyn Fyn tyrosine kinase 10
Gad1 Glutamnic acid decarboxylase 65 9
Gdi-1 Guanosine diphosphate dissociation inhibitor X
Girgeo-22 Gene trap ROSA-b-Geo 22 10
(Continued)
TABLE 19.2
Genes and Male Offense (Continued)
Gene Name Chromosome
Gnai2 Guanine nucleotide binding protein, Alpha 9
inhibiting 2
Hrh1 Histamine 1 receptor 6
Htr1b 5-HT1B receptor 9
Il-6 Interleukin-6 5
Maoa Monoamine oxidase A X
Mm2 Membrane metalo enopeptidase 3
Ncam Neural cell adhesion molecule 9
Nos1 Nitric oxide synthase 1 5
Nos3 Nitric oxide synthase 3 5
Nr3e1 Nuclear receptor family2 group E member 10
Oxt Oxytocin 2
Penk Enkephalin 4
Pet-1 Pet-1 ETS factor 1
Reg-2 Regulator of G-protein signaling 1
Slc6a3 Dopamine transporter 13
Slc6a4 Serotonin transporter 11
Sts Steroid sulfatase X/Y pseudoautosomal
Tacr1 Tachnykinine-1 receptor 6
Tgfa Transforming growth factor 6
Trpc2 TRP ion channel 7
steroid sulfatase and neurosteroids,15 serotonergic system,9 and effects of hormones

on neuropeptide gene transcription24 in mouse offense.
19.3 HUMANS
Most researchers believe that, as with animals, there are many types of human
aggression. Although some have suggested that offense, defense, and predation are
distinct types of aggression in humans, it is widely recognized that it is difficult to
define the different types of aggression in humans. In part, this is because different
motor patterns of offense, defense, and predation in humans may be masked by the
use of extrinsic weapons. Other distinctions among types of human aggression
include hostile vs. instrumental aggression3 and, within the hostile category, between
impulsive and nonimpulsive aggression.19 Also, there are further distinctions within
any of these categories. These are physical vs. verbal, active vs. passive, and direct
vs. indirect. These different parsing of the aggressive phenotype could have an effect
on the results of genetic analyses. But distinctions are rarely made in research on
the genetics of human aggression. Rather, aggression often is treated as a single
unitary trait. This lumping together of different types of aggression may render
difficult genetic analyses. However, it has recently been suggested but not proven
that all aggression in humans is of the defensive type.1
Aggression 287
The methods for investigating human aggression are described by Baron and
Richardson3 and by Volavka.28 The actual behavior may be directly observed or
inferred from self- or archival reports. Observed behavior is frequently studied in
controlled experiments in the laboratory. More recently, it has also been observed
in normal social and cultural settings. Human aggression is not only investigated as
behavior (for example, the Child Behavior Check List) but also as a personality trait.
Some tests used are the Minnesota Multiphasic Personality Invertory (MMPI) and
Buss-Durkee Hostility Inventory. The implication from results of these tests is that
in the same interpersonal situation, some personality types are more prone than
others to be aggressive.
Many genetic studies of human aggression have been for traits inferred from
personality tests. These have often focused on impulsive or antisocial personality traits
that are associated with aggression.7,26 There is indication that some of the individual
differences in such traits are due, at least in part, to genetic variation. Also, variants
of the genes for monoamine oxidase A (MAOA), tryptophan hydroxylase, and the
5HT1B receptor may be associated with some aspects of impulsivity, including aggres-
sion. Recently, the effects of MAOA variants on measures of male aggression have
been shown to be dependent on maltreatment in childhood8 or heroin addiction.12
Other genetic studies have been concerned with antisocial or criminal behavior
based on archival data. However, the data do not prove that variations in murder,
assault, rape, or other offenses associated with physical attacks are substantially
heritable. The failure to detect a nonzero heritability for criminal behavior involving
physical attacks may be a function of the low incidence of criminal violence. Large
sample sizes are required to detect nonzero heritability for behaviors of low inci-
dence. These have not been available to date for criminal behaviors involving
physical attacks. However, this may also be due to interactions of genotype and
environment.8,12
All of this research should be critically evaluated in the context of the chapters
in this volume on twin studies and family studies. Regardless, it is clear that in
humans as in mice only part of the individual differences in any type of aggression
is due to genetic variants.
19.4 MICE AND HUMANS

Candidate genes for effects on human aggression may be identified from research
on animals, especially mice. We have written several times about the use of genetic
variants with effects on mouse aggression as models for human aggression.16,18,19,20,22
For these to be behavioral models, the aggressive trait in mice and humans must be
the same. For example, genetic variants for offense in mice would be relevant to
genetics of offense but not defense in humans. Thus, there is a great need to identify
the categories of aggression that are the same in mice and humans as well as to
investigate the genetics not only of escalated encounters but also conflict resolution.22
Once this is done, it may be possible to use genetic variants, including single genes,
with effects in mice to search for the genes with effects on each type of aggression
in humans. Similar approaches in other invertebrate and vertebrate animals may also
be of use in developing genetic models of human aggression.22
REFERENCES
1. ALBERT DJ, WALSH ML, JONIK RH: Aggression in humans: What is its biological
foundation? Neuroscience and Biobehavioral Reviews, 1993, 17, 405425.
2. ALLEVA E: Assessment of aggressive behavior in rodents. In: PM Conn (Ed):
Paradigms for the study of behavior. Methods in Neurosciences Vol. 14. Academic
Press, New York, 1993, pp 111137.
3. BARON RA, RICHARDSON DR: Human Aggression 2nd ed. Plenum Press, New
York, 1994.
4. BJORKQVIST K, NIEMELA P: Of Mice and Women: Aspects of Female Aggression.
Academic Press, New York, 1992.
5. BOCK GR, GOODE JA: Genetics of Criminal and Antisocial Behaviour. John Wiley
& Sons, New York, 1996.
6. BRAIN PF, HAUG M, KAMIS A: Hormones and different tests for aggression with
particular reference to the effects of testosterone metabolites. In: J Balthazart, E Prove,
R. Gilles (Eds): Hormones and Behaviour in Higher Vertebrates. Springer-Verlag,
Berlin, 1983, pp 290304.
7. CAREY G: Genetics and violence. In: AJ Reiss Jr., KA Miczek, JA Roth (Eds)
Understanding and Preventing Violence Vol. 2: Biobehavioral Influence. National
Academy Press, Washington, 1994, pp 2158.
8. CASPI A, MCCLAY J, MOFFITT TE, MILL J, MARTIN J, CRAIG I W, TAYLOR
A, POULTON R: Role of genotype in the cycle of violence in maltreated children.
Science, 2002, 297, 851854.
9. CHIAVEGATTO S, NELSON RJ: Interaction of nitric acid and serotonin in aggres-
sive behavior. Hormones and Behavior, 2003, 44, 233241
10. CRAWLEY JN, BELKNAP JK, COLLINS A, CRABBE JC, FRANKEL W, HEND-
ERSON N, HITZEMANN RJ, MAXSON SC, MINER LL, SILVA AJ, WEHNER
JM, WYNSHAW-BORIS A, PAYLOR R: Behavioral phenotypes of inbred mouse
strains. Psychopharmacology, 1998, 132, 107124.
11. Feldker DE, Datson N.A., Veenema AH, Proutski V, LATHOUWERS D , DE KLOET
E R, VREUGDENHIL E: (2003b). Gene chip analysis of hippocampal gene expres-
sion profiles of short- and long-attack-latency mice: technical and biological impli-
cations. Journal of Neuroscience Research, 2003, 74, 701716.
12. GERRA G, GAROFANO L, BOSARI S, PELLEGRINI C, ZAIMOVIC A, MOI G,
BUSSANDRI M, MOI A, BRAMBILLA F, MAMELI A, PIZZAMIGLIO M, DON-
NINI C: Analysis of monoamine oxidase A (MAO-A) promoter polymorphism in
male heroin-dependent subjects: behavioural and personality correlates. J Neural
Transm. 2004, 111, 61121.
13. HAHN ME, SCHANZ N: Issues in the genetics of social behavior: revisited. Behavior
Genetics, 1996, 26, 463470.
14. JONES SE, BRAIN PF: Performance of inbred and outbred laboratory mice in
putative tests of aggression. Behavior Genetics, 1987, 17, 8796.
15. LE ROY I, MORTAUD S, TORDJMAN S, DONSEZ-DARCEL E, CARLIER M,
DEGRELLE H, ROUBERTOUX PL: Genetic correlation between steroid sulfatase
concentration and initiation of attack behavior in mice. Behavior Genetics, 1999, 29,
131136.
16. MAXSON SC: Potential genetic models of aggression and violence in males. In: P
Driscoll (Ed) Genetically Defined Animal Models of Neurobehavioral Dysfunctions.
Birkhauser, Boston, 1992, pp 174188.
Aggression 289
17. MAXSON SC: Methodological issues in genetic analyses of an agonistic behavior

(offense) in male mice. In: D Goldowitz, D Wahlsten, RE Wimer (Eds): Techniques
for the Genetic Analysis of Brain and Behavior: Focus on the Mouse. Elsevier,
Amsterdam, 1992, pp 349373.
18. MAXSON SC: Issues in the search for candidate genes in mice as potential animal
models of human aggression. In: GR Bock and JA Goode (Eds): Genetics of Criminal
and Antisocial Behaviour. John Wiley & Sons, New York, 1996, pp 2130.
19. MAXSON SC: Homologous genes, aggression, and animal models. Developmental
Neuropsychology, 1998, 14, 143156
20. MAXSON SC: Mouse genes and animal models of aggression in humans. In: M
Haug and R Whalen (Eds): Brain, Behavior and Cognition. American Psychological
Association, Washington, 1998, pp 273282.
21. MAXSON SC, CANASTAR A: Conceptual and methodological issues in the genetics
of mouse agonistic behavior. Hormones and Behavior, 2003, 44, 258262.
22. MAXSON SC, CANASTAR A: Genetic aspects of aggression in non-human animals.
In, RJ Nelson (Ed.): Biology of Aggression. Oxford University Press, New York, pp
319.
23. MICZEK KA, MAXSON SC, FISH EW, FACCIDOMO, S: Aggressive behavioral
phenotypes in mice. Behavioral Brain Research, 2001, 125, 167181.
24. NOMURA M, MCKENNA E, KORACH KS, PFAFF, D, OGAWA, S: Estrogen
receptor- regulates transcript levels for oxytocin and arginine vasopressin in the
hypothalamic paraventricular nucleus of male mice. Molecular Brain Research, 2002,
109, 8494.
25. NORLIN EM, GUSSING F, BERGHARD A: Vomeronasal phenotype and behavioral
alterations in G alpha i2 mutant mice. Curr Biol., 2003, 13, 12141219.
26. RHEE SH, WALDMAN ID: Genetic and environmental influences on antisocial
behavior: a meta-analysis of twin and adoption studies. Psychol Bull., 2002, 128,
490529.
27. RODRIGUIZ RM, CHU R, CARON MG, WETSEL WC: Aberrant responses in
social interaction of dopamine transporter knockout mice. Behav Brain Res., 2004,
148, 185198.
28. VOLAVKA, J: Neurobiology of Violence. American Psychiatric Press, Washington,
D.C., 1995.
20 Genetic Analysis of
Emotional Behaviors Using
Animal Models
Andr Ramos and Pierre Mormde
CONTENTS
Summary ................................................................................................................ 291
20.1 Introduction................................................................................................. 291
20.2 Definitions .................................................................................................. 292
20.3 Measurements ............................................................................................. 292
20.3.1 Unconditioned Stress Responses, Forced Exposure...................... 293
20.3.2 Unconditioned Stress Responses, Choice Tests............................. 295
20.3.3 Conditioning Experiments.............................................................. 295
20.3.4 Interpreting Emotionality through a Multivariate Approach......... 296
20.4 Genetic Analysis......................................................................................... 298
20.4.1 Selection Experiments.................................................................... 298
20.4.2 Quantitative Genetic Analysis........................................................ 299
20.4.3 Genetic Links between Different Traits......................................... 299
20.4.4 The QTL Chase .............................................................................. 300
20.4.5 Gene Targeting and Transgenesis .................................................. 303
References.............................................................................................................. 303
SUMMARY
The aim of this chapter is to introduce the reader to the universe of experimental
approaches available for the genetic study of emotionality, the focus being on the
use of animal models. First, some preliminary conceptual issues will be discussed
and then an overview of the main methods of measuring and interpreting the so-
called emotional behaviors will be presented. Finally we present the classical and
modern strategies to determine the genetic basis of variance in emotionality-related
traits.
20.1 INTRODUCTION
Studying the genetic aspects of emotions, in either humans or animal models, is a
particularly challenging task for three main reasons: emotions are difficult to define;
emotions are difficult to measure; and emotions are difficult to genetically analyze.
291
These three types of difficulties, as well as some of the experimental methods

designed to overcome them, will be discussed here. Without intending to be exhaus-
tive, we will introduce throughout the chapter some findings related to the search
for genes influencing emotional responses of laboratory rodents.
20.2 DEFINITIONS
The interpretation of terms derived from the word emotion may vary among authors,
but most researchers agree that emotions are subjective experiences, that they appear
under nonordinary circumstances and that they involve behavioral and physiological
changes.46 In the field of experimental psychobiology, the concept of emotionality
has been used widely for laboratory animals, as defined by Hall.28
The term emotionality is defined as the state of being emotional. This state consists of
a group of organic, experiential and expressive reactions and denotes a general upset or
excited condition of the animal. Emotionality can be thought of as a trait The reader
is warned against interpreting emotionality as a thing or faculty. It is merely a convenient
concept for describing a complex of factors. the term emotion is conventionally used
to designate the experience which results from emotional stimulation.
For being subjective, emotion and emotionality cannot be directly measured.

Hence, scientists often attempt to infer such subjective states from their measurable
behavioral manifestations. In operational terms, the defecation rate in the open field
test (see below) has long been used as the measure of emotionality in the rat. Because
locomotor activity in this same test often covaries negatively with the defecation
rate,17 an animal with high defecation rate and low locomotion in the open field is
usually considered as highly emotional. Nevertheless, a number of experimental data
show that the correlation between these two variables is not always confirmed and
that individual animals or strains may present, for example, a high defecation rate
together with a high locomotor activity.45 In addition, pharmacological studies do
not provide unequivocal support for the use of open field measures as suitable indices
of anxiety,36 a psychological trait or state which is related to emotionality.
Therefore, although the term emotionality, like stress, may be convenient to
describe a broad category of psychological and biological processes, interindividual
or interstrain differences must be defined in a more analytical way. Those differences
are not only quantitative (more or less emotional), but their behavioral and neuroen-
docrine expressions also may vary in nature (qualitative differences). For instance,
highly emotional animals may display either freezing or fleeing behavior, depending
on the wiring between the central emotional state and behavioral outputs. Therefore,
there is a growing perception among researchers that emotionality has in fact several
different dimensions, instead of being a unitary and linear construct.46
20.3 MEASUREMENTS
A number of paradigms have been used to measure emotional behaviors of exper-
imental animals. Basically, these experimental paradigms consist of exposing
Genetic Analysis of Emotional Behaviors Using Animal Models 293
animals to aversive stimuli while observing and analyzing their behavior. The aver-
sive stimuli may vary in nature from being essentially physical (electrical footshocks,
food deprivation, submersion in water, etc.) to those considered to be mainly psy-
chological (novel environments, strongly illuminated areas, open spaces, exposure
to the smell of predators, heights, social instability, etc.). Another important variable
is the ability of the animal to avoid or escape the aversive stimulus. Most experi-
mental situations use unconditioned responses to a number of novel environments,
or learning paradigms.
20.3.1 UNCONDITIONED STRESS RESPONSES, FORCED EXPOSURE

The open field test is most widely used in behavioral research. The apparatus
consists of a large arena, round or square, typically 1 m wide for rats, surrounded
by a wall to prevent escape; the floor is usually marked with lines to allow the
quantification of locomotion. The illumination is variable, and frequently intense to
increase its aversiveness. Behavioral measures include ambulation (peripheral and
in the center), rearings (on the walls and in the open space), grooming, freezing, as
well as defecation and urination (measures of autonomic nervous system activation).
The association of low ambulation with a high rate of defecation, as mentioned
above, has been the operational definition of emotionality in the rat, but this unitary
measure of emotionality has been criticized. Indeed, many studies show that the
emotionality construct, as measured in the open field, is made of several independent
factors or components, the two main factors being motor discharge or activity
(including locomotion and rearing), and autonomic balance or emotional reac-
tivity (measured by the number of fecal boli). Urination primarily would be an
index of territorial marking.54,46 Locomotion can be further divided in central (away
from the walls) and peripheral (when the animals can touch the walls), the former
being more related with fear and anxiety and the latter with general activity.46 Strain
differences in defecation and locomotor activity were described in mice53,54,62 and
in rats45 (Figure 20.1).
Many other novel environment tests are used to characterize the emotional
reactivity of animals. Some of them are smaller with a low level of illumination to
reduce the aversiveness of the environment, which favors the locomotor activity of
the animals. These are known as activity cages, and are usually equipped to record
automatically the movements of the animals. General motor activity can also be
recorded by various devices without moving the animals out of their home cage (for
instance with infrared beams, induction coils, or implanted devices).
Alternatively, the environment can be made more complex to favor other aspects
of behavior and allow the analysis of the different components of activity in novel
environments, such as exploratory activity (hole-board, introduction of objects in
the open field, perforated walls, etc.), or motivational conflicts (presence of food,
water, or conspecifics). One such test is the social interaction test21 based on the
observation that the time spent by pairs of male rats in performing social behaviors
varies with the aversiveness of the environmental conditions. Hyponeophagia is a
conflict test where hungry rats are exposed to food placed in the center of a brightly
lit novel environment. Highly emotional rats are expected to show longer latency to
100
Outer locomotion
75
50
25
0
SHR WKY BN WF LEW FIS
15
Inner locomotion
10
0
5
Fecal boli
0
FIGURE 20.1 Open field. Locomotion in the peripheral (outer) and central (inner) part and
defecation score measured in 12 males each from six rat strains (SHR = Spontaneously
Hypertensive Rat; WKY = Wistar Kyoto; BN = Brown Norway; WF = Wistar Furth; LEW =
Lewis; FIS = Fischer 344). (Reprinted from Behav. Brain Res., 85, Ramos, A. et al., A multiple
test study of anxiety-related behaviours in six inbred rat strains, 5769, Copyright 1997, with
permission from Elsevier.)
approach food and have a lower food consumption.42,62 These tests have been devel-
oped mainly for the screening of anxiolytic drugs.
Among other tests of reactivity, we can cite the acoustic startle reflex26,62 and
ultrasonic pup vocalizations, elicited during the first 2 weeks of life by isolating
the pups from their mother,52 but this list is far from complete.
20.3.2 UNCONDITIONED STRESS RESPONSES, CHOICE TESTS

These tests consist of two (or more) compartments differing in their level of aver-
siveness. The animal can freely move from one compartment to the other(s). For
instance, the black and white box is made of a small, dark compartment and a
large, brightly illuminated area connected by a small passage or door through which
the animal can move freely. Transitions between compartments and general loco-
motion are recorded. Anxiolytic drugs usually increase both measures, antidepres-
sants being inactive.12 Many variants of this test have been described as emergence
tests, the animal being introduced in a safe compartment (usually small and dark,
where the animal can be left for some time to habituate) and the time to emerge
into a more aversive compartment (usually larger, brightly illuminated and unex-
plored previously) is measured.
The elevated plus maze is based on studies showing that rats seem to experience
more fear when exposed to open, elevated alleys than when exposed to enclosed
alleys, as the result of fear naturally induced in rodents by open spaces and/or
elevated areas. This is the most popular test used nowadays for the screening of
anxiolytic drugs in rats.29,61 The apparatus has four elevated arms 50 cm long and
10 cm wide arranged in a crosslike disposition, with two opposite arms enclosed in
high walls and two arms open. At their intersection, a central platform gives access
to any of the four arms. Rats are placed on the central platform, and for 5 min, total
locomotion or exploration is measured as the total number of arm entries, whereas
the percentage of entries in the open arms is inversely related to fear, since it is
increased by some anxiolytic treatments (benzodiazepine, but not serotonin-related
drugs) and reduced by anxiogenic drugs such as caffeine, yohimbine, and amphet-
amine. Indeed, rats spend more time in the enclosed arms and show more behavioral
and physiological signs of fear when confined to the open arms than when confined
to the closed arms: decreased locomotion and higher defecation rate and plasma
corticosterone response. Strain differences have been described in mice62 and in
rats45 (Figure 20.2).
20.3.3 CONDITIONING EXPERIMENTS

Most of these experiments are based on the administration of aversive stimuli, most
frequently foot shocks. In active avoidance paradigms, experimental animals can
escape or avoid the unconditioned stimulus by performing the adequate response
programmed by the experimenter: for instance bar-pressing or shuttling from one
side of the cage to the other. The learning of the response can be impaired by previous
administration of unavoidable shocks. This paradigm, known as learned helpless-
ness, is used as a model of depression for pathophysiological and pharmacological
20
Time open arms (s)

15
10
0
15
Total arm entries
10
0
FIGURE 20.2 Elevated plus maze. Time spent in the open arms and number of total arm
entries measured in 12 males each from six rat strains. (Reprinted from Behav. Brain Res.,
85, Ramos, A. et al., A multiple test study of anxiety-related behaviours in six inbred rat
strains, 5769, Copyright 1997, with permission from Elsevier.)
studies. Another popular paradigm for psychopharmacological studies is the behav-

ioral despair, or forced swimming, test, where a rat learns in a first exposure that
there is no escape possible from a small water tank and eventually displays
depressed behavior upon subsequent testing.43
20.3.4 INTERPRETING EMOTIONALITY THROUGH A MULTIVARIATE

APPROACH
Understanding the psychological significance of a given emotional behavior is not
simple. Very often, different individuals vary for a series of behaviors in a seemingly
inconsistent way, that is, an animal (or group of animals) may appear highly fearful
in relation to one specific behavior (e.g., defecation) but not to another (e.g., ambu-
lation). For an example, see the behavior of BN rats in Figures 20.1 and 20.2.
Therefore, to increase the reliability of the study of emotional reactivity in laboratory
animals, different behaviors can be measured in one single test, or a set of different
tests can be applied to the same group of animals. In both cases, results normally
suggest that different behavioral reactions and different experimental contexts may
be associated to different dimensions of emotionality.
Single test situations. The multivariate or factorial analysis has been successfully
applied to distinguish several dimensions of emotional reactivity in experimental
animals in single experimental designs. The elevated plus maze paradigm is taken as
an example because it is probably the best documented. Cruz and collaborators15
measured 13 behavioral variables in the elevated plus maze from 30 male Wistar rats.
The first and most important two factors are seen as representing variables of anxiety
and locomotion. Percentage of open arm entries, time in open arms, and time in the
closed arms (negatively related) are the variables with the highest loadings on the first
factor. These measures were changed to one direction by administration of anxiolytic
drugs and to the opposite direction by anxiogenic drugs and were therefore thought
to measure anxiety. The second factor reflects mainly the number of entries in the
closed arms (locomotion). The total number of arm entries, which is normally used
as the main index of locomotion in the plus maze, is an ambiguous measure since
this variable loads simultaneously on both anxiety and locomotion factors, appearing
to be more contaminated by emotional responses than the variable number of closed
arm entries. Finally, time spent in the center of the plus maze loaded on a third factor.
Multiple test situations. This kind of analysis is most powerful to investigate the
data obtained from the same animals in different test situations. For instance, Trullas
and Skolnick62 studied the behavior of 16 inbred mouse strains in the open field and
the plus maze. The variables that saturated factor 1 (44.5% of total variance) were
all related to time spent or number of crosses into the open arms of the elevated plus
maze. Since the time spent in the open arms correlated with both the acoustic startle
reflex and hyponeophagia in a novel environment, factor 1 was interpreted as reflect-
ing the level of anxiety of the animal. Variables related to ambulation, both in the
open field and plus maze, loaded on factor 2 (27.5% of total variance), which appears
to measure general activity. Here again, a third factor (10.3% of total variance) was
defined almost exclusively as time spent in the central platform of the plus maze. A
similar analysis was conducted in our laboratory on the data obtained for 19 variables
from four behavioral tests: open field, plus maze, black and white box, and social
interaction.45 The subjects were male and female rats from six inbred strains. The
first factor (36.6% of total variance) was also defined by variables classically asso-
ciated with anxiety (locomotion in the central part of the open field, entries and time
spent in the open arms of the plus maze, behavior in the black and white box). The
second factor (30.3% of total variance) was defined by measures of locomotion (total
and outer locomotion in the open field, locomotion in the closed arms of the plus
maze). Interestingly enough, time of social interaction in a novel environment loaded
on a third factor (18.3% of total variance), together with defecation rates in the open
field and black and white box. Such an independence between behaviors measured
in different anxiety tests, such as the plus maze and social interaction, has already
been described by File.21 These data show that the three measures of open field
behaviorouter locomotion, inner locomotion and defecationload on different
factors and are therefore independent for the most part.
This multifactorial approach brings a number of interesting conclusions. For

example, in many of these experiments, the first two principal axes of variability
are related, respectively, to the level of emotional stimulation and to the amount of
locomotor activation in the apparatus. This reminds us of the two main factors
extracted by Eysenck19 from numerous descriptive studies of human personality:
emotional stability vs. instability (neuroticism) and extraversion vs. introversion. Of
course, this approach is still an oversimplification of the reality, as shown, for
instance, by the independence of the results obtained in different tests of anxiety.
Complexity pops up as soon as it is looked for.
20.4 GENETIC ANALYSIS

A great variability can be seen in emotional behaviors among individual animals.
Different experimental strategies have been used throughout the past 50 years to
analyze this variability and identify its causes. For the remainder of this chapter,
these strategies will be presented while some illustrative experimental data will be
used as examples. Differences between inbred strains raised in the same environment,
selection experiments and quantitative genetic studies have shown that genetic factors
have a major influence on interindividual differences, although they interact with
environmental influences, especially during the prenatal and neonatal period. Molec-
ular genetic studies, on the other hand, are now blossoming and should soon unravel
the molecular bases for these differences in the various aspects of emotional reactivity.
20.4.1 SELECTION EXPERIMENTS

We have shown examples of strain differences for emotional behaviors in Figures
20.1 and 20.2. If these traits are heritable, selective breeding should be effective to
modify the value of the character in the population. Many examples show that indeed
this is the case for numerous reactivity traits as measured:
In the open field: Maudsley reactive (MR) and Maudsley nonreactive

(MNRA) rat strains, selected on their defecation rate;3 Floripa H and L rats
selectively bred for high and low locomotion in the central aversive area;50
strains of mice selected for their divergent locomotor activity from an F3
generation of a cross between BALB/c and C57BL/6.17
In the elevated plus-maze: High anxiety-related behavior (HAB) and low
anxiety-related behavior (LAB) rat lines.33
In the ultrasonic vocalization (USV) test: High- and low-USV rat lines selected
for contrasting rates of USV response to isolation in 10-day old pups.4
In activity tests: Naples high-excitability (NHE) and Naples low-excitability
(NLE) rat lines,57 Wistar-Kyoto hyperactive (WKHA) rat strain,32 Tsukuba
high-emotional (THE) and Tsukuba low-emotional (TLE) rat strains.25
In active avoidance tests: Roman high and low avoidance (RHA/RLA),2
Syracuse high and low avoidance (SHA/SLA)5 lines of rats, Koltushi high
and low avoidance (KHA/KLA),56 high and low avoidance animals
(HAA/LAA)40 strains of rats.
Those strains/lines have been used extensively for quantitative genetic analysis
and are now being used in molecular genetic studies.
20.4.2 QUANTITATIVE GENETIC ANALYSIS

Most phenotypes related to emotional and stress responses are polygenically regu-
lated. The aim of quantitative genetic analysis is to give information about the
structure of genetic effects on the trait under study. One frequently calculated
parameter is heritability, the fraction of the observed variance which is caused by
differences in heredity. Much more information can be obtained, as it can be seen
in the chapters on quantitative genetics. These analyses are based on the study of
populations from various genetic crossings. The study of families takes advantage
of the similarities between related individuals that share part of their genes.
The classical analysis method involves the parental, F1, and F2 generations (and
eventually backcross) used by Mendel. Parental populations are frequently inbred
strains, produced by brothersister mating for more than 20 generations to obtain a
high degree of homozygosity. This was done for instance for open field and plus
maze behavior in rats,47 open field behavior in pigs,18,38 exploratory behavior in
mice,13 ultrasonic vocalizations in mice27,51 and avoidance behavior in rats.8
The diallel design compares several inbred strains and all possible F1 hybrid
crosses. Examples can be found for open-field behavior in mice.13,14,54
20.4.3 GENETIC LINKS BETWEEN DIFFERENT TRAITS

In highly inbred strains and their F1 hybrids, individuals are genetically identical to
one another. Variability within these populations must be caused by nongenetic
factors. In the F2+ and backcross populations, genes assort and the individuals differ
from each other genetically as well as environmentally. We have seen before that
this property is used to study the genetic architecture of the traits under study.
Another use of recombinant populations is to study the genetic links between dif-
ferent traits. If two traits are correlated in both nonsegregating and segregating
populations, they have a high probability of being controlled by the same segregating
unit (and vice versa).
A classic example is the study by Hendley and co-workers3032 of the relationship
between hypertension and hyperactivity in the spontaneously hypertensive rat (SHR).
This strain was selected from the Wistar-Kyoto strain (WKY) for high blood pres-
sure, and is also behaviorally more active in novel as well as familiar environments.
Although many reasons could be put forward to justify a genetic link between these
phenotypes,63 Hendley showed that these traits segregated independently30 and she
was able to raise two new strains of ratsone hypertensive but normoactive
(WKHT), and another normotensive but hyperactive (WKHA)by inbreeding
selected brothersister pairs from the segregating F2 population.31,3 As compared
with the WKY, the WKHA strain also shows a lower emotional reactivity, as mea-
sured by neuroendocrine responses to novel environment exposure.7 However, when
animals from the segregating F2 population (WKYWKHA) were studied, no cor-
relation between these different parameters was found.10 These data show (1) that
the selection for a high locomotor activity in a novel environment independently

coselected animals with a low emotional reactivity, and (2) that the various measures
of emotional reactivity are not necessarily linked genetically.
Therefore, one should be careful when interpreting simple associations between
traits even in selectively bred lines (especially if no replicate lines are available),
because independent traits may appear to be correlated (false positives) due to
undesirable factors such as inbreeding and genetic drift.11 Nevertheless, selective
breeding is certainly a powerful method to either suggest or corroborate genetic
correlations between emotional traits. For example, behaviors measured in the open
field and elevated plus maze seem to be genetically linked in some studies but not
in others. Accordingly, the inbred rat strains Lewis and SHR (which were not
intentionally selected for any emotional behavior) differ for their scores of explora-
tion in both the central area of the open field and the open arms of the plus maze.
In both cases, Lewis rats seem more fearful than their SHR counterparts.45,47,49 These
consistent findings suggest that the anxiety-related measures of the two tests are
genetically linked. In one factor analysis involving various inbred strains, such a
link was indeed confirmed.45 However, another study using F2 rats derived from the
LewisSHR intercross showed open field and plus maze measures to be unrelated.47
Following these seemingly contradictory results, two independent selection experi-
ments came to corroborate the existence of a genetic correlation between open field
and plus maze variables. HAB and LAB rats selected for high and low avoidance
of the plus maze open arms also showed high and low avoidance of the central area
of the open field, respectively.35 The same pattern was observed in the Floripa H
and L rat lines, selected for central locomotion in the open field and differing also
in their plus-maze behavior.50
In conclusion, genetic links between emotional behaviors do not follow an all
or nothing type of situation. Instead, of all genes having some effect on emotionality,
some are likely to influence specific responses in specific tests, whereas others have
pleiotropic influences on a variety of behaviors. Consequently, the genetic overlap
between these behaviors will depend on the populations involved and are expected
to be most frequently partial only.
20.4.4 THE QTL CHASE

The principles and methods of QTL (quantitative trait loci) analysis are presented
in Chapter 5 of this book. Briefly, the aim of the strategy is to map chromosome
regions that contain genes affecting a quantitative trait (e.g., behavior) through the
use of polymorphic genetic markers, the most popular ones being the microsatellites,
which are simple-sequence nucleotide repeats regularly spaced on the genome and
highly polymorphic. The populations normally used are segregating generations (F2
or backcross) derived from two inbred strains, or a panel of recombinant inbred (RI)
strains. The detection of QTLs associated with differences in emotional behaviors
is a first step to the identification of the genes responsible for genetic variation. This
knowledge will allow the analysis of the functional pathways involved in behavioral
variability.
Flint et al.24 were the first to report, in 1995, QTLs associated with emotional
behaviors in mice. In rats, the first QTLs for emotionality were mapped in 1996 by
Moisan et al.37 (locomotor activity in a novel environment, WKY WKHA F2
intercross) and in 1999 by Ramos et al., using an F2 intercross from Lewis and SHR
inbred strains.48 These initial studies very often identified loci that were rather
specific for one or a few behavioral responses and that were not replicated across
different strains. Both these aspects brought some doubts about the psychological
significance of these traits and the biological interest of these loci. More recent
studies showed that some QTLs are likely to harbor genes with more general
emotional effects, whereas others affect specific behaviors that may correspond to
particular forms of fear and anxiety or even reflect other unrelated traits. For exam-
ple, an elegant work by Turri et al.64 showed that a QTL on mouse chromosome 15
has a broad range of behavioral effects that are consistent with a presumed influence
on fear or anxiety, but not on general locomotion. On the other hand, a QTL on
chromosome 1, which was thought to be anxiety-related after having been replicated
in different studies,23 was shown in fact to have an overall influence on exploratory
behaviors in both aversive and nonaversive situations. Finally, one locus on chro-
mosome 6 showed a very specific effect on a single variable from one particular
test, the elevated plus maze. In rats, a multivariate approach applied to QTL mapping
pointed to similar conclusions by identifying two loci (on chromosomes 1 and 19)
for very specific behaviors and one locus (on chromosome 5) affecting several
anxiety-related measures, such as avoidance responses, fear conditioning, and behav-
ior in the plus maze and open-field tests.20 QTLs with pleiotropic effects on numerous
fear-related measures are seen as good candidates for harboring genes expected to
be relevant for human anxiety.23
Some QTLs for emotionality may have pleiotropic effects on other types of
behaviors, potentially related to pathologies that can be comorbid with anxiety
disorders, such as alcoholism. In 1998, the group of Lucinda Carr, using the selected
rat lines P (alcohol-preferring) and NP (alcohol-nonpreferring), identified a powerful
QTL for alcohol consumption on the mid-portion of chromosome 4.1,6 A few months
later, a QTL located near this chromosome region was shown to have a strong
influence on the locomotion in the inner part of the open field, an index of emo-
tionality, in Lewis/SHR F2 rats.48 In spite of having their peaks more than 15 cM
apart, the confidence intervals of these two QTLs overlapped. The existence of a
QTL on chromosome 4 for inner locomotion in the open field (named Ofil1) was
corroborated by a second study involving two selected rat lines derived from the
Lewis and SHR strains.39 More recently, another whole-genome QTL analysis based
on the F2 intercross of two other rat strains that contrast for alcohol intake, namely
HEP (high-ethanol preferring) and WKY (Wistar-Kyoto) led to the identification of
a significant locus on chromosome 4, approximately midway between the two above-
mentioned QTLs (Figure 20.3), that strongly affects ethanol preference.59
Considering that anxiety is one of the predisposing factors involved in alcohol
abuse,41 the combination of findings described above led to the hypothesis that these
different QTLs might correspond in fact to one single locus with pleiotropic effects
on alcohol drinking and emotionality. This hypothesis has been corroborated so far.
The latest data from two rat lines derived from HEP and WKY, which were selected
Alcohol preference in P and NP

10
Alcohol preference in HEP and WKY
9
Emotionality in Lewis and SHR
8
7
LOD score
6
5
4
3
2
1
0
0 20 40 60 80 100 120 140 160 180 200
Mb
FIGURE 20.3 QTL map. Approximate positions of three QTLs independently mapped on
rat chromosome 4 for: inner locomotion in the open field (emotionality) in Lewis and SHR
rats;48 preference for 10% ethanol in P and NP rats;1,6 preference for 5% ethanol in HEP and
WKY rats.59 Physical map positions were estimated by comparing linkage data from the
original publications and marker coordinates (in base pairs) from the Rat Genome Database
(http://rgd.mcw.edu/).
according to their genotypes (HEP/HEP or WKY/WKY) at this QTL, showed that

this locus influenced not only free or forced alcohol intake, but also locomotion in
the center of the open field.60 Interestingly, when the same experimental approach
was applied to a Lewis/SHR intercross (unpublished data), this chromosome region
was also found to affect both alcohol- and open-fieldrelated behaviors. Moreover,
another QTL study has associated this same genomic region with corticosterone
levels in rats, a well-known stress-related hormone.44 Considering that Lewis and
SHR rats also differ in alcohol drinking,16 further studies shall demonstrate whether
the QTLs revealed by these three distinct genetic models (P/NP, Lewis/SHR,
HEP/WKY) represent different loci that happen to pertain to the same linkage group
or if they correspond to real pleiotropic genes. If the latter situation were true, then
the close investigation of this chromosome region may shed some light on the
molecular relationship between emotional reactivity and alcoholism.
Besides the classical methods for identifying QTLs for emotional behaviors,
which involve phenotyping and genotyping of very large segregating populations, a
new strategy recently has been applied with success. It is based on a panel of
chromosome substitution strains (CSSs), each of which carries a single chromosome
from a donor strain on the genome background of a host strain. The effectiveness
of this strategy was shown, for example, by the replication of a QTL for open-field
behavior located on mouse chromosome 1.58 Two other alternative strategies, high-
resolution mapping using commercially available outbred animals and quantitative
complementation testing, were recently used to dissect this specific mouse QTL into
smaller, more precise components as well as to identify one of its underlying genes.65
20.4.5 GENE TARGETING AND TRANSGENESIS

A vast literature on the behavioral characterization (including emotionality) of
knockout, knockin and transgenic animals was produced in the past 10 years. In
fact, the manipulation of individual genes that are potentially involved in the control
of fear-related responses became one of the main tools used in the genetic study of
anxiety. Since a large number of genes coding for neuronal messengers, receptors,
enzymes, and signaling molecules started to be either inactivated, overexpressed, or
modified, new integrative hypotheses have been proposed to explain the molecular
pathways involved in the pathogenesis of anxiety. Discussing that approach is beyond
the scope of this chapter and the interested readers are encouraged to read the recent
reviews on this topic.9,22,34,55,66
REFERENCES
1. BICE P, FOROUD T, BO R, CASTELLUCCIO P, LUMENG L, LI TK, CARR LG:
Genomic screen for QTLs underlying alcohol consumption in the P and NP rat lines.
Mamm Genome, 1998, 9:949955.
2. BIGNAMI G: Selection for high rates and low rates of avoidance conditioning in the
rat. Anim Behav, 1965, 13:220221.
3. BROADHURST PL: The Maudsley reactive and nonreactive strains of rats : A survey.
Behav Genet, 1975, 5:299319.
4. BRUNELLI SA, HOFER MA, WELLER A: Selective breeding for infant vocal
response: a role for postnatal maternal effects? Dev Psychobiol, 2001, 38:221228.
5. BRUSH FR: Genetic determinants of individual differences in avoidance learning:
behavioral and endocrine characteristics. Experientia, 1991, 47:10391050.
6. CARR LG, FOROUD T, BICE P, GOBBETT T, IVASHINA J, EDENBERG H,
LUMENG L, LI T-K. A quantitative trait locus for alcohol consumption in selectively
bred rat lines. Alcohol Clin Exp Res, 1998, 22:884887.
7. CASTANON N, HENDLEY ED, FAN X-M, MORMDE P: Psychoneuroendocrine
profile associated with hypertension or hyperactivity in spontaneously hypertensive
rats. Am J Physiol, 1993, 265:R1304R1310.
8. CASTANON N, PEREZ-DIAZ F, MORMEDE P: Genetic analysis of the relation-
ships between behavioral and neuroendocrine traits in roman high and low avoidance
rat lines. Behav Genet, 1995, 25:371384.
9. CLMENT Y, CALATAYUD F, BELZUNG C: Genetic basis of anxiety-like behav-
iour: A critical review. Brain Res Bull, 2002, 57:5771.
10. COURVOISIER H, MOISAN MP, SARRIEAU A, HENDLEY ED, MORMEDE P:
Behavioral and neuroendocrine profile of stress reactivity in the WKHA/WKY inbred
rat strains : a multifactorial and genetic analysis. Brain Res, 1996, 743:7785.
11. CRABBE JC: Animal Models in Neurobehavioral Genetics: Methods for Estimating
Genetic Correlation. In BC JONES, P MORMDE (Eds): Neurobehavioral Genetics:
Methods and Applications. CRC Press, Boca Raton, 1999, pp 121138.
12. CRAWLEY JN: Neuropharmacologic specificity of a simple animal model for the
behavioral actions of benzodiazepines. Pharmacol Biochem Behav, 1981, 15:695699.
13. CRUSIO WE, SCHWEGLER H, VAN ABELEEN JHF: Behavioral responses to
novelty and structural variation of the hippocampus in mice. I. Quantitativegenetic
analysis of behavior in the open-field. Behav Brain Res, 1989, 32:7580.
14. CRUSIO WE, VAN ABELEEN JHF: The genetic architecture of behavioural
responses to novelty in mice: a quantitative-genetic analysis. Heredity, 1986, 56:5563.
15. CRUZ APM, FREI F, GRAEFF FG: Ethopharmacological analysis of rat behavior
on the elevated plus-maze. Pharmacol Biochem Behav, 1994, 49:171176.
16. DA SILVA GE, RAMOS A, TAKAHASHI RN: Comparison of voluntary ethanol
intake by two pairs of rat lines used as genetic models of anxiety. Braz J Med Biol
Res, 2004, 37:15111517.
17. DEFRIES JC, GERVAIS MC, THOMAS EA: Response to 30 generations of selection
for open-field activity in laboratory mice. Behav Genet, 1978, 8:313.
18. DSAUTS C, BIDANEL JP, MORMEDE P: Genetic study of behavioral and
pituitary-adrenocortical reactivity in response to an environmental challenge in pigs.
Physiol Behav, 1997, 62:337345.
19. EYSENCK HJ: Psychopathology and Personality: Extraversion, Neuroticism and
Psychoticism. In A GALE, JA EDWARDS (Eds): Individual Differences and Psy-
chopathology. Academic Press, London, 1997, pp 1330.
20. FERNNDEZ-TERUEL A, ESCORIHUELA RM, GRAY JA, AGUILAR R, GIL L,
GIMNEZ-LLORT L, TOBEA A, BHOMRA A, NICOD A, MOTT R, DRISCOLL
P, DAWSON GR, FLINT J: A quantitative trait locus influencing anxiety in the
laboratory rat. Genome Res, 2002, 12:618626.
21. FILE SE: The biological basis of anxiety. In HY MELTZER, D NEROZZI (Eds):
Current Practices and Future Developments in the Pharmacotherapy of Mental Dis-
orders. Elsevier Science Publishers B.V. Amsterdam, 1997, pp 159165.
22. FINN DA, RUTLEDGE-GORMAN MT, CRABBE JC: Genetic animal models of
anxiety. Neurogenetics, 2003, 4:109135.
23. FLINT J: Analysis of quantitative trait loci that influence animal behavior. J Neuro-
biol, 2003, 54:4677.
24. FLINT J, CORLEY R, DEFRIES JC, FULKER DW, GRAY JA, MILLER S, COL-
LINS AC: A simple genetic basis for a complex psychological trait in laboratory
mice. Science, 1995, 269:14321435.
25. FUJITA O, ANNEN Y, KITAOKA A: Tsukuba High- and Low-emotional strains of
rats (rattus norvegicus) : an overview. Behav Genet, 1994, 24:389415.
26. GLOWA JR, HANSEN CT: Differences in response to an acoustic startle stimulus
among forty-six rat strains. Behav Genet, 1994, 24:7984.
27. HAHN ME, HEWITT JK, ADAMS M, TULLY T: Genetic influences on ultrasonic
vocalizations in young mice. Behav Genet, 1987, 17:155166.
28. HALL CS: Emotional behavior in the rat I. Defecation and urination as measures of
individual differences in emotionalitly. J Comp Psychol, 1934, 18:385403.
29. HANDLEY SL, MCBLANE JW: An assessment of the elevated X-Maze for studying
anxiety and anxiety-modulating drugs. J Pharmacol Toxicol Method, 1993, 29:129138.
30. HENDLEY ED, ATWATER DG, MYERS MM, WHITEHORN D: Dissociation of
genetic hyperactivity and hypertension in SHR. Hypertension, 1983, 5:211217.
31. HENDLEY ED, OHLSSON WG: Two new inbred rat strains derived from SHR :
WKHA, hyperactive, and WKHT, hypertensive, rats. Amer J Physiol, 1991,
261:H583H589.
32. HENDLEY ED, WESSEL DJ, VAN HOUTEN J: Inbreeding of Wistar-Kioto rat
strain with hyperactivity but without hypertension. Behav Neural Biol, 1986, 45:116.
33. LANDGRAF R, WIGGER A: High vs. low anxiety-related behavior rats: an animal
model of extremes in trait anxiety. Behav Genet, 2002, 32:301314.
34. LESCH K-P: Molecular foundation of anxiety disorders. J Neural Transm, 2001,
108:717746.
35. LIEBSCH G, MONTKOWSKI A, HOLSBOER F, LANDGRAF R: Behavioural
profiles of two Wistar rat lines selectively bred for high or low anxiety-related
behaviour. Behav Brain Res, 1998, 94:301310.
36. LISTER RG: Ethologically-based animal models of anxiety disorders. Pharmacol
Ther, 1990, 46:321340.
37. MOISAN MP, COURVOISIER H, BIHOREAU MT, GAUGUIER D, HENDLEY
ED, LATHROP M, JAMES MR, MORMEDE P: A major quantitative trait locus
influences hyperactivity in the WKHA rat. Nat Genet, 1996, 14:471473.
38. MORMEDE P, DANTZER R, BLUTHE R, CARITEZ J: Differences in adaptive
abilities of three breeds of chinese pigs. Genet Sel Evol, 1984, 16:85102.
39. MORMDE P, MONEVA E, BRUNEVAL C, CHAOULOFF F, MOISAN M-P:
Marker-assisted selection of a neuro-behavioural trait related to behavioural inhibition
in the SHR strain, an animal model of ADHD. Genes Brain Behav, 2002, 1:111116.
40. OHTA R, MATSUMOTO A, HASHIMOTO Y, NAGAO T, MIZUTANI M: Behav-
ioral characteristics of rats selectively bred for high and low avoidance shuttlebox
responses. Cong Anom, 1995, 35:223229.
41. PANDEY SC: Anxiety and alcohol abuse disorders: a common role for CREB and
its target, the neuropeptide Y gene. Trends Pharmacol Sci, 2003, 24:456460.
42. PARE WP: Hyponeophagia in Wistar Kyoto (WKY) rats. Physiol Behav, 1994,
55:975978.
43. PORSOLT RD, BERTIN A, JALFRE M: Behavioural despair in rats and mice:
strain differences and the effects of imipramine. Eur J Pharmacol, 1978, 51:291294.
44. POTENZA MN, BRODKIN ES, JOE B, LUO X, REMMERS EF, WILDER RL,
NESTLER EJ, GELERNTER J: Genomic regions controlling corticosterone levels
in rats. Biol Psychiatry, 2004, 55:634641.
45. RAMOS A, BERTON O, MORMEDE P, CHAOULOFF F: A multiple test study of
anxiety-related behaviours in six inbred rat strains. Behav Brain Res, 1997, 85:5769.
46. RAMOS A, MORMEDE P: Stress and emotionality : a multidimensional and genetic
approach. Neurosci Biobehav Rev, 22:3357, 1998.
47. RAMOS A, MELLERIN Y, MORMDE P, CHAOULOFF F: A genetic and multi-
factorial analysis of anxiety-related behaviours in Lewis and SHR intercrosses. Behav
Brain Res, 1998, 96:195205.
48. RAMOS A, MOISAN M-P, CHAOULOFF F, MORMDE C, MORMDE P: Iden-
tification of female-specific QTLs affecting an emotionality-related behavior in rats.
Mol Psychiatry, 1999, 4:453462.
49. RAMOS A, KANGERSKI AL, BASSO PF, SANTOS JES, ASSREUY J, VEN-
DRUSCOLO LF, TAKAHASHI, RN: Evaluation of Lewis and SHR rat strains as a
genetic model for the study of anxiety and pain. Behav Brain Res, 2002, 129:113123.
50. RAMOS A, CORREIA EC, IZDIO GS, BRSKE GR: Genetic selection of two new
rats lines displaying different levels of anxiety-related behaviors. Behav Genet, 2003,
33:657668.
51. ROUBERTOUX P, CARLIER M, DEGRELLE H, HAAS-DUPERTUIS MC, PHIL-

LIPS J, MOUTIER R: Co-segregation of intermale aggression with the pseudoauto-
somal region of the Y chromosome in mice. Genetics, 1994, 135:225230.
52. ROUBERTOUX P, MARTIN B, LEROY I, BEAU J, MARCHALAND C, PEREZ-
DIAZ F, COHEN-SALMON C, CARLIER M: Vocalizations in newborn mice:
genetic analysis. Behav Genet, 1996, 26:427437.
53. ROYCE JR, CARRAN A, HOWARTH E: Factor analysis of emotionality in ten
inbred strains of mice. Multivariate Behav Res, 1970, 5:1948.
54. ROYCE JR, HOLMES TM, POLEY W: Behavior genetic analysis of mouse emo-
tionality. III.The diallel analysis. Behav Genet, 1975, 5:351372.
55. RUDOLPH U, MHLER H: Analysis of GABAA receptor function and dissection
of the pharmacology of benzodiazepines and general anesthetics through mouse
genetics. Annu Rev Pharmacol Toxicol, 2004, 44:475498.
56. RYZHOVA LY, KULAGIN DA, LOPATINA NG: Correlated variability of the loco-
motor activity and emotionality in rats under selection program for high and low
active avoidance value. Genetika, 1983, 19:121125.
57. SADILE AG, CERBONE A, LAMBERTI C, CIOFFI LA: The Naples high (NHE)
and low excitable (NLE) rat strains: progressive report. Behav Brain Res, 1984,
12:228229.
58. SINGER JB, HILL AE, BURRAGE LC, OLSZENS KR, SONG J, JUSTICE M,
OBRIEN WE, CONTI DV, WITTE JS, LANDER ES, NADEAU JH: Genetic dis-
section of complex traits with chromosome substitution strains of mice. Science,
2004, 304: 445448.
59. TERENINA-RIGALDIE E, MOISAN M-P, COLAS A, BEAUG F, SHAH KV,
JONES BC, MORMDE P: Genetics of behaviour: phenotypic and molecular study
of rats derived from high- and low-alcohol consuming lines. Pharmacogenetics, 2003,
13:543554.
60. TERENINA-RIGALDIE E, JONES BC, MORMEDE P: Pleiotropic effect of a locus
on chromosome 4 influencing alcohol drinking and emotional reactivity in rats. Genes
Brain Behav, 2003, 2:125131.
61. TREIT D: Animal models for the study of anti-anxiety agents : A review. Neurosci
Biobehav Rev, 1985, 9:203222.
62. TRULLAS R, SKOLNICK P: Differences in fear motivated behaviors among inbred
mouse strains. Psychopharmacology, 1993, 111:323331.
63. TUCKER DC, JOHNSON AK: Behavioral correlates of spontaneous hypertension.
Neurosci Biobehav Rev, 1981, 5:463471.
64. TURRI MG, DATTA SR, DEFRIES J, HENDERSON ND, FLINT J: QTL analysis
identifies multiple behavioral dimensions in ethological tests of anxiety in laboratory
mice. Curr Biol, 2001, 11:725734.
65. YALCIN B, WILLIS-OWEN SAG, FULLERTON J, MEESAQ A, DEACON RM,
RAWLINS JNP, COPLEY RR, MORRIS AP, FLINT J, MOTT R: Genetic dissection
of a behavioral quantitative trait locus shows that Rgs2 modulates anxiety in mice.
Nat Genet, 2004, 36:11971202.
66. WOOD SJ, TOTH M: Molecular pathways of anxiety revealed by knockout mice.
Mol Neurobiol, 2001, 23:101119.
21 Genetic Analysis of Food

Search Behavior in the
Fruit Fly (Drosophila
melanogaster)
Amsale T. Belay and Marla B. Sokolowski
CONTENTS
Introduction............................................................................................................ 307
21.1 The Fruit Fly D. melanogaster as a Model Organism................................ 308
21.2 Polygenic and Single-Gene Analysis .......................................................... 308
21.2.1 Polygenic Analysis ........................................................................... 309
21.2.2 Single-Gene Analysis ....................................................................... 309
21.2.3 Polygenic vs. Single-Gene Analysis of Behavior ............................ 310
21.3 Food Search Behavior.................................................................................. 311
21.4 Role of the Foraging Gene in Other Organisms......................................... 312
References.............................................................................................................. 314
INTRODUCTION
Food search behavior is a complex trait influenced by many internal and external
factors. It can involve sensory modalities such as smell, taste, and vision; motor
functions such as locomotion; internal environmental cues such as the degree of
starvation or satiation; and external environmental cues such as the quality, quantity,
and distribution of the food supply. All of this information is likely integrated in
the brain and an output (e.g., feed or search more) is then generated. In the
Sokolowski laboratory, we use the fruit fly Drosophila melanogaster as our model
organism to genetically dissect the components of food search behavior. In partic-
ular, we are interested in understanding the biological basis for naturally occurring
variation in larval and adult food search behaviors as well as the biochemical
pathways that contribute to these variations. In this chapter, we discuss the fruit
fly D. melanogaster as a model organism for behaviorgenetic analysis. We com-
pare and contrast the polygenic analysis of naturally occurring behavioral variation
307
and the single-gene mutant analysis approaches and discuss food search behavior
in the larval and adult fruit fly D. melanogaster. We conclude by discussing how
this research can be extended to additional organisms.
21.1 THE FRUIT FLY D. MELANOGASTER AS A MODEL

ORGANISM
D. melanogaster offers many advantages for behavior and neurogenetic analysis.
It expresses complex patterns of behavior including learning, courtship, circadian
rhythms, food search, and locomotion.23,45 In addition, a superb genetic database
of information has been collected over more than 80 years (see reference 21 for
an excellent concise tour of Drosophila genetic technology and reference 2 for all
you may want to know about the fly). In addition to its fully sequenced genome,
Drosophila has a wonderful array of molecular, neurobiological, and genomic tools
(see http://flybase.bio.indiana.edu/ for database of the Drosophila genome and
related tools). Moreover, it has relatively few neurons and neural networks, making
it feasible to identify neural substrates underlying behavioral variation. Finally,
findings from Drosophila are readily extended to additional organisms because
mechanisms underlying behavior often show functional homologies between
species.24
The life cycle of the fly is rapid (1014 days at 25C), and thousands of
genetically identical flies are bred in less than a month, making it amenable for
behavioral analysis of genetically identical individuals reared under controlled envi-
ronmental conditions. The mated female lays hundreds of eggs on the food (fly
medium in the laboratory, and fruit in nature) that hatch after 24 h (at 25C). The
larva (all three instars) forages by feeding on yeast with its mouth hooks and moves
by extending its anterior and retracting its posterior end. Midway through the third
instar, the larva leaves the food in search of a pupation site. After 5 days of larval
life, the larva pupates for 4 days during which it metamorphoses into an adult fly.
The fly then emerges from the pupal case and searches for food and mates.
21.2 POLYGENIC AND SINGLE-GENE ANALYSIS

Polygenic and single-gene analyses are useful approaches for the genetic dissection
of behavior in Drosophila. Historically, single-gene mutant analysis addresses ques-
tions concerning the mechanisms involved in behavior, whereas polygenic analysis
concerns itself with questions about the evolution of behavior. However, the two
approaches are valid and complementary, as we illustrate later in the chapter. Using
Drosophila food search behavior as an example, we describe how the genetic con-
tribution to a naturally occurring behavior polymorphism was first determined using
a polygenic strategy and later refined using the single-gene mutant approach. Below
we discuss some widely used techniques for polygenic analysis of behavioral traits
and follow this with a discussion of single-gene mutant analysis. We then compare
and contrast these approaches.
Genetic Analysis of Food Search Behavior in the Fruit Fly 309
21.2.1 POLYGENIC ANALYSIS

The type of polygenic inheritance involved in naturally occurring variation ranges
from a single major gene with minor modifiers to that of many genes, all with small,
equal, and additive effects on the phenotype. When few genes are involved, it is
possible in D. melanogaster to determine their chromosomal location. When many
genes are involved quantitative trait loci (QTL) mapping and whole genome micro-
arrays are used to identify genes underlying behavioral variations.
Until recently, the identification of these genes was considered a daunting task.
However, recent advances make the fly an attractive organism for identifying the
genes underlying QTL. These include a collection of highly polymorphic molecular
markers with known physical map locations, deficiency chromosome stocks made
on an isogenic background, as well as numerous strains with mutations in a single
locus. Detailed methods for the analysis of quantitative traits are found in Falconer
and Mackay,13 Mackay,26 and Mather and Jinks.27 QTL analysis has been applied to
many Drosophila traits such as olfactory behavior,14 courtship,19 longevity,28 bristle
number,29 environment-dependent survival,50 and wing shape.51 The QTL intervals
are first narrowed down using quantitative deficiency complementation tests and
finally through complementation tests with strains carrying mutations in candidate
genes.1 Association analysis between the single nucleotide polymorphisms (SNPs)
and the behavioral variation enables determination of the within-gene molecular
alterations responsible for the natural variation in behavior. Together these
approaches identified dopa decarboxylase (ddc)12 and shuttle craft (stc)33 as genes
underlying QTLs for Drosophila life span.
Whole genome transcriptional profiling using micoarrays is another approach
used to identify genes involved in natural variation in behavioral traits.46 Toma
et al.49 investigated the transcription profiles of two artificially selected strains of
flies that differ in their response to gravity. Microarray analyses of these strains
revealed that 250 genes differed in the RNA expression between the positive and
negative geotaxis strains. Preexisting single-gene mutants were then used to test and
validate the functional relevance of these genes to geotaxis behavior. Studies such
as these link polygenic and single-gene mutant approaches in the analysis of complex
behaviors.
21.2.2 SINGLE-GENE ANALYSIS

The genetic dissection of behavior in Drosophila using single-gene mutant analysis
provides insight into the mechanistic bases of behavior. In single-gene analysis,
mutations are induced in normal (wild-type) behaving strains using mutagens such
as ionizing radiation, ethyl methane sulfonate (EMS), or P-element mutagenesis,
each of which has its own advantages and disadvantages (discussed in Reference
21). After mutagenesis is performed, the behavior of the larva or fly is screened
with the aim of later identifying X-linked or autosomal mutations affecting the
phenotypes of interest. The intent is to alter normal function by mutation (mutate
away from the wild-type phenotype) and determine, through comparisons of wild-
type and mutant function, the genetic factors essential for the expression of the
normal wild-type phenotype. Mutant screens can also generate conditional muta-
tions that are dependent on factors such as temperature, light, or diet. Conditional
mutations are useful for studies of the developmental and/or acute functions of a
gene. In subsequent analyses, mutants can also be mutated (mutation can be gen-
erated in a mutant genetic background) to find new mutations that suppress or
enhance the original phenotype.35
Numerous single-gene mutations that affect behavior (e.g., learning and memory,
courtship, circadian rhythms, olfaction, phototaxis) have been identified and char-
acterized in D. melanogaster.22,32,45,48 In the past, it was unusual for single-gene
behavioral mutants to be generated on defined genetic backgrounds. Once a mutant
is generated, it is important to know (1) how the genetic background affects the
expression of the behavior, and (2) whether the mutant phenotype is selected against
in the laboratory.11 If so, careful control of genetic background, whether through
making strains homozygous or continuous backcrossing to a well-defined control
strain, should be carried out.
One area of research that has been virtually ignored is the study of whether a
gene originally identified by mutant analyses is variable in natural populations (an
exception is the work of Costa and colleagues6). This can be accomplished by
analyzing natural populations to determine if identified SNPs show associations with
natural variation in behavior.
21.2.3 POLYGENIC VS. SINGLE-GENE ANALYSIS OF BEHAVIOR

Most naturally occurring individual differences in behavior that have a genetic
component are polygenically based; that is, the behavioral differences can be attrib-
uted to more than one genetic factor. This contrasts with mutant forms that are
generated and screened for in the laboratory. Mutant screens enable us to identify
more loci that affect a given phenotype than studying natural variants.
Most single-gene mutants would not survive in nature because their phenotypes
are usually more severe than natural behavioral variants. Many of these induced
mutants have deleterious effects on fitness due to the pleiotropic effects of their
mutation. One can get around this problem by selectively studying hypomorphic
mutations that cause a small reduction in the amount of gene product. These kinds
of mutations tend to exhibit the behavioral alterations but not the other pleiotropic
phenotypes.20 Furthermore, such subtle mutations are likely representative of natural
variants that have been selected under natural conditions. It is important to note that
in the laboratory, an induced mutant strain would accumulate heritable differences
in minor genes as a result of genetic drift and/or natural selection in the laboratory.
These minor genes may not be directly related to the expression of the mutant
phenotype. In contrast, naturally occurring variants likely have more subtle effects
on the phenotype and have been incorporated into the genome over many gener-
ations through selection. Thus, by definition, naturally occurring variants have a
polygenic rather than a single-gene basis. Little is known about the molecular basis
of genes implicated in natural behavioral variation. Questions for the future include:
Do differences in normal behavior result from alterations in the structural or regu-
latory part of the proteins encoded by these genes? Which behavioral genes are
likely to vary in nature? How, at the molecular level, does genetic background
influence the expression of the behavioral phenotype?
21.3 FOOD SEARCH BEHAVIOR

Most of our studies on food search behavior have focused on the larval stage of
Drosophila. The larvas sensory system47 and behaviors are simple relative to those
of the adult fly. The larva can sense light,36 smell5, and taste47 as well as modify its
locomotion depending on abiotic factors such as humidity, food gradients, and light.
As mentioned previously, the larva spends most of its life feeding and moving in
search of food. Sokolowski41 developed a larval assay to measure locomotion in the
presence of food (a yeast and water paste) that she called a foraging behavior assay.
Generally speaking, when animals determine that the food source is a good one,
they slow down their rate of locomotion and exhibit high turning rates.3 In contrast,
when food is absent or when the food quantity is not adequate, the rate of locomotion
is high and larvae move in more of a straight-line pattern. Foraging behavior, of the
larval18 and the adult fly,3 can be modified by how satiated the animal is prior to the
behavioral test. The behavior can also be modified by external environmental factors
such as food quality, food patch size, and interpatch distance. The manner in which
these factors modify the food search behavior is predictable. If a fly or larva is
starved, it will accept lower-quality food (or smaller patch sizes).
In the early 1980s Sokolowski41 discovered a naturally occurring polymorphism
in larval food search behavior. Third instar D. melanogaster larvae collected from
single fruits in a pear orchard show a bimodal distribution in their movement patterns
while foraging on yeast: 70% of the larvae forage as rovers, exhibiting foraging
trails that are longer and straighter than those of the sitter larvae (30% of the
population).41,44 Importantly, this difference in behavior is only exhibited in the
presence of food (yeast paste). In the absence of food, the foraging paths of both
morphs are long, relatively straight, and do not significantly differ in length.34 Thus,
the rover/sitter difference is conditional on the presence of a nutritive environment
and is not due to a general sluggishness in the sitter larvae.43 When food is unlimited,
their development times and growth curves do not differ.18 The length of the foraging
trails are modifiable by the environment (e.g., by modifying patch quality or the
degree of starvation prior to the test). However, the relative differences in behavior
between the rover and sitter morphs are maintained within a given environment.18
Food search behavior in the adult fly is assayed by measuring the distance the
female walks postfeeding in an arena with a sucrose drop.34 After feeding, rover
adults walk significantly farther than do sitters, and like larvae they do not differ in
their locomotion in the absence of food. When more than one food patch is present,
both rover larvae and adults move between patches in search of food, whereas sitters
move to the nearest patch and remain feeding nearby. We have shown that there is
a selective advantage for the rovers in crowded conditions, whereas sitters do better
in uncrowded conditions. Thus, density-dependent selection may have played a role
in the evolution of the rover/sitter polymorphism. The olfactory behavior of rover
and sitter larvae is wild-type and does not differ in response to yeast.40
As behavioral geneticists, we are, of course, interested in determining whether

there is a genetic component to the difference in behavior between the rover and
sitter morphs. Polygenic analysis (including chromosome substitutions, 16 reciprocal
cross-analysis, and compound autosome analysis42) shows that this trait is affected
by a single major gene named foraging (for) located on the second pair of chromo-
somes in D. melanogaster.79 Further localization using standard recombination
mapping techniques was extremely difficult due to the overlapping nature of this
quantitative behavioral phenotype, and the pleiotropic effects that marker genes and
genetic background have on these behavioral phenotypes.42 To map for, we devised
a new version of the single-gene mutant approach which we term lethal tagging.
Briefly, we tagged the behavioral trait (difficult to map) with a recessive lethal
mutation (easy to map), proved that the tag was very close to for, and then mapped
the lethal tag, which enabled subsequent mapping of for. In this way, for was
localized to region 24A35 on the polytene chromosome map.810 Osborne et al.30
then cloned for and proved that for is synonymous with the dg2 gene in Drosophila
that encodes one of two cGMP-dependent protein kinases (PKG).25 They showed
that for mutants map within dg2, rover fly heads have significantly higher PKG
enzyme activity than do naturally occurring sitters or sitter mutants, and that rover
larvae have significantly higher PKG activity in their central nervous systems (CNS)
than do sitters. The abundance of dg2 RNA is also higher in the naturally occurring
rovers than in sitters, indicating that the difference between the morphs likely results
from differences in dg2 expression. The strongest proof that for is dg2, however,
came when using transgenic sitter larvae that had four copies of rover CDNA from
dg2. The behavior of these transgenic sitter larvae did not differ from rover, nor did
the PKG enzyme activities in their larval CNS. This was the final proof that for is
dg2. Our future work involves studying where and when in the fly higher levels of
for-PKG need to be expressed to produce rover behavior. We are currently mapping
the tissue distribution of for-PKG in order to identify cell types where higher level
of for-PKG is sufficient for rover-like behavior.
Osborne et al.30 provided the first rigorous demonstration of the molecular basis
for a naturally occurring difference in behavior. The significance of this study is the
demonstration that small changes in a kinase (rovers have only a 12% increase in
PKG enzyme activity in their heads compared with sitters) can account for large
differences in naturally occurring behavior. These subtle differences at the molecular
level may be representative of a way that behavior is modulated in natural populations.
21.4 ROLE OF THE FORAGING GENE IN OTHER

ORGANISMS
The fact that many genes identified in the fly have structural and functional homol-
ogies to vertebrate genes suggests that genetic discoveries in the fruit fly can con-
tribute to our general understanding of evolutionarily conserved developmental,
behavioral, and physiological processes. As such, the mammalian counterparts of
PKG might play a role in the regulation of eating in mammals. Indeed, for has a
human counterpart called PRKG1 that has high sequence homology to for and
conservation of most of the for splice sites.15,31 We are currently investigating whether
a genetic polymorphism found in the 3UTR of the PRKG1 mRNA is associated
with obesity in a sample of morbidly obese and normal-weight humans. In the future,
it will be of interest to investigate the association of other recently identified PRKG1
SNPs in humans with alterations in food-related behaviors.
Carrying on with the hypothesis that PKGs role in foraging is conserved across
the animal kingdom, Ben-Shahar et al.4 investigated the role of for-PKG in the
honeybee, Apis mellifera. Following the cloning of Amfor, an ortholog of the Droso-
phila foraging gene, they showed that the age-related transition by honeybees from
in-hive work (nursing) to out-of-hive work (foraging) is associated with an increase
in the expression of the honeybee foraging gene, Amfor. Furthermore, they demon-
strated that this effect is not a byproduct of the nurses being 2 to 3 weeks younger
than foragers. They generated same-age nurses and foragers by colony manipulation
and found that precocious foragers still had higher expression of Amfor RNA and
PKG activity. Additionally, cGMP treatment of in-hive workers elevated PKG activ-
ity and caused precocious foraging behavior. Thus, the same gene is involved in
food-related behaviors in the honeybee and the fly; however, the regulation of the
gene differs. Drosophila for is involved in allelic variation in behavior in the fly,
whereas the honeybee Amfor gene is involved in behavioral plasticity during the
lifetime of the individual. While other genes are involved in the switch from nurse
to forager in the honeybee,52 it was sufficient to manipulate the levels of for-PKG
to cause a change in the percentage of forager honeybees.
Scheiner et al.37,38 used lessons learned from the honey bee to ask function-
related questions in the fruit fly. Interestingly, sensory responsiveness in honey bees
has been correlated with different aspects of foraging behavior and different types
of learning. Scheiner et al.39 hypothesized that the fly for gene might also influence
sensory responsiveness and nonassociative learning in Drosophila. They used tech-
niques originally developed for honeybee studies to investigate the role of for-PKG
in adult fruit fly sensory responsiveness. A fruit fly extends its proboscis when its
foreleg is stimulated with sucrose. This is called a proboscis extension response
(PER). Scheiner and colleagues measured sucrose responsiveness of food, but not
water-deprived flies. They elicited PER following stimulation of sucrose receptors
found on the terminal segment of the foreleg known as the tarsus. Habituation of
the PER occurred when, after repeated stimulations, the flies no longer gave a PER
in response to sucrose stimulation. The results showed that rovers are more respon-
sive to sucrose than sitters and sitter mutants, and that this was true for flies exposed
to short or long periods of food deprivation. rovers also showed less habituation
than sitters and sitter mutants; rovers and sitters with similar sucrose responsiveness
were used for these habituation studies. Thus for-PKG affects sensory responsiveness
to sucrose and habituation in D. melanogaster. Like rover flies, forager bees show
higher levels of sensory responsiveness than do nurse bees and sitter flies.
Genetic manipulation of PKG in the nematode C. elegans, reveals two modes
of locomotion behavior in the presence of food: roaming and dwelling.17 Roamers
have high-speed locomotion with infrequent turns, while dwellers show low-speed
locomotion with frequent turns. The worm ortholog of the foraging gene, egl-4,
affects this feeding behavior. Interestingly, contrary to the case in Drosophila and
the honey bee, lower levels of PKG in C. elegans is associated with increased
roaming behavior. It is of interest to know whether PKG plays a role in feeding
behavior of other organisms. Recently, Fitzpatrick and Sokolowski15 constructed
protein phylogenies using 32 PKG sequences that include 19 species and proposed
five different evolutionary histories that can explain the link between PKG and
feeding behaviors in fruit flies, honeybees and worms. Studies such as these start
to bridge the gap between genetic model organisms and organisms not traditionally
used for genetic analyses.16
Sokolowski and collaborators have implicated a cGMP signal transduction path-
way in the regulation of food search behavior in flies and honeybees. The next step
is to identify additional components of this food search pathway. This can be
accomplished by using: (1) the single-gene mutant approach to generate mutations
in genes that interact with for, and (2) the QTL analysis approach to localize naturally
occurring variants that act as minor modifiers of foraging behavior in Drosophila.
We are currently mapping genes identified from these types of studies.
The rover/sitter model system is being developed to understand both the mech-
anistic and evolutionary significance of behavioral variation. With this in mind, we
are on the road to developing an understanding of how food search behavior is
affected at the level of the gene, molecule, nervous system, organism, population,
and speciesin both the laboratory and in nature.
ACKNOWLEDGMENTS
This chapter is dedicated to the memory of Dr. Lionel Peypelut, a dear friend and
colleague. The research described here from the Sokolowski laboratory was funded
by NSERC, CIHR and the Canada Research Chairs Program to M.B.S.
REFERENCES
1. Anholt, R.R. and Mackay, T.F.C., Quantitative genetic analysis of complex behaviors
in Drosophila, Nat. Rev. Genet., 5:838849, 2004.
2. Asburner, M., Hawley, S., and Golic, K., Drosophila: A Laboratory Handbook, Cold
Spring Harbor Laboratory, Cold Spring Harbour, New York, 2004.
3. Bell, W.J., Searching Behavior: The Behavioral Ecology of Finding Resources, Chap-
man & Hall, New York, 1991.
4. Ben-Shahar, Y., Ribichon, A., Sokolowski, M.B., and Robinson, G.E., Influence of
gene action across different time scales on behavior, Science, 296:741744, 2002.
5. Carlson, J., Olfaction in Drosophila: from odor to behavior, Trends. Genet.,
12:175180, 1996.
6. Costa, R., Peixoto, A.A., Barbujani, G., and Kyriacou, C.P., A latitudinal cline in
Drosophila clock gene, Proc. R. Soc. Biol. Sci. 250:4349, 1992.
7. de Belle, J.S. and Sokolowski, M.B., Heredity of rover/sitter: alternative foraging
strategies of Drosophila melanogaster, Heredity, 59:7383, 1987.
8. de Belle, J.S. and Sokolowski, M.B., Heredity of rover/sitter foraging behavior in
Drosophila melanogaster, genetic localization to chromosome-2L using compound
autosomes, J. Insect Behav., 2:291299, 1989.
9. de Belle, J.S., Hilliker, A.J., and Sokolowski, M.B., Genetic localization of foraging
(for): a major gene for larval behavior in Drosophila melanogaster, Genetics,
123:157164, 1989.
10. de Belle, J.S., Sokolowski, M.B., and Hilliker, A.J., Genetic analysis of the foraging
microregion of Drosophila melanogaster, Genome, 36:94101, 1993.
11. de Belle, J.S. and Heisenberg, M., Expression of Drosophila mushroom body muta-
tions in alternative genetic backgrounds: a case study of the mushroom body miniature
gene (mbm), Proc. Natl. Acad. Sci. U.S.A., 93(18):98759880, 1996.
12. De Luca, M., Roshina, N.V., Geiger-Thornsberry, G.L., Lyman, R.F., Pasyukova,
E.G., and Mackay, T.F.C., Dopa-decarboxylase affects variation in Drosophila lon-
gevity, Nat. Genet. 34:429433, 2003.
13. Falconer, D.S. and MacKay, T.F.C., Introduction to Quantitative Genetics, 4th ed.,
Longman, New York, 1996.
14. Fanara, J.J., Robinson, K.O., Rollmann, S.M., Anholt, R.R., and Mackay, T.F.C,
Vanaso is a candidate quantitative trait gene for Drosophila olfactory behavior, Genet-
ics, 162(3):13211328, 2002.
15. Fitzpatrick, M.J. and Sokolowski, M.B., In search of food: exploring the evolutionary
link between cGMP-dependent protein kinase (PKG) and behavior, Integr. Comp.
Biol., 44:2836, 2004.
16. Fitzpatrick, M.J., Ben-Shahar, Y., Smid, H.M., Vet, L.E.M., Robinson, G.E., and
Sokolowski, M.B., Candidate genes for behavioral ecology, Trends Ecol Evol,
20:96104, 2005.
17. Fujiwara, M., Sengupta, P., and McIntire, S.L., Regulation of body size and behavioral
state of C. elegans by sensory perception and the EGL-4 cGMP-dependent protein
kinase, Neuron, 36:10911102, 2002.
18. Gleason, J.M., Nuzhdin, S.V., and Ritchie, M.G., Quantitative trait loci affecting a
courtship signal in Drosophila melanogaster, Heredity, 89:16, 2002.
19. Graf, S.A. and Sokolowski, M.B., The rover/sitter Drosophila foraging polymorphism
as a function of larval development, food patch quality and starvation, J. Insect Behav.,
2:301313, 1989.
20. Greenspan, R.J., A kinder, gentler genetic analysis of behavior: dissection gives way
to modulation, Curr. Opin. Neurobiol, 7:805811, 1997.
21. Greenspan, R.J., Fly Pushing, The Theory and Practice of Drosophila Genetics, Cold
Spring Harbor Laboratory Press, Cold Spring Harbour, New York, 2004.
22. Greenspan, R.J., E Pluribus Unum, Ex Uno Plura: Quantitative- and single-gene
perspectives on the study of behavior, Annu. Rev. Neurosci., 27:79105, 2004.
23. Hall, J.C., The mating of a fly, Science, 264: 17021714, 1994.
24. Harmer, S.L., Panda, S., and Kay, S.A., Molecular bases of circadian rhythms, Annu.
Rev. Cell Dev. Biol., 17:215253, 2001.
25. Kalderon, D and Rubin, G.M., cGMP-dependent protein kinase genes in Drosophila,
J. Biol. Chem., 264: 1073810748, 1989.
26. Mackay, T.F.C., Quantitative trait loci in Drosophila, Nat. Rev. Genet. 2:1120, 2001.
27. Mather, K. and Jinks, J.L., Biometrical Genetics, 3rd ed., Chapman & Hall, New York,
1982.
28. Nuzhdin, S.V., Pasyukova, E.G., Dilda, C.L., Zeng, Z.-B., and Mackay, T.F.C., Sex-
specific quantitative trait loci affecting longevity in Drosophila melanogaster, Proc.
Natl. Acad. Sci. U.S.A., 94:97349739, 1997.
29. Nuzhdin, S.V., Dilda, C.L. and Mackay, T.F.C., The genetic architecture of selection
response: inference from fine-scale mapping of bristle number quantitative trait loci
in Drosophila melanogaster, Genetics, 153:13171331, 1999.
30. Osborne, K., Robichon, A., Burgess, E., Butland, S., Shaw, R.A., Coulthard, A.,
Pereira, H.S., Greenspan, R.J., and Sokolowski, M.B., Natural behavior polymor-
phism due to a cGMP-dependent protein kinase of Drosophila, Science, 277:834836,
1997.
31. Ostravik, S., Natarajan, V., Tasken, K., Jahnsen, T., and Sandberg, M., Characteriza-
tion of the human gene encoding the type I alpha and type I beta cGMP-dependent
protein kinase (PRKGI) 1997, Genomics, 42:311318, 1997.
32. Pak, W.L. and Leung, H.T., Genetic approaches to visual transduction in Drosophila
melanogaster, Receptors Channels, 9(3):149167, 2003.
33. Pasyukova, E.G., Roshina, N.V., and Mackay, T.F.C., Shuttle craft: a candidate quan-
titative trait gene for Drosophila melanogaster lifespan, Aging Cell, 3:297307, 2004.
34. Pereira, H.S. and Sokolowski, M.B., Mutations in the larval foraging gene affect adult
locomotory behavior after feeding in Drosophila melanogaster, Proc, Natl. Acad. Sci.
U.S.A., 90:50445046, 1993.
35. Pereira, H.S., MacDonald, D.E., Hilliker, A.J., and Sokolowski, M.B., Chaser (Csr),
a new gene affecting larval foraging behavior in Drosophila melanogaster, Genetics,
140:263270, 1995.
36. Sawin, E.P., Harris, L.R., Campos, A.R., and Sokolowski, M.B., Sensorimotor trans-
formation from light reception to phototactic behavior in Drosophila larvae, J. Insect
Behav., 7:553567, 1994.
37. Scheiner, R. Erber, J., and Page, R.E., Tactile learning and the individual evaluation
of the reward in honey bees (Apis mellifera L.), J. Comp. Physiol., 185:110, 1999.
38. Scheiner, R. Erber, J., and Page, R.E., The effects of genotype, foraging role and
sucrose responsiveness on the tactile learning performance of honey bees (Apis
mellifera L.), Neurobiol. Learn. Mem., 76:138150, 2001.
39. Scheiner, R., Sokolowski, M.B. and Erber, J., Activity of cGMP-dependent protein
kinase (PKG) affects sucrose responsiveness and habituation in Drosophila melano-
gaster, Learn. Mem., 11:303311, 2004.
40. Shaver, S.A., Varnam, C., Hilliker, A.J., and Sokolowski, M.B., The foraging gene
affects adult but not larval olfactory related behavior in Drosophila, Behav. Brain
Res., 95:2329, 1998.
41. Sokolowski, M.B., Foraging strategies of Drosophila melanogaster: a chromosomal
analysis, Behav. Genet., 10:291302, 1980.
42. Sokolowski, M.B., Genetic analysis of behavior in the fruit fly, Drosophila melan-
gaster. In Techniques for the Genetic Analysis of Brain and Behavior, D. Goldwitz,
D. Wahlsten, and R.E. Winer, Eds., Elsevier Science Publishers, 1992.
43. Sokolowski, M.B. and Hansell, K.P., The foraging locus: behavioral tests for normal
muscle movement in rover and sitter Drosophila melanogaster larvae, Genetica,
85:205209, 1992.
44. Sokolowski, M.B., Pereira, H.S., and Hughes, K., Evolution of foraging behavior in
Drosophila by density dependent selection, Proc. Natl. Adac. Sci. U.S.A.,
94:73737377, 1997.
45. Sokolowski, M.B., Drosophila: genetics meets behavior, Nat. Rev. Genet., 2:879890,
2001.
46. Sokolowski, M.B. and Wahlsten, D., Gene-Environment interaction and complex
behavior. In Methods in Genomic Neuroscience, H. Chin, and S. Moldin, Eds., CRC
Press, New York, 2002.
47. Stocker, R.F., The organization of the chemosensory system in Drosophila melano-
gaster: a review, Cell Tissue Res., 273:326, 1994.
48. Stocker, R.F., Drosophila as a focus in olfactory research: mapping of olfactory

sensilla by fine structure, odor specificity, odorant receptor expression, and central
connectivity, Microsc. Res. Tech., 55(5):284296, 2001.
49. Toma, D.P., White K.P., Hirsch, J., and Greenspan, R.J., Identification of genes
involved in Drosophila melanogaster geotaxis, a complex behavioral trait, Nat.
Genet., 31:349353, 2002.
50. Wang, M.H., Lazebny, O., Harshman, L.G., and Nuzhdin, S. Environment-dependent
survival of Drosophila melanogaster: a quantitative genetic analysis, Aging Cell,
3:133140, 2004.
51. Weber, K., Eisman, R., Morey, L., Patty, A., Sparks, J., Tausek, M., and Zeng, Z.B.,
An analysis of polygenes affecting wing shape on chromosome 3 in Drosophila
melanogaster, Genetics, 153(2):773786, 1999.
52. Whitefield, C.W., Cziko, A.M., and Robinson, G.E., Gene expression profiles in the
brain predict behavior in individual honey bees, Science, 302(5643):296299, 2003.
22 Genetic and Molecular

Analyses of Drosophila
Courtship Behavior
Jean-Marc Jallon
CONTENTS
22.1 Introduction ................................................................................................... 319
22.2 Mating Behavior of Drosophila melanogaster,
Visual and Acoustic Stimuli ........................................................................ 320
22.3 Pheromones and Biosynthetic Genes............................................................ 321
22.4 Gustatory and Olfactory Receptors............................................................... 324
22.5 Sex Determination Cascade and the Courtship
Behavior Master Gene Fruitless .................................................................. 326
References.............................................................................................................. 328
22.1 INTRODUCTION
Drosophila melanogaster is a marvelous material for behavioural genetic anal-
ysis. First, it has been shown to be highly amenable for genetic analysis. It has
a reduced developmental time (about 9 days at 25C) and a compact genome
(four pairs of chromosomes including three pairs of autosomes and one pair of
sex chromosomes). The classic genetic tools such as visible markers and balancer
chromosomes can be efficiently used together with new technology to elucidate
molecular mechanisms. The complete genome sequence has been known for 5
years1 and can be compared with those of other model organisms that appear in
the literature. Large parts of those of other Drosophila species are also available2
at http://flybase.bio.indiana.edu.
As far as behavior is concerned, Drosophila behaves in two states, which follow
each other in postembryonic development, as a crawling larva first and then as a
flying insect. The imaginal behaviors are quite sophisticated, involving instinctive
and plastic facets. Mating behavior is the most complex.
319
22.2 MATING BEHAVIOR OF DROSOPHILA

MELANOGASTER, VISUAL AND ACOUSTIC
STIMULI
Contrary to butterflies, which attract each other from a distance, Drosophilae usually
meet on the feeding spot attracted by decomposing fruit volatiles. There the mating
behavior of D. melanogaster is a sequence of fixed action patterns including orien-
tation by a male toward a female, its tapping of the female cuticle, male wing display
generating the courtship song, following of the female by the male, its licking the
females genitalia, rejection or acceptance of the male by the female, male mounting
and copulation upon females acceptance.3,4
The orientation by males appears to be guided primarily by visual cues as shown
in studies with several types of visual mutants.5 Among them, a most spectacular
deviation from wild-type behavior was observed with the blind no retention potential
A males: once they have touched a female, they get excited but are unable to follow
the partner and go on singing in the place where she had been.6 In species of the
auraria complex, the orientation behavior can be quantified more objectivelymale
neck movementsthan in D. melanogaster but in all cases it is triggered across a
window by moving flies whichever their sex.7,8 Moreover, experiments described in
the next section strongly suggest that chemical cues stimulate courtship behavior in
synergy with visual cues.9
Acoustic signals are mainly produced by the vibration of one or two male wings
in which indirect flight muscles play a major role thoroughly studied by Ewing.10
They are perceived by the Johnstons organ, on the antennal second segment, which
is stimulated by movements of the terminal arista, as shown by surgical and muta-
tional experiments. The love song of D. melanogaster consists of two different
elements known as sine song and pulse song. In the pulse song, the time between
successive pulses, called the interpulse interval (IPI), is approximately 35 msec for
D. melanogaster and 50 msec for the sibling species D. simulans. Playback exper-
iments have shown the importance of the IPI duration to stimulate the mating speed
of pairs involving a wingless male and a winged female.11 But other acoustic
parameters that differ between species, such as the pulse song burst duration or sine
song bout length, may play a role. Actually, at early ages, the two species share
similar IPI values, close to those of D. simulans, but maturation establishes the
species specificity of this character.
Single-gene mutations affecting song have led to interesting results. In the old
mutant cacophony induced by the chemical mutagen EMS, pulse song showed a
longer average IPI (45 msecvery different from the wild-type value of 35 msec
and close to that of D. simulans). Males, then, are less successful at mating when
they bear this mutation in an X gene.12 Dissonance, another mutation induced by
the chemical mutagen EMS was found also to affect pulse songpolycyclic and irregular
and was shown to be an allele of the mutant no-on transient, identified for visual
defects and which encodes an RNA binding protein.13 Another song mutation,
croaker, was induced by P insertion mutagenesis. Such mutant males initiate court-
ship after a long latency and also generate polycyclic pulse songs with long IPI
values. Moreover their flying ability is much reduced, suggesting a general defect
Genetic and Molecular Analyses of Drosophila Courtship Behavior 321
in the motor system.14 Other P insertions created mutations affecting the sine song
frequency including Beethoven in a chromosome II gene encoding a protein with
ATPase activity. Several sex determination genes, when mutated, also showed song
abnormalities. Altogether, 17 single genes were involved.15 Another approach with
quantitative genetics took advantage of the natural variation within and between
species.16,17 Several studies of D. melanogaster interstrain variations showed additive
effects of chromosomes II and III to control IPI, with a larger effect of the third
chromosome. But only one candidate gene, tipE, another channel gene on 3L, was
clearly included in a quantitative trait locus (QTL)actually, in the major QTL of
chromosome III. A recent QTL study of IPI differences in backcross hybrids between
D. simulans and D. sechellia, another species of the melanogaster subgroup, has
shown similar contributions of autosomal chromosomes to IPI and localized four
QTL on 2R and two on 3R. In this study, three candidate genes, maleless, croaker,
and fruitless are related to QTL.15
The courtship behavior may be subject to associative modification. Immature
virgin females and recently fertilized females stimulate males as much as mature
virgin females, but both reject courters. While virgins usually escape from the
courting male by decamping, the copulated females extrude their ovipositor which
blocks genital contact. The rejection behavior in fertilized females is triggered by a
peptide which is produced in the male accessory gland and transferred to females
during copulation.18 Males that have previously courted fertilized females show a
reduced level of courtship toward a virgin female.19
22.3 PHEROMONES AND BIOSYNTHETIC GENES

Pheromones have been involved in Drosophila courtship for a long time,20 but
analytical experiments have started more recentlymainly in my group. Olfacto-
metric tests have shown that Drosophila melanogaster CantonS males were not
attracted from a distance by conspecific females, as for moths, but that females could
be attracted by males.9 This led us to look first for female pheromones similar to
the aphrodisiac contact pheromones discovered in Musca domestica,21 and Venard
and Jallon22 showed that one male did not react to a female cuticular extract but that
two males courted each other in the presence of a female extract, more or less
intensively depending on the dose of extract. This suggested that female cuticular
pheromones could indeed stimulate male courtship, but that such chemical cues
acted in synergy with visual cues.6
In the CantonS strain of D. melanogaster, female contact pheromones were
identified in a complex mixture of long-chain hydrocarbons present in the cuticle
of both sexes: (Z)-7, (Z)-11, heptacosadiene (7,11 HD), and homologous dienes are
characteristic of mature females.23 However, while such abundant dienes can stim-
ulate CantonS male courtship to copulation,24,25 a larger spectrum of long-chain
hydrocarbons, present in mature or immature flies, is able to stimulate early courtship
behaviors such as male wing vibrations.26 This is especially the case of unsaturated
hydrocarbon (HC) with 27+/2 carbons bearing at least one double bond in position
(Z)-7, including (Z)-7 pentacosene present in mature flies of both sexes but not of
(Z)-7 tricosene.27 Thus chemical sex appeal involves more than one type of molecule
and is rather a complex mixture of synergistic compounds, HC signatures, whose

proportions might vary between populations like the pheromone blend of moths.28,29
But the more volatile pheromones hypothetized by Shorey and Bartell20 and Savarit
et al.30 have not yet been identified (see also next section).
As far as the male attractivity evidenced in olfactometric tests is concerned, it
should be due to male rich compounds, which are more volatile than female cuticular
compounds. Two of them have to be considered: 7T, the most abundant HC in mature
CantonS males,23 and cis-vaccenyl acetate (cVA), a non-HC compound produced in
the ejaculatory bulb.31 There are two opposite behavioral roles played by cVA: at
low dose it is an aggregation pheromone for both sexes,32 and at high dose it is a
male courtship inhibitor.33 Further behavioral experiments in CantonS have shown
that 7T was also able to stimulate females and inhibit other male courtships.3438
These cuticular HC also play a critical role in the sexual isolation of the sympatric
D. melanogaster and D. simulans and related species.39,40 Indeed 7,11 HD is absent
from the cuticle of D. simulans females, whereas 7T is abundant in both sexes of
most of its populations. 7T stimulates D. simulans male courtship in a dose-depen-
dent way,34 which is inhibited by 7,11 HD.40 7T inhibits CantonS male courtship,
which is stimulated by 7,11 HD.
Most of the HC information about D. melanogaster presented above was deduced
from studies performed with flies from the well-known laboratory strain CantonS.
Later, a rich polymorphism of these cuticular HCs was discovered during a large-
scale study involving 85 populations collected around the world. This study showed
that sub-Saharan African females markedly differed from those of most other pop-
ulations, including CantonS. Such females, like those collected in the primary forest
of Tai (Ivory Coast) also have 7,11 HDbut 10 times less, plus a large amount of
its position isomer (Z)-5, (Z)-9 heptacosadiene (5,9 HD), which is rare in Canton S
females. This chemical difference is recognized by CantonS males which prefer
homotypic females while Tai males make little difference between the two types of
females.34 Moreover males of the various populations showed marked differences in
their content of 7T which increased with the latitude of their collection site, while
the content of the homologous (7P) decreased.41 Females, both equatorial and tem-
perate showed a clear preference for males of their population.42 It is thus suggested
that behavioural responses of males or females of different populations might have
dose-response curves to specific chemical component differing with genome differ-
ences between populations.34,35,43,44
Chromosome exchanges between both diene morphs showed that the 7,11
HD/5,9 HD ratio was mainly controlled by third chromosome loci.45 Later recom-
binations experiments with marker genes and detailed deficiency studies linked this
variation to a single segregating factor in cytological position 87 C-D of chromosome
III,46 located close to gene desat1.47 On the other hand, metabolic studies using
radiolabelled potential precursors had shown that cuticular HC could be synthetized
de novo by flies which could synthetize, elongate, desaturate and decarboxylate fatty
acids to make their pheromones.48 Taking advantage of the high degree of conser-
vation observed among known desaturase genes, Wicker-Thomas et al.47 cloned
desat1, the first gene of desaturase enzyme in Drosophila. This gene with four introns
shows only one open reading frame, although it has a complex pattern of transcription
with some sex differences.49,50 Its heterologous expression in yeast could complement
the mutant ole1, deficient in desaturase activity, and established the D9 specificity
of the Drosophila enzyme as well as its preference for palmitic acid as substrate.51
Insertions of various P transposons in the promoter region of desat1 showed more
or less decrease in the level of all 7 unsaturated HC in both sexes including 7T and
7,11HD. As their excision often reverted the wild-type HC phenotype, the desat1
enzyme is clearly involved in a common step of biosynthesis, the desaturation which
transforms palmitic acid into palmitoleic acid (C16:1,D9/w7).25
Although the 7,11 HD/5,9 HD polymorphism was first hypothetized to be linked
to an allelic variation in the desat1 gene with each allele producing a different
desaturase isoform, the reality was more complex. Although the Tai desat1 gene
coded a polypeptide chain with 3 amino acid substitutions compared to that of
CantonS, the yeast transformation results with either coding sequence were similar,
infirming the simple hypothesis. Later a mini-dissection of the locus around the
desat1 gene revealed the presence of two more desaturase genes, desat2 and desat3.
Desat2 was characterized in the Tai strain. Its uncorrupted open reading frames
(ORF) encodes a polypeptide with 65% identity with that of desat1 but a smaller
amino domain. The heterologous expression of desat2 in yeast also complemented
the deficient mutant ole1, suggesting again a D9 specificity, but a different substrate;
myristoleic acid (C14:1,D9/w5) was mainly produced instead of palmitoleic acid.51
This position w5 for HC is abundant mainly in the African strains and there only
in females. Actually desat2 can be expressed only in African females that make
5,9HD. These facts support a critical role for both desat1 and desat2 in the first
desaturation of linear saturated fatty acids. Another study has shown that the knock-
down of the desat2 gene in ancestral populations during their migrations in Africa
was linked to the deletion of 16 bp in the promoter region.52
To finally make dienes, monoenic fatty acids have to go through a second
desaturation, taking place in females of the D. melanogaster species and probably
not in D. simulans. As the enzyme specificity is usually restricted to a certain type
and position of bond along the chain and a substrate either saturated or monoenic,
other desaturase genes were searched for. Indeed the published sequence of
D. melanogaster genome suggested the existence of more desaturase genes, one on
chromosome II supposed to be more specific for males and four more on chromo-
some III.1,49 Among these latter genes, one in the cytological region 68 had been
shown to be possibly important to make female dienes.53 Chertemps et al.54 then
completely characterized a new desaturase gene, called desatF, which is expressed
in D. melanogaster females but not in conspecific males. Using the interference
RNA technique, they were able to knock down this gene almost completely. A
dramatic result was observed in the cuticle of transformed females: the level of 7,11
HD was decreased by more than 80% while that of all 7-monoenes was much
increased. This clearly shows that desatF adds a second double bond to monoenic
fatty acids to make dienic ones. The description of another group of biosynthetic
genes has been initiated, those coding for fatty acid elongases.55
Another approach to try to localize genes controlling HC production was quan-
titative genetics QTL analyses of HC were performed in flies of recombinant lines
produced between D. simulans and D. sechellia, another member of the melanogaster
subgroup restricted to the Seychelles archipelago. As already mentioned, the former

species has 7T as its major cuticular compound in both sexes of most populations
while the latter species shows a marked sexual dimorphism with much 7,11HD in
mature females as in temperate D. melanogaster females. Separate QTL analyses
of 7T and 7,11HD evidenced two regions for each trait. Both are associated with
one region on the right arm of chromosome III, while the others are either on the
left arm of that chromosome (7,11HD) or on chromosome X (7T). Interestingly, it
was noted that desat1/2 on one hand and desatF on the other hand were found to
be located near the QTL shared by 7T and 7,11HD and the QTL specific for 7,11HD,
respectively.15
22.4 GUSTATORY AND OLFACTORY RECEPTORS

The sex pheromones of female moths, which are very volatile, are perceived by male
antennal olfactory sensillae. In D. melanogaster male antennae had been considered
less important for the female excitatory messages, as antennaless homozygous males
were known to court intensively wild type females whichever the number of their
antennae.56 But it has been suggested that they might receive male inhibitory inputs
as a large percentage of the mutant males without antennae courted each other.8
Moreover it has been shown with mutants that the main class of antennal olfactory
sensillae basiconicaewere not necessary for the detection of female phero-
mones.57 On the other hand, there were evidences that gustatory organs could perceive
excitatory female compounds, when the male raises his foreleg to tap the epicuticle
and then vibrates one wing or when he licks the female genitalia with his proboscis
before attempting copulation. For example Venard et al.58 have coated male legs with
a thin layer of dental wax; when all legs had been coated, no wing vibration toward
females could be observed, but if only five legs had been treated, the response varied
in intensity depending on the untreated leg; it was quasi-null if a hindleg was untreated
and strong if it was a foreleg. This suggested that some of the gustatory hairs present
on the foreleg and not in the other pairs of legs might perceive cuticular compounds.
Scanning electron microscopy and crystal violet diffusion experiments did detect a
set of typical gustatory sensillae displaying a marked sexual dimorphism.58 Other
gustatory sensillae, which have been described on the proboscis, are used when the
male licks the female genitalia in a later courtship step.59
Many new and exciting information resulted from the molecular studies which
followed the discovery of olfactory receptors by Buck and Axel60 and the completion
of the Drosophila genome sequencing.1 Two large families of about 60 genes coding
for chemoreceptors with seven transmembrane domains and coupled with G proteins
were discovered by several groups,6163 about half involved in olfaction Or, and half
involved in gustation Gr. Then an important functional work has been initiated using
the GAL4-UAS method designed by Brand and Perimon64 and which makes use of
a yeast transcription factor GAL4 which directs the expression of a gene of inter-
est*for example a given Gr-or a reporter gene only in the target cells defined by
the gene* promoter. In this way Bray and Amrein65 have characterized Gr68a which
is expressed only in males and only in about 20 of the male specific chemosensillae
borne by the forelegs. Then the knockdown of this Gr68a gene by RNA interference
in the very neurons where their expression is restricted and the inactivation of the
same neurons by tetanus toxin have both led to a marked decrease of male courting
activity; moreover, both manipulations reduced occurrences of specific courtship
steps such as wing vibrations and attempted copulations. This strongly suggested
that Gr68a was indeed a good candidate receptor to attach a female contact phero-
mone such as 7,11 HD; direct binding experiments still have to be done. The fact
that these transformed males are still able to perform part of the courtship suggests
the involvement of other types of Gr, possibly involved in binding of other contact
pheromones.66 Other elements of the pheromone transduction cascade have also been
described, two chemosensory proteins CheA29a and CheB42a which share some
characteristics with oderant pheromone binding protein (OBP/PBP) and are expressed
only in the front legs of males as Gr68a67 and one ion channel subunit, ppk25. This
gene is expressed not only in legs, but also in antennae and several of its mutations
strongly affect the courtship behaviour of males.68
In the olfactory system, it has been clearly shown that most of the 62 Or genes
were expressed alone in olfactory sensory neurons, while the situation looked more
complex for Gr in gustatory neurons.6971 Another difference is the projection
pattern of the sensory neurons. While most olfactory sensory neurons expressing
a given Or converge to a single glomerulus, one functional subunit of the brain
antennal lobes, a similar structure to the vertebrate olfactory bulb, gustatory sensory
neuron primary projections are not in the brain but in different parts of the nervous
system and thus much less well characterized. For example all gustatory neurons
of both forelegs converge to the anterior part of the thoracic ganglion. In a pioneer
study, Possidente and Murphey72 have described the sexual dimorphism of these
projections.
Only 2 out of about 50 glomeruli7374 have been found to display a dimorphism
between males and females VA1v and mainly DA1, which is 62% larger in males
than in females.75 These glomeruli are anatomically homologous to the more dra-
matically dimorphic glomeruli of moths. In males of these insects there is a macro-
glomerulus whose function is to integrate the complexity of the female pheromone
bouquet. Interestingly these two large structures are targets for the fruitless tran-
scription factor which also target only one or two other glomeruli.69 The olfactory
sensory neurons which converge toward the Drosophila dimorphic glomeruli start
in trichoid sensillae like those sensitive to pheromones in moths;2 they express
olfactory receptors which have been identified as respectively Or47b and Or67d.
Could cVA not be a possible ligand for one of these olfactory receptors? cVA
has been shown to be detected by a subset of trichoid sensillae of the antennal third
segment58,61 and one piece of its transduction machinery has been identified. Flies
lacking the gene lush, which codes for its OBP, display much less aggregation
ability.67 This gene has been originally identified because some of its mutations lead
to specific attraction toward high concentrations of ethanol and other short-chain
alcohols which are avoided by wild-type flies. Actually, lush is expressed in about
150 trichoid sensillae on the ventral-lateral surface, in both males and females, whose
sensory neurons project into the VA1 and DA1 glomeruli.76 It is possible that another
ligand might be the volatile female pheromone hypothetized by Shorey and Bartell.20
22.5 SEX DETERMINATION CASCADE AND THE

COURTSHIP BEHAVIOR MASTER GENE FRUITLESS
The chromosomal sex of D. melanogaster is determined by the relative number of
X and A chromosomes (A being any autosome, either II, III, or IV). The presence
of a Y chromosome is not as important as it is in higher vertebrates. But there is a
set of genes which control each other in a hierarchical way as shown in the Baker
group and the Nothiger group. The X-linked gene, sex-lethal, controls a series of
autosomal genes including two transformer genes (tra on chromosome III and tra-2
on the second) and the doublesex gene (III). Female somatic development requires
first the activation of sex-lethal whose products regulate the splicing of tra main
transcript. This tra product is necessary for femaleness during one critical period
during development. Tra encodes another RNA binding protein which regulates the
expression of a very small number of genes, among them doublesex which is then
expressed in a female specific transcript. In the absence of the femizing tra product,
a male specific doublesex transcript is produced. The misexpression of doublesex
might result in the transformation of both chromosomal males and females into
intersexes.7779
The use of gynandromorphs has suggested that neural circuits for male and
female behaviors might coexist in the brain. Such sex mosaics are produced with
unstable ring-X chromosome (Xr), the early loss of which produces clones of cells,
either female (X*/Xr) or male (X*/O). Using sex-linked markers (such as the aster-
isk,*) to infer the sex genotypes of internal tissues and a detailed analysis of the
sexual behaviors either typically male or female, it was possible to identify the tissue
the sex of which correlates with the sex of behavioral patterns. They were called
behavioral foci by Hotta and Benzer.80 The focus for male behavior includes a limited
part of the head. A more refined mapping using internal histochemical markers
localized this focus in the dorsal posterior brain, an era which includes the mushroom
bodies, structures already involved in the sensory information integration. But the
focus controlling female behaviors, such as receptivity and oviposition, was mapped
in the dorsal anterior part of the brain which contains the neuroendocrine cells.81
The two foci commanding sex specific behaviors thus seemed distinct.81 Another
female important focus, the sex-appeal focus, was very far from the brain; this
internal structure was in the ventroposterior region of the blastoderm fate map,
pointing to fat bodies and oenocytes.82
More recently it became possible to produce sex mosaics where in a fly of one
sex the other sex was expressed only in a brain tissue, thanks to the GAL4-UAS
method introduced earlier.64 Actually, a group of brain cells could be feminized in
males through a tissue restricted transcription of the major feminizing gene, tran-
former.83 Some of the transformed males courted wild-type males as continuously
as wild-type females. The parallel activation of the reporter gene B-galactosidase
allowed the identification of the target tissues, a few functional glomeruli of the
antennal lobes and part of their projections in the mushroom bodies.84 The observed
bisexual behavior might be linked either to the dysfunction of the male inhibition
center or to the switching on of a female specific activation center, as a consequence
of the tra induced feminization.
Flies with a female caryotype but which are mutant for tra develop and behave
as males. In such flies, a female phenotype could be restored by a transgene that
carried the female-specific cDNA of tra under the control of a heat-shock promoter.
This transgene could also transform caryotypic male animals into sterile females.
Raised at 25C, both such transformed XX and XY displayed typical male courtship
while at the same time, rich in dienes, they were attractive to males. When the
expression of tra was forced by heat shock, applied during a limited period around
puparium formation, male behavior was abolished and replaced by female behavior.
Thus sexual behavior is irreversibly programmed during a critical period as a result
of the activity switch of tra.85
Another strategy used a collection of mutations which affected all male courtship
behavior and mapped in a small region of the third chromosome, at the fruitless(fru)
locus. The first fru mutant (fru 1) was induced by x-rays and described behaviorally
first as a male-sterile mutant, then as a male courter of both sexes.86 Actually, it
corresponded to a small deletion together with a chromosomal inversion in the
interval 90C91B.87 Later four additional alleles (fru 2, fru 3, fru 4 and fru-satori)
have been obtained by P element insertion mutagenesis and showed different behav-
ioral characteristics. For example fru-satori males do not court females but can court
males; fru2-4 males court both males and females; fru 4 males do not perform love
song whichever the courted sex.4,8,88 Although a recognizable insertion like P made
this gene cloning possible, molecular studies were laborious as fru had several
transcripts and several promoters were identified. Homologies with members of the
BTB family of transcriptional regulators that also contains zinc-finger motifs were
found. In the regulatory region, the presence of three putative binding sites for the
transformer gene product, similar to those found in doublesex clearly showed that
fru was a neural sex-determination switch downstream of transformer, occupying
in the sex determination cascade a position parallel to doublesex.89,90
While transcripts from three promoters are common for both sexes, transcripts
from promoter P1 are controlled by the sex determination cascade and produce in
males a specific form fruM that is longer. A series of elegant experiments have
allowed Demir and Dickson91 to demonstrate that male splicing specifies the main
part of male courtship behavior. They also showed that this mode of splicing is
sufficient to have an otherwise normal female court as a male.91 But how is the gene
organizing the male neurocircuit which performs the male courtship program?
A transgene, in which the GAL4 coding sequence had been introduced into the
fruM coding sequence, was constructed by two groups and used to direct a detailed
search of fru target cells.92,93 Most of the peripheral sensory systems were concerned,
that is about 20 to 30 gustatory neurons of both the prothoracic legs and the proboscis
but also more than 100 to 200 neurons of the olfactory and acoustic parts of the
antennae. The olfactory neurons projected primarily to 34 glomeruli of the antennal
lobes as reported earlier.
In the central part of the nervous system, fruM is expressed in a few scattered
groups of cells. One group is a bilateral cluster of about 60 neurons in the suboe-
sophageal ganglion where fruM expression could be blocked with the RNA inter-
ference technique.94 Surprisingly, such transformed males started to court wild-type
females ten times more quickly than wild types, apparently bypassing the sensory
steps of early courtship such as orientation and tapping usually occurring in wild
type males; but they quickly displayed all later steps, often all at the same time,94
suggesting that these suboesophageal neurons might inhibit the courtship start until
sufficient and adequate sensory stimuli allow it.
The question of sexual dimorphism in the number of and positions of brain cell
clusters involved in courtship behavior has also been investigated. Until recently,
clear differences had been found only in the abdominal ganglion.72,95 In 2005,
Stockinger et al.,93 using a fluorescent reporter gene directed by a specific P [GAL4]
element inserted in the fru gene, have only observed larger clusters in the male
superior protocerebrum. But, with a slightly different technique, Kimura et al.96
could identify two more groups of sexually dimorphic neuron clusters: in the medulla
of the optic lobes there were a few more interneurons only in males; marked
differences in a brain region dorsal to the antennal lobes were also characterized.
There, on each side, a cluster of about 30 neurons was characterized in males while
there were not more than 5 in females. Moreover, these neurons also displayed clear
sex differences in their projection patterns. These interneurons seem to start in the
suboesophageal ganglion where gustatory informationspossibly contact phero-
mone inputscollected by tarsal and labellar gustatory neurons might be integrated
and finish in the superior lateral protocerebrum.
Transformer and Fruitless are thus remarkable examples of switch genes able
to transform a complex behavioral program such as courtship. The behavioral trans-
formation is linked to structural modifications in the nervous system. In both cases,
they are not fully understood but seem limited to a few clusters of neurons. The
action of such master genes can be studied in more depth thanks to all the sophis-
ticated tools that Drosophila provides for cellular and molecular analyses.
REFERENCES
1. Myers, E.W., Sutton, G.G., Delcher, A.L. et al., A whole genome assembly of
Drosophila, Science, 287, 2196, 2000.
2. Ashburner, M., Golic, K. and Hawley, S., Drosophila, A Laboratory Handbook, 2nd
ed., Cold Spring Harbor Laboratory Press, Cold Spring Harbor, 2005.
3. Bastock, M. and Manning, A., The courtship of Drosophila melanogaster, Behaviour,
8, 85,1955.
4. Hall, J.C., The mating of a fly, Science, 264, 1702, 1994.
5. Markow, T.A., Behavioral and sensory basis of courtship success in Drosophila
melanogaster, Proc. Natl. Acad. Sci. USA, 84, 6200, 1987.
6. Jallon, J.M. and Hotta, Y., Nonchemical messages of the female Drosophila melan-
ogaster, in Taniguchi Symposia in Biophysics No7, Hotta, Y., Tokyo University Press,
Tokyo, 1983, 136.
7. Oguma, Y. et al., Courtship behavior and sexual isolation between Drosophila auraria
and D. triauraria in darkness and in light, J. Evol. Biol., 9, 803, 1996.
8. Yamamoto, D., Jallon, J.M. and Komatsu, A., Genetic dissection of sexual behavior
in Drosophila melanogaster, Annu. Rev. Entomol., 42, 551, 1997.
9. Venard, R. and Jallon, J.M., Evidence for an aphrodisiac pheromone of female
Drosophila, Experientia, 36, 211, 1980.
10. Ewing, A., Functional aspects of Drosophila courtship, Biol. Rev., 58, 275, 1983.
11. Rybak, F., et al., Acoustic communication in Drosophila melanogaster courtship: are
pulse- and sine-frequencies important for courtship success? Can. J. Zool., 80, 987,
2002.
12. Schilcher, F.V., A mutation which changes courtship song in Drosophila melano-
gaster, Behav. Genet., 7, 251, 1977.
13. Kulkarni, S.J., Steinlanf, A.F. and Hall, J.C., The dissonance mutant of courtship
song in Drosophila melanogaster, Genetics, 118, 267, 1988.
14. Yokokura, T., Ueda, R. and Yamamoto, D., Phenotypic and molecular characterization
of croaker, a new mating behavior mutant of Drosophila melanogaster, Japan J.
Genet., 70, 103, 1995.
15. Gleason, J. and Ritchie, M., Do quantitative trait loci for a courtship song difference
between Drosophila simulans and D. sechellia overlap with candidate genes and
intraspecific QTLs?, Genetics, 166, 1303, 2004.
16. Mackay, T.F. et al., Genetics and genomics of Drosophila mating behavior, Proc.
Natl. Acad. Sci. USA, 102, 6622, 2005.
17. Gleason, J. et al., A quantitative trait loci for cuticular hydrocarbons associated with
sexual isolation between Drosophila simulans and D. sechellia, Genetics, 171, 1, 2005.
18. Chen, P.S., Stumm-Zollinger, E., Aigaki, T., Balmer, J., Bienz, M. and Bohlen, P., A
male accessory gland peptide that regulates reproductive behavior of female Droso-
phila melanogaster, Cell, 54, 291, 1988.
19. Gailey, D.A., Jackson, F.R. and Siegel, R.W., Conditioning mutations in Drosophila
melanogaster affect an experience-dependent behavioral modification in courting
males, Behav. Genet., 106, 613, 1984.
20. Shorey, H.H. and Bartell, R., Role of volatile sex pheromone in stimulating male
courtship behavior in Drosophila melanogaster, Anim. Behav., 18, 159, 1970.
21. Carlson, D.A. et al., Sex attractant pheromones of the house fly: isolation, identifi-
cation and synthesis, Science, 174, 76, 1971.
22. Venard, R., Attractants in the courtship behavior of Drosophila melanogaster, in
Development and Neurobiology of Drosophila, Siddiqi, O., Babu, P., Hall, L.M. and
Hall, J.C., Plenum Press, New York, 1980, 457.
23. Antony, C. and Jallon, J.M., The chemical basis for sex recognition in Drosophila
melanogaster, J. Insect. Physiol., 28, 873, 1982.
24. Arienti, M., Analyse de la variabilit de quelques mcanismes impliqus dans le
comportement sexuel de populations differentes de Drosophila melanogaster, Ph.D.
thesis, Universit de Paris XI, Orsay, 1993.
25. Ueyama, M. et al., Mutations in the desat1 gene reduces the production of courtship-
stimulatory pheromones through a marked effect on fatty acids in Drosophila mela-
nogaster, Insect Biochem. Mol. Biol., 35, 911, 2005.
26. Pechine, J.M., Antony, C. and Jallon, J.M., Precise characterization of cuticular
compounds in young Drosophila, J. Chem. Ecol., 14, 1071, 1988.
27. Antony, C. et al., Compared behavioural responses of male Drosophila melanogaster
(Canton S) to natural and synthetic aphrodisiacs, J. Chem. Ecol., 11, 1617, 1985.
28. Klun, J.A. and Maini, S., Genetics basis of an insect communication system in the
European corn borer, Environ. Entomol., 11, 1084, 1982.
29. Jallon, J.M. and Pechine, J.M., Une autre race chimique de Drosophila melanogaster
en Afrique, CR. Acad. Sci., Paris, 309, 1551, 1989.
30. Savarit, F. et al., Genetic elimination of known pheromones reveals the fundamental
chemical basis of mating and isolation in Drosophila, Proc. Natl. Acad. Sci. USA,
96, 9015, 1999.
31. Butterworth, F.M., Lipids in Drosophila, a newly detected lipid in the male, Science,
163, 1256, 1969.
32. Bartelt, R.J. et al., Cis-vaccenyl acetate acts as an aggregation pheromone in Droso-
phila melanogaster, J. Chem. Ecol., 11, 1747,1986.
33. Jallon, J.M., Antony, C. and Benamar, O., Un anti-aphrodisiaque produit par les males
de Drosophila melanogaster, C. R. Acad. Sci., Paris, 292, 1147, 1981.
34. Jallon, J.M., A few chemical words exchanged by Drosophila during courtship, Behav.
Genet., 14, 441, 1984.
35. Scott, D. and Jackson, L.L., Interstrain comparison of male predominant aphrodisiacs
in Drosophila melanogaster, J. Insect Physiol., 34, 863, 1988.
36. Scott, D., Genetic variation for female mate discrimination in Drosophila melano-
gaster, Evolution, 48, 112, 1994.
37. Ferveur, J.F. and Sureau, G., Simultaneous influence on male courtship of stimulatory
and inhibitory pheromones produced by live sex-mosaics Drosophila melanogaster,
Proc. R. Soc. London B, 263, 967, 1996.
38. Grillet, M., Dartevelle, L. and Ferveur, J.F., A Drosophila male pheromone affects
female sexual receptivity, Proc. R. Soc. London B, 273, 315323, 2006.
39. Cobb, M. and Jallon, J.M., Pheromones, mate recognition and courtship stimulation
in the Drosophila melanogaster species sub-group, Anim. Behav., 39, 1058, 1990.
40. Coyne, J.A., Crittenden, A.P. and Mah, K., Genetics of a pheromonal difference
contributing to reproductive isolation in Drosophila, Science, 265, 1461, 1994.
41. Rouault, J., Capy, P. and Jallon, J.M., Variations of male cuticular hydrocarbons with
geoclimate variables: an adaptative mechanism in Drosophila melanogaster?, Genet-
ica, 110, 117, 2001.
42. Haerty, W. et al., Reproductive isolation in natural populations of Drosophila mela-
nogaster, Genetica, 116, 215, 2002.
43. Sureau, G. and Ferveur, J.F., Co-adaptation of pheromone production and behavioural
responses in Drosophila melanogaster males, Genet. Res., 74, 129, 1999.
44. Marcillac, F., Grosjean, Y. and Ferveur, J.F., A single mutation alters production and
discrimination of Drosophila pheromones, Proc. R. Soc. London B, 272, 303, 2004.
45. Ferveur, J.F. et al., World-wide variation in Drosophila melanogaster sex pheromones:
behavioural effects, genetic bases and potentially evolutionary consequences, Genet-
ica, 97, 73, 1996.
46. Coyne, J.A.,Wicker-Thomas,C. and Jallon, J.M., A gene responsible for a cuticular
hydrocarbon polymorphism in Drosophila melanogaster, Genet. Res., 73, 189, 1999.
47. Wicker-Thomas, C., Henriet, C. and Dallerac, R., Partial characterization of a fatty
acid desaturase gene in Drosophila melanogaster, Insect Biochem. Mol. Biol., 27,
963, 1997.
48. Pennanech, M. et al., Incorporation of fatty acids into cuticular hydrocarbons of male
and female Drosophila melanogaster, J. Insect Physiol., 43, 1111, 1997.
49. Jallon, J.M. and Wicker-Thomas, C., Genetic studies on pheromone production in
Drosophila, in Insect Pheromone Biochemistry And Molecular Biology, Blomquist,
G.J. and Vogt, R.G, Elsevier, Amsterdam, 2003, 253.
50. Marcillac, F. et al., A mutation with major effects on Drosophila melanogaster sex
pheromones, Genetics, 171, 1617, 2005.
51. Dallerac, R.C. et al., Delta 9 desaturase gene with a different substrate specificity is
responsible for the cuticular diene hydrocarbon polymorphism in Drosophila mela-
nogaster, Proc. Natl. Acad. Sci. USA, 97, 9449, 2000.
52. Takahashi, A.S. et al., The nucleotide changes governing cuticular hydrocarbon vari-
ation and their evolution in Drosophila melanogaster, Proc. Natl. Acad. Sci. USA,
98, 3920, 2001.
53. Wicker-Thomas, C. and Jallon, J.M., Role of Enhancer of zeste on the production of
Drosophila melanogaster pheromonal hydrocarbons, Naturwissenschaften, 87, 76,
2000.
54. Chertemps, T. et al., Female-specific desaturase gene responsible for diene hydrocar-
bon biosynthesis and courtship behaviour in D. melanogaster, Insect Molec. Biol., in
press, 2006.
55. Chertemps, T. et al., A new elongase selectively expressed in Drosophila male
reproductive system, Biochem. Biophys. Res. Commun., 333, 1066, 2005.
56. Begg, M. and Hogben, L., Chemoreceptivity of Drosophila melanogaster, Proc. R.
Soc. London B, 133, 1, 1946.
57. Stocker, R.F. and Gendre, N., Courtship behavior of Drosophila, genetically and
surgically deprived of basichonic sensilla, Behav. Genet., 19, 371, 1988.
58. Venard, R., Antony, C. and Jallon, J.M., Drosophila Chemoreceptors, in Neurobiology
Of Sensory Systems, Singh, N. and Strausfeld, N., Eds., Plenum Press, New York,
1989, 377.
59. Nayak, S.V. and Singh, R.N., Sensilla on the tarsal segments and the mouthparts of
adult Drosophila melanogaster, Int. J. Insect Morph. Embryol., 12, 273, 1983.
60. Buck, L. and Axel, R., A novel multigene family may encode odorant receptors: a
molecular basis for odor recognition, Cell, 65, 175, 1991.
61. Clyne, P.J., Warr, C.G. and Carlson, J.R., Candidate taste receptors in Drosophila,
Science, 287, 1830, 2000.
62. Scott, K. et al., A chemosensory gene family encoding candidate gustatory and
olfactory receptors in Drosophila, Cell, 104, 661, 2001.
63. Dunipace, L., Meister, S., McNealey, C. and Amrein, H., Spatially restricted taste
receptors in the Drosophila gustatory system, Curr. Biol., 11, 822, 2001.
64. Brand, A.H. and Perimon, N., Targeted gene expression as a means of altering cell
fates and generating dominant phenotypes, Development, 117, 401, 1993.
65. Bray, S. and Amrein, H., A putative Drosophila pheromone receptor expressed in
male-specific yaste neurons is required for efficient courtship, Neuron, 39, 1019, 2003.
66. Thorne, N. et al., Taste perception and coding in Drosophila, Cur. Biol., 14, 1065,
2004.
67. Xu, P.X. et al., Drosophila OBP LUSH is required for activity of pheromone-sensitive
neurons, Neuron, 45, 193, 2005.
68. Lin, H. et al., A Drosophila DEG/EnaC channel subunit is required for male response
to female pheromones, Proc. Natl. Acad. USA, 102, 12831, 2005.
69. Fishilevitch, E. and Vosshall, L.B., Genetic and functional subdivision of the Droso-
phila antennal lobe, Cur. Biol., 15, 1548, 2005.
70. Couto, A., Alenius, M., and Dickson, B.J., Molecular, anatomical and functional
organization of the Drosophila olfactory system, Cur. Biol., 15, 1535, 2005.
71. Amrein, H. and Thorne, N., Gustatory perception and behavior in Drosphila melan-
ogaster, Cur. Biol., 15, 673, 2005.
72. Possidente, D.R. and Murphey, R.K., Genetic control of sexually dimorphic axon
morphology on Drosophila sensory neurons, Dev. Biol., 132, 448, 1989.
73. Stocker, R., The organization of the chemosensory system in Drosophila melano-
gaster, Cell Tissue Res., 275, 3, 1994.
74. Laissue, P.P. et al., Three-dimensional reconstruction of the antennal lobe in Droso-
phila melanogaster, J. Comp. Neurology, 405, 543, 1999.
75. Kondoh, Y. et al., Evolution of sexual dimorphism in the olfactory brain of Hawaiian
Drosophila, Proc. R. Soc. Lond., 270, 1005, 2003.
76. Kim, M.S., Repp, A. and Smith, D.P., LUSH Odorant-Binding Protein mediates
chemosensory responses to alcohols in Drosophila melanogaster, Genetics, 150, 711,
1998.
77. Nothiger, R. and Steinman-Zwicky, M., Sex determination in Drosophila, Trends in
Genetics, 1, 209, 1985.
78. Baker, B., Sex in flies: the splice of life, Nature, 340, 521, 1989.
79. Christiansen, A.E. et al., Sex comes in from the cold: the integration of sex and
pattern, Trends in Genetics, 18, 510, 2002.
80. Hotta, Y. and Benzer, S., Courtship in Drosophila mosaics: sex-specific foci for
sequential action patterns, Proc. Natl. Acad. Sci. USA, 73, 4154, 1976.
81. Hall, J.C., Control of male reproductive behavior by the central nervous system of
Drosophila: dissection of a courtship pathway by genetic mosaics, Genetics, 92, 437,
1979.
82. Jallon, J.M. and Hotta, Y., Genetic and behavioural studies of female sex appeal in
Drosophila, Behav. Genet., 9, 257, 1979.
83. McKeown, M., Belote, J.M. and Boggs, R.T., Ectopic expression of the female
transformer gene leads to female transformation of chromosomally male Drosophila,
Cell, 53, 887, 1988.
84. Ferveur, J.F. et al., Genetic feminization of brain structures and changed sexual
orientation in male Drosophila, Science, 267, 902, 1995.
85. Arthur, B.I. et al., Sexual behavior in Drosophila is irreversibly programmed during
a critical period, Cur. Biol., 8, 1187, 1998.
86. Hall, J.C., Courtship among males due to a male-sterile mutation in Drosophila
melanogaster, Behav. Genet., 8, 125, 1978.
87. Gailey, D. and Hall, J.C., Behavior and cytogenetics of fruitless in Drosophila mel-
anogaster: different courtship defects caused by separate closely linked lesions,
Genetics, 121, 773, 1989.
88. Villella, A., Extended reproductive roles of the fruitless gene in Drosophila melan-
ogaster by behavioural analysis of new fruitless mutants, Genetics, 147, 1011, 1997.
89. Ito, H. et al., Sexual orientation in Drosophila is altered by the satori mutation in the
sex-determination cascade gene fruitless that encodes a zinc finger protein with a
BTB domain, Proc. Natl. Acad. Sci. USA, 93, 9687, 1996.
90. Ryner, L.C. et al., Control of male sexual behavior and sexual orientation in Droso-
phila by the fruitless gene, Cell, 87, 1079, 1996.
91. Demir, S. and Dickson, B.J., fruitless splicing specifies male courtship behaviour in
Drosophila, Cell, 121, 785,2005.
92. Manoli, D.S. et al., Male-specific fruitless specifies the neural substrates of Droso-
phila courtship behaviour, Nature, 436, 395, 2005.
93. Stockinger, P. et al., Neural circuitry that governs Drosophila male courtship behavior,
Cell, 121, 795, 2005.
94. Manoli, D.S. and Baker, B., Median bundle neurons coordinate behaviours during
Drosophila male courtship, Nature, 430, 564, 2004.
95. Lee, G. and Hall, J.C., Abnormalities of male-specific FRU protein and serotonin
expression in the CNS of fruitless mutants in Drosophila, J. Neurosc., 21, 513, 2001.
96. Kimura, K.I. et al., Fruitless specifies sexually dimorphic neural circuitry in the
Drosophila brain, Nature, 438, 229, 2005.
23 A New Era for Drosophila

Learning and Memory
Studies
Daniel Comas, Guillaume Isabel and
Thomas Prat
CONTENTS
Introduction............................................................................................................ 333
23.1 Visualizing Neuronal Circuits Thanks to Genetics ..................................... 338
23.2 Imaging Brain Activity ................................................................................ 339
23.3 Learning and Memory Mutants ................................................................... 340
23.4 Temporal Control of Gene Expression........................................................ 341
23.5 A Tool to Build Anatomo-Functional Maps ............................................... 343
23.6 Localization of Olfactory Memory in Drosophila MBs ............................. 343
23.7 Dynamic of Olfactory Memory Phases in Drosophila ............................... 345
References.............................................................................................................. 348
INTRODUCTION
Drosophila melanogaster is one of the most intensively studied organisms in biology,
and it serves as a model system for the investigation of many cellular, developmental
and behavioral processes common to other species, including humans. This holds
true in particular for brain studies, as Drosophila central nervous system is made of
neurons and glia that operate on the same fundamental principles as their mammalian
counterparts. Thus, most neurotransmitters are identical in flies and humans, and
despite the fact that the Drosophila brain has only 100,000 cells,1 it produces complex
behaviors and sustains various forms of learning and memory. Besides studies of
fundamental brain properties, fly models are being developed for a variety of neu-
rodegenerative disorders, and the field is beginning to harness the power of Droso-
phila genetics to dissect pathways of disease pathogenesis and identify potential
targets for therapeutic intervention. This approach is possible because about 50%
of human genes have a Drosophila ortholog.2 In addition, transgenic strategies that
allow to introduce human genes into Drosophila continue to expand the list of
modeled diseases, which now includes Parkinsons disease, Alzheimers disease,
Huntingtons disease and several spinocerebellar ataxias.3,4
333
During the past few decades, the pool of molecular genetics techniques that
apply to the fruit fly has increased enormously. It has been a long time since the
selection of randomly and chemically produced mutants was the most common
technique to analyze genes function. Nowadays, the extensive use of transposable
element-induced mutations allows a much faster identification and study of genes
involved in a given developmental or physiological pathway. Once combined with
the systematic generation of high-resolution deletions, this global approach should
eventually provide researchers with a complete mutant collection (see Table 23.1).5,6
Moreover, new techniques have arisen that permit one to directly disturb the
expression of a defined gene, without the need of performing a mutagenesis. Rong79
and colleagues have developed the technique of homologous recombination in
Drosophila. This technique, however, remains heavy. The development of the inter-
ference RNA (RNAi) combined with the P[GAL4] enhancer-trap system allows one
to inhibit the expression of any gene of interest in almost any group of cellsfor
example, in a particular brain circuit.10
Drosophila genome is one of the first multicellular eukaryotic genomes to be
completely sequenced, together with Caenorhabditis elegans.11,12 Genes prediction has
been carried out by two different consortiums (Berkeley Drosophila Genome Project,
BDGP, and Heidelberg FlyArray) with some interesting small differences in the total
gene number and limits.13,14 The results of these international collaborations have been
made public in different Web sites (Table 23.1). Flybase offers most of the information
available for each of the 14,000 to 16,000 predicted Drosophila genes (Humans are
predicted to have about 20,000 to 25,000 protein-coding genes).15,16 One can find in
Flybase not only the sequence of the different transcripts and proteins, but also the
different mutant alleles and phenotypes, predicted protein domains, molecular function,
similarities within other species, publications referring to this gene, and more.
The whole-genome sequencing strategy has opened a new era of molecular strate-
gies such as the studies of the transcriptome and protein interactome.17,18 Several papers
report the use of microarrays to study physiological pathways involved in circadian
rhythms, immunity, sex differences, or memory.1924 Recently, White and collaborators
used probe sequences tiled throughout the genome.14 More than 170,000 oligonucle-
otides were designed against each exon, intron, and intergenic region. Their purpose
was not only to detect the differential expression of all genes and their alternative
mRNAs throughout development, but also to identify new, transcribed regions.
Giot and colleagues25 published a two-hybridbased, proteinprotein interaction
map of the fruit fly proteome. They produced a draft map of 7,048 proteins and
20,405 interactions, each with an associated rate of confidence. The complete
description of the proteome (expression pattern of all proteins and their interactions)
is thought to provide the mechanistic basis for much of the physiology.
The exponential growth in the volume of accessible biological information has
generated a plurality of voices surrounding the annotation of genes and their
products. The Gene Ontology project (see Table 23.1) seeks to provide a unified
set of structured vocabularies to describe gene products in any organism. This
work includes building three extensive ontologies to describe molecular function,
biological process, and cellular component, and providing a community database
resource that supports the use of these ontologies. The annotated genome sequence
A New Era for Drosophila 335
TABLE 23.1
Drosophilas Web Resources
Web Site Brief Description What to Search?

FLYBASE A database of the Drosophila genome. The Genes, mutants, stocks,
http://www.flybase.org core Internet resource for Drosophila researchers,
researchers. publications, etc.
BDGP (BERKELEY Consortium of the Drosophila Genome Drosophila sequence, new
DROSOPHILA GENOME Center to generate and maintain biological releases, expression
PROJECT) annotations of its genome sequence. patterns, cDNA and EST,
http://www.fruitfly.org 1. Producing gene disruptions using natural transposable
P element-mediated mutagenesis. elements, gene
2. Characterizing the sequence and disruption, comparative
expression of cDNAs. genomics, methods, etc.
3. Developing informatics tools that
support the experimental process, identify
features of DNA sequence, and allow
them to present up-to-date information
about the annotated sequence to the
research community.
The Heidelberg FlyArray A database of the Drosophila genome. A Genes, expression
http://www.hdflyarray.zmbh. comparison between BDGP, Heidelberg patterns, phylogenetics,
uni-heidelberg.de gene prediction, and the in situ P-element insertions, Fly
hybridisation patterns for Heidelberg Arrays, etc.
unique genes.
Drosophila Genomics 1. Collecting and distributing DNA clones, Cell lines, clones, vectors,
Resource Center 2. collecting and distributing cell lines, microarrays, etc.
http://www.dgrc.cgb.indiana. 3. manufacturing and distributing
edu microarrays, 4. assisting scientists to use
these materials, 5. developing and testing
emerging genomics technologies.
DroSpeGe Drosophila species genomes. A service of Comparison between
http://www.species.flybase. FlyBase and Genome Sequencing Drosophila species
net/ Centers. This service provides a preview genomes.
of Drosophila genome data, with genome
maps and BLAST sequence search, for
several Drosophila species.
Gene Ontology Collaborative effort to address the need for Biological process and
http://www.geneontology.org consistent descriptions of gene products cellular compartment of
in different databases. Consortium to your protein of interest.
produce a controlled vocabulary that can
be applied to all organisms even as
knowledge of gene and protein roles in
cells is accumulating and changing. There
are three organizing principles of GO:
molecular function, biological process
and cellular component.
(Continued)
TABLE 23.1
Drosophilas Web Resources (Continued)
Bloomington The main Drosophila stock center: with a Mutant Drosophila
http://www.flystocks.bio.indi collection of P-element strains, EP strains, strains.
ana.edu EMS, natural produced mutants, etc.
Szeged Drosophila Stock EPs Drosophila stock collection. EPs strains.

Center Drosophila strains in order to overexpress
http://www.expbio.bio. genes.
u-szeged.hu/fly
Tucson Stock Center The Tucson Drosophila stock center Drosophila species.
http://www.stockcenter.arl.ar provides subcultures of approximately
izona.edu/ 270 different Drosophila species.
DrosDel Drosophila deletion stock center. Drosophila deletion
http://www.drosdel.org.uk/ Construction of a new Drosophila collection
deletion collection using RS
(re-arrangement screen) P-elements and a
European Drosophila network.
Flytrap HTML-based gene expression database. A database of P(Gal4)
http://www.fly-trap.org/ strains and its expression
within the adult brain.
GETDB Drosophila stock center. GETDB is an Gal4 enhancer-trap strains

http://www.flymap.lab.nig.ac integrated database compiling insertion and its expression
.jp/%7Edclust/getdb.html site, expression pattern and mutant pattern.
phenotype of a large collection (6966) of
Gal4 enhancer-trap lines generated and
analyzed by NP Consortium.
Exelixis Drosophila Stock Drosophila stock center. New collection Drosophila P-element
Collection stocks generated by Exelixis and strains.
http://www.Drosophila.med. distributed to Bloomington and Harvard.
harvard.edu
GENEXEL Drosophila stock center. GeniSys database EP strains (charge).
http://www.genexel.com/eng of Drosophila EP lines.
/htm/ genisys.htm
P-SCREEN Database Another P-element Drosophila database Drosophila P-element
http://www.flypush.imgen.bc from the BDGP gene disruption project. strains.
m.tmc.edu/pscreen
Flytrap GFP Protein Trap database. Flytrap acts as Micrographs, DNA
http://www.flytrap.med.yale. a central data repository for Drosophila sequence and mapping
edu transgenic lines being created using an information from the
intron protein-trap strategy. inserted P-element, and
expression patterns of
the tagged proteins.
FlyMine An integrated database for Drosophila and Gene comparisons

http://www.flymine.org Anopheles genomics. The FlyMine between
project is building an integrated database D. melanogaster, A.
of genomic, expression and protein data gambiae and C. elegans.
for Drosophila and Anopheles and making
this available to the worldwide research
community.
FlyRNAi Drosophila RNAi screening center. A A facility center to do
http://www.flyrnai.org library of double-stranded RNAs directed genome-wide RNAi
against all predicted open reading frames. screens and the creation
This resource can now be used to conduct of a database of
high-throughput cell-based RNAi screens information.
to identify genes involved in various
assays.
Flybrain Drosophilas brain. An online atlas and Atlas of the Drosophila
http://www.flybrain. database of the Drosophila nervous Brain, Genetic
uni-freiburg.de system. Dissection of the
Brain, Developmental
Studies, The Dissectable
Brain, 3D Models of the
Brain.
FlyView Drosophila Image database. Image Compare images on the
http://www.flyview. database on Drosophila development and computer screen and
uni-muenster.de genetics, especially on expression patterns search for special
of genes (enhancer-trap lines, cloned patterns at different
genes). developmental stages.
FlyMove Development image database. Internet Images and videos of

http://www.flymove. resource to study Drosophila development. Drosophila development.
uni-muenster.de
Interactive fly Drosophila development. A cyberspace Information about gene
http://www.sdbonline.org/fly guide to Drosophila genes and their roles role during development.
/aimain/1aahome.htm in development. Biochemical pathways,
organs, images and brain
genes.
CURAGEN Corporation The worlds first comprehensive protein Interactions between

http://www.portal.curagen.com/ interaction map for a multicellular Drosophila proteins.
cgibin/interaction/ organism. Description of the Drosophila
flyHome.pl two-hybrid interactome pathways.
Homophila Human disease to Drosophila gene Drosophila homologs of
http://www.superfly.ucsd. database. Homophila utilizes the sequence human diseases genes.
edu/homophila information of human disease genes in
order to determine if sequence homologs
of these genes exist in the current
Drosophila sequence database.
of Drosophila melanogaster and of Drosophila pseudoobscura, together with the

associated biology, provides the foundation for a new era of sophisticated in silico
and in vivo functional studies.
23.1 VISUALIZING NEURONAL CIRCUITS THANKS

TO GENETICS
A powerful mean of revealing neuronal architecture in the Drosophila brain (Figure
23.1) is the GAL4/UAS enhancer-trap technique2628 that enables a selective activation
of any cloned gene in a wide variety of tissues and cells (Figure 23.2). An enhancer-
trap element is a transposon containing an exogenous gene, such as the yeast tran-
scriptional activator GAL4. Insertions that occur close to a transcriptional enhancer
cause the GAL4 gene to be expressed in a pattern reflecting the enhancers spa-
tiotemporal regulatory properties. Thousands of lines carrying an insertion of the
P[GAL4] element are available. With a single cross, flies are created that carry a
second P-element with a reporter gene inserted downstream of GAL4 binding sites:
the upstream activating sequences (UASs). The reporter gene is expressed in the same
cells as GAL4. This flexible system allows, for example, one to label a particular
group of cells if the reporter gene encodes the green fluorescent protein (GFP) (Figure
23.2). Using this approach, Yang and colleagues29 characterized intrinsic cells of the
Drosophila memory center as the mushroom bodies (MBs) (Figure 23.1). Rather than
KC KC
Ca Ca
LH
LH
AGT AGT

AL AL
FIGURE 23.1 (SEE COLOR INSERT FOLLOWING PAGE 236) Drosophila olfactory
memory system. Olfactory sensory neurons project to the antennal lobes (ALs). From there,
projection neurons project through the antenno-glomerular tract (AGT) and connect mushroom
body (MB) dendrites localized in the calyx (Ca), as well as the lateral horn (LH). Each MB
is composed of about 2,500 neurons, the Kenyon cells (KC). Three types of KC project in
five lobes: /, / and .
Mushroom body + Enhancer-trap transposable element Mushroom body-expressed

enhancer 5IR 3IR gene
GAL4
GAL4 w+ ampRR
amp
AAAAA
GAL4-mRNA
GFP expressed
within MB
AAAAA
GFP-mRNA
GAL4
+
UAS
UAS
UAS
UAS
UAS
GFP
5IR 3IR
FIGURE 23.2 (SEE COLOR INSERT) The GAL4/UAS system. When an enhancer-trap
transposable element is inserted near transcription enhancers that control expression in a given
structure, the GAL4 gene that is contained in the P-element is expressed in the same structure.
If this fly contains also a reporter gene downstream of the UAS sequences, GAL4 will
transcribe the reporter.
seeing homogenous neurons, they found the 5,000 MB neurons to be compound

neuropils in which parallel subcomponents exhibit discrete patterns of gene expres-
sion.30,31 Parallel channels of information flow, perhaps with different computational
properties, subserve different roles (see below).
Using the enhancer-trap system, Ito and colleagues32 searched for MBs extrinsic
neurons. They used a reporter construct encoding the presynaptic protein neuronal
synaptobrevin-green fluorescent protein. They showed that output MB neurons are
scarce, and that, surprisingly, few MB extrinsic neurons project to the deutocerebrum,
the premotor pathway immediate modifier of behavior.
23.2 IMAGING BRAIN ACTIVITY

One major caveat of Drosophila central brain studies is that direct electrophysiological
analysis is scarce,33 due mainly to the small size of neuron cell bodies (less than 5
m in diameter). To circumvent this difficulty, two partially alternative approaches
have been followed: analysis of learning memory mutants at the neuromuscular
junction,34,35 which leaves out the complexity of brain physiology, as well as the
analysis of isolated MB neurons in culture.36 Those experimental systems can provide
interesting molecular and cellular information, but they are inadequate for assessing
neuronal function at the level necessary for a global understanding of memory systems.
Odor processing occurs in a complex-tissue environment, and identification of
the repertoire of brain cell assemblies involved in olfactory memory requires
visualization of the network activity at high spatial and temporal resolution, in
preparations as intact as possible. Optical neural activity recordings allow study of

active brains with micrometer-spatial resolution, and activity-sensitive fluorescent
probes have been recently used in Drosophila. Interestingly, those sensors are
proteins and therefore their expression can be driven to specific subsets of neurons
with the GAL4/UAS system. Several sensors have been successfully brought to
Drosophila that monitor the local change of pH that accompanies neurotransmitter
release,37 or changes in the intracellular calcium concentration that provide a valuable
indicator of electrical activity (Aequorin;38 Cameleon;39 Camgaroo;40 G-CaMP41,42).
Using the G-CaMP reporter and two-photon microscopy, stereotyped odor-evoked
patterns have been observed in the antennal lobe glomeruli41 and in the MBs.42
The ultimate goal of imaging studies is to build a functional map of cell assem-
blies encoding memory in different regions of Drosophila brain, by comparing the
activity of trained and nave animals, in normal flies or memory mutants. A first
step achieved recently as a transient change in the spatial code was observed in the
antennal lobe of wild-type flies 3 minutes after olfactory associative conditioning.37
The field of Drosophila brain imaging is only emerging, and the extensive use of
various activity-sensors and diverse optical setups should provide us with major
information within the next few years.
23.3 LEARNING AND MEMORY MUTANTS

We do not intend here to review in detail all Drosophila learning and memory genes
or the various conditioning protocols.4346 Rather, we would like to show how recent
genetic technologies greatly facilitate the identification and study of learning and
memory genes. The first Drosophila memory mutants were issued from random
chemical mutagenesis using ethyl methane sulfonate (EMS), a potent mutagen. But
because EMS induces mostly subtle DNA alterations, the identification of the
mutated gene is often difficult. The first step consists of genetically mapping the
mutation by complementation test, using Drosophilas deletion collection. However,
this only defines a broad area covering many other genes. On the contrary, the use
of a marked transposable P-element allows one to easily identify the gene of interest.
In the first behavioral screen, about 4,000 fly stocks carrying EMS-induced
mutations were tested for their ability to learn in an olfactory conditioning assay.47,48
The first Drosophila learning and/or memory mutants were dunce (dnc) and rutabaga
(rut). Biochemical defects were observed in rut (reduced cAMP level) and dnc
(increased cAMP level).49 Nevertheless, the cloning of the responsible gene took
years and was facilitated by the use of new P-induced alleles. For example, descrip-
tion of P-element insertions within 200 nucleotides of where the rut transcription
started, identified rut as the structural gene for the Ca2+/Calmodulin-responsive
adenylate cyclase.50,51 New EMS-induced dnc mutants came from a screen for female
sterility mutants.52 The dnc gene was identified by recombinational mapping of dnc
mutations with restriction site polymorphisms as genetic markers.53
The first amnesiac (amn) mutant was identified in a screen for flies with affected
memory.48 However, the amn gene was cloned as a second site suppressor of the dnc
female sterility phenotype from a P-element-induced allele, and has been repeatedly
isolated since. 5456 The amn encodes a putative adenylate-cyclase activating peptide.
Another learning and/or memory mutant is radish (rsh).57 The rsh gene was
localized within a 180-kb interval in the 11D-E region of the X chromosome, and
several candidate genes were identified.57 Recently, Chiang and colleagues58 reported
that the responsible gene for rsh phenotype was a phospholipase A2. However a
second team has reported that rsh encodes a novel protein with possible nuclear
localization motifs.46,59 This discrepancy illustrates the difficulty linked to working
with EMS-induced behavioral mutants.
P-elementbased behavioral screens for learning and memory mutants have also
been performed. Various mutants were issued from these mutagenesis, including
nalyot (a myb-related Adf1 transcription factor),60,61 leonardo (a zeta isoform of the
14-3-3 protein),62 and volado (two splice variants of an -integrin).63
Dubnau and colleagues24 recently performed a behavioral screen for long-term
memory mutants, in parallel to microarray experiments aimed to select genes with
altered expression after long-term memory training. This work led to the identifica-
tion of proteins involved in mRNA processing and translation.24
We recently described crammer (cer), a gene involved specifically in the setting
up of long-term memory.64 The P[GAL4] cer strain has a reduced long-term memory
but a normal short-term and middle-term memory. Interestingly, in the wild-type
strain, cer is transiently underexpressed three hours after long-term memory training.
As the Cer peptide is an inhibitor of cysteine proteinases, the decrease in its expres-
sion shortly after intensive training must lead to a transient activation of its cysteine
proteinase(s) target(s).64 Altogether these works demonstrate the power of the for-
ward-genetic approach: learning or memory mutants are first characterized without
knowing in advance the molecular function of the gene involved.
Last, some genes have been shown to be involved in Drosophila learning or
memory due to function similarity to well-described genes in other species. For
example, NF1, a GTPase-activating protein for Ras, is linked to the human disease
neurofibromatosis that sometimes leads to learning disabilities.65 Drosophila NF1
mutants show a defect in olfactory learning.66 Notch, a critical component of an
evolutionarily conserved signaling mechanism that regulates neural development, is
involved in hippocampal synaptic plasticity and in learning and memory in mice.67,68
It was recently shown that Notch signaling is required for long-term memory for-
mation in adult Drosophila.69,70
23.4 TEMPORAL CONTROL OF GENE EXPRESSION

When a gene is mutated, the biochemical function of the protein may be affected,
or the transcription level of the normal mRNA may be decreased. The resulting
biochemical defect may occur during development and/or adult life. Thus, a learning
or memory defect or both may be due to structural brain defects rather than to a
specific physiological alteration. This issue is crucial as even a faint developmental
defect may have major consequence on brain function.
Two recent techniques, developed by the laboratory of Ronald Davis, enable one
to discriminate a role in brain development from a role in memory formation per
se. TARGET and Gene-Switch71 are variants of the invaluable GAL4/UAS expres-
sion system. TARGET, which stands for Temporal and Regional Gene Expression
Targeting, uses a modified temperature-sensitive yeast, GAL80 transcription factor

(GAL80ts), to repress GAL4 activity (Figure 23.3). At low temperature (typically
18C), the functional GAL80ts protein inhibits GAL4 transcriptional activity. At
30C, GAL80ts becomes inactive and allows the GAL4-mediated transcription (Fig-
ure 23.3). This tool can be used to rescue the defect of a memory mutant by
expressing the normal protein in specific regions of the adult brain, while preventing
its expression during development. Conversely, if the UAS downstream gene
expresses a double-stranded RNA (RNAi), one will be able to lower the expression
of a given gene in specific adult circuits (Figure 23.3). Thus, the TARGET technique
allows one to increase or decrease the expression of a precise gene in a temporal
and regional manner.
Gene-Switch uses a different mechanism to temporally control GAL4 activity.72
The GAL4 DNA-binding domain was fused to a mutated progesterone receptor
ligand-binding domain and part of NF-B p65 activation domain. This synthetic
transcription factor is active only when the antiprogestin mifeprestone (RU486) is
present. When flies are fed RU486, Gene-Switch is on and the UAS-transgene
is expressed. As with the GAL4/UAS system, region-specific Gene-Switch activity
is accomplished by the use of a specific enhancer.73,74
Finally, we recently developed a new GAL4 tool using an inducible chimerical
protein that carries the suppressor-of-hairy-wing repressor domain fused to the
Development Adulthood Phenotype

A
GAL4, UAS-RNAi
MB-specific MB-specific Memory defects possibly

GAL 4 RNAi GAL 4 RNAi due to structural defects
RNAi RNAi
+ + (RNAi activity
during development)
UAS
UAS
UAS
UAS
UAS
UAS
UAS
UAS
UAS
UAS
RNAi RNAi
RNAi RNAi
GAL 80ts
B Low temperature High temperature
GAL4, UAS-RNAi
MB-specific MB-specific Specific Memory defects

GAL 80ts
ts
GAL 4 GAL 80 GAL 4 RNAi
RNAi
(RNAi activity during
+ adult life)
UAS
UAS
UAS
UAS
UAS
UAS
UAS
UAS
UAS
UAS
RNAi RNAi
RNAi RNAi
FIGURE 23.3 The TARGET technique helps in discriminating between a developmental

gene action from an adult physiological effect. A sequence encoding an inverted repeat RNA
(RNAi) is placed under upstream activating sequence (UAS) control. (A) GAL4 activates
RNAi expression during development and adulthood. A memory defect could be caused
indirectly by structural defects occurred during development. (B) The TARGET system solves
this problem. Flies are maintained at low temperature during their developmental stages. At
this temperature, GAL80ts inhibits the transcription activity of GAL4, so no RNAi is expressed.
When adults emerge, they are transferred at permissive temperature (30C); the GAL80ts
protein becomes ineffective and the GAL4 activates the transcription of the RNAi in adult
MBs. The putative memories defects are due to the decreased concentration of the mRNA
targeted by the RNAi.
GAL4 DNA-binding domain. This tool allows also a temporally controlled repres-
sion of genes located near a P[UAS] insertion.75
23.5 A TOOL TO BUILD ANATOMO-FUNCTIONAL MAPS

A sophisticated approach to disturb neuronal circuits, based on a rapid and reversible
blockage of synaptic transmission was developed by Kitamoto.76 The shibire (shi)
gene encodes a microtubule-associated GTPase, Dynamin, which is involved in
endocytosis and is essential for synaptic vesicle recycling and maintenance of the
readily releasable pool of synaptic vesicles.77 The temperature-sensitive allele, shits1,
is defective in vesicle recycling at restrictive temperature (>29C), resulting in a
rapid blockage of synaptic transmission. The shits1 mutation has a dominant effect
because it blocks chemical synapses even in the presence of a normal shi+ allele.
Expression of shits1 tool can therefore be used within the GAL4/UAS system. The
GAL4/UAS-shits1 approach is very powerful as it allows inhibiting particular brain
circuits at a precise time, for example during conditioning or memory retrieval. Thus,
it has been established that abolishing vesicle-mediated secretion from the dorsal-
paired medial (DPM) cells, which normally express amn, phenocopies amn mutant
memory defect.78
The tools described so far allow a temporal control of gene expression and
neuronal activity. Their recent development has opened the way for a dynamic
analysis of memory systems.
23.6 LOCALIZATION OF OLFACTORY MEMORY

IN DROSOPHILA MBS
As mentioned, the insect brain contains a pair of prominent and characteristically
shaped neural centers known as MBs. In Drosophila, MBs are composed of three
main classes of neurons whose axons divide to form two vertical lobes ( and
) and three median ones (, and )30 (Figure 23.1). MBs are a specialized
neuropile involved in processing and storing multimodal sensory information.79
In the 1970s and 1980s, the function of the MBs was assessed in different insect
species with the classical interventionism approachescooling and ablation.80,81
Over the years, MBs have been implicated in olfactory learning and memory, and
in a variety of complex functions including courtship, motor control, and spatial
recognition.
In Drosophila, a single associative-learning trial (named below the short
protocol), consisting of an odor (the conditional stimulus, or CS) accompanied by
electric shocks (the unconditional stimulus, or US), induces olfactory learning and
memory. Brain mutants with MBs structural defects were isolated in the 1980s.
The use of an unlimited number of mutant animals with the same anatomical defect
without interventionism helped in highlighting the role of MBs in olfactory learning
and memory.82 However, one caveat of this approach is that the anatomical defects
are not specific to the MBs.83 An alternative approach was used to generate
flies without MBs.83 Hydroxyurea, an antimitotic agent, was fed to newly hatched
wild-type larvae. At that early developmental stage, only five neuroblaststhe

neurons precursor cellsare mitotically active within each brain hemisphere: the
four that generate the MBs, and one in the antennal lobe. Thus, hydroxyurea
treatment leads to viable adult flies with almost no MBs. These flies showed no
olfactory memory.83 The essential role of MBs in olfactory memory is further
outlined by the the high expression in the MBs of many of the proteins involved
in learning and memory such as Dnc,84 Rut51 or DCO.85 Even though Amn48 is not
expressed in the MBs, it is expressed by the dorsal-paired median neurons that
project onto the MBs.78
Thus, Drosophila MBs are required for olfactory learning and memory, and
insights from the first mutants have indicated that the cAMP pathway plays a key
role for memory establishment. Nevertheless, it remained to be proven that cAMP
metabolism was strictly required in MBs for correct learning and memory. Thanks
to the GAL4/UAS system,26 Connolly and colleagues86 disrupted MBs cAMP sig-
naling by expressing a constitutively activated G-protein. Permanent adenylyl cyclase
activation led to an impaired associative memory.
That memory is impaired after ablation or functional disruption of a brain
structure does not necessarily imply that the memory trace itself is localized within
this structure. To localize short-term memory (STM), it was necessary to rescue the
memory abilities of a mutant by expressing the corresponding protein in a specific
brain structure. In a first step, Zars and colleagues87 expressed Rut Ca2+/Calmodulin
adenylyl cyclase in MBs to restore a normal learning capacity. However, in this
experiment, Rut was expressed in the MBs at the adult stage and also during
development. Thus, the behavioral rescue might have been due to the correction of
an MB developmental defect, indirect cause of the learning defect. Thanks to the
TARGET and SWITCH methods (see above), McGuire and colleagues88 showed
that the presence of Rut in adult MBs alone was sufficient to rescue rut memory
defect. The current view is that Rut adenylyl cyclase could act during learning as a
coincidence detector for the US and CS.
Are the MBs required during the acquisition, consolidation, or retrieval phase?
It was possible to address this question by expressing the temperature sensitive Shits1
protein specifically in MBs neurons to transiently disrupt synaptic neurotransmis-
sion.89,90 It was shown that the synaptic outputs of MB neurons are required during
retrieval of the STM but not during acquisition or consolidation. All together, these
data indicate that a STM trace can be localized in MBs.
In addition to STM, Drosophila can display long-term memory (LTM) after a
spaced and repeated conditioning (long protocol).91 LTM can last several days and
depends on de novo protein synthesis. Are MBs implicated in Drosophila LTM?
To answer this question we analyzed the alpha-lobes absent (ala) mutant that shows
a peculiar MB phenotype. Ten percent of ala individuals possess all five MBs lobes,
36% lack the horizontal and lobes, and 4.5% lack vertical and lobes (the
remaining sub-populations presented different MB phenotypes in the left and the
right hemispheres).92 ala mutant flies were trained with the short or the long protocol,
and we analyzed separately the brains of flies that had made the correct and
the wrong choices during the memory test, to calculate the memory score of
each class of ala mutants. Flies lacking / lobes displayed no LTM although their
STMs were wild-type. Flies with all lobes present or without / lobes had a normal
STM and LTM. Thus, MBs are necessary to perform LTM, and more particularly
the / vertical lobes.92 By expressing Shits1 in / lobes, it was further shown that
lobes outputs are required during LTM retrieval.93
If MBs play a pivotal role in Drosophila olfactory learning and memory, a recent
study suggests that brain structures located outside the MBs also participate in
olfactory memory.94 We have recently shown that the Drosophila brain is asymmet-
ric, as a small structure expressing Fasciclin IIthe asymmetric bodyis present
only in the right hemisphere. Interestingly, about 7% of Canton-Special wild-type
flies present a bilateral structure. Those symmetric flies showed no LTM at 4 days,
suggesting that asymmetry may be required for generating, maintaining, or retrieving
LTM.94 The asymmetric body is located near the central complex, a structure that
connects brain hemispheres. However, the functional links between the asymmetric
body and other brain circuits remain to be sorted out.
23.7 DYNAMIC OF OLFACTORY MEMORY PHASES

IN DROSOPHILA
The short protocol induces two labile phases: STM, which is disrupted in mutants
affected for cAMP metabolism and lasts about 30 min; and middle-term memory
(MTM), which is disrupted in amn and lasts for a few hours. STM and MTM are
anesthesia sensitive as they are erased if flies are cooled down to 4C after condi-
tioning. This property suggests that STM and MTM are sustained by electrical brain
activities.
Drosophila also displays two long-lasting memory phases, anesthesia-resistant
memory (ARM)95 and LTM.96 ARM is generated by the short protocol or several
massed trials. ARM can still be detected after a few days.91 This memory is disrupted
in rsh mutants.57 The molecular pathways involved in ARM formation remain largely
unknown, in part because rsh gene identity is debated (see above). The atypical
protein kinase M (aPKM) is a persistently active truncated isoform of atypical protein
kinase C (aPKC). Over-expression of aPKM enhances memory after massed condi-
tioning, but not after spaced training.97 Moreover, inhibition of aPKM disrupts
consolidated memory after massed conditioning.97 It is therefore conceivable that
aPKC is a molecular support of ARM.
LTM is induced by the long protocol and measured for at least one week.91 Like
ARM, LTM is anesthesia resistant.96 It is disrupted by a cAMP response binding
protein (dCREB2-b) repressor,98,99 likely due to the inhibition of gene expression
required to establish LTM. Yin and colleagues100,101 reported that flies over-expressing
a particular CREB isoform (dCREB2-a) generated LTM after a single training cycle.
However, this work could not be replicated, and it was shown that the original
dCREB2-a transgenic flies carried an accidental mutation that produced a truncated
protein with no DNA binding domain.99 Moreover, adult induction of the correct
CREB2-a isoform led to lethality.99
What is the dynamic of memory phases in Drosophila? We proposed recently
a model that involves two parallel memory pathways, one with cAMP dependent
STM/MTM, and the other with ARM (Figure 23.4A). Indeed, dnc and rut retain a
significant level of early memory,102 suggesting that an adenyl cyclase-Rut indepen-
dent learning might exist. Moreover, ARM levels in rut and amn are near normal,57,93
while their labile memories are strongly affected. Thus ARM does not seem to
depend on STM/MTM. Instead a second learning process could give rise to STM2
phases and later to ARM (Figure 23.4).
What are the relationships between ARM and LTM? To answer that question,
the ala mutant was trained with the long protocol, and the memory of flies lacking
vertical / lobes was measured at 30 minutes and 5 hours after the training. Thirty-
minute memory was normal, but, surprisingly, 5-hour memory was close to zero.
Memory performance was normal at 5 hours when flies without vertical lobes were
trained with the short protocol93 (Figure 23.5). Why does a longer training give rise
to a weaker memory? ala flies display no LTM because they lack the vertical lobes,
the center for LTM. These flies show a normal ARM at 5 hours after the short
protocol, but they show no ARM after the long protocol. This result suggests that
ARM is erased after LTM conditioning. Thus the consolidated memory phases
generated by olfactory conditioning are exclusive (Figure 23.6).93 Why is ARM
erased after LTM conditioning? We propose that ARM could act as a gating mech-
anism for LTM formation, avoiding a heavy cascade of gene expression in absence
of intensive, spaced conditioning. Despite the relative simplicity of the Drosophila
dunce
A rutabaga amnesiac
DCO
LRN 1 STM 1 MTM
c A MP-dependent way
? ? radish
LRN 2 STM 2 ARM
B
STM 1
STM 2
MTM
ARM
Memory level
1h 2h 3h 4h 5h 6h
FIGURE 23.4 (A) Model of associative memory phases and (B) temporal dynamics of
memory phases generated by a single cycle of conditioning (short protocol). LRN: learning;
STM: short-term memory; MTM: middle-term memory; ARM: anesthesia-resistant memory.
60
50
1X
40
PI
30
20 5X
10
0
30 min 3h 5h
FIGURE 23.5 In flies without MB alpha lobes that normally sustain long-term memory, the
long protocol decreases memory performance at 5 hours in comparison with the short protocol.
Grey line: short protocol; black line: long protocol.
A dunce
rutabaga
amnesiac
DCO
LRN 1 STM 1 MTM LTM
cAMP-dependent way
? ? radish
LRN 2 STM 2 ARM
B
B
STM
MTM
LTM
level
Memory level
Memory
1h 2h 3h 4h 5h 6h
FIGURE 23.6 (A) Model of associative memory phases and (B) temporal dynamics of
memory phases generated by five spaced cycles of conditioning (long protocol). LRN: learn-
ing; STM: short-term memory; MTM: middle-term memory; ARM: anesthesia-resistant
memory; LTM: long-term memory.
brain, this model suggests a cognitive complexity more frequently associated with
mammalian models. It supports the idea that Drosophila is a valid model in which
to study some of the molecular and cellular mechanisms involved in normal or
pathological human memory.3
REFERENCES
1. Shimada, T. et al. Analysis of the distribution of the brain cells of the fruit fly by an
automatic cell counting algorithm, Physica A 350, 144, 2005.
2. Rubin, G.M. et al. Comparative genomics of the eukaryotes, Science 287 (5461),
2204, 2000.
3. Shulman, J.M. et al. From fruit fly to bedside: translating lessons from Drosophila
models of neurodegenerative disease, Curr Opin Neurol 16 (4), 443, 2003.
4. Bier, E. Drosophila, the golden bug, emerges as a tool for human genetics, Nat Rev
Genet 6 (1), 9, 2005.
5. Parks, A.L. et al. Systematic generation of high-resolution deletion coverage of the
Drosophila melanogaster genome, Nat Genet 36 (3), 288, 2004.
6. Thibault, S.T. et al. A complementary transposon tool kit for Drosophila melanogaster
using P and piggyBac, Nat Genet 36 (3), 283, 2004.
7. Rong, Y.S. and Golic, K.G. Gene targeting by homologous recombination in Droso-
phila, Science 288 (5473), 2013, 2000.
8. Rong, Y.S. et al. Targeted mutagenesis by homologous recombination in D. melan-
ogaster, Genes Dev 16 (12), 1568, 2002.
9. Rong, Y.S. and Golic, K.G. A targeted gene knockout in Drosophila, Genetics 157
(3), 1307, 2001.
10. Piccin, A. et al. Efficient and heritable functional knock-out of an adult phenotype
in Drosophila using a GAL4-driven hairpin RNA incorporating a heterologous spacer,
Nucleic Acids Res 29 (12), E55, 2001.
11. Adams, M.D. et al. The genome sequence of Drosophila melanogaster, Science 287
(5461), 2185, 2000.
12. Misra, S. et al. Annotation of the Drosophila melanogaster euchromatic genome: a
systematic review, Genome Biol 3 (12), R1, 2002.
13. Hild, M. et al. An integrated gene annotation and transcriptional profiling approach
towards the full gene content of the Drosophila genome, Genome Biol 5 (1), R3, 2003.
14. Stolc, V. et al. A gene expression map for the euchromatic genome of Drosophila
melanogaster, Science 306 (5696), 655, 2004.
15. Venter, J.C. et al. The sequence of the human genome, Science 291 (5507), 1304,
2001.
16. Consortium, I.H.G.S. Finishing the euchromatic sequence of the human genome,
Nature 431 (7011), 931, 2004.
17. Callahan, C.A. et al. Control of neuronal pathway selection by a Drosophila receptor,
Nature 376 (6536), 171, 1995.
18. Hoskins, R.A. et al. Heterochromatic sequences in a Drosophila whole-genome
shotgun assembly, Genome Biol 3 (12), R1, 2002.
19. Claridge-Chang, A. et al. Circadian regulation of gene expression systems in the
Drosophila head, Neuron 32 (4), 657, 2001.
20. McDonald, M.J. and Rosbash, M. Microarray analysis and organization of circadian
gene expression in Drosophila, Cell 107 (5), 567, 2001.
21. De Gregorio, E. et al. Genome-wide analysis of the Drosophila immune response by
using oligonucleotide microarrays, Proc Natl Acad Sci USA 98 (22), 12590, 2001.
22. Girardot, F., Monnier, V. and Tricoire, H. Genome wide analysis of common and
specific stress responses in adult Drosophila melanogaster, BMC Genomics 5 (1),
74, 2004.
23. Jin, W. et al. The contributions of sex, genotype and age to transcriptional variance
in Drosophila melanogaster, Nat Genet 29 (4), 389, 2001.
24. Dubnau, J. et al. The staufen/pumilio pathway is involved in Drosophila long-term

memory, Curr Biol 13 (4), 286, 2003.
25. Giot, L. et al. A protein interaction map of Drosophila melanogaster, Science 302
(5651), 1727, 2003.
26. Brand, A.H. and Perrimon, N. Targeted gene expression as a means of altering cell
fates and generating dominant phenotypes, Development 118 (2), 401, 1993.
27. Wilson, C. et al. P-element-mediated enhancer detection: an efficient method for
isolating and characterizing developmentally regulated genes in Drosophila, Genes
Dev 3 (9), 1301, 1989.
28. Bellen, H.J. et al. P-element-mediated enhancer detection: a versatile method to study
development in Drosophila, Genes Dev 3 (9), 1288, 1989.
29. Yang, M.Y. et al. Subdivision of the Drosophila mushroom bodies by enhancer-trap
expression patterns, Neuron 15 (1), 45, 1995.
30. Crittenden, J.R. et al. Tripartite mushroom body architecture revealed by antigenic
markers, Learn Mem 5 (12), 38, 1998.
31. Strausfeld, N.J., Sinakevitch, I. and Vilinsky, I. The mushroom bodies of Drosophila
melanogaster: an immunocytological and golgi study of Kenyon cell organization in
the calyces and lobes, Microsc Res Tech 62 (2), 151, 2003.
32. Ito, K. et al. The organization of extrinsic neurons and their implications in the
functional roles of the mushroom bodies in Drosophila melanogaster Meigen, Learn
Mem 5 (12), 52, 1998.
33. Wilson, R.I., Turner, G.C. and Laurent, G. Transformation of olfactory representations
in the Drosophila antennal lobe, Science 303 (5656), 366, 2004.
34. Zhong, Y. and Wu, C.F. Altered synaptic plasticity in Drosophila memory mutants
with a defective cyclic AMP cascade, Science 251 (4990), 198, 1991.
35. Renger, J.J. et al. Role of cAMP cascade in synaptic stability and plasticity: ultra-
structural and physiological analyses of individual synaptic boutons in Drosophila
memory mutants, J Neurosci 20 (11), 3980, 2000.
36. Wright, N.J. and Zhong, Y. Characterization of K+ currents and the cAMP-dependent
modulation in cultured Drosophila mushroom body neurons identified by lacZ expres-
sion, J Neurosci 15 (2), 1025, 1995.
37. Yu, D., Ponomarev, A. and Davis, R.L. Altered representation of the spatial code for
odors after olfactory classical conditioning; memory trace formation by synaptic
recruitment, Neuron 42 (3), 437, 2004.
38. Rosay, P. et al. Synchronized neural activity in the Drosophila memory centers and
its modulation by amnesiac, Neuron 30 (3), 759, 2001.
39. Fiala, A. et al. Genetically expressed cameleon in Drosophila melanogaster is used
to visualize olfactory information in projection neurons, Curr Biol 12 (21), 1877,
2002.
40. Yu, D. et al. Detection of calcium transients in Drosophila mushroom body neurons
with camgaroo reporters, J Neurosci 23 (1), 64, 2003.
41. Wang, J.W. et al. Two-photon calcium imaging reveals an odor-evoked map of activity
in the fly brain, Cell 112 (2), 271, 2003.
42. Wang, Y. et al. Stereotyped odor-evoked activity in the mushroom body of Drosophila
revealed by green fluorescent protein-based Ca2+ imaging, J Neurosci 24 (29), 6507,
2004.
43. Waddell, S. and Quinn, W.G. Flies, genes, and learning, Annu Rev Neurosci 24, 1283,
2001.
44. Dubnau, J., Chiang, A.S. and Tully, T. Neural substrates of memory: from synapse
to system, J Neurobiol 54 (1), 238, 2003.
45. Heisenberg, M. Mushroom body memoir: from maps to models, Nat Rev Neurosci 4
(4), 266, 2003.
46. Davis, R.L. Olfactory learning, Neuron 44 (1), 31, 2004.
47. Quinn, W.G., Harris, W.A. and Benzer, S. Conditioned behavior in Drosophila mel-
anogaster, Proc Natl Acad Sci USA 71 (3), 708, 1974.
48. Quinn, W.G., Sziber, P.P. and Booker, R. The Drosophila memory mutant amnesiac,
Nature 277 (5693), 212, 1979.
49. Livingstone, M.S., Sziber, P.P. and Quinn, W.G. Loss of calcium/calmodulin respon-
siveness in adenylate cyclase of rutabaga, a Drosophila learning mutant, Cell 37 (1),
205, 1984.
50. Levin, L.R. et al. The Drosophila learning and memory gene rutabaga encodes a
Ca2+/Calmodulin-responsive adenylyl cyclase, Cell 68 (3), 479, 1992.
51. Han, P.L. et al. Preferential expression of the Drosophila rutabaga gene in mushroom
bodies, neural centers for learning in insects, Neuron 9 (4), 619, 1992.
52. Mohler, J.D. Developmental genetics of the Drosophila egg. I. Identification of 59
sex-linked cistrons with maternal effects on embryonic development, Genetics 85 (2),
259, 1977.
53. Davis, R.L. and Davidson, N. Isolation of the Drosophila melanogaster dunce chro-
mosomal region and recombinational mapping of dunce sequences with restriction
site polymorphisms as genetic markers, Mol Cell Biol 4 (2), 358, 1984.
54. Feany, M.B. and Quinn, W.G. A neuropeptide gene defined by the Drosophila mem-
ory mutant amnesiac, Science 268 (5212), 869, 1995.
55. Moore, M.S. et al. Ethanol intoxication in Drosophila: Genetic and pharmacological
evidence for regulation by the cAMP signaling pathway, Cell 93 (6), 997, 1998.
56. Toba, G. et al. The gene search system. A method for efficient detection and rapid
molecular identification of genes in Drosophila melanogaster, Genetics 151 (2), 725,
1999.
57. Folkers, E., Drain, P. and Quinn, W.G. Radish, a Drosophila mutant deficient in
consolidated memory, Proc Natl Acad Sci USA 90 (17), 8123, 1993.
58. Chiang, A.S. et al. radish encodes a phospholipase-A2 and defines a neural circuit
involved in anesthesia-resistant memory, Curr Biol 14 (4), 263, 2004.
59. Folkers, E., Waddell, S. and Quinn, W.G. Personal communication, 2004.
60. Boynton, S. and Tully, T. latheo, a new gene involved in associative learning and
memory in Drosophila melanogaster, identified from P element mutagenesis, Genet-
ics 131 (3), 655, 1992.
61. DeZazzo, J. et al. Nalyot, a mutation of the Drosophila myb-related Adf1 transcription
factor, disrupts synapse formation and olfactory memory, Neuron 27 (1), 145, 2000.
62. Skoulakis, E.M. and Davis, R.L. Olfactory learning deficits in mutants for leonardo,
a Drosophila gene encoding a 14-3-3 protein, Neuron 17 (5), 931, 1996.
63. Grotewiel, M.S. et al. Integrin-mediated short-term memory in Drosophila, Nature
391 (6666), 455, 1998.
64. Comas, D., Petit, F. and Preat, T. Drosophila long-term memory formation involves
regulation of cathepsin activity, Nature 430 (6998), 460, 2004.
65. Shen, M.H., Harper, P.S. and Upadhyaya, M. Molecular genetics of neurofibromatosis
type 1 (NF1), J Med Genet 33 (1), 2, 1996.
66. Guo, H.F. et al. A neurofibromatosis-1-regulated pathway is required for learning in
Drosophila, Nature 403 (6772), 895, 2000.
67. Costa, R.M., Honjo, T. and Silva, A.J. Learning and memory deficits in Notch mutant
mice, Curr Biol 13 (15), 1348, 2003.
68. Wang, Y. et al. Involvement of Notch signaling in hippocampal synaptic plasticity,

Proc Natl Acad Sci USA 101 (25), 9458, 2004.
69. Presente, A. et al. Notch is required for long-term memory in Drosophila, Proc Natl
Acad Sci USA 101 (6), 1764, 2004.
70. Ge, X. et al. Notch signaling in Drosophila long-term memory formation, Proc Natl
Acad Sci USA 101 (27), 10172, 2004.
71. Leung, B. and Waddell, S. Four-dimensional gene expression control: memories on
the fly, Trends Neurosci 27 (9), 511, 2004.
72. Burcin, M.M. et al. Adenovirus-mediated regulable target gene expression in vivo,
Proc Natl Acad Sci USA 96 (2), 355, 1999.
73. Osterwalder, T. et al. A conditional tissue-specific transgene expression system using
inducible GAL4, Proc Natl Acad Sci USA 98 (22), 12596, 2001.
74. Roman, G. et al. P[Switch], a system for spatial and temporal control of gene
expression in Drosophila melanogaster, Proc Natl Acad Sci USA 98 (22), 12602,
2001.
75. Pascual, A., Huang, K.L. and Preat, T. Conditional UAS-targeted repression in Droso-
phila, Nucleic Acids Res 33 (1), e7, 2005.
76. Kitamoto, T. Conditional modification of behavior in Drosophila by targeted expres-
sion of a temperature-sensitive shibire allele in defined neurons, J Neurobiol 47 (2),
892, 2001.
77. Chen, M.S. et al. Multiple forms of dynamin are encoded by shibire, a Drosophila
gene involved in endocytosis, Nature 351 (6327), 583, 1991.
78. Waddell, S. et al. The amnesiac gene product is expressed in two neurons in the
Drosophila brain that are critical for memory, Cell 103 (5), 805, 2000.
79. Li, Y. and Strausfeld, N.J. Morphology and sensory modality of mushroom body
extrinsic neurons in the brain of the cockroach, Periplaneta americana, J Comp
Neurol 387 (4), 631, 1997.
80. Menzel, R., Erber, J. and Masuhr, T. Learning and memory in the honeybee, in
Experimental Analysis of Insect Behaviour, Barton-Brown, L. Springer, Berlin, Ger-
many, 1974, pp. 195.
81. Erber, J., Masuhr, T. and Menzel, R. Localization of short-term memory in the brain
of the bee Apis mellifera, Physiol Entomol 5, 343, 1980.
82. Heisenberg, M. et al. Drosophila mushroom body mutants are deficient in olfactory
learning, J Neurogenet 2 (1), 1, 1985.
83. de Belle, J.S. and Heisenberg, M. Associative odor learning in Drosophila abolished
by chemical ablation of mushroom bodies, Science 263 (5147), 692, 1994.
84. Nighorn, A., Healy, M.J. and Davis, R.L. The cyclic AMP phosphodiesterase encoded
by the Drosophila dunce gene is concentrated in the mushroom body neuropil, Neuron
6 (3), 455, 1991.
85. Skoulakis, E.M., Kalderon, D. and Davis, R.L. Preferential expression in mushroom
bodies of the catalytic subunit of protein kinase A and its role in learning and memory,
Neuron 11 (2), 197, 1993.
86. Connolly, J.B. et al. Associative learning disrupted by impaired Gs signaling in
Drosophila mushroom bodies, Science 274 (5295), 2104, 1996.
87. Zars, T. et al. Localization of a short-term memory in Drosophila, Science 288 (5466),
672, 2000.
88. McGuire, S.E. et al. Spatiotemporal rescue of memory dysfunction in Drosophila,
Science 302 (5651), 1765, 2003.
89. Dubnau, J. et al. Disruption of neurotransmission in Drosophila mushroom body
blocks retrieval but not acquisition of memory, Nature 411 (6836), 476, 2001.
90. McGuire, S.E., Le, P.T. and Davis, R.L. The role of Drosophila mushroom body
signaling in olfactory memory, Science 293 (5533), 1330, 2001.
91. Tully, T. et al. Genetic dissection of consolidated memory in Drosophila, Cell 79 (1),
35, 1994.
92. Pascual, A. and Preat, T. Localization of long-term memory within the Drosophila
mushroom body, Science 294 (5544), 1115, 2001.
93. Isabel, G., Pascual, A. and Preat, T. Exclusive consolidated memory phases in Droso-
phila, Science 304 (5673), 1024, 2004.
94. Pascual, A. et al. Neuroanatomy: brain asymmetry and long-term memory, Nature
427 (6975), 605, 2004.
95. Quinn, W.G. and Dudai, Y. Memory phases in Drosophila, Nature 262 (5569), 576,
1976.
96. Isabel, G. and Preat, T. Unpublished results, 2005.
97. Drier, E.A. et al. Memory enhancement and formation by atypical PKM activity in
Drosophila melanogaster, Nat Neurosci 5 (4), 316, 2002.
98. Yin, J.C. et al. Induction of a dominant negative CREB transgene specifically blocks
long-term memory in Drosophila, Cell 79 (1), 49, 1994.
99. Perazzona, B. et al. The role of cAMP response element-binding protein in Drosophila
long-term memory, J Neurosci 24 (40), 8823, 2004.
100. Yin, J.C. et al. A Drosophila CREB/CREM homolog encodes multiple isoforms,
including a cyclic AMP-dependent protein kinase-responsive transcriptional activator
and antagonist, Mol Cell Biol 15 (9), 5123, 1995.
101. Yin, J.C. et al. CREB as a memory modulator: induced expression of a dCREB2
activator isoform enhances long-term memory in Drosophila, Cell 81 (1), 107, 1995.
102. Tully, T. and Quinn, W.G. Classical conditioning and retention in normal and mutant
Drosophila melanogaster, J Comp Physiol [A] 157 (2), 263, 1985.
24 Behavioral Genetics in the

Nematode Caenorhabditis
elegans
Gert Jansen and Laurent Sgalat
CONTENTS
24.1 Summary........................................................................................................ 353
24.2 C. elegans as a Model Organism .................................................................. 353
24.2.1 C. elegans Has a Very Simple Nervous System .............................. 353
24.2.2 C. elegans Is Well Suited for Genetic Studies................................. 355
24.3 Behaviors of C. elegans and Mutations Affecting Them............................. 355
24.3.1 Chemosensory Behavior and Its Plasticity....................................... 356
24.3.2 Egg Laying........................................................................................ 359
24.3.3 Regulation of Locomotion in Response to Food ............................. 361
24.3.4 Ethanol Resistance............................................................................ 362
24.3.5 Social Feeding Behavior................................................................... 363
24.3.6 Memory............................................................................................. 364
24.4 Conclusion ..................................................................................................... 367
References.............................................................................................................. 367
Authors Note ........................................................................................................ 369
24.1 SUMMARY
This chapter provides an overview of the behavioral studies that can be undertaken
on an extremely simple organism, the nematode Caenorhabditis elegans. First, we
will briefly describe this animal model, which has become increasingly popular for
molecular and cellular biology studies, and then we will present a few examples of
behavioral studies that are conducted on this organism. We aim to show that this
model provides paradigms for questions central to many behaviors, and that they
can be addressed at a single cell/single gene resolution.
24.2 C. ELEGANS AS A MODEL ORGANISM

24.2.1 C. ELEGANS HAS A VERY SIMPLE NERVOUS SYSTEM
The nematode Caenorhabditis elegans is a free-living round worm present in the
soils of temperate climates (Wood, 1988; Riddle et al., 1997). For laboratory use, it
353
can be grown easily on petri dishes seeded with Escherichia coli (Figure 24.1). Its
length is about 1 mm as an adult and it has 959 somatic cells. Among those, 302
of them are neurons, which can be divided into 116 classes on the basis of their
morphology (Figure 24.2). With a laser beam, it is possible to destroy a given neuron
in an anaesthetized animal to assess the involvement of this neuron in a behavior or
a function. Moreover, thealmost invariantconnectivity of all neurons has been
established by 3D reconstruction of serially sectioned animals.
At the molecular level, the nervous system of C. elegans shares a common
organization with the far-distant vertebrates: its main excitatory transmitter is
FIGURE 24.1 C. elegans on a culture plate. Worms of various stages can be seen, including
eggs, larvae and adults. Adults are approximately 1 mm long.
FIGURE 24.2 The nervous system of C. elegans. Most or all neurons of C. elegans are
visualized by expressing a gpc-2::GFP construct. Weakly, staining of muscle cells can be
seen. Arrows indicate the ventral nerve cord, a triangle indicates staining in the anterior
ganglia, an asterisk indicates the posterior ganglia.
Behavioral Genetics in the Nematode Caenorhabditis elegans 355
glutamate, its main inhibitor is GABA, the neuromuscular synapse is mainly cho-
linergic, and neuronal activity is regulated in part by two major modulators that
display a wide spectrum of effects in the CNS of vertebrates: dopamine and sero-
tonin. A variety of peptidesalbeit less numerous than in mammalsalso act as
neuromodulators.
24.2.2 C. ELEGANS IS WELL SUITED FOR GENETIC STUDIES

Two intrinsic properties make C. elegans an excellent system for genetics: its
rapidity to grow (3 days/generation) and the small size of its genome (100 Mbases).
Animals are generally hermaphrodites with the capability to self-fertilize, an
invaluable advantage when one has to perform large genetic screens. Males can
also be generated at will, for the purpose of crosses. Over the years, hundreds of
genetic markers useful for genetic mapping have been described (Riddle et al.,
1997). Besides the classical genetics approach (from mutant to gene), one can also
use reverse genetics (from gene to mutant), including techniques to generate
knockout worms, RNAi and transgenesis. The cloning of mutations is facilitated
by the fact that the genome of C. elegans is entirely covered by yeast artificial
chromosomes (YACs) and cosmids. Furthermore, in 1998, C. elegans became the
first animal of which the genome was entirely sequenced (The C. elegans Sequenc-
ing Consortium, 1998). The analysis of the sequence revealed an extensive con-
servation of protein sequences between this organism and vertebrates. In 1998, the
RNAi techniquewhich is a fast and efficient way to inactivate a gene of interest
has been introduced in C. elegans (Fire et al., 1998). This technique has now been
extended to most animal models.
In 2002, Sydney Brenner, John Sulston and Bob Horvitz were awarded the Nobel
Prize for Medicine and Physiology for the introduction of C. elegans as a model
organism in biology and for their studies that have elucidated key molecules that
regulate programmed cell death.
24.3 BEHAVIORS OF C. ELEGANS AND MUTATIONS

AFFECTING THEM
Despite its simplicity, more than 20 behaviors can be studied on the worm C. elegans.
For the sake of simplicity, we will focus in this chapter on only six behaviors,
representative of the work that can be done with C. elegans, and we will try to point
out their relevance to other models. The behaviors described below are:
Chemosensory behavior and its plasticity
Egg laying
Regulation of locomotion in response to food
Ethanol resistance
Social feeding behavior
Memory
24.3.1 CHEMOSENSORY BEHAVIOR AND ITS PLASTICITY

The detection of various compounds in our environment is essential for life, it
protects us from toxic compounds and enables us to find food and judge its quality.
These principles also hold true for C. elegans. The nematode responds to many
chemical cues from its environment, including water soluble and volatile attrac-
tants and repellents (reviewed in Mori and Oshima, 1997). Detection of these
compounds is primarily mediated by 11 pairs of amphid chemosensory neurons
in the head. Ablation of these neurons with a laser microbeam has identified the
specific sensory functions for each of these neurons (Figure 24.3). Olfactory cues
are detected by the AWA, AWB, AWC, ADL, and ASH neurons; the nose of the
worm, of which the AWA and AWC cells mediate chemo-attraction. Chemotaxis
to water soluble attractants, such as salts, is mediated by five pairs of amphid
neurons, the ASE, ASG, ASI, ASK, and ADF neurons, of which the ASE neurons
are most important. The ADL and ASH neurons mediate avoidance of water-
soluble repellents.
The development of a robust odorant chemotaxis assay has made it possible to
identify many genes involved in olfactory signaling using forward genetic screens.
More recently, this approach has been complemented by reverse genetic studies.
Together, these studies have given us a very good idea about the basic signaling
mechanisms used in odorant detection in the AWA and AWC neurons. Since
C. elegans has only few olfactory neurons, expression of only one olfactory receptor
per cell (as is the case in mammals) would not be sufficient. Instead, the nematode
expresses many G protein coupled receptors (GPCRs) in each olfactory neuron.
With these receptors C. elegans can discriminate between different odorants detected
by the same cells and adapt to an odorant while remaining responsive to another.
FIGURE 24.3 (SEE COLOR INSERT FOLLOWING PAGE 236) DIC photograph of the
head of a worm. The outline of the pharynx (p) can be seen and the beginning of the gut (g).
The approximate positions of the amphid sensory neuron cell bodies are indicated with colored
circles. The dendrite (d) and an axon (a) of one of the cells are drawn. Below, the names of
the corresponding cells are given, and the compounds perceived by these cells.
The exact molecular mechanisms that make this possible are not completely clear,
but part of the specificity arises from a functional asymmetry between the left and
the right AWC cell (Wes and Bargmann, 2001). In addition, odorant detection seems
modulated by a complex G protein signaling network, involving two to three redun-
dant stimulatory G subunits and an inhibitory G in each neuron (Lans
et al., 2006).
As in mammals, binding of an odorant to a GPCR and activation of heterotri-
meric G proteins is thought to activate downstream effector molecules. In the AWA
cells, G proteins probably activate TRPV channels resulting in a Ca2+ influx. In the
AWC cells, G proteins induce an increase in intracellular cGMP, mediated by
guanylyl cyclases, leading to opening of a cyclic nucleotide-gated channel. These
studies have shown that C. elegans uses quite similar sensory signaling pathways
as mammals.
Prolonged exposure of nematodes to an odorant reduces chemotaxis to that
odorant (Mori and Oshima, 1997). This behavior is odorant specific, modulated by
food and stress and depends on the odorant concentrations used during pre-exposure
and chemotaxis. This behavior provides a good basis to study olfactory adaptation,
desensitization or perhaps even olfactory learning. However, although several genes
have been identified that play a role in olfactory plasticity, it has been quite difficult
to identify additional genes. The main difficulty is that this behavior is dependent
on many factors and hence relatively variable. However, the careful dissection of
the behavioral variables of olfactory plasticity over the past decade, the current
knowledge of olfactory signaling, and the development of novel techniques to study
gene function and identify cells involved should bring novel insights into the molec-
ular mechanisms of olfactory plasticity.
C. elegans shows strong chemo-attraction to salts. In nature, the detection of
salts is essential for salt homeostasis and salts are probably cues for food. In the
laboratory, salt detection is mainly tested using two assays. In one assay, animals
are placed on a salt gradient and tested for chemotaxis to the highest concentration
(Ward, 1973). This single worm assay has enabled the identification of the sensory
neurons involved in salt detection using laser ablation techniques. The detection of
NaCl is mainly mediated by the ASE chemosensory neurons, but a residual response
is mediated by the ADF, ASG, and ASI neurons. A more recent study confirmed
the importance of the ASE neurons for chemotaxis to NaCl, and revealed a functional
asymmetry between the left and the right cell: ASE left is primarily sensitive to
sodium, whereas ASE right is primarily sensitive to chloride and potassium (Pierce-
Shimomura et al., 2001). In an alternative assay, the quadrant assay, animals are
given a choice between two salt concentrations, by putting them on a plate divided
into four quadrants that can each be filled independently with salt containing agar
(Wicks et al., 2000). This assay provides a robust method to screen for additional
chemotaxis mutants.
In rodents and flies, the main mechanism of salt detection involves ion influx
through degenerin/epithelial Na+ channels (DEG/ENaC), leading to membrane
depolarization and neurotransmitter release. These ENaC channels can be blocked
specifically with the drug amiloride. In rodents, 75% of the response to salt can be
blocked with amiloride; in humans the amiloride sensitivity is less pronounced,
indicating that also other salt detection mechanisms exist. In C. elegans, chemotaxis
to salts is not blocked by amiloride. Perhaps the worm can help to elucidate the
molecular mechanisms of salt detection.
C. elegans is attracted to salt concentrations ranging from 0.1 to 100 mM. High
salt concentrations result in avoidance behavior. This latter response is probably due
to a general avoidance of high osmotic strength, mediated by the nociceptive neurons
ASH and ADL. Forward and reverse genetic approaches have identified several
genes involved in salt detection in C. elegans. These include a cGMP-gated channel
and several Ca2+ activated proteins, indicating that cGMP and Ca2+ are important
second messengers in salt detection. Surprisingly, our recent analysis of these
mutants in the quadrant assay showed that these animals still respond to 25100
mM salt, indicating that another independent high salt detection pathway exists
(Hukema et al., 2006). Thus far, no molecules involved in this pathway have been
identified.
Recently, we have adapted the quadrant assay to study the plasticity of the
response to salt. In this assay, animals are pre-exposed to relatively high, but attrac-
tive concentrations of NaCl (100 mM), and subsequently tested for chemotaxis to
NaCl. Animals pre-exposed to NaCl are no longer attracted to NaCl but even avoid
it. This behavior, which we call gustatory plasticity, is time and concentration
dependent, reversible and partially salt specific (Jansen et al., 2002). The finding
that prolonged exposure to NaCl results in avoidance behavior suggests that gustatory
plasticity is more than adaptation or desensitization. We used two methods to identify
sensory neurons involved in this process. We tried to rescue the plasticity defect of
a G protein subunit mutant, gpc-1, by expressing the gpc-1 gene in specific cells.
In addition, we interfered with cell function by expressing a dominant mutant version
of an ion channel in specific sensory neurons. Using these techniques, we have
shown that gustatory plasticity requires a balance of antagonistic sensory inputs,
involving attractive signals via the ASE salt detection neurons, and aversive signals
via the ASI chemosensory neurons and the nociceptive neurons ASH and ADL
(Hukema et al., 2006).
In a candidate-gene approach, we analyzed gustatory plasticity of 66 mutant
strains (Hukema et al., submitted). In this study, we identified several signaling
cascades and neurotransmitters involved in gustatory behavior. Taking into account
the cellular circuitry and the expression patterns of the various signaling molecules,
we propose a model for the molecular and cellular mechanisms that regulate gusta-
tory plasticity. We are currently performing additional cell-specific rescue experi-
ments to test whether the different signaling molecules indeed function in the cells
as proposed in our model. First, the gustatory response requires salt detection
transmitted by a cGMP and Ca2+ pathway and another unknown high salt detection
pathway. These salt-attraction signals are probably mediated by the ASE neurons.
This signal is balanced by an unknown signal detected by the ASI neurons, trans-
mitted by G proteins, probably activating a thus far unknown guanylyl cyclase.
cGMP subsequently activates a cGMP-gated channel resulting in a Ca2+ influx and
activation of several proteins, ultimately leading to a salt-aversion signal. Another
aversion signal is provided by the nociceptive neurons. At least two G alpha subunits
and a G gamma subunit function in these neurons, transmitting an unknown envi-

ronmental signal to activate heteromultimeric TRP channels.
At present it is unclear in which cells these signals are integrated, but we show
that intercellular signaling requires glutamate, serotonin and dopamine. Integration
of these signals in interneurons, leading to the activation of motoneurons, requires
the Go/Gq signaling network. Also the environmental signals that modulate the
response to NaCl have not been identified, but our results show that gustatory
plasticity depends on salt concentration and exposure time. In addition, our prelim-
inary data suggest that also food signals modulate gustatory plasticity. We feel that
the analysis of gustatory plasticity in C. elegans provides a very good model to
unravel the molecular mechanisms of behavioral plasticity and the integration of
several different sensory cues.
24.3.2 EGG LAYING

How can a variety of neuronal inputs of different weights be converted in an all-or-
none decision? The egg-laying behavior of C. elegans is an interesting paradigm to
study this form of integration.
The anatomy of the egg-laying system is extremely simple. The opening of the
vulva muscles is controlled by a pair of bilateral motoneurons (called HSN) running
from the head to the vulva, where they innervate the vulva muscles. These neurons
are serotonergic, and deliver pulses of serotonin directly on the vulva muscles.
However, even under favorable conditions, eggs are not laid continuously, but in
bursts. This is caused by fluctuations between an inactive state, during which eggs
are retained in the uterus, and an active state, during which eggs are laid at short
intervals. This constitutes a higher order of organization, which is controlled by a
neuropeptide. This neuropeptide is related to FMRF-amide; it is not known if its
physiological concentration varies over time.
Like most animals, worms have developed throughout evolution a behavior to
secure their progeny at their best; they will lay their eggs only if they perceive the
environment as favorable (Figure 24.4). This means that the inputs of the sensory
neurons critical for the decision-making process have to be integrated to deliver an
all-or-none output on the neurons driving the liberation of eggs. The sensory neurons
sensing the parameters critical for this behavior are almost all known. They are
olfactory, chemosensory, thermosensory, and touch sensory neurons, mainly located
in the head. Thanks to the description of all neuronal connections, it is possible to
infer in which head interneurons integration probably takes place. In other systems,
one would successfully dissect this behavior through electrophysiology. Unfortu-
nately, despite recent progress, this technique is still rarely used because of technical
difficulties. But genetic tools can circumvent the lack of electrophysiological data.
To study the regulation of egg laying by the environmental clues, a genetic approach
is to perturb the regulation by appropriate mutations. One can imagine mutations in
which animals would not lay eggs when they normally should, or mutations in which
animals would lay eggs when they should not. In fact, both classes of mutations
exist (called egl-d and egl-c, respectively).
Odors Odors
Osmolarity (?) Osmolarity (?)
Touch Touch
Temperature Temperature
Humidity (?) Humidity (?)
Food Food
Egg laying Egg laying
FIGURE 24.4 Egg laying as a model of integration. The egg-laying behavior is the result of
the integration of various sensory input. Negative stimuli (such as bad odors, vibrations, cold,
etc.) block the egg-laying process (right side of the figure). When all the sensory inputs are
appropriate (left side) egg-laying occurs. The question marks indicate stimuli that are thought
to influence egg laying although they have not been experimentally demonstrated. The advan-
tage of the C. elegans model is that it is amenable to the dissection of the stimuli at a one
cell/ one gene resolution.
Many causes can lead to an egl-d phenotype, including the inability of the animal
to lay eggs (for instance, morphological defects of the vulva). To distinguish which
egl-d mutants are truly unable to lay eggs, from those who could lay eggs but suffer
from misregulation, these mutants are tested for their response to serotonin, a strong
stimulator of egg laying, which acts directly on the vulva muscles. The egl-d mutants
who are responsive to serotonin must possess a functional egg-laying apparatus.
Therefore, their egl-d phenotype is supposed to be a misregulation mechanism.
Almost a dozen genes produce such mutants; unfortunately, few have been identified
yet. Interestingly, several egl-c mutants have abnormally short HSN processes, which
prevents the formation of synapses between HSNs and head neurons. This suggests
that the HSN neurons are indeed in a repressed (hyperpolarized) state under
unfavorable conditions and are derepressed in favorable conditions. The molecular
mechanisms involved are still unknown. The use of cameleon proteins, which are
GFP-based indicators of intracellular calcium, will certainly be helpful tools to
further understand this process.
24.3.3 REGULATION OF LOCOMOTION IN RESPONSE TO FOOD

Dopamine and serotonin are important modulators of various behavioral responses
of many different animals. Also in humans these biogenic amines have been
implicated in many behaviors and diseases. In C. elegans serotonin and dopamine
modulate several behaviors, including egg laying and locomotion. Careful genetic
dissection of these behaviors has led to the identification of several genes that
mediate the effects of serotonin and dopamine.
In 2000, Sawin et al. described a new paradigm to analyze behavioral plasticity.
They found that C. elegans modulates its locomotory rate when it encounters food.
Well-fed animals slow down when they encounter food. This behavior was called
the basal slowing response. Surprisingly, starved animals slow down even more,
called the enhanced slowing response. From previous studies it was known that
many responses to food are mediated by serotonin and dopamine. Therefore, the
authors tested if animals with defects in serotonin and/or dopamine synthesis were
affected in either the basal or the enhanced slowing responses. Testing several of
such mutants revealed that the basal slowing response required dopamine but not
serotonin. Applying exogenous dopamine to the mutant C. elegans cultures could
restore these defects. Next, laser ablation experiments were used to identify the
neurons involved in this response. C. elegans has 8 dopaminergic neurons, 4 CEP
neurons, 2 ADEs, and 2 PDEs. Laser ablation of subsets of these neurons showed
that these 3 types of neurons function redundantly in the regulation of the basal
slowing response. Interestingly, only the neurons that are in contact with bacteria
are required for this behavior. Sawin et al. found that mechanosensory stimuli
probably activate the dopaminergic neurons.
Sawin et al. (2000) showed that the enhanced slowing response required sero-
tonin, synthesized by the NSM neurons. The authors found no indication that addi-
tional serotonergic neurons were involved, but a residual enhanced slowing response
could be observed in animals in which the NSMs were ablated, indicating that also
other cells are involved in this response. It is unclear what signals induce the
enhanced slowing response. Interestingly, the enhanced slowing response could be
blocked by using serotonin antagonists and potentiated by the drug fluoxetine
(Prozac), a selective serotonin reuptake inhibitor.
Using the enhanced slowing response, the group of Horvitz has performed
genetic screens to identify genes that regulate this behavior. Thus far, two genes
have been reported, mod-1 and mod-5 (mod stands for modulation of locomotion
defective). MOD-1 is a serotonin-gated chloride channel and MOD-5 is a serotonin
reuptake transporter (Ranganathan et al., 2000, 2001). Clearly, these proteins play
important roles in the modulation of behavior by serotonin in C. elegans. The
importance of serotonin reuptake transporters (SERT) was already known from
previous studies in other organisms, including humans. The importance of the sero-
tonin gated chloride channel in humans remains to be determined. Characterization
of such channels from mammals might provide novel targets for the developmental
of human pharmaceutical drugs.
24.3.4 ETHANOL RESISTANCE

Invertebrates have recently emerged as additional models of drug abuse. Although
the range of testable behaviors is more limited (for instance, invertebrates cannot
be tested in addiction tests in which animals self-administer cocaine or other stim-
ulatory agents), small animals such as C. elegans are clearly sensitive to the effects
of molecules such as ethanol, cocaine, and nicotine. Since the molecular architecture
of vertebrates and invertebrates is quite similar, it is therefore tempting (and poten-
tially rewarding) to use the powerful approach of genetics to understand the molec-
ular effects of psychoactive agents at the gene level. Since growing these animals
is cheap and easy, robust statistics can be obtained by testing large numbers.
When exposed to ethanol, C. elegans reacts in two phases. A first rapid phase
of excitation consists of increased head movements and locomotion. Then, there is
a progressive lack of coordination, followed by immobility and unresponsiveness to
prodding.
Using the classical genetics approach of phenotypic screens, Morgan and Sed-
ensky (1995) identified nine genes in which mutations modified the resistance to
ethanol. Interestingly, these mutations also changed the sensibility to the volatile
anesthetics enflurane and isoflurane. One of the genes (slo-1) has been cloned and
turned out to encode a potassium channel of the BK family. When the slo-1 gene
is inactivated, animals are apparently normal, but they become highly resistant to
ethanol (Davies et al., 2003). Conversely, mutations in other genes lead to increased
ethanol sensitivity (such as gas-1, which encodes a component of the mitochondrial
electron transport chain).
Natural isolates of C. elegans display various tolerance to ethanol. Interestingly,
in these strains ethanol sensitivity correlates with sensitivity to anesthetics, and
surprisingly, to the social behavior called clumping (see Section 24.3.5). Since the
clumping behavior is mediated in part by the neuropeptide Y receptor (NPY), it is
tempting to speculate that NPY may also be involved in ethanol resistance. Indeed,
three laboratory strains carrying mutations in the NPY gene have been shown to be
more resistant than the reference strain to ethanol in an acute tolerance assay (Davies
et al., 2004). This is of particular interest since NPY is known to be a modulator of
alcohol consumption in rodents and humans. In rats, abnormal or low NPY activity
can promote high alcohol drinking. In humans, a population study recently showed
that polymorphisms on the peptide sequence of NPY could be associated with high
average alcohol consumption.
Recently, microarray analysis was applied to ethanol response in C. elegans.
This technique allows a global survey of genes whose expression levels are modified
upon experimental conditions. 230 out of the 19,000 genes of C. elegans are either
up- or down-expressed following exposure to ethanol. Although such global exper-
iments are not devoid of background noise, they provide a list of candidate genes,
which can later be investigated in more details.
In conclusion, the combination of these various approaches should improve the
knowledge of the mechanisms underlying ethanol intoxication in C. elegans, and
provide a molecular model for better understanding the effects of ethanol in mammals.
24.3.5 SOCIAL FEEDING BEHAVIOR

Many animal species live in groups, or aggregate. This simple form of social behavior
depends on environmental conditions, such as the presence of food or predators or
the season, and varies between individuals of a species and between species. Dif-
ferent natural isolates of C. elegans feed on bacteria either in groups or alone, named
social or solitary feeders, respectively. This finding not only enables the dissection
of the molecular mechanisms that govern social behavior, but also how these mech-
anisms have evolved.
Solitary feeders slow down when they encounter food and disperse on a bacterial
lawn, while social feeders move rapidly on a bacterial lawn and aggregate where
food is most abundant. These natural occurring behavioral differences depend on a
one amino-acid difference in a neuropeptide receptor, NPR-1 (de Bono and Barg-
mann, 1998). The receptor contains a phenylalanine at position 215 (215F) in the
social strains and a valine at this position (215V) in solitary strains. Loss-of-function
of the npr-1 gene resulted in strong aggregation, indicating that NPR-1 represses
social feeding. Sequencing the npr-1 gene from three different C. elegans species
showed that the social NPR-1 215F must have been the ancestral form, indicating
that the standard laboratory C. elegans strain N2, commonly referred to as wild type
actually shows mutant solitary behavior due to mutation of the npr-1 gene (Rogers
et al., 2003).
The most likely candidate ligands of the NPR-1 receptor are encoded by 22
FMR Famide and related peptide (FaRP) genes, flp-1 to flp-22. Two groups
independently tested the 59 neuropeptides encoded by these FaRP genes for activa-
tion of the NPR-1 215F or NPR-1 215V receptors, either in Xenopus oocytes or the
C. elegans pharynx (Rogers et al., 2003) or in mammalian CHO cells (Kubiak et al.,
2003). These assays showed that flp-18 and flp-21 encoded FaRPs could activate
the NPR-1 receptor. The NPR-1 215V receptor variant, found in solitary nematodes,
displayed a stronger response to the peptides than the 215F variant. These findings
are in agreement with the behavioral data, which show that the NPR-1 215V receptor
is more active in repressing social feeding, resulting in solitary feeding behavior.
These findings were confirmed by testing flp-21 loss-of-function mutants and over-
expression animals (Rogers et al., 2003).
Using various genetic methods de Bono and colleagues (2002) identified several
other proteins that regulate this behavior and they identified cells involved. First,
they tested mutants with defective responses to various chemical compounds. Muta-
tions in the TRPV cation channel subunit genes osm-9 and ocr-2 and mutations in
the odr-4 and odr-8 genes, required for the correct localization of particular receptors
to cilia, abolished the social feeding behavior of npr-1 animals. Using cell-specific
expression of the ocr-2 and odr-4 genes in subsets of the chemosensory neurons
and ablation of these neurons using a laser microbeam, de Bono et al. implicated
the ADL and ASH nociceptive neurons in social feeding. Finally, their genetic
analysis showed that the signal transduced by the nociceptive neurons is antagonized
by other sensory neurons. Together, these experiments suggest a model in which
aversive signals from bacteria stimulate the ADL and ASH nociceptive neurons,
which promote social feeding, while signals from unidentified neurons repress social
feeding.
Second, Coates and de Bono (2002) follow up on the finding that NPR-1
regulates social feeding. By carefully analyzing the expression pattern of the npr-1
gene, using a npr-1::GFP fusion construct, and subsequent cell specific expression
of the NPR-1 215V gene in candidate cells in npr-1 animals they implicated the
AQR, PQR and URX neurons in the regulation of social feeding. This finding was
confirmed by genetically blocking the electrical activity of these neurons. In addition
to inhibition of social feeding via NPR-1 in these cells, social feeding is stimulated
by signaling through a cGMP-gated channel. The AQR, PQR and URX neurons are
directly exposed to the body fluid, which seems to function analogously to blood.
Together, these findings suggest a model in which the body cavity neurons, AQR,
PQR and URX, regulate the choice between solitary and social feeding by integrating
NPR-1 mediated signals, which inhibit aggregation, and cGMP mediated signals,
which stimulate aggregation (Coates and de Bono, 2002). At present, it is unclear
where these signals originate.
24.3.6 MEMORY
Although it may sound incredible to many mammalian neurobiologists, worms can
display some forms of associative memory. The most popular assay, described in
this section, is based on the storage of temperature information. Thirty years ago,
when C. elegans genetics was still in its infancy, Hedgecock and Russel (1975)
showed that when worms are placed in a thermal gradient in absence of food, they
move to the temperature where they last were in presence of food (isothermal
tracking). It is thought that, when in the wild, recollection of environmental param-
eters such as temperature and hygrometry can help them to find their food. On the
laboratory bench, it is just a paradigm showing that the animals are able to associate
the presence of food with the sensation of temperature and to store this information.
In absence of food, the isothermal tracking behavior is kept for several hours (Figure
24.5); then the worms will cross isotherms randomly to seek food at other temper-
atures. If food is encountered, a new acquisition period begins.
By laser-ablation experiments, Mori and collaborators (1995) have identified two
(possibly three) neurons as being critical for the memory process. These neurons (AIY
and AIZ) are integrating interneurons, which receive synapses from most head sensory
neurons (including AFD, a sensory neuron specialized in sensing temperature). Ani-
mals in which AIY or AIZ are deleted have peculiar phenotypes: AIY-deleted worms
are cryophilic (they are attracted by colder temperatures) and AIZ-deleted worms are
thermophilic, indicating that these neurons probably have antagonistic effects in wild-
type worms. Since both AIY and AIZ synapse on RIA (another integrating interneuron
downstream of AIY and AIZ), this suggests a mechanism by which the temperature
information could be encoded by the relative strength of AIY and AIZ connections
on RIA. When the sensory neuron AFD is ablated, the worms are either cryophilic,
or atactic (do not feel the temperature). Thanks to the numerous mutants available in
C. elegans, some of these ablation experiments can be reproduced genetically, which
then provide the immense advantage of having unlimited numbers of modified animals
A B
25C 25C
15C 15C
wild type atactic mutant
C D
25C 25C
15C 15C
cryophilic mutant thermophilic mutant
FIGURE 24.5 Associative learning assay and mutant classes. When placed in a temperature
gradient without food (here 15C at the center and 25C at the edge of the dish), worms will
migrate to the temperature they last encountered food (A). Several mutants have been isolated
which are atactic (B), cryophilic (C), or thermophilic (D). Some of these mutations mimic
phenotypes obtained after key neurons have been ablated, disturbing the balance between
cold-driving and heat-driving neurons.
(Figure 24.5). For example, mutations in the ttx-3 gene specifically disable the AIY
neuron due to axonal defects.
How does the worm compare the ambiant temperature (Tamb) to the cultivation
temperature (Tcult)? To test whether AFD has a role in this process, an easy way
would be to record the activity of AFD (and other neurons) in various conditions.
Unfortunately, it is not possible to stick an electrode in a single worm neuron (cell
diameter = a few micrometers) without severely perturbing it. Samuel and collabo-
rators (2003) have elegantly got around this difficulty by using a pH-sensitive green
fluorescent protein localized in the synaptic vesicles of AFD. Since the internal side
of the synaptic vesicles is more acidic than the extracellular environment, the fluo-
rescence levels can be correlated with synaptic release. By doing so on immobilized
live animals, the authors could come to the conclusion that AFD synaptic release is
high if Tamb and Tcult are different, and lower if they are close. Therefore, the single
AFD neuron encodes a direct comparison between actual and memorized tempera-
ture. It is likely that this process involves regulation of gene expression in AFD, but
the molecular targets are still unknown. A schematic view of the neuronal circuitry
associated with this form of learning is shown in Figure 24.6.
The picture is clearer in the AIY interneuron, which naturally expresses the
neuron-specific calcium sensor (NCS-1) protein. In mammals, NCS-1 is involved
in long-term potentiation. When worms are deprived of NCS-1 by a mutation, their
behavior becomes cryophilic and resembles that of AIY-deleted animals, indicating
that NCS-1 is a key component of AIY function. More interestingly, overexpression
of NCS-1 in AIY increases performance levels, accelerates learning, and produces
a memory with slower extinction (Gomez et al., 2001).
Serotonin is likely to be the main inductor of this form of memory in C. elegans
for several reasons: (1) serotonin levels are high when animals are fed and low when
animals are starved; (2) serotonin is released in the head by two neurosecretory cells,
the NSM, which have sensory endings in the pharynx, or the muscle that pumps
food in C. elegans. Thus, NSM release in the body cavity is directly related to the
presence of food; (3) exogenous serotonin can mimic the presence of food in the
sensory input
(temperature)
Temperature
assessment
Serotonin AFD X and comparison
AIY AIZ
Information
storage
RIA
RIB
RIM
Behavior
FIGURE 24.6 Neuronal circuitry underlying associative behavior. Neurons are represented
according to their class (triangle: sensory, hexagon: interneuron, circle: motorneuron). Neuron
X (dashed) has been postulated but is not identified. Only the major connections are drawn.
During the conditioning phase, the signal is memorized by the action of serotonin, which
may act on sensory or interneurons. See main text for details.
learning process: animals grown without food but in the presence of serotonin will
go to this temperature when tested. It is particularly interesting to note that serotonin,
one of the oldest neurohormones in evolution, has conserved its role of memory
inductor from roundworms (C. elegans) to mollusks (Aplysia) and mammals
(reviewed in Mayford and Kandel, 1999).
Over the years, several mutants have been identified in phenotypic screens after
random mutagenesis. These mutants have been picked because they are defective
for the memory assay. Some of them are cryophilic (e.g., ttx-1) and some of them
are thermophilic (e.g., tax-6), suggesting again that the memory process results from
a balance between thermo- and cryophilic inputs. The genes corresponding to these
mutants have been identified in some cases, and this has revealed that proteins
involved in information storage in C. elegans are common to neuronal activity. For
instance, tax-2 and tax-4 encode both subunits of a cyclic nucleotide-gated ion
channel and are expressed in the AFD neuron. Such channels have been implicated
in olfaction and photoreception in mammals. Electrophysiological properties of the
TAX-2/TAX-4 channel suggest that calcium influx is part of the signal transduction
cascade that encodes the temperature information.
24.4 CONCLUSION
Despite its simplicity, the nematode C. elegans possesses in its CNS many of the
neuronal features that have been selected and conserved throughout evolution
because they confer a selective advantage for the bearer. These are the internal
clocks, the possibility to habituate or to sensitize to a stimulus the need for coordi-
nation and integration, and, of course, the faculty to learn from experience, that is
memory. The molecular description of the mechanisms underlying these neuronal
properties is still far from complete in any organism.
As a model, C. elegans offers the advantage of being integrated. This means
that it is possible to study a behavior at the level of the whole animal, and yet to
describe it with a cellular and molecular resolution. The possibility to destroy a
single cell to assay its role in a function is a powerful tool. But the strength of the
C. elegans model resides in the power of genetics, which unambiguously relates
molecules and biological functions, and which is an invaluable way to identify the
genes relevant to a behavior.
In the few examples described in this chapter, it is striking to discover how much
neuronal molecular mechanisms have been conserved through evolution.
REFERENCES
Coates, JC and de Bono, M. (2002) Antagonistic pathways in neurons exposed to body fluid
regulate social feeding in Caenorhabditis elegans. Nature, 419, 925929.
Davies, AG, Pierce-Shimomura, JT, Kim, H, VanHoven, MK, Thiele, TR, Bonci, A, Barg-
mann, CI, and McIntire, SL. (2003) A central role of the BK potassium channel in
behavioral responses to ethanol in C. elegans. Cell., 115, 65566.
Davies, AG, Bettinger, JC, Thiele, TR, Judy, ME, and McIntire, SL. (2004) Natural variation
in the npr-1 gene modifies ethanol responses of wild strains of C. elegans. Neuron.,
42, 73143.
de Bono, M and Bargmann, CI. (1998) Natural variation in a neuropeptide Y receptor homolog
modifies social behavior and food response in C. elegans. Cell., 94, 679689.
de Bono, M, Tobin, DM, Davis, MW, Avery, L and Bargmann, CI. (2002) Social feeding in
Caenorhabditis elegans is induced by neurons that detect aversive stimuli. Nature,
419, 899903.
Fire, A, Xu, S, Montgomery, MK, Kostas, SA, Driver, SE and Mello, CC. (1998) Potent and
specific genetic interference by double-stranded RNA in Caenorhabditis elegans.
Nature, 391, 806811.
Gomez, M, De Castro, E, Guarin, E, Sasakura, H, Kuhara, A, Mori, I, Bartfai, T, Bargmann,
CI, Nef, P. (2001) Ca2+ signaling via the neuronal calcium sensor-1 regulates asso-
ciative learning and memory in C. elegans. Neuron, 30, 241248.
Hedgecock, EM and Russel, RL (1975) Normal and mutant thermotaxis in the nematode
Caenorhabditis elegans. Proc. Nat. Acad. Sci., 72, 40614065.
Hukema, RK, Rademakers, S, Burghoorn, J and Jansen, G. Antagonistic sensory cues generate
gustatory plasticity in Caenorhabditis elegans. Submitted.
Jansen, G, Weinkove, D and Plasterk, RHA. (2002) The G-protein gamma subunit gpc-1 of
the nematode C. elegans is involved in taste adaptation. EMBO J., 21, 986994.
Kubiak, TM, Larsen, MJ, Nulf, SC, Zantello, MR, Burton, KJ, Bowman, JW, Modric, T and
Lowery, DE. (2003) Differential activation of social and solitary variants of the
Caenorhabditis elegans G protein-coupled receptor NPR-1 by its cognate ligand AF9.
J. Biol. Chem., 278, 3372433729.
Lans, H, Rademakers, S and Jansen, G. (2006) A network of stimulatory and inhibitory G
subunits regulates olfaction in Caenorhabditis elegans. Genetics, 167, 16771687.
Mayford, M, and Kandel, ER. (1999) Genetic approaches to memory storage. Trends Genet.,
15, 46370.
Morgan, PG, and Sedensky, MM. (1995) Mutations affecting sensitivity to ethanol in the
nematode, Caenorhabditis elegans. Alcohol Clin Exp Res., 19:14239.
Mori, I, and Oshima, Y. (1995) Neural regulation of thermotaxis in Caenorhabditis elegans.
Nature, 376, 344348.
Mori, I, and Oshima, Y. (1997) Molecular neurogenetics of chemotaxis and thermotaxis in
the nematode Caenorhabditis elegans. Bioessays, 19, 10551064.
Pierce-Shimomura, J.T., Faumont, S., Gaston, M.R., Pearson, B.J. and Lockery, S.R. (2001).
The homeobox gene lim-6 is required for distinct chemosensory representations in
C. elegans. Nature, 410, 694698.
Ranganathan, R, Cannon, SC and Horvitz, HR. (2000). MOD-1 is a serotonin-gated chloride
channel that modulates locomotory behaviour in C. elegans. Nature, 408, 470475.
Ranganathan, R, Sawin, ER, Trent, C and Horvitz, HR. (2001). Mutations in the Caenorhab-
ditis elegans serotonin reuptake transporter MOD-5 reveal serotonin-dependent and
independent activities of fluoxetine. J. Neurosci., 21, 58715884.
Riddle, TBDL, Meyer, BJ and Priess, JR eds. (1997) C. elegans II. Cold Spring Harbor
Laboratory Press, Plainview, New York.
Rogers, C, Reale, V, Kim, K, Chatwin, H, Li, C, Evans, P and de Bono, M. (2003) Inhibition
of Caenorhabditis elegans social feeding by FMRFamide-related peptide activation
of NPR-1. Nat. Neurosci., 6, 11781185.
Samuel, AD, Silva, RA, and Murthy, VN. (2003) Synaptic activity of the AFD neuron in
Caenorhabditis elegans correlates with thermotactic memory. J Neurosci., 23, 3736.
Sawin, ER, Ranganathan, R and Horvitz, HR. (2000) C. elegans locomotory rate is modulated
by the environment through a dopaminergic pathway and by experience through a
serotonergic pathway. Neuron., 26, 619631.
The C. elegans Sequencing Consortium (1998) Genome sequence of the nematode C. elegans:
A platform for investigating biology. Science, 282, 20122018.
Ward, S. (1973) Chemotaxis by the nematode Caenorhabditis elegans: identification of
attractants and analysis of the response by use of mutants. Proc. Natl. Acad. Sci. USA
70, 817821.
Wes, PD and Bargmann, CI. (2001) C. elegans odour discrimination requires asymmetric
diversity in olfactory neurons. Nature, 410, 698701.
Wicks, SR, de Vries, CJ, van Luenen, HGAM and Plasterk, RHA (2000) CHE-3, a cytosolic
dynein heavy chain, is required for sensory cilia structure and function in Caenorhab-
ditis elegans. Dev. Biol,. 221, 295307.
Wood, W.B, ed. (1988) The Nematode Caenorhabditis elegans. Cold Spring Harbor Labora-
tory Press, Plainview, New York.
AUTHORS NOTE
We apologize to the authors whose work could not be cited due to lack of space.
25 Genetics, Behavior, and

Brain Dopamine Systems
Robert Hitzemann, Shannon McWeeney, and
John Belknap
CONTENTS
25.1 Introduction .................................................................................................. 371

25.2 A Behavioral ApproachThe Catalepsy Phenotype.................................. 373
25.3 From Transporter and Receptor Density to Behavioral Phenotypes .......... 376
25.4 D2 Dopamine Receptor Density and Drd2 Expression ............................... 379
25.5 Regulation of Drd2 Expression ................................................................... 385
25.6 Conclusions .................................................................................................. 386
References.............................................................................................................. 387
25.1 INTRODUCTION
Even for the reader unfamiliar with the topic of this chapter but familiar with
dopamine and behavior, it is obvious that the topic of genes, behavior and brain
dopamine systems cannot be covered in detail. Rather, the goal of the chapter is
to provide the reader with an introduction and some examples, largely extracted
from work in the authors laboratories. Previous reviews (Hitzemann et al., 1995;
Hitzemann, 1998) can be referenced for important background material. The focus
of this chapter will be largely on D2 dopamine receptors. The reasons for this
emphasis include the longstanding emphasis on D2 receptors in the etiology
and/or expression of a variety of behaviors including schizophrenia, substance
abuse and alcoholism (Hitzemann, 1998). The other reason for this emphasis is
the amazing variability among individuals in D2 receptor density and in the response
to drugs which either stimulate or block these receptors. The reasons for this vari-
ability remain unclear, but there is ample evidence to suggest that genetic factors
have an important role. Here we provide a clinical example to illustrate the variation
in receptor density and drug response. Twenty-three normal controls were adminis-
tered 0.5 mg/kg of methylphenidate iv. after being told they would receive either
placebo or methylphenidate; behavioral effects were measured before and after
the injection and included self-ratings of pleasant and unpleasant drug effects.
On another day, the subjects were scanned for D2 /D3 receptor density using positron
371
emission tomography (PET) and 11C-raclopride as the ligand. Twelve of the 23

subjects (52%) reported the drug effects as very pleasant, 9 subjects (39%) reported
the drug effects as unpleasant and 2 subjects (9%) reported no marked behavioral
effects (Volkow et al., 1999). Importantly, the individuals who reported the drug
effects as pleasant had lower levels of D2 receptor availability when compared with
the unpleasant group (Figure 25.1). Although the average difference was small
(~20%), it was similar to the persistent differences reported between normal controls
and withdrawn chronic alcoholics, cocaine, methamphetamine and heroin addicts
(see, e.g., Volkow et al. 1993). These differences were nominally thought to be
associated with receptor down-regulation as a result of drug-induced dopamine
release. However, the data described in Figure 25.1 suggest that low D2 receptor
availability could be a risk factor for drug abuse since presumably, the pleasant
group would be more likely to try the drug again. The data in Figure 25.1 also
illustrate that in this sample the range of D2 receptor availability was more than
100%. For other clinical samples, including postmortem samples, the range of
D2 receptor availability has been similar and in some cases even greater (Hitzemann
et al., 1998).
The question of whether these large differences are genetic or environmental
cannot be easily answered in human populations; however, among mice it is clear
that the genetic effect is substantial (Jones et al., 1999; Hitzemann et al., 2003). The
3.8
3.6
3.4
3.2
BMax /KD
3.0
2.8
2.6
2.4
2.2
2.0
Pleasant Unpleasant Neutral
Perception of Methylphenidate
FIGURE 25.1 The relationship between D2 dopamine receptor binding potential and the
response to a methylphenidate challenge (data adapted from Volkow et al., 1999). The binding
potential (BMax/KD) for the receptor was estimated in 23 normal controls using positron
emission tomography (PET) and 11C-raclopride as the PET ligand. Note that with the binding
potential calculation a value of 1.0 indicates that there is no specific binding; thus, the range of
binding potentials in this figure varies by >100%. Subsequently, the subjects were challenged
with methylphenidate (0.5 mg/kg, iv.) and were asked to rate their response to the challenge
on a visual analog scale; responses varied from very pleasant to neutral to unpleasant.
Genetics, Behavior, and Brain Dopamine Systems 373
remainder of this chapter focuses on these and related data which have contributed
to our genetic understanding of brain dopamine systems and behavior.
25.2 A BEHAVIORAL APPROACHTHE CATALEPSY

PHENOTYPE
The typical antipsychotic drugs such as haloperidol exert both their therapeutic
and untoward side effects by blocking the D2 class of receptors, which includes the
D2, D3, and D4 receptors. The major untoward side-effects (largely, if not entirely
generated by blockade of the D2 receptors) are known as extrapyramidal symptoms
(EPS) which have many of the same features as Parkinsons disease. In mice cata-
lepsy is the murine equivalent of EPS; although different approaches have been used
to measure catalepsy (e.g., Kanes et al., 1993; Fowler et al., 2001), all approaches
assess some aspect of drug-induced immobility in the absence of sedation. Fink
et al. (1982) appear to have been the first to note that two inbred mouse strains
(BALB/cJ and CBA/J) differed markedly in terms of haloperidol-induced catalepsy
(BALB/cJ was more sensitive). These authors also noted that the BALB/cJ mice had
a higher D2 receptor density which was counter to their expectation that a lower
receptor density would make the animals more sensitive to the effects of haloperidol.
Kanes et al. (1993) expanded the number of strains examined to 8 and was able
to show a >30-fold difference in sensitivity between the most sensitive strain
(BALB/cJED50 = 0.3 mg/kg) and least sensitive strain (LP/J ED50 = 10 mg/kg);
it was also observed that there was very little variation among the inbred strains in
terms of the catalepsy response induced by the D1 receptor antagonist, SCH 23390.
These data suggested that the variation in haloperidol response was not due to a
general deficit in extrapyramidal function. Pharmacokinetic differences were also
eliminated as a potential cause. The number of standard inbred strains examined
was subsequently expanded to 15 strains; in addition, 25 strains of the BXD recom-
binant inbred panel were also phenotyped (Figure 25.2) (Hitzemann et al., 1995;
Kanes et al., 1996). The expansion of the standard inbred panel only placed additional
strains between the extreme genotypes. The development of the BXD RI panel is
described elsewhere in this volume. Figure 25.2 illustrates that in the BXD RI panel
no RI strain was more sensitive than the DBA/2J strain, but several strains were
significantly less sensitive than the C57BL/6J strain. This phenomenon, sometimes
called transgressive segregation, must in the RI panel be associated with genetic
effects. For example, the DBA/2J strain may contribute some nonresponsive alleles.
Thus, in a mapping study using a C57BL/6J DBA/2J intercross, it would not be
unexpected to detect a QTL in which the DBA/2J allele is associated with non-
response, and, indeed, such a QTL has been detected (Kanes et al., 1996; Patel and
Hitzemann, 1999).
The RI panel is useful to look for genetically correlated traits and these corre-
lations may in turn be informative about the underlying mechanisms. However,
before looking at these correlations it is necessary to introduce WebQTL. WebQTL
is an assembly of databases and tools that exploits sophisticated gene mapping
and related statistical methods to rapidly perform analyses of gene expression and
10
14
12 Standard Inbred Strains
10
8
8
6
Haloperidol ED50 - mg/kg
4
6 2
0
A
BALB/c
AKR
DBA/2
PL
CE
SJL
C3H/He
A/He
C57Bl/6
CBA
S129
LP
C57L
P
4
0
21
11
31
1
DBA
C57
5
30
22
24
15
27
13
14
19
28
20
12
6
25
32
16
18
23
29
2
8
BXD RI Strains
FIGURE 25.2 The ED50 values for haloperidol-induced catalepsy among a panel of stan-
dard-inbred mouse strains and among a panel of the BXD recombinant inbred strains (data
adapted from Hitzemann et al. 1995). A full description of the catalepsy response and the
methods used to determine the ED50 value is found in Kanes et al. (1993).
behavioral data (Chesler et al., 2004). The databases focus on the original BXD RI
panel (Taylor 1978), although data for some of the newer BXD RI strains are
available. Important for the immediate discussion is that the database contains more
than 650 published phenotypes, with the largest single segment focusing on drug-
related behavioral phenotypes. The catalepsy data illustrated above have been entered
into the database (ID# 10336). Querying the database for phenotypes that are
correlated with catalepsy revealed significant correlations with cocaine-induced
(15 mg/kg) stereotypy (ID# 10297) and ventral midbrain D2 dopamine receptor
density (ID# 10270) (Jones et al., 1999). These data are illustrated in Figures 25.3
and 25.4. The data in Figure 25.3 show that the RI strains most sensitive to the
haloperidol effect are also the strains most sensitive to the effects of cocaine
(r = 0.68, p < 0.0002); we interpret these data to show that if we could determine
the traits associated with increased haloperidol sensitivity, some of these would also
be associated with increased cocaine sensitivity (or vice versa). Given that both drugs
100
Cocaine - Induced Stereotyped Behavior

(change in repeated movements/15 min)
r = - 0.68 ( p < 0.0002)
80
60
40
20
-20
0 2 4 6 8 10
Haloperidol-Induced Catalepsy
(ED50 mg/kg)
FIGURE 25.3 The relationship between haloperidol-induced catalepsy (ID# 10336) and
cocaine (15 mg/kg)-induced stereotypy (ID# 10297) in a panel of the BXD recombinant
inbred strains. The ID values are those that reference the phenotypes at WebQTL
(www.WebQTL.org); the data entered into the analysis may be downloaded from this site.
The catalepsy data are taken from Kanes et al. (1996) and the cocaine data are taken from
Jones et al. (1999).
exert their pharmacological effects largely via effects on brain dopamine systems,
such an association cannot be unexpected. Figure 25.4 illustrates a potential
mechanism. Here we see that the animals most sensitive to haloperidol are those
that have the highest density of D2 receptorsa result consistent with that reported
by Fink et al. (1982). Data not shown is the correlation between receptor density
and cocaine response (r = 0.68, p < 0.001). Thus, we conclude that the density of
ventral midbrain D2 receptors (presumably autoreceptors) is associated both with
haloperidol and cocaine response. These relationships are not entirely intuitives, but
the data illustrate the power of the approach and the cumulative value of the RI data.
Furthermore, we again see the wide variation in D2 receptor density and the apparent
association with a drug response. It could be argued that the variation in receptor
density is simply an epiphenomenon reflecting a difference in the number of dopam-
ine neurons in the ventral midbrain. Data on the number of neurons in the ventral
tegmental area are also found in the database (ID# 10223). For the RI panel there
is no significant correlation between receptor density and neuron number (p > 0.3).
Finally, it should be noted that for the striatum there are 12 different measures
of D2 receptor density in the database; none of these are significantly correlated
to catalepsy, confirming earlier results from selective breeding studies (Hitzemann
et al., 1991; Qian et al., 1992; 1993).
80
D2 Receptor Density - Ventral Midbrain r = - 0.68 ( p < 0.001)

60
(BMAX - fmole/mg protein)
40
20
0 2 4 6 8 10
Haloperidol-Induced Catalepsy
(ED50 mg/kg)
FIGURE 25.4 The relationship between haloperidol-induced catalepsy (ID# 10336) and
ventral midbrain D2 dopamine receptor density (ID# 10270) in a panel of the BXD recom-
binant inbred strains. Details as in the legend to Figure 3. Receptor density data taken from
Jones et al. (1999).
25.3 FROM TRANSPORTER AND RECEPTOR DENSITY

TO BEHAVIORAL PHENOTYPES
In the previous section we began with a behavioral phenotype (catalepsy) that
genetically led us to another behavioral phenotype (cocaine response) and one
measure of D2 receptor density. In this section, we begin with dopamine transporter
and receptor density and ask what behaviors show strong genetic correlations.
Perhaps, not surprisingly we are again led to cocaine response. Janowsky et al.
(2001) examined the DA transporter (DAT) binding of RTI-55 to 21 of the BXD RI
strains. Using a data-mining approach similar to that described here, the authors
concluded that DAT density was significantly correlated with cocaine and metham-
phetamine-induced locomotor activation and thermic responses (both hypo- and
hyperthermia); however, the binding was not correlated with stimulant behaviors
related to sensitization, reward, voluntary consumption, stereotypy or seizures. One
example of a significant association is shown in Figure 25.5; RTI-55 binding (ID#
10234) is significantly (r = 0.69, p < 0.0005) correlated with the cocaine-induced
(30 mg/kg) difference score (cocaine treatment saline treatment) in vertical rearing
movements (ID #10310) (Jones et al., 1999). Under the conditions used by these
authors cocaine decreased rearing, with the most significant decrease found in those
RI strains with the lowest DAT density. In contrast, the strains with the lowest DAT
density are the strains that exhibit the most marked cocaine-induced locomotor
activation (ID# 10489) (Phillips et al., 1998). For the data shown in Figure 25.6, the
2 r = 0.69 ( p < 0.0005)

(change in rearing movement/15min)
Cocaine - Induced Rearing
0
-2
-4
-6
-8
-10
-12
-14
1 2 3 4 5 6
Dopamine Transporter Density
(BMAX - pmoles/mg protein)
FIGURE 25.5 The relationship between striatal dopamine transporter density (ID# 10234)
and cocaine (30 mg/kg)-induced changes of vertical rearing movements (ID# 10310) in a
panel of the BXD recombinant inbred strains. Transporter density data taken from Janowsky
et al. (2001) and the cocaine data are taken from Jones et al. (1999).
18000
Cocaine - Induced Locomotor Activity
16000
14000
(cm traveled/15min)
12000
10000
8000
6000
4000 r = 0.55 ( p < 0.005)

2000
0
1 2 3 4 5 6
Dopamine Transporter Density
(BMAX - pmoles/mg protein)
FIGURE 25.6 The relationship between striatal dopamine transporter density (ID# 10234)
and cocaine (40 mg/kg)-induced changes in locomotor activity (ID# 10224) in a panel of the
BXD recombinant inbred strains. Cocaine data are taken from Phillips et al. (1998).
cocaine (40 mg/kg) response was also indexed as a difference score. Figures 25.5
and 25.6 illustrate a familiar pattern in behavioral researchas some behaviors
increase, others decrease. Figures 25.5 and 25.6 also illustrate another theme of this
chapter, the marked genetic variation within the brain dopamine system; DAT den-
sity, here calculated as the BMax differs by more than 300% among the RI strains.
This variation is not significantly associated with the number of DA neurons in the
ventral tegmental area (r = 0.13) (ID# 10223) or the substantia nigra zona compacta
(r = 0.07) (ID# 10224) (data not shown). Here it should be noted that unlike DAT
density, among the RI strains variation in the number of dopamine neurons is
relatively modest (~40%).
Measures of RI strain striatal D2 receptor density in the WebQTL database were
obtained independently in two studies (Jones et al., 1999; Hitzemann et al., 2003)
that used different techniques and ligands to assess receptor binding. Given that
there were only 14 RI strains in common to both studies, the agreement between
the studies is relatively good. Figure 25.7 illustrates the agreement for the dorsal
striatum (r = 0.67, p < 0.008) (ID# 10254 vs. ID# 10220); the strains with low and
high density were readily identified in both studiesthe strains with intermediate
receptor density were not as reliable. Also note the marked variation in receptor
density among the RI strains: ~100% in Hitzemann et al. (2003) and >300% in Jones
et al. (1999). Using the binding data for the dorsomedial striatum (Hitzemann et al.,
2003), the database was queried as was done for DAT. The most significant behav-
ioral correlation (r = 0.74, p < 0.0004) was for a cocaine (45 mg/kg) response (here
400
D2 Receptor Density - (fmoles/mg)
350
300
r = 0.67 ( p < 0.008)
250
200
150
100
50
0
160 180 200 220 240 260 280 300 320
D2 Receptor Binding (fmoles/mg)
FIGURE 25.7 A comparison of two independent measures of striatal D2 dopamine receptor

density in a panel of the BXD recombinant inbred strains. Data were taken from Hitzemann
et al. (2003) [ID#10220] and Jones et al. (1999) [ID# 10254]. The two studies were completely
independent; both studies used 125I-epidepride as the ligand but differed in strategyauto-
radiography (Hitzemann et al., 2003) vs. membrane binding (Jones et al., 1999).
10000
Cocaine - Induced Locomotor Activity
9000 r = 0.74 ( p < 0.004)
(cm traveled/15min)
8000
7000
6000
5000
4000
160 180 200 220 240 260 280 300 320
FIGURE 25.8 The relationship between striatal D2 dopamine receptor density (ID# 10220)
and cocaine (45 mg/kg)-induced locomotor activation (ID# 10318) in a panel of the BXD
recombinant inbred strains. Receptor binding data are taken from Hitzemann et al. (2003);
the cocaine response data are taken from Jones et al. (1999).
a difference score measure of locomotor activity- ID# 10318) (Jones et al., 1999)
(Figure 25.8); the analogous data for Phillips et al. (1998)see abovewas also
significantly correlated to receptor density (r = 0.55, p < 0.03) (data not shown).
Overall, the examples cited above illustrate how it is possible to use the RI
strains to dissect the genetic relationships among dopamine systems and behavioral
phenotypes. Importantly, it was found that individual cocaine responses are associ-
ated with different aspects of the brain dopamine system.
25.4 D2 DOPAMINE RECEPTOR DENSITY

AND DRD2 EXPRESSION
In the previous sections we illustrated the marked differences among the RI strains
in D2 receptor density. It would be natural to assume that the variance (>300% in
one study) is due to marked differences in Drd2 expression. Some years ago, we
and others concluded that there was a significant mismatch between receptor density
and gene expression (e.g., Qian et al., 1992; 1993). Individual differences in D2
receptor density were not associated with differences in Drd2 expression; however,
it was possible to show that the receptor gradients within the striatum (rostral- caudal,
ventral-dorsal and lateral-medial) did parallel to a large extent the pattern of gene
expression.
With the availability of brain gene expression data for the RI strains, it has been
possible to return to the paradox of D2 receptor density and gene expression with
substantially greater statistical power. However, regardless of which measures of
receptor density and gene expression from the WebQTL database are compared, the
results are the same: there appears to be no relationship (see, e.g., Figure 25.9).
Thus we again conclude that most of the variance in receptor density is asso-
ciated with some factor or factors different from Drd2 expression. Mapping all 12
measures of D2 receptor density revealed moderate associations (p < 0.005) on Chr
15 and Chr 16 but none on Chr 9 near the Drd2 locus (data not shown); in contrast,
mapping Drd2 expression reveals a moderate signal on Chr 9 near the gene locus
(LOD ~2) and weaker signals on other chromosomes but none on Chr 15 and 16
(data not shown).
Given the mismatch between receptor density and gene expression, the question
arises as to whether Drd2 gene expression data per se is of some value to our
understanding of the relationships among genetics, brain dopamine systems and
behavior. We answer this question positively for several reasons. First, the possibility
exists that receptor density and gene expression do tract together in some brain
region that has not been measured. Areas that come to mind with potentially very
salient behavioral effects include the prefrontal cortex, the central nucleus of the
amygdala and the bed nucleus of the stria terminalis. Our attempts to measure
receptor binding in these areas with quantitative autoradiography have (to date) not
been successful. Second, as noted above, there is evidence that Drd2 expression is
important for explaining the receptor gradients within the brain and this perhaps is
part of a larger pattern of brain organization that is important to the issue at hand
genetics, dopamine and behavior. Third, there is compelling evidence to show that
320
300
280
260
240
220
200
180
r = 0.05 ( p < 0.5)
160
140
100 150 200 250 300 350 400
Drd2 Expression
(relative units)
FIGURE 25.9 The relationship between Drd2 expression and D2 dopamine receptor density
(ID# 10220) in a panel of the BXD recombinant inbred strains. Receptor binding data are
taken from Hitzemann et al. (2003); Drd2 expression data are from the Affymetrix U74A
array and are posted at the WebQTL site (www.WebQTL.org).
Drd2 expression is strongly associated with behaviors for which there is strong
evidence for dopamine modulation (Hitzemann et al., 2003). A query to the WebQTL
database for phenotypes correlated to Drd2 expression reveals that the strongest
association (r = 0.74, p < 0.00004) is for ethanol-induced conditioned place prefer-
ence (CPP)time spent on the ethanol paired floor (ID# 10090) (Cunningham,
1995). CPP is considered to be a measure of a drugs hedonic or rewarding properties.
These data are illustrated in Figure 25.10. Other behavioral phenotypes within the
top 10 correlations include morphine two-bottle choice (ID# 5968) (r = 0.60,
p < 0.003) and ethanol two-bottle choice (ID# 10479). Although the database is
over-populated with drug-related phenotypes, the chance that 3 drug reward/con-
sumption phenotypes would appear in the top 2% of all possible correlations is very
small. The possibility must be considered that the associations with Drd2 expression
result from the close linkage of Drd2 to other genes. In this regard, we have
previously reported that ethanol two-bottle choice is highly correlated to Ncam
expression (Hitzemann et al., 2003); Ncam is 96 kb from Drd2.
An alternative hypothesis is that Drd2 is coexpressed with a family of genes and
that it is the collective action of the gene family which modulates behavior. To test
the first component of this hypothesis we turned to a C57BL/6J DBA/2J F2 brain
gene-expression dataset. The data set contains expression results (Affymetrix 430
A and B arrays) for 56 F2 animals formed from the reciprocal F1 hybrids and is
balanced for males and females. Genotypic data for this F2 dataset have been obtained
for >300 microsatellite markers (or on average 1 marker/5 cM). Both the genotypic
80
75 r = 0.74 ( p < 0.0004)

(percent time on ethanol grid floor)
Conditioned Place Preference
70
65
60
55
50
45
100 150 200 250 300 350 400
Drd2 Expression
(relative units)
FIGURE 25.10 The relationship between Drd2 expression and ethanol conditioned place
preference (ID# 10090) in a panel of the BXD recombinant inbred strains. See legend to
Figure 25.9 for details on expression data. Conditioned place preference data are taken from
Cunningham (1995).
and phenotypic data have been posted at the WebQTL site. Drd2 expression was
significantly (p < 106) associated with 38 unique transcripts (Table 25.1); assuming
a total of 30,000 unique transcripts on both the A and B arrays, one would predict
only 1 chance association. The SymAtlas database at http://www.gnf.org was queried
to determine if the transcripts listed in Table 25.1 were specifically expressed in the
striatum. For some transcripts data were available only for human brain expression.
Expression in the striatum was rated on a four-point scale ranging from under-
expressed to markedly and uniquely over-expressed in the striatum (compared to
other brain regions). Fourteen of the 38 transcripts (37%) were moderately to sig-
nificantly over-expressed in the striatum and these were by and large the transcripts
most significantly associated with Drd2 expression. Figure 25.11 illustrates the
correlation between Drd2 and preproenkephalin (Penk1) expression and Drd2 and
regulator of G-protein signaling 9 (Rgs9) expression. Drd2 expression varied 49%
across the F2 sample while Penk1 and Rgs9 expression varied 94 and 73%, respec-
tively. It should be noted (Table 25.1) that none of the genes significantly associated
with Drd2 expression were significantly differentially expressed between the B6 and
D2 strains. There are numerous implications to this paradox but importantly it
suggests that screening gene lists for significant differences between the parental
strains will significantly underestimate the number of genes showing differential
expression in a segregating population.
The data in Figure 25.11 were collected from a whole brain sample; the question
arises as to whether similar results would be obtained for striatal gene expression.
Striatal data for an F2 sample are not currently available; however, striatal data are
available for a subset of the BXD RI strains. Figure 25.12 illustrates the correlation
between Drd2 and Penk1 striatal expression in the RI strains; the correlation (r =
0.6) is similar to that seen in the F2 sample, likely reflecting the balance between
the increased sample power in the F2 sample vs. the increased specificity of the
striatal sample.
One could argue that strongest associations seen in Table 25.1 are simply an
artifact related to differences in the size and/or number of cells in the striatum. Since
only half the brain was used for the expression analysis, it was possible to determine
if striatal size covaried with expressionno significant association was detected
(data not shown). To determine if there is a relationship between gene expression
and cell density requires a detailed stereological analysis. However, data are available
in the BXD RI series for the number of cholinergic neurons in the striatum (Dains
et al., 1996). A significant correlation (r = 0.61, p < 0.001) was detected between
Drd2 expression and the number of cholinergic neurons (ID# 10106) but only in the
most rostral aspect of the striatum (data not shown); in more caudal areas, no
significant correlation was detected but it should be noted that gene expression was
not measured in a parallel fashion to measures of the number of cholinergic neurons.
Thus, in lieu of other data, we conclude that cell density may partially account for
the associations seen in Table 25.1.
Walker et al. (2004) have analyzed a rat brain gene expression dataset using a
strategy somewhat similar to that described here. These authors asked the question
what genes are coexpressed with Drd1. The results of their analysis (summarized
in figure 2 of Walker et al., 2004) indicated that there was a large group of genes
TABLE 25.1
Transcripts Significantly (p < 106) Associated with Drd2 Expressiona
Affymetrix ID# Array Gene Symbol r Striatal Expression
1455725 A H3f3b 0.62

1449373 A Dnajc3 0.61
1449314 A Zfpm2 0.60
1448113 A 4930403J22Rik 0.60
1435250 A 2810013E07Rik 0.60
1425023 A Usp3 0.60
1433582 A 1190002N15Rik 0.60
1451864 A Cacng8 0.60
1455085 B 1700086L19Rik 0.60
1424490 A 2410005H09Rik 0.61 +
1419066 A Ier5l 0.61 +
1418782 A Rxrg 0.62 +
1448790 A Sema6b 0.62 +
1450944 A Cspg4 0.62 +
1424132 A Hras1 0.62
1434153 B Shb 0.62 +
1455629 B Drd1a 0.63 +++
1455609 B C030025P15Rik 0.64 +
1417804 A Rasgrp2 0.64 +
1456640 B Sh3rf2 0.64
1455564 B Bcr 0.64 ++
1416050 A Scarb1 0.64
1455701 B Snx26 0.65 +
1422705 A Tmepai 0.65 +++
1418881 A Efcbp2 0.66 +++
1448327 A Actn2 0.67
1442166 B Cpne5 0.68 ++
1432184 B 2610204M08Rik 0.69 +++
1460710 A Adora2a 0.73 +++
1427343 A Rasd2 0.74 +++
1455296 B Adcy5 0.74
1427523 A Six3 0.74 ++
1455190 A Gng7 0.75 +++
1451331 A Ppp1r1b 0.76 +++
1418691 A Rgs9 0.77 +++
(Continued)
TABLE 25.1
Transcripts Significantly (p < 106) Associated with Drd2 Expressiona
(Continued)
Affymetrix ID# Array Gene Symbol r Striatal Expression
1423544 A Ptpn5 0.79 +++

1427038 A Penk1 0.80 +++
1449420 A Pde1b 0.88 ++
1418950 A Drd2 1.00 +++
a Brain transcript expression was measured in a sample of 56 C57BL/6J DBA/2J F animals using
2
the Affymetrix 430A and 430B arrays. Data were analyzed using position dependent nearest
neighbor algorithm (Zhang et al. 2003). Drd2 expression was then correlated with the expression
of all other transcripts on the arrays; the threshold for significance was set at p < 106. At this
threshold, the expectation of a chance correlation is <1, assuming that there are 30,000 unique
transcripts on the A and B arrays. Striatal expression data was taken from the GNF database and
was semi-quantitatively judged from under-expressed () to highly over-expressed (+++).
9.2
9.1 r = 0.77 ( p < 10-6)
Rgs9 Expression
9.0
8.9
8.8
8.7
8.6
10.3
10.2
Penk1 Expression
r = 0.80 ( p < 10-6)

10.1
10.0
9.9
9.8
9.7
9.6
9.5
8.0 8.1 8.2 8.3 8.4 8.5
Drd2 Expression
FIGURE 25.11 The relationship between Drd2 expression and the expression of Penk1 or Rgs9
in a sample of C57BL/6J DBA/2J F2 animals. Details are found in the legend to Table 25.1.
coexpressed with Drd1 in the striatum (including the nucleus accumbens) and per-
haps not surprisingly, their list overlaps with the genes listed in Table 25.1. Common
genes in both lists were the following: Drd1, Drd2, Penk 1, Ppp1r1b, Pde1b, Adcy5,
Adora2a and Rgs9. Thus, data from two independent sources and two different
species, mutually confirm a set of genes that includes Drd2 which are coexpressed
in the striatum.
18.0
Penk1 Expression 17.8 r = 0.61 ( p < 0.001)
17.6
(striatum)
17.4
17.2
17.0
16.8
10.0 10.2 10.4 10.6 10.8 11.0 11.2 11.4 11.6
Drd2 Expression
(striatum)
FIGURE 25.12 The relationship between striatal Drd2 expression and striatal Penk1 expres-
sion in panel of the BXD recombinant inbred strains. Expression data were obtained using
the Affymetrix 430 2.0 array; data provided by Rosen G. and Williams R. (unpublished
observations).
25.5 REGULATION OF DRD2 EXPRESSION

As noted by Lee et al. (2003) detailed analysis of the transcription control mech-
anisms that regulate dopamine receptor genes has revealed complex patterns involv-
ing both activators and repressors (Minowa et al. 1992a, 1992b, 1993, 1994, Lee
et al., 1996, 1997). Hwang et al. (2001) have identified a dopamine receptor regu-
lating factor gene (Drrf) (alias: Kruppel-like factor 16 [Klf16]; basic transcription
element binding protein 4 [Bteb4I]) which binds to GC and GT boxes in the pro-
moters of both Drd1 and Drd2 and displaces Sp1 and Sp3 from these binding sites.
Querying the databases noted above to determine if there is an association between
Drrf and Drd2 expression reveals no significant association (r = 0.08). Since Drd1
and Drd2 are co-expressed, factors known to regulate Drd1 expression such as
Mrf1(Meis2), TGIF, Zic2 and Sp3 (Yang et al., 2000a, 2000b) also become candi-
dates for the regulation of Drd2 expression. Of special note is the role Zic2 may
have in the tissue specific distribution of Drd1 expression (Yang et al., 2000b). On
the basis of existing data, we have been unable to detect an association between the
expression of these transcription factors and Drd1 or Drd2.
An alternative approach to understanding gene co-expression is to ask if the key
genes listed in Table 25.1 have common transcription factor binding motifs. There are
several algorithms available to address this issue. Here we summarize the results using
the oPOSSUM algorithm which is publicly available at http://www.cisreg.ca/cgi-
bin/oPOSSUM/opossum. oPOSSUM is a system for determining the over-representa-
tion of transcription factor binding sites within a set of co-expressed genes as compared
with a pre-compiled background set. The background data set was compiled by iden-
tifying all strict one-to-one human/mouse orthologs from the Ensembl database. The
orthologs were aligned and the conserved noncoding regions were identified. The
conserved regions 5 kb upstream and downstream from the transcription start site are
scanned for common transcription factor sites among a group of coexpressed genes.
Some genes from a coexpressed group are excluded from the analysis, e.g., if the
conserved promoter regions could not be identified in a satisfactory manner. Of the
14 striatal coexpressed genes noted above, 7 were included in the analysis: Drd2, Rgs9,
Pde1b, Nptp, Penk1, CopineV and Adora2a. The transcription factor binding site most
enriched in this group (as compared with the reference group) was NF-kappaB [NF-
] (p < 0.0001); four of the top ten enriched motifs were from the Rel/NF-kappaB
family. These data cannot be unexpected. Previous studies have shown that both Drd2
and Penk1 expression are regulated by NF- (ONeill and Kaltschmidt, 1997; Fioren-
tini et al., 2002). Other genes expressed in brain and known to be regulated by NF-
include neural cell adhesion molecule (Simpson et al., 2000), inducible nitric oxide
synthase (Madrigal et al., 2001), -opioid receptors (Kraus et al., 2003), brain derived
neurotrophic factor (Lipsky et al., 2001), calcium/calmodulin dependent protein kinase
II (Kassed et al., 2004) and amyloid presursor protein (Grilli et al., 1995). Thus, we
must conclude that although the striatal coexpressed genes all have the NF- motif,
the cause of the striatal specific expression must be associated with other factors.
However, as noted by Meffert and Baltimore (2005), there is increasing evidence that
the NF- family of transcription factors are involved in the regulation of neuronal
activity-dependent transcription and behavior.
25.6 CONCLUSIONS
The intent of this chapter was to provide the interested reader with an introduction
to the topic of genetics, brain dopamine systems and behavior. The focus was almost
entirely on data collected in mice; however, some data suggest that these results can
be extrapolated to other species, including man. Key findings which have been
replicated numerous times are (a) the marked genetic differences in behavioral
responses to dopamine agonists and antagonists, (b) the marked genetic variation in
at least some elements of the brain dopamine systemsnotably the D2 dopamine
receptor and the dopamine transporter and (c) the relationships between a and b.
The reasons for the receptor and transporter variation(s) are unclear but some strat-
egies for dissecting these remarkable genetic effects are presented.
ACKNOWLEDGMENTS
The author would like to thank the many students, fellows and collaborators that
have contributed in some way to the research described in this chapter. Special thanks
to Glen Rosen, Rob Williams and colleagues for the BXD recombinant inbred striatal
gene expression data. This work was supported in part by grants from the US Public
Health ServiceAA 11034, AA 13484 and MH 51372and support from the
Department of Veterans Affairs Medical Research Program. The Gene Network
WebQTL site featured prominently in the review was funded in part by MH P20-
62009 to Rob Williams.
REFERENCES
Chesler, E.J. et al., WebQTL: rapid exploratory analysis of gene expression and genetic
networks for brain and behavior, Nat Neurosci, 7, 485, 2004.
Cunningham, C.L., Localization of genes influencing ethanol-induced conditioned place pref-
erence and locomotor activity on BXD recombinant inbred mice, Psychopharmacol,
120, 28, 1995.
Dains, K., Hitzemann, B., and Hitzemann, R., Genetics, neuroleptic response and the orga-
nization of cholinergic neurons in the mouse striatum, J Pharmacol Exp Ther, 279,
1430, 1996.
Fink, J.S., Swerdloff, A., and Reis, D.J., Genetic control of dopamine receptors in the mouse
caudate nucleus: relationship on cataleptic response to neuroleptic drugs, Neurosci
Lett, 32, 301, 1982.
Fiorentini, C. et al., Nerve growth factor regulates dopamine D2 receptor expression of
prolactinoma cell lines via p75NGFR-mediated activation of nuclear factor -B, Molec-
ular Endocrinology, 16, 353, 2002.
Fowler, S.C., Zarcone, T.J., and Vorontsova, E., Haloperiodol-induced microcatalepsy differs
in CD-1, BALB/c, and C57BL/6 mice, Exp Clin Psychopharmacol, 9, 277, 2001.
Grilli, M. et al., Identification and characterization of a kappa B/Rel binding site in the
regulatory region of the amyloid precursor protein gene, J Biol Chem, 270, 26774,
1995.
Hitzemann, R.J. et al., On the selection of mice for haloperidol response and non-response,
Psychopharmacol, 103, 244, 1991.
Hitzemann, R. et al., Genetics and the organization of the basal ganglia, Int Rev Neurobiol,
7, 485, 1995.
Hitzemann, R. The regulation of D2 dopamine receptor expression, Mol Psych, 3, 198, 1998.
Hitzemann, R. et al., Dopamine D2 receptor binding, Drd2 expression and the number of
dopamine neurons in the BXD recombinant inbred series: genetic relationships to
alcohol and other drug associated phenotypes, Alcohol Clin Exp Res, 27, 1, 2003.
Hwang, C.K. et al., Dopamine receptor regulating factor, DDRF: a zinc finger transcription
factor, Proc Natl Acad Sci USA, 98, 7558, 2001.
Janowsky, A. et al., Mapping genes that regulate density of dopamine transporters and
correlated behaviors in recombinant inbred mice, J Pharmacol Exp Ther, 298, 634,
2001.
Jones, B.C. et al., Quantitative-trait loci analysis of cocaine-related behaviors and neurochem-
istry, Pharmacogenetics, 9, 607, 1999.
Kanes, S., Hitzemann, B.A., and Hitzemann R.J., On the relationship between D2 receptor
density and neuroleptic-induced catalepsy among eight inbred mouse strains, J Phar-
macol Exp Ther, 267, 538, 1993.
Kanes, S. et al., Mapping the genes for haloperidol-induced catalepsy, J Pharmacol Exp Ther,
277, 1016, 1996.
Kassed, C.A. et al., Injury-induced NF-B activation in the hippocampus: implications for
neuronal survival, FASEB J, 18, 723, 2004.
Kraus, J. et al., The role of nuclear factor kappaB in tumor necrosis factor-regulated tran-
scription of the human :-opioid receptor gene, Mol Pharmacol, 64, 876, 2003.
Lee, S-H., Minowa, M.T., and Mouradian, M.M., Two distinct promoters derive transcription
of the human D1A dopamine receptor gene, J Biol Chem, 271, 25292, 1996.
Lee, S-H. et al., Tissue-specific promoter usage in the D1A dopamine receptor gene in brain
and kidney, DNA Cell Biol, 16, 1267, 1997.
Lee, S-H. et al., Genomic organization and promoter characterization of the murine dopamine
receptor regulator factor (DRRF) gene, Gene, 304, 193, 2003.
Lipsky, R.H. et al., Nuclear factor kappaB is a critical determinant in N-methyl-D-aspartate
receptor-mediated neuroprotection, J Neurochem, 78, 254, 2001.
Madrigal, J.L. et al., Inducible nitric oxide synthase expression in brain cortex after acute
restraint stress in regulated by nuclear factor B-mediated mechanisms, J Neurochem,
76, 532, 2001.
Meffert, M.K. and Baltimore, D., Physiological function for brain NF-kappaB, Trends in
Neurosci, 28, 37, 2005.
Minowa, M.T., Characterization of the 5' flanking region of the human D1A dopamine receptor
gene, Proc Natl Acad Sci USA, 89, 3045, 1992a.
Minowa, M.T., Characterization of the 5' flanking region of the human D2 dopamine receptor
gene, Biochemistry, 31, 8389, 1992b.
Minowa, M.T., Minowa, T., and Mouradian, M.M., Activator region analysis of the human
D1A dopamine receptor gene, J Biol Chem, 268, 23544, 1993.
Minowa, M.T., Minowa, M.T., and Mouradian, M.M., Negative modulator of the rat D2
dopamine receptor gene, J Biol Chem, 269, 11656, 1994.
ONeill, L.A.J. and Kaltschmidt, C., NF-kappaB: a crucial transcription factor for glial and
neuronal cell function, Trends Neurosci, 20, 252, 1997.
Patel, N.V. and Hitzemann, R.J., Detection and mapping of quantitative trait loci for halo-
peridol-induced catalepsy in a C57BL/6J x DBA/2J F2 intercross, Behav Genet, 29,
303, 1999.
Phillips, T.J., Huson, M.G., and McKinnon, C.S., Localization of genes mediating acute and
sensitized locomotor responses to cocaine in BXD-By recombinant inbred mice, J
Neurosci, 18, 3023, 1998.
Qian, Y., Hitzemann, B., and Hitzemann, R., D1 and D2 dopamine receptor distribution in
the neuroleptic non-responsive (NNR) and neuroleptic responsive (NR) lines of
micea quantitative receptor autoradiographic study, J Pharmacol Exp Therap, 261,
341, 1992.
Qian, Y. et al., D1 and D2 dopamine receptor turnover and D2 mRNA levels in the neuroleptic
responsive (NR) and neuroleptic non-responsive (NNR) lines of mice, J Pharmacol
Exp Ther, 267, 1582, 1993.
Simpson, C.S. and Morris, B.J., Regulation of neuronal cell adhesion molecule expression
by NF-kappaB, J Biol Chem, 275, 16879, 2000.
Taylor, B.A., Recombinant inbred strains: use in gene mapping, in Origins of Inbred Mice,
Morse, H.C., Ed., Academic Press, New York, 1978, pp 428438.
Volkow, N.D. et al., Decreased dopamine-D(2) receptor availability is associated with reduced
frontal metabolism in cocaine abusers, Synapse, 14, 169, 1993.
Volkow, N.D. et al., Prediction of reinforcing responses to psychostimulants in humans by
brain dopamine D2 receptor levels, Am J Psych, 156, 1440, 1999.
Walker, J.R. et al., Applications of a rat multiple tissue gene expression data set, Genome
Research, 14, 742, 2004.
Yang, Y. et al., Three-amino acid extension loop homeodomain proteins Meis2 and TGI
differentially regulate transcription, J Biol Chem, 275, 20734, 2000a.
Yang, Y. et al., ZIC2 and Sp3 repress Sp1-induced activation of the human D1A dopamine
receptor gene, J Biol Chem, 275, 38863, 2000b.
Zhang, L., Miles, M.F., and Aldape, K.D., A model of molecular interactions in short oligo-
nucleotide array, Nat Biotechnol, 21, 818821, 2003.
26 Natural Genetic Variation

of Hippocampal Structures
and Behavioran Update
Hans-Peter Lipp, Irmgard Amrein,
Lutz Slomianka, and David P. Wolfer
CONTENTS
Abstract .................................................................................................................. 389

26.1 Introduction................................................................................................. 390
26.2 The Hippocampal Mossy Fiber Projection ................................................ 391
26.3 Selectively Bred Strains as Tools for Testing BrainBehavior
Relations ..................................................................................................... 392
26.4 Verifying Association between Traits: Genetic and
Phenotypic Correlations ............................................................................. 393
26.5 Solutions for the Linkage Problem ............................................................ 394
26.6 Extending the Behavioral Analysis: Correlations with Other
Hippocampus-Dependent Tasks ................................................................. 396
26.7 Searching for Non Hippocampal Behavioral Correlates ....................... 399
26.8 Conflicting Data ......................................................................................... 401
26.9 Emerging Hypotheses................................................................................. 402
26.10 Some More Problems................................................................................. 403
26.11 The Real World Test for Genetic BrainBehavior Relations: Natural
Selection and Evolution ............................................................................. 403
26.12 Conclusions and Outlook ........................................................................... 406
References.............................................................................................................. 408
List of Abbreviations in Text ................................................................................ 410
ABSTRACT
This chapter analyzes the natural genetic variability of the so-called intra/infrapy-
ramidal mossy fiber (IIP-MF) projection in mice and rats, which terminates upon
the basal dendrites of hippocampal pyramidal cells, and its relations to behavior.
The analysis included thus far the following steps: (i) identification of structural
traits sensitive to selective breeding for extremes in two-way avoidance, (ii) testing
389
the robustness of the associations found by studying individual and genetic corre-
lations between hippocampal traits and behavior, (iii) confirming causal relationships
by manipulating the structural variable in inbred (isogenic) strains, thereby elimi-
nating the possibility of genetic linkage, (iv) further ruling out the possibility of
spurious associations by studying the correlations between the hippocampal trait and
other behaviors known to depend on hippocampal functioning, (v) searching for
behavioral correlates of the IIP-MF variation in behaviors thought to be unrelated
to hippocampal function, (vi) testing the effects of natural selection on differential
IIP-MF distributions in naturalistic environments, and (vii), studying the variability
of the IIP-MF distribution in small mammals with differential lifestyles. Taken
together, the data imply that the hippocampus mediates a considerable number of
behaviorally distinct processes, a view in accordance with the connectivity of this
structure. The common denominator may be that variations of the IIP-MF entail
differences in the stability of parallel hippocampal processing.
The advantages of this approach are: (i) hippocampal traits can be manipulated
simply by controlled breeding (no manipulative approach can match this experimen-
tal finesse thus far); (ii) it identifies natural regulation sites for behavior; and (iii) it
offers an easy way to discover possible behavioral functions of the hippocampus.
The disadvantages are: (i) the study of the underlying causality depends on the
detection of genes responsible for the trait; (ii) the behavioral effects are typically
smaller than after invasive manipulations, and this entails a greater sensitivity to
environmental influences; (iii) correlative studies often face traditionally more skep-
ticism than invasive manipulations; and (iv) the rising costs of maintaining breeding
colonies of mice for classical behavior genetic studies is slowly suffocating this
approach.
26.1 INTRODUCTION
In any mammalian species, no brain is like another, and every individual behaves
differently. Some of this variability obviously is dependent on the environment. On
the other hand, a considerable portion must be caused by biochemical and structural
variations of the brain that originate from both exogenous influences and actions of
genes during brain development. This chapter will deal with the biology of traits,
meaning the study of individual structural or behavioral differences that lie within
the quantified normal range of a population, in contrast to abnormal, i.e., patholog-
ical, differences that are most obvious in so-called neurological mutants.
Since behavioral traits appear to be complex, they have been attributed to many
genes, more precisely, to mutated alleles, which were thought to contribute small
effects adding to a phenotype. More recently, gene-targeting techniques allow for
selective inactivation of genes thought to be important for memory and learning and
other aspects of behavior. Such studies have revealed a remarkable capacity of the
genome and the brain to compensate for targeted mutations. However, such com-
pensation is variable, and this variable penetrance of targeted deletions often results
in changes resembling a quantitative trait. Hence, it appears reasonable to assume
that at least part of the observed behavioral variability of genetic origin is not based
Natural Genetic Variation of Hippocampal Structures and Behavior 391
on the additive action of hundreds of genes but on the effects of a few major genes,
and, perhaps, on single locus effects.
A theoretical problem is whether such single locus effects could be specific for
behavioral talents and motivational traits. Since activation of genes is thought to
precede brain development, this would seem unlikely. Most genes regulating brain
development are probably affecting many different neuronal populations, and their
effects must thus be pleiotropic. Likewise, mutation of genes controlling the basic
physiology of neurons and glial cells must show pleiotropic effects. There is an
exception however. Genes that are activated late during development can modulate
only the final steps of brain differentiation,1 a period during which only a few
forebrain systems undergo their final formation. The known late-forming systems
include portions of the olfactory bulb, the hippocampal formation, in particular the
dentate gyrus, and parts of the cerebellum. Thus, late-acting alleles that influence
the connective organization, receptor regulation, and neurochemistry of these struc-
tures are likely to be fairly specific for learning, cognition, and motivation. Such
mutated alleles will cause behavioral changes not because of a particular nucleic
acid sequence but by virtue of timing. Another mechanism for behavioral selectivity
is possible if brain structures known to modulate behavior show morphological or
biochemical features not found elsewhere in the brain.
26.2 THE HIPPOCAMPAL MOSSY FIBER PROJECTION

A structure that shows both late differentiation and morphological/biochemical spec-
ificity is the dentate gyrus of the hippocampal formation. It contains a secondary
proliferation zone for neurons, remote from the normal periventricular matrix. The
neurons (the granule cells) send axons called mossy fibers (MFs) to the pyramidal
cell layer of hippocampal region (CA3), to the hilus of the dentate gyrus (CA4),
and recurrent collaterals to the granule cell layer (Figure 26.1). The axons to CA3
terminate in a layer above the pyramidal cells (the suprapyramidal mossy fiber layer,
SP-MF ), and in a much narrower band and more patchily on the basal dendrites,
the intra/ infrapyramidal mossy fiber projection (IIP-MF). Barber et al. first reported
genetic variation of the IIP-MF projection for inbred mouse strains.2 Since the
terminals of the mossy fibers can be visualized with clarity by means of Timms
stain, the area of their distribution on brain sections can be measured easily by means
of stereology or digital image analysis.
Structural differences between inbred mouse strains imply that variations of the
IIP-MF projection must have reduced adult plasticitya desirable feature for study-
ing structural variations related to behavior. One should notice that this reduced
plasticity appears to be more pronounced in the CA3 region, where the mossy fiber
axons terminate in highly specialized synaptic complexes. Recurrent collaterals of
the same mossy fibers making more simple synapses on granule and basket cells of
the dentate gyrus are known for reactive sprouting after injuries in rats and humans
and for lifelong growth in guinea pigs. Clearly, systems showing pronounced adult
growth and plasticity are less convenient if one wishes to discover hereditary cova-
riations between behavioral and brain traits. Thus, the IIP-MF projection to CA3
FIGURE 26.1 Schematic view of a cross-section of the rodent hippocampal formation. The
terminal fields of the hippocampal mossy fiber projection appear in black. Abbreviations:
CA1, CA3, and CA4, hippocampal subfields; FIM, fimbria fornicis; GC, granule cells; LM,
stratum lacunosum-moleculare; MOL, molecular layer of dentate gyrus; OR, stratum oriens;
PYR, stratum pyramidale; RAD, stratum radiatum; SGL, supragranular layer of dentate gyrus.
appears to be a promising candidate structure for testing whether its variability might
be related to hippocampus-dependent behaviors.
26.3 SELECTIVELY BRED STRAINS AS TOOLS

FOR TESTING BRAINBEHAVIOR RELATIONS
A methodological problem in finding behavioral correlates of a brain trait is selecting
appropriate behavioral tests. Unfortunately, they are numerous and can be varied
infinitely. Thus, finding correlations remains a matter of chance. A more economical
approach is to use mouse or rat strains that have been selectively bred for extremes
in a behavior thought to be governed by a particular brain region. There are many
examples of behavioral divergence after so-called artificial selective breeding (for a
short review see Lipp and Wolfer3). For example, rats have been bred for high vs.
low two-way avoidance conditioning, a behavior often improved by hippocampal
lesions. These strains, in this case the roman high and low avoidance rats,4 are useful
to decide quickly whether a particular trait (which must be known to show genetic
variation) differentiates after selective breeding. If so, it is a candidate for a brain
system underpinning the behavior; if not, the system investigated is probably not
relevant for causing individual differences in behavior. Investigations of a sample
of RHA and RLA rats showed, indeed, a significant difference in two of the three
MF projection fields, namely the size of the CA4 projection and the IIP-MF pro-
jection, which appeared enlarged in the poor avoiders (Figure 26.2).
These results confirmed a potential role of the IIP-MF variations, but posed the
question whether it might not be the size difference of CA4, which was somehow
FIGURE 26.2 Selective breeding for extremes in two-way avoidance learning of rats and
differentiation of the mossy fiber projections (modified after Lipp and Wolfer, 1995).3 Note
that the percentage of the IIP-MF in Figs. 26.3, 26.4, and 26.8 have been calculated in relation
to the entire area of CA3/CA4 as measured by manual planimetry used for the older studies.
In more recent publications, the use of digital morphometry permitted to measure only the three
mossy fiber subfields. Hence, Figs. 26.5, 26.6, 26.7, and 26.9 give only the percentage of IIP-
MF in relation to the size of the suprapyramidal mossy fiber layer, correcting in a simple way
for differences in hippocampal size. Since the size of the suprapyramidal MF projection is a
good covariate of the total area of CA3/CA4, the two indices are strongly correlated, however.
related to behavior, or another hidden covariate in the brainthe eternal question

in correlative brain research. Because selective breeding for a trait acts presumably
on more than one brain mechanism, the observed association might simply be
dependent on another unknown system in the brain that shows genetic covariation
with the MFs, the latter being irrelevant for the behavior. In the worst case, differ-
ences between selectively bred strains might be based simply on genetic drift. In
such a case, small breeding populations will become increasingly homozygous for
alleles. With bad luck, one line will then be homozygous for alleles facilitating
avoidance learning, and the other line for alleles impairing it, without any functional
relation to the selective breeding process. Hence, brainbehavior associations found
in selectively bred strains must be analyzed further in order to test whether the
association was spurious.
26.4 VERIFYING ASSOCIATION BETWEEN TRAITS:

GENETIC AND PHENOTYPIC CORRELATIONS
The most easy way to delineate potential correlations between behavioral and brain
traits is to study a sample of inbred mouse strains known to show differences in
these traits. Genetic correlations can then be computed between the strain means of
either trait. These values can be obtained from publications or can be assessed by
studying a group of individuals. In the case of the IIP-MF and two-way avoidance
learning, it was found that the rank orders of the two variables showed a very strong
negative correlation [r = 0.96, p < 0.01, (Figure 26.3a)]. Again, poorly avoiding
strains showed the largest IIP-MF projections, but there was no correlation with the
size of the other mossy fiber fields. Genetic correlations between strain means are
useful when the behavioral variable shows large epigenetic (environment-dependent)
variability, blurring the influence of the brain variable. Their weakness is that they
depend on a small number of strains that appear as data points in the correlations:
one outlier strain can annihilate an existing real correlation, or an unlucky choice
of strains can fake it.
This problem can be solved by studying phenotypic correlations, that is, corre-
lations using individuals in which both traits are measured. Two extreme strains in
terms of IIP-MF and two-way avoidance learning were intercrossed, namely DBA/2
and C3H and the resulting F2 generation of 51 animals was then tested and their
hippocampi morphometrized.4 In such a sample, other alleles potentially affecting
two-way avoidance or IIP-MF projections are scrambled during meiosis and there-
fore might mask a weak phenotypic correlation. Nonetheless, the analysis revealed
again a strong negative correlation [r = 0.82, p < 0.001, (Figure 26.3b)]. Thus, the
association between avoidance learning and IIP-MF as observed in rats and inbred
strains was confirmed again, and turned out to be insensitive against potentially
confounding alleles. In other words, the correlation was functionally robust. This
step of the analysis clarified also causal relations: the distribution of the IIP-MF is
established in the postnatal period between days 10 to 20 and remains rather stable
afterward. Because the variability of the IIP-MF was obtained experimentally (by
means of meiotic scrambling of alleles), and because the effects of this manipulation
predict adult behavior, one is no longer dealing with a simple correlation without
causal implications (for example length of arm vs. length of leg), but with a regres-
sion of behavior on a structural variable. Nonetheless, phenotypic correlations have
two weak points. One is genetic linkage (more precisely, linkage disequilibrium),
in which case a chromosomal segment may contain both an allele relevant for
avoidance learning and another one for the size of the IIP-MF projection. One may
note that this is also a problem in gene targeting.5 A more subtle point is that
producing hybrids between strains with behavioral extremes often results in hybrid
vigor: the filial generations outperform both parental strains. The reasons for hybrid
vigor are unknown, but it may improve complex learning behavior to a point where
it can mask strain-specific factors. See also Wolfer et al. for an example of this
problem in knockout mice.6
26.5 SOLUTIONS FOR THE LINKAGE PROBLEM

This issue cannot always be solved. The prerequisite is to have an inbred strain in
which the brain trait can be manipulated experimentally, or nearly identical substrains
that show differences in the brain trait. For the IIP-MF behavior relation, both
approaches were feasible. The IIP-MF distribution can be enlarged experimentally
by means of postnatal thyroxin injections during the critical growth period of the
FIGURE 26.3 Variations of the infrapyramidal mossy fiber projection and avoidance learning in mice (modified after Lipp et al., 1989).4 (a) genetic
correlation between the strain means of the IIP-MF projection and avoidance learning at the last day of training (day 5); (b) phenotypic correlation
between IIP-MF projection and avoidance learning in individual mice from an F2 cross between the inbred strains DBA/2 and C3H; (c) developmentally
Natural Genetic Variation of Hippocampal Structures and Behavior
induced variation of the IIP-MF by means of thyroxin injections and adult two-way avoidance learning.
395
MFs. This treatment was applied to pups of the mouse strain DBA/2, a good
avoidance learner with scanty IIP-MF projections. It resulted in considerable, par-
tially dose-dependent variability of the IIP-MF projection that remained strongly
and negatively correlated with the avoidance learning of the adult individuals [r =
0.75, p < 0. 00014 (Figure 26.3c)]. Because all animals of this strain are isogenic,
chromosomal linkage cannot account for the correlation between the two traits.
Rather, it appears that size variations of the IIP-MF predict two-way avoidance
learning, regardless of whether the structural variation has been caused by genetic
or epigenetic factors. A further confirmation was then found by the group of Crusio
reporting subline differentiation in radial maze learning (that is, other hippocampus-
dependent behavior, see below) in the mouse strain C57BL/6.7 This difference was
found to be associated with a difference in the extent of the IIP-MF projection. Since
the sublines can differ at only a few loci at best, these findings are a strong argument
against chromosomal linkage.
26.6 EXTENDING THE BEHAVIORAL ANALYSIS:

CORRELATIONS WITH OTHER HIPPOCAMPUS-
DEPENDENT TASKS
The studies in avoidance learning suggested that this talent was somehow dependent
on intrahippocampal size variations of the mossy fibers, but it was not clear whether
there might be a hidden variable in the brain outside the hippocampal formation that
was responsible for the learning differences. This variable might be concomitantly
activated by genetic factors or thyroxin fluctuations, both of them activating the
growth of the hippocampal IIP-MF. Such pleiotropy appeared not very likely, but
nonetheless made control studies necessary. Thus, samples of mice were tested for
other so-called hippocampal behavior, that is, lesion-induced impairment of learning
or exploration in spatial test situations. If they would show correlations with the IIP-
MF projection again, a spurious correlation with a nonhippocampal variable should
become an untenable hypothesis.
Correlations between the extent of the IIP-MF projection and performance
in several of these paradigms indeed were found. In most, superior performance
was now associated with extended IIP-MF projections. For example, mice with
extended IIP-MF projections made fewer working memory errors in a relatively
small radial maze8 (also see below). This was found in form of a genetic correlation
(Figure 26.4a) and after developmental stimulation of IIP-MF growth by means of
thyroxin (Figure 26.4 b, c). Likewise, an inbred rat strain with superior radial maze
performance showed enlarged IIP-MF projection, while another strain with smaller
IIP-MF was superior in multiple T-maze learning.9
Selective breeding for differential rearing (a form of mouse exploratory behavior
in the open field assumedly mediated by the hippocampus also) showed an associated
increase in the extent of the IIP-MF,8 and similar observations were made in mice
selectively bred for high and low open-field activity10 (see Figure 26.5).
FIGURE 26.4 (a) Genetic correlation between IIP-MF and radial maze learning (strain means S.E.M.); (b) variability of the IIP-MF projection as
observed after postnatal saline (sham) injections. Note that the total error scores are higher than those of the thyroxin-injected mice, and that a correlation
Natural Genetic Variation of Hippocampal Structures and Behavior
can be found; (c) reduced error scores of mice having received postnatal injections of thyroxin showing a similar correlation (modified after Lipp and
Wolfe, 1995).3
397
FIGURE 26.5 Activity and IIP-MF in mice selectively bred for differential open-field activ-
ity. Highly active mice show larger IIP-MF projections.10 Abbreviations: HI, two mouse lines
selectively bred for high activity in the open field, both of them pigmented; LO, two mouse
lines selectively bred for low activity in the open field, both of them albinotic; CTL, two
randomly bred mouse lines, except that one was selected for pigmented coat color, the other
for albinism.
In the Morris water maze, phenotypic correlations were found, too.11,12 However,
the variation of the IIP-MF was correlated not with the learning of the task, but with
the ability of relearning a new platform location, which was superior in mice with
extended IIP-MF projections (Figure 26.6). Curiously, the correlation appeared to
be better in the left hippocampus. Subsequent studies showed that the asymmetry
of the IIP-MF projection was contributing to this kind of performance as well; mice
with larger IIP-MF projections at left crossed over the former platform position more
during the first day of reversal, but reoriented faster toward the new platform location
the following day. The worst relearners were animals with small symmetrical IIP-
MF projections.11,12
These studies indicated that variations of the IIP-MF projections (or of an
intrahippocampally linked structural covariate) were probably associated with a
physiological process inside the hippocampus causing behavioral differences
in spatial learning and exploration. On the other hand, the results did not blend
well into unitary concepts of hippocampal function, such as spatial reference13 vs.
working memory.14 Depending on the task, they appeared to correlate once with
working memory, once with spatial reference memory, and seemed to be also
involved in behavioral flexibility and perhaps hemispheric lateralization.
FIGURE 26.6 Correlation between IIP-MF and escape performance from the Morris water
maze in the left hippocampus of 19 mice from a randomly bred stock generated by means of
a 4-way diallel cross. Day 5 is the second day after platform reversal. Note superior perfor-
mance of mice with enlarged IIP-MF projections. Along the ordinate typical swim paths
(modified after Bernasconi-Guastalla et al. 1994).12
26.7 SEARCHING FOR NON HIPPOCAMPAL

BEHAVIORAL CORRELATES
Initially, the term hippocampus-dependent behavior denoted a behavior sensitive to or
abolished by lesions of the hippocampus but not other brain structures. This distinction
has never been validated thoroughly because in most lesion studies, the number of
control structures (mostly neocortex) was limited. Nonetheless, it was adopted by many
studies using targeted disruption of genes and shortened to the term hippocampal
implying a category of behavior exclusively mediated by the hippocampus. The most
frequent example is swimming-navigation learning, which is impaired or abolished
after hippocampal lesions, and with it the ability to remember the location of or to
find a hidden target platform. Since hippocampal lesions usually spare the capacity of
finding a cued (visually marked) platform, it is assumed that this form of escape
learning is mediated by nonhippocampal structures and is thus frequently labeled as
nonhippocampal. This distinction is not logical, since there is no evidence that the
hippocampus is not participating in comediating nonhippocampal behavior, such as
visually cued water escape learning. The only safe conclusion is that an intact hippoc-
ampus is more important for complex learning than for a simpler form. In addition,
there is enough evidence that other brain regions can impair swimming navigation,
particularly its procedural components. Nevertheless, it has become current practice
to contrast behavioral effects of targeted gene deletions as hippocampal, usually with
the connotation cognitive, vs. nonhippocampal and noncognitive.
FIGURE 26.7 IIP-MF variations and strength of paw preference. (a) Mice with strong paw
preference (HI) show the largest IIP-MF projections, the smallest are found in mice with
weak paw preference. Random-bred controls (RND-2) lie in-between; (b) weak genetic (strain
mean) correlation between IIP-MF and strength of paw preference in inbred strains of mice
(modified after Lipp et al., 1996).15
FIGURE 26.8 IIP-MF variations and attack latencies toward an intruder into the home cage.
(a) Mice selectively bred for long attack latencies show larger IIP-MF projections;16 (b)
percentage of attacking males is higher in mouse strains with scanty IIP-MF projections
(modified after Guillot et al., 1994).17
With the appearance of IIP-MF variations as good markers for variations of

intrahippocampal physiological processing, it became possible to test whether the
hippocampus was also comediating some of the so-called nonhippocampal behav-
iors, which are thought to be nonlearned and nonspatial. Lipp et al.15 were testing
whether the IIP-MF distribution would correlate with paw preference of mice. They
found that selective breeding for strong vs. weak paw preference was associated with
larger IIP-MF in the mice with consistent paw preference (Figure 26.7) and that
asymmetries of the IIP-MF distribution were weakly but significantly correlated with
ipsilateral paw preference. Sluyter et al.16 found that selective breeding for short vs.
long attack latencies in the male intruder paradigm was associated with smaller IIP-
MF projections and found a similar genetic correlation in inbred strains17 (Figure 26.8).
26.8 CONFLICTING DATA

This short and incomplete review has focused mainly on our own studies and on
some data from the groups of Crusio and Schwegler. Other laboratories have reported
partially different results whose detailed evaluation, however, would go beyond the
scope of this chapter. There appears to be agreement that variations of the IIP-MF
correlate positively but sometimes weakly with exploratory behavior in mice18,19
Prior et al.20 found, by comparing two lines, that mice bred for high male aggression
levels had larger IIP-MF projections yet were more fearful and less explorative. On
the other hand, in the more exploratory line with smaller IIP-MF projections, the
size of the IIP-MF projection correlated positively with exploratory-like behavior
while no such within-line correlation was found in the aggressive animals. Such
strain dependencies of correlations of the IIP-MF projections were also observed in
other studies.21,22
On the other hand, there were studies reporting missing genetic correlations
between IIP-MF and radial maze learning abilities,23 and mouse strains with differ-
ential avoidance learning abilities did not show differences of the IIP-MF distri-
bution.24 Occasionally, size variations of the other MF fields also were associated
or correlated with behavioral scores,12,18,25 but these relations cannot be discussed
here.
The discrepancies found for radial maze learning may reflect procedural varia-
tions. Correlations as reported by Crusio and others were observed in a comparatively
small radial maze at ground level, while the negative reports from the laboratory of
Roullet and Lassalle23 were based on the use of a much larger elevated radial maze.
Possibly, such differences could entail a differential use of cues and favor differential
strategies, thus masking or unmasking participation of factors related to the IIP-MF
projection. Also, it should be kept in mind that IIP-MF variations are only one
cerebral trait among many others that might influence behavior.
In the past decade, there have been several reports of altered MF distributions
in genetically engineered mice. In many cases, these alterations appear to be prima-
rily markers of anomalous brain development, since the final adult distribution of
hippocampal MFs is the result of proper timing in generation, migration and axonal
growth of pyramidal cells, and, separately, of granule cells that are generated in an
ectopic proliferation zone in the hilus of the dentate gyrus. Obviously, behavioral
phenotypes of such mutant mouse lines may have multiple causations and it is
difficult to disentangle the possible contributions from the targeted systems and
altered MF projections. For example, mice lacking the mineral corticoid receptor
show a strongly extended IIP-MF projection and many behavioral symptoms of
hippocampal lesions, but it is difficult to judge whether their rather moderate impair-
ment in swimming-navigation learning reflects partial compensation by the IIP-MF
projection, or a minor role of the hormone receptors in this task. Apart from devel-
opmental changes being reflected in relative permanent changes of the MF distri-
bution (as observed in normal strain differences), mutant mice may show adult MF
sprouting due to epilepsy, apoptosis of target neurons and/or changes in adult neu-
rogenesis of the granule cells themselves that can produce both altered patterns of
mossy fiber distribution and behavioral changes. Finally, interpretations should be
made carefully when the changes in MF distribution in mutant mice are subtle and
have been assessed only in a limited portion of the hippocampus. The extent of the
IIP-MF decreases gradually from rostral (septal) to ventral (temporal) levels of the
hippocampus, while recurrent MF collaterals in the dentate granule cell layer are
abundant in ventral parts and almost missing in the rostral portion of the hippocam-
pus. Thus, subtle mutation-induced changes in hippocampal size and form can fake
a difference in MF distribution when compared at one hippocampal plane only.
26.9 EMERGING HYPOTHESES

The findings of correlations between IIP-MF and noncognitive behaviors did and
do not fit well into prevailing theories about hippocampal functions, which require
considerable bending to explain hippocampal contributions to paw preference and
attack behavior. A less theoretical explanation is simply to accept multifunctionality
of the hippocampus as it is reflected by the diversity of inputs and outputs at the
cortical and subcortical levels. Thus, one would assume that the behavioral function
of the dorsal hippocampus with its connections to the prefrontal cortex is related to
movement planning and spatial processing, while the ventral-most parts that project
to the ventromedial nucleus of the hypothalamus and the amygdala might be more
relevant to emotional/aggressive behavior. Such parallel processing of multiple
inputs and outputs requires temporal stabilization of activity as well as some form
of shielding against interference from neuronal activity in neighbored hippocampal
zones involved in different activities. Neuroanatomically, the simplest explanation
would be that extended IIP-MF projections provide better protection against inter-
ference arising from neighboring regions or other afferent subcortical and cortical
systems, simply because they occupy a prominent position in controlling the firing
behavior of the CA3 pyramidal cells, diminishing the impact of entorhinal and
intrahippocampal connections. Thus, activity in a given channel or segment along
the septotemporal axis remains confined according to the amount of MF terminals
on both apical and basal dendrites, IIP-MF reflecting the genetic and environmentally
more variable part of the CA3-projection. Thus, small mossy fiber projections with
entirely missing IIP-MF would characterize a hippocampus in which neuronal activ-
ity from any input segment could spread quickly through the entire hippocampal
formation, while massive suprapyramidal and IIP-MF projections would prevent
such spread.
At the behavioral level, individuals with small IIP-MF projections would be
expected to be hyper-reactive, quickly aggressing or fleeing, good two-way avoiders,
but inferior in complex tasks and memory because they are easily distracted. On the
other hand, individuals with large MF input to CA3 would be characterized by more
predictable (longer ongoing) behavior, regardless of what they are doing. For exam-
ple, they might be expected to use a preferred paw more insistently, showing
prolonged attack latency (as mice usually show an ethological foreplay before
launching an attack). They would also be expected to appear more explorative and
less anxious (because of better buffering or even ignoring moderate external or
motivational stimuli), and to perform better in complex learning and memory tasks
requiring a certain stress tolerance and concentration. On the other hand, they might
also show prolonged inappropriate responses such as in two-way avoidance learning

(Figure 26.3), or prolonged searching over old target locations in the water maze.
Taken together, genetic variations of the MF projection to CA3 would be a mecha-
nism for tuning a basic hippocampal function, namely, the control of parallel pro-
cessing in various parts of the forebrain that are all interconnected with the hippoc-
ampal formation. Lesions of the entire hippocampus must then appear as a
combination of both extremes, namely hyper-reactivity and hyperemotionality com-
bined with stereotyped and rigid behavior. This in fact is the case, at least in rodents.
26.10 SOME MORE PROBLEMS

This explanation has an intuitive appeal but poses another methodological problem:
how can this be tested behaviorally? Multifunctionality of the hippocampus also implies
that any change in this structure probably will be manifested in behavior, but the
outcome may depend on the behavioral context, genetic background, and other factors.
Another problem is to explain why genetic variation of the IIP-MF in mice
occurs at all. Having scarce IIP-MF projections equivalent to a mild hippocampal
lesion syndrome would seem an undesirable property for mice, and one is left
wondering why natural selection has not eliminated such an unhealthy feature,
even if it were only of partial importance for animal cognitive behavior. Nonetheless,
it occurs in a couple of mouse strains, and investigations of wild house mice
have consistently revealed individuals with fairly scanty IIP-MF projections (unpub-
lished observations by H.-P. Lipp and H. Schwegler).
A disturbing possibility is that MF variations somehow influence behavior, but
the behavioral tests applied to mice fail to measure biologically meaningful abilities
of the animalsan argument often proposed by ethologists. Likewise, it can be
argued that laboratory mice are unrealistic models because of their long domestica-
tion and behavioral degeneration and one might dismiss the observed correlations
as laboratory artifacts. One approach to solve this problem would be a broad screen-
ing for genes underlying MF variations, and trying to elucidate the path from gene
to behavior. But if the former arguments were true, such (costly) approaches would
be doomed for failure.
26.11 THE REAL WORLD TEST FOR GENETIC

BRAINBEHAVIOR RELATIONS: NATURAL
SELECTION AND EVOLUTION
The arguments above can only be answered by studying correlations between MF
variations and behavior in an ecological context. If MF variations are of any relevance
for natural behavior, they must respond to natural selection. Likewise, species with
different lifestyles, but comparable brains must show differences in their IIP-MF
projections. To study this issue, a field station was established in Russia in collab-
oration with behavioral geneticists.
In order to observe natural selection within a species, laboratory mice with

genetically different IIP-MF projections (C57BL/6, DBA/2, C3H and NZB) were
crossed in order to obtain a sample of hybrid mice characterized by large genetic
variability of this trait. These were released to two large outdoor pens and left there
for 2 years by now. Samples of mice were taken in yearly intervals and compared
with control mice from the same crosses that were kept under standard laboratory
conditions and random mating. An analysis after 2 years revealed that the hippoc-
ampi of the feralized mice were morphologically adaptating. In both pens, there was
a significant increase of the suprapyramidal MF projection, which was correlated
with a general increase in brain weight of about 5%. Independently of this increase,
there was a concomitant reduction of the IIP-MF projection (Figure 26.9).
Further follow-up over 4 years showed the divergence of the IIP-MF to be
consistent.26 Moreover, samples of mice caught at the end of the second year outdoors
were transferred to the laboratories (Zrich and Moscow) and bred there, the
reduction in the IIP-MF projections in Zrich being largely conserved in the seventh
generation, even after embryo transfer. Behavioral studies in Zrich showed no
parallel selection effects in the water maze and open field,27 but naturally selected
lines showed neophobia toward a novel object in an open field.28 Studies in
Moscow with the naturally selected lines showed increased defecation in the open
field, reduced activity in exploratory arenas after repeated exposures, and, interest-
ingly, an improved ability of extrapolating the movements of a reward toward a
hidden consumption site.26 Taken together, this study confirmed the prediction of an
association between reduced IIP-MF and behavioral reactivity, and showed, most
importantly, an extremely rapid natural selection effect on both hippocampal cir-
cuitry and behavior. Obviously, the association between IIP-MF and behavior is of
FIGURE 26.9 Reduction of the IIP-MF projection in two demes of feralized mice after 2
years in Russian outdoor pens as compared with control mice kept indoors. The time span
includes 7 to 8 generations of mice. Note that the reduction in pen 2 was observed after 1
year, while it took 2 years in the other pen.
ecological relevance. The mice had to live in the shelters throughout the year.
Intentional exploration of the pen territory (as expected for carriers with expanded
IIP-MF projections) bears a high risk of predation by owls, while the social structure
in the shelter (a few highly aggressive dominant males and many quickly escaping
subordinates) favors again high behavioral reactivity (either for flight or fight), at
least in males. As the lifestyle in these types of pens is not demanding in terms of
spatial abilities and complex learning, the lacking selection effects on performance
in classic cognitive tests such as the water maze is not unsurprising. Most likely,
other genetically dependent brain traits supporting the observed phenotype were
subject to selection as well. Nonetheless, the speed of genetic adaptation of the IIP-
MF projection implies that selection acted upon a few major loci only.
These results received support from studying the hippocampi of small wild
mammals in Russia and elsewhere. From the results of experimental natural selec-
tion, it could be predicted that small rodent species living in ecological niches
requiring high behavioral reactivity or readiness for flight would show small IIP-
MF projections. An example of such adaptation is given in Figure 26.10, which
shows large IIP-MF projections in the bank vole (Clethrionomys glareolus), a species
adapted to a wide range of different habitats, and extremely reduced IIP-MF pro-
jections in the root vole (Microtus oeconomus), living in homogeneous grassland
habitats with small home ranges. Interestingly, the two species showed about equal
performance in water maze learning but used very different strategies, the bank voles
showing controlled spatial searching behavior, the root voles searching inefficiently
but swimming at high speed, reacting to any change in the setup, such as platform
reversal, with frantic but ill-directed swimming.29 Other studies showed that wood
mice (Apodemus ssp.), known for roaming large territories, have large IIP-MF
projections as well, which were also found in patrolling insectivores (Sorex ssp.).
On the other hand, species living chiefly underground, such as the vole Microtus
subterraneus and the mole (Talpa europea), have shown much reduced or missing
IIP-MF projections.
FIGURE 26.10 IIP-MF projections in wild voles. (a) Root vole (Microtus oeconomus), living
in small monotonic habitats and showing rapid yet chaotic swim strategies in the water maze;
(b) Bank vole (Clethrionomys glareolus), a species capable of adapting to many different
habitats and showing predictable and efficient learning in the water maze (modified after
Pleskacheva et al. 2000).29
While these findings confirmed the prediction of MF variations underlying

behavioral reactivity, we also noted that rodents with relatively large IIP-MF pro-
jections had larger home ranges. Thus, a link of the size of the IIP-MF projection
with spatial abilities remained an alternative hypothesis. In order to test this question,
we were looking for a species with excellent spatial abilities and found it represented
by nectar-eating bats that must patrol large territories with constantly changing
locations of flowers delivering nectar. If large IIP-MF projections would be a pre-
requisite for mastery of spatial short- and long-term memory, these bats should show
them. If, on the other hand, behavioral reactivity would be the main factor associated
with MF variations, freely and rapidly flying species should be characterized by
scanty MF projections enabling them to react instantaneously to stimuli encountered
during flight. Thus far, three South American species have shown the smallest MF
projections among all species observed thus far.30 Pending further investigations,
this would imply that the behavioral dimension associated with the size of the MF
projections is indeed the degree of behavioral reactivity, that is, sensitivity to dis-
traction. If so, one would also expect that the species with the highest degree of
parallel processing and multitasking should show the most extreme adaptation of
the IIP-MF projection. Thus far, this seems to be the case in the human hippocampus,
where every pyramidal cell in CA3 receives both suprapyramidal and infrapyramidal
MF afferents (unpublished data). Further predictive comparative studies will be
necessary to confirm this conclusion. In any case, it is now almost certain that natural
selection is operating massively on MF traits, within and between species. Thus,
variations of the IIP-MF are not simply epiphenomena irrelevant for behavior but
may reflect gene actions important not only for the regulation of individual behavior,
but also for the evolutionary selection of traits relevant at the population level.
26.12 CONCLUSIONS AND OUTLOOK

This short review has shown that an analysis of natural genetic variation can lead
to new perspectives and hypotheses about hippocampal functions that would be
difficult to obtain by any invasive technique. Its power is based on the fact that a
correlative approach using available genetically defined mouse and rat populations
can identify in the brain natural regulation sites for setting permanent differences in
behavioral responsiveness and particular talents. Once discovered, such natural reg-
ulation sites can be used to study the function of a brain structure beyond the gross
deficits that follow its destruction or inactivation. There is probably no other tech-
nique by which brain structures can be varied so elegantly, and it is difficult to see
how an interfering technique, tissue destruction or gene targeting, could alter behav-
ior, memory and learning in such a natural way.
On the other hand, the goal of neurogenetics is to understand the pathway from
gene to brain and to behavior. The example here is a typical phenotype-to-genotype
approach. It models the end-product of the genotype-to-phenotype pathway quite
nicely and offers a relevant target to be investigated further. However, it lends itself
not easily to the discovery of candidate genes, perhaps best by using quantitative
trait locus (QTL) mapping techniques (see Chapter 5). Even then, it will be difficult
to identify by molecular techniques those major genes that determine the genetic
variation of the IIP-MF projection because these regulatory genes might act locally
and their products occur in low concentrations during a limited developmental
period. But it is not impossible.
Another point is that such a correlation analysis is time consuming, requiring
many control studies. This is because working with brain traits means working
with relatively minor behavioral differences as compared with the effects of struc-
tural destruction. Minor differences are more sensitive to environmental effects and
genetic interactions. One must notice, however, that the natural genetic differences
in behavior as found between mouse strains are often much larger than those
observed after targeted disruption of genes.31 Thus, the most straightforward
genotype-to-behavioral phenotype approach faces similar problems but tends to
overlook them.
Correlative approaches are often dismissed because of the difficulties in disen-
tangling pleiotropic effects of unknown genes causing the traits in brain and behavior.
These difficulties should not be denied. Anyone familiar with the effects of gene
targeting on memory, learning, and cognition is aware of the fact that there is no
way to avoid pleiotropy of the targeted mutation. Even worse, there is a common
belief that elimination of a particular locus is revealing the function of that gene.
What is really observed, however, is the function of the entire genome without the
targeted locus. Again, the basic problems remain always the samethey are just
perceived differently.
The unique strength of this approach is that it starts with individual, within-
species variations of brain traits and behavior that are sensitive to artificial selection
and manipulation by classical breeding and eventually helps predict species-specific
adaptations. This is a fundamental difference to the traditional comparative approach,
which tries to match a special talent of a species with the particular development of
a brain structure. While fruitful in linking sensory abilities to corresponding devel-
opment of brain parts, this strategy faces problems in matching cognitive abilities
with altered patterns of neuronal circuitry. It would seem that the example described
in this chapter is, thus far, the only rational approach in finding natural regulatory
sites for complex behavior that is not depending on luck and speculation.
Last, one would expect that such a simple and relatively cost-effective approach
might win more followers. Unfortunately, the popularity of the mouse as a main
species for genetic engineering has created a ballooning veterinarian industrial and
administrative complex imposing excessive (and often unnecessary) costs of mouse
care that threaten to suffocate traditional behavioral genetic methods in this species.
With steadily rising costs, the insight may grow that neither classical Mendelian
crosses nor artificial selection for behavioral traits needs pathogen-free conditions
and veterinarian management and supervision.
ACKNOWLEDGMENTS
This article was supported by Swiss National Science Foundation and the NCCR
Neural Plasticity and Repair.
REFERENCES
1. Lipp, H.-P., Genetic variability, individuality and the evolution of the mammalian-
brain, Behavioural Processes 35, 1933, 1995.
2. Barber, R.P., Vaughn, J.E., Wimer, R.E., and Wimer, C.C., Genetically-associated
variations in the distribution of dentate granule cell synapses upon the pyramidal cell
dendrites in mouse hippocampus, Journal of Comparative Neurology 156, 417434,
1974.
3. Lipp, H.-P., and Wolfer, D.P. New paths towards old dreams: microphrenology or the
study of intact brains in intact worlds. In Behavioural Brain Research in Naturalistic
and Semi-Naturalistic Settings: Possibilities and Perspectives, Alleva, E., Fasolo, A.,
Lipp, H.-P., Nadel, L. and Ricceri, L., eds. (Kluwer, Dordrecht, the Netherlands),
pp. 139, 1995
4. Lipp, H.-P., Schwegler, H., Crusio, W.E., Wolfer, D.P., Heimrich, B., Driscoll, P., and
Leisinger-Trigona, M.-C., Using genetically-defined rodent strains for the identifica-
tion of hippocampal traits relevant for two-way avoidance learning: a non-invasive
approach, Experientia 45, 845859, 1989.
5. Gerlai, R., Gene-targeting studies of mammalian behavior: Is it the mutation or the
background genotype? Trends in Neurosciences 19, 177181, 1996.
6. Wolfer, D.P.P., Mller, U., Stagliar-Bozizevic, M., and Lipp, H.-P., Assessing the
effects of the 129/Sv genetic background on swimming navigation learning in trans-
genic mutants: a study using mice with a modified beta-amyloid precursor protein
gene, Brain Research 771, 113, 1997.
7. Jamot, L., Bertholet, J.Y., and Crusio, W.E., Neuroanatomical divergence between
two substrains of C57BL/6J inbred mice entails differential radial-maze learning,
Brain Research 644, 352356, 1994.
8. Crusio, W.E., Schwegler, H., and Brust, I., Covariations between hippocampal mossy
fibres and working and reference memory in spatial and non-spatial radial maze tasks
in mice, European Journal of Neuroscience 5, 14131420, 1993.
9. Prior, H., Schwegler, H., and Dcker, G., Dissociation of spatial reference memory,
spatial working memory, and hippocampal mossy fiber distribution in two rat strains
differing in emotionality, Behavioural Brain Research 87, 183194, 1997.
10. Hausheer-Zarmakupi, Z., Wolfer, D.P., Leisinger-Trigona, M.-C., and Lipp, H.-P.,
Selective breeding for extremes in open-field activity of mice entails a differentiation
of hippocampal mossy fibers, Behavior Genetics 26, 167176, 1996.
11. Schpke, R., Wolfer, D.P., Lipp, H.-P., and Leisinger-Trigona, M.-C., Swimming
navigation and structural variations of the infrapyramidal mossy fibers in the hippoc-
ampus of the mouse, Hippocampus 1, 315328, 1991.
12. Bernasconi-Guastalla, S., Wolfer, D.P., and Lipp, H.-P., Hippocampal mossy fibers
and swimming navigation in mice: correlations with size and leftright asymmetries,
Hippocampus 4, 5364, 1994.
13. OKeefe, J., and Nadel, L. The Hippocampus as a Cognitive Map, Clarendon Press,
Oxford, 1978.
14. Olton, D.S., and Feustle, W.A., Hippocampal function required for nonspatial working
memory, Experimental Brain Research 41, 380389, 1981.
15. Lipp, H.-P., Collins, R.L., Hausheer-Zarmakupi, Z., Leisinger-Trigona, M.-C., Crusio,
W.E., Nosten-Bertrand, M., Signore, P., Schwegler, H., and Wolfer, D.P., Paw later-
alization and intra/infrapyramidal mossy fibers in the hippocampus of the mouse,
Behavior Genetics 26, 167176, 1996.
16. Sluyter, F., Jamot, L., Van Oortmerssen, G.A., and Crusio, W.E., Hippocampal mossy
fiber distributions in mice selected for aggression, Brain Research 646, 145148,
1994.
17. Guillot, P.-V., Roubertoux, P.L., and Crusio, W.E., Hippocampal mossy fiber distri-
butions and intermale aggression in seven inbred mouse strains, Brain Research 660,
167169, 1994.
18. Belzung, C., Hippocampal mossy fibres: implication in novelty reactions or in anxiety
behaviours? Behavioural Brain Research 51, 149155, 1992.
19. Roullet, P., and Lassalle, J.-M., Genetic variation, hippocampal mossy fibres distri-
bution, novelty reactions and spatial representation in mice, Behavioural Brain
Research 41, 6170, 1990.
20. Prior, H., Schwegler, H., Marashi, V., and Sachser, N., Exploration, emotionality, and
hippocampal mossy fibers in nonaggressive AB/Gat and congenic highly aggressive
mice, Hippocampus 14, 135140, 2004.
21. Dimitrieva, N.I., Gozzo, S., Dimitriev, Y., and Ammassari-Teule, M., Mossy fiber
distribution in four lines of rats: a correlative study with avoidance abilities and
excitability thresholds, Physiological Psychology 12, 3034, 1984.
22. Lipp, H.-P., Schwegler, H., Heimrich, B., Cerbone, A., and Sadile, A.G., Strain-
specific correlations between hippocampal structural traits and habituation in a spatial
novelty situation, Behavioural Brain Research 24, 111123, 1987.
23. Roullet, P., and Lassalle, J.M., Behavioural strategies, sensorial processes and hip-
pocampal mossy fibre distribution in radial maze performance in mice, Behavioural
Brain Research 48, 7785, 1992.
24. Hoffmann, H.J., Wenkel, R., Schicknick, H., Bernstein, H.-G., and Schwegler, H.,
Hippocampal mossy fiber distribution does not correlate with two-way active avoid-
ance performance in backcross lines derived from inbred mouse strains DBA/2 and
C3H, Brain Research 589, 171174, 1992.
25. Isgor, C., Slomianka, L., and Watson, S.J., Hippocampal mossy fibre terminal field
size is differentially affected in a rat model of risk-taking behaviour, Behavioural
Brain Research 153, 714, 2004.
26. Poletaeva, I.I., Pleskacheva, M.G., Markina, N.W., Perepiolkina, O.W., Scheffrahn,
H., Wolfer, D.P.P., and Lipp, H.-P., [Environmental habitat-related pressure: behav-
ioral alterations and morphological changes in the brain of the house mouse] Russian,
Ecologia 3, 231236, 2001.
27. Rissi, S. Moosfasersystem und Verhalten bei der Hausmaus nach Embryonentransfer.
M.D. thesis, University of Zrich, Zrich (2002).
28. Ceschi, A. Effects of experimental natural selection on exploratory activity and
anxiety in mice: assessment by means of three behavioral paradigms. M.D. thesis,
University of Zrich, Zrich (2004).
29. Pleskacheva, M.G., Wolfer, D.P., Kupriyanova, I.F., Nikolenko, D.L., Scheffrahn, H.,
Dell'Omo, G., and Lipp, H.-P., Hippocampal mossy fibers and swimming navigation
learning in two vole species occupying different habitats, Hippocampus 10, 1730,
2000.
30. Amrein, I. Functional and neuroanatomical correlates of adult neurogenesis in the
dentate gyrus of domesticated and wild rodents. Ph.D. thesis, University of Zrich,
Zrich (2004).
31. Lipp, H.-P., and Wolfer, D.P., Genetic background problems in the analysis of
cognitive and neuronal changes in genetically modified mice, Clinical Neuroscience
Research 3, 221232, 2003.
LIST OF ABBREVIATIONS IN TEXT

CA1, CA3 Subregions of hippocampus, CA stands for cornu
ammonis (Ammons horn)
CA4 Hilus of the dentate gyrus
IIP-MF Intra/infrapyramidal mossy fibers
MF Mossy fibers
QTL Quantitative trait locus
RHA Roman high avoidance
RLA Roman low avoidance
SP-MF Suprapyramidal mossy fibers
27 Expression and Brain

Structure: Black Boxes
between Genes and
Behaviors
Jeremy L. Peirce and Robert W. Williams
CONTENTS
27.1 Introduction................................................................................................. 412

27.2 Genes and Development............................................................................. 412
27.3 Brains and Behavior ................................................................................... 412
27.4 Dealing with Complexity ........................................................................... 413
27.5 Complex Trait Analysis and Quantitative Trait Loci (QTL) Mapping..... 413
27.6 The Value of Genetic Reference Populations............................................ 415
27.7 Microarray Data: The Cheapest, Highest Throughput Phenotype ............ 416
27.8 Microarray Data: Caveats from Complex Tissue ...................................... 417
27.9 mRNA Abundance QTLs........................................................................... 417
27.10 Cis-Acting QTLs and Trans-Acting QTLs ................................................ 418
27.11 mRNA Is Not protein................................................................................. 419
27.12 From QTL to Gene..................................................................................... 419
27.13 Data Sets in GeneNetwork......................................................................... 419
27.14 Genetic Correlation .................................................................................... 421
27.15 Clustering and Networks............................................................................ 423
27.16 Cluster Tree ................................................................................................ 423
27.17 Network Graph ........................................................................................... 423
27.18 More to Do with a Gene List: WebGestalt and Related Tools ................. 423
27.19 Back to the BrainsMouse Brain Library (MBL) ................................... 424
27.20 Putting the Pictures Together ..................................................................... 425
27.21 Allen Brain Atlas (ABA) ........................................................................... 426
27.22 GENSAT ..................................................................................................... 426
27.22.1 Example: Mapping Dentate Gyrus Volume................................. 426
27.22.2 Example: Ror2 and Msx2 as Candidate Genes for DGV13a ...... 428
27.23 A Side Note on the Problems of Specificity and Power........................... 429
27.24 Conclusion .................................................................................................. 430
References.............................................................................................................. 430
411
27.1 INTRODUCTION
The operation of the adult mammalian brain involves a large fraction of the
genome.1,2 Upward of 50% of the gene complement is expressed in even single
discrete brain regions such as the cerebellum, striatum, and hippocampus. Transcrip-
tional complexity in the brain is unusually high compared with other organs, and
roughly 40% of all genes are transcribed in two or more alternative forms.3 Devel-
opment adds another major axis of complexity. In small mammals such as mice,
assembling a brain is a beautifully choreographed dance that involves the prolifer-
ation, migration, differentiation, and interconnection of 75 million neurons,
25 million glial cells, and a web of 10 million or more blood vessel cells over a
30-day period. In humans, this process involves three orders of magnitude more
neurons and their supporting cast and takes a decade or two.4 It seems conservative
to estimate that 75% or more of the mammalian genome is engaged at some stage
of making or maintaining the brain.
The major topic of this chapter is the study the genetic influences on brain
organization at various levels. Variations in genes affect either transcription or
function, and both of these can influence brain structure. Brain structure, in turn,
affects behavior both in variations in gross organization and variation in the ways
circuits are formed and utilized. Studying each level of organization is valuable and
gives us the opportunity to treat transcription and morphology in the brain as
intermediate phenotypes and not simply a black box between brain and behavior.
27.2 GENES AND DEVELOPMENT

Most of what we have to say is based on data collected exclusively from sexually
mature adult mice. We recognize that genes act as part of transitory developmental
processes. Much of what we see in adults is therefore the downstream consequence
of embryonic expression. This is an important caveat to keep in mind. However, it
is equally important to keep in mind that variation in adult brain structure is the
final and most important product of development. The variation among brains of
adults can be exploited using genetic methods to lead us straight to the causative
genes, even when those genes have been in the full-off position for months or years.
In fact, the genetic analysis of adult brain variation is now becoming a reasonably
effective and objective way to define key developmental genes responsible for
variation in brain development, structure, and even function.
There are several important complementary large-scale research efforts, notably
GENSAT, that have examined prenatal and early postnatal gene expression patterns
in the brain. These resources can give us insight into these transitory developmental
expression patterns. In addition, many genes are likely to have selective and/or partially
redundant functions in different brain regions.5,6 At every level, of course, these genetic
effects will exist alongside and interact with complex environmental effects.
27.3 BRAINS AND BEHAVIOR

Differences in brain structure generate differences in behavior. Even a crude
measurement such as brain weight is associated with behavioral variation, most
Expression and Brain Structure 413
prominently when the comparisons are made among species.711 Of course, the more
specific the region, cell population, or synaptic subcircuit being compared, the more
specific the related behavior differences are likely to be. In songbirds, for example,
variation in the size of neuron populations in the system is correlated with various
features of song learning and production.12,13 A more specific relationship is the
variation in size of the infrapyramidal mossy fibers in the hippocampus, which
correlates with differences in open-field exploration and novelty-induced fear,14 as
well as with avoidance learning.15,16 Other aspects of hippocampal morphology also
have behavioral correlates,1720 a subset of which is likely to be causal. We expect
that numerous other structural variants will be linked to behavioral differences.
27.4 DEALING WITH COMPLEXITY

The most common approaches to studying the remarkable complexities of brain
development and cellular architecture are reductionist. Approaches are focused on
isolated molecules, single circuits, or nuclei in simple model systems. In the case
of mice, we often study effects of single-gene modifications on a single inbred
genetic backgroundoften that of the inbred mouse strain C57BL/6J. This approach
has led to many impressive stories and has provided numerous insights into key
mechanisms. However, reductionist approaches often provide only a partial picture
and will not necessarily provide results that are readily applicable to other systems.
This is one of the complaints leveled against mouse modelsthat we can cure a
single particular mouse of a disease, but that these results do not mean that we will
make equivalent progress in humans. The reason for the failure has nothing to do
with the particular mouse, but with the assumption that results from any single,
defined system will generalize well. Biological processes and the generation of
behavior are somewhat messy and contingent affairs, and behavior in particular is
often the result of complex nonlinear interactions that are hard to model or predict.
The phrase behavioral mechanism is in many cases an oxymoron. Reductionist
methods may provide unequivocal answers to specific questions in a specific context.
We may be able to understand the role of perturbations of Camk2a expression level
at one particular age, in one particular sex, and in one particular strain or species.
But how successful will this data make us in prediction of structure or function at
another stage, in another sex, and on a different genetic background? The unfortunate
answer is that our predictive powers are woefully limited. We certainly cannot predict
effects of gene knockouts with any assurance before we do the experiment, and once
we have results from one strain, we still usually cannot predict the effects of the
same allele in another strain. Results of this type may lead to high-impact papers,
but may not lead to real understanding of a biological system.
27.5 COMPLEX TRAIT ANALYSIS AND QUANTITATIVE

TRAIT LOCI (QTL) MAPPING
Complex trait analysis is an effective approach for studying sets and systems of
central nervous system (CNS) phenotypes. This approach has been commonly used
to study behavioral and pharmacological traits of relevance to drug and alcohol use
and abuse.2127 Complex trait analysis has also been applied to a wide variety of
CNS divisions and cell populations: total brain weight,28,29 neuron and glial cell
number,30 the size and structure of the hippocampus31 and dentate gyrus,32 olfactory
bulb,33 striatum,34 retina,35 neocortex,36 cerebellum,37 as well as cytoarchitectonic
fields including the barrel cortex.36,38
These and other quantitative trait locus (QTL) mapping studies are intermediaries
between purely reductionist tools such as single-background gene knockouts, and
large, human-genetic studies, which may have much lower power (certainly per
participant!) but whose results, hopefully, are more general. A typical mouse QTL
study falls somewhere between these extremes. It addresses the complexity of genetic
variation but limits its scope to improve the power to detect differences. When
applied to a genetic reference population and to molecular phenotypes, however,
QTL mapping provides an invaluable tool for building observations that can bridge
different levels of complexity.
Instead of studying a small number of isogenic animals or using a single animal
model, we usually study a segregating population that consists of 20 to 2,000
genetically unique individuals or strains. Studies of the genetic basis of common
diseases in human populations often exploit this approach. For example, much of
the population of Iceland has been harnessed to find common gene variants that
contribute to pervasive human diseases. Rodent geneticists have their own, much
simpler, version of a population-based approach that usually involves breeding two
inbred strains known to differ significantly in a trait of biological interest. These
parental strains are used to generate heterozygous F1 offspring. The F1s are used
to make an experimental crosseither an intercross (F2) or a backcross to one of
the parental strains (usually abbreviated N2). Both F2 and N2 populations are
genetically complexno two animals are alike. We therefore refer to these popula-
tions as a segregating cross because gene variants segregate following Mendels
laws. For example, there may be a twofold difference between the parental strains
in the number of dopaminergic neurons in the substantia nigra, horizontal cells in
the retina, or the volume of the hippocampal dentate gyrus.32 The question we would
like to answer with the F2 and N2 progeny is what set of gene variants contribute
to the variation among members of the cross. Can we track down the QTLs that are
jointly responsible for the genetic fraction of the variation?
A quantitative trait can generally be measured continuously, often on a ratio
scale. While an essentially dichotomous Mendelian trait potentially can be converted
into a quantitative measurement, most quantitative traits vary more continuously.
Generally, this is because multiple genetic and environmental factors affect the
phenotype. QTL mapping is a process in which we tease apart the individual and
joint contributions. We test whether there is a statistically significant association
between differences in genotype at a particular gene or marker and differences in
the phenotype. The question is, Is there an association between a trait, such as the
volume of the dentate gyrus, and any gene (a gene variant can act as a marker) or
genetic marker in our experimental cross? In the case of a Mendelian trait such as
albinism in mice or Huntingtons disease in humans, we will find an almost perfect
linkage between a gene variant (tyrosinase and huntingtin, respectively) and variation
in the phenotype.
In the case of QTL mapping when a marker is very near a gene that modulates
the trait, there will also be a comparatively high level of association between the
phenotype and the genotype, though far less than for a Mendelian trait. When a trait
is genuinely polygenic, however, a single QTL may account for only 5 to 10% of
the variance. In this case, no single marker will be highly predictive of phenotype.
A QTL map is a simple graph of the degree of association between chromosomal
intervals (usually on the x-axis) and trait values. These plots are usually presented
with degree of association (often given as likelihood ratio score (LRS), log of odds
(LOD), or log P) plotted on the y-axis as a function of chromosomal location.
A chromosomal region with a high degree of phenotypic association marks each
QTL. While we call this a genetic linkage or association, a significant linkage is
actually a directional claim of causality (though not necessarily of mechanistic
interaction): a QTL map is a search for the set of genuine genetic causes of variation
in phenotype, and in that respect is no different from a study of Mendelian mutations,
although loci with small effects on the phenotype are, of course, harder to detect.
Another way to think of a QTL map is as an association heat map by position,
generated when we fit a model that allows for one marker at a time to have an
influence on the phenotype. Considered this way, it is easy to imagine extending
the model to allow two genes and their interaction to have an effect on the phenotype.
This two-gene model, or pair scan, is valuable for finding pairs of QTLs with
significant interactions, with or without individual effects. Higher-order models are
easy to imagine, but they rapidly become prohibitive due to the large numbers of
cases needed to rigorously test more complex models.
Most of the common tools for QTL mapping, including pair scans, are integrated
into GeneNetwork (GN, www.genenetwork.org), a Web-accessible resource for sys-
tems genetics. Many of the examples in the rest of this chapter are drawn from this
resource and can be easily replicated and extended from any browser. GN includes
tools for QTL mapping and analysis of large-scale phenotypes and array data, a
variety of analyses utilizing genetic correlation, clustering, network building and
other exploratory tools, and extensive connections to external systems genetics
related tools. The remainder of this chapter will focus on the capabilities of this tool
set (and associated external data sets built by other investigators) and approaches to
analyzing morphological phenotypes.
27.6 THE VALUE OF GENETIC REFERENCE

POPULATIONS
For a highly heritable phenotype with little environmental noise, (noise in this sense
includes any nongenetic influence such as maternal environment and technical error
of the measurement) a standard F2 intercross (F2; stage marked on Figure 27.1)
offers an extremely efficient means of mapping genetic variation. The environmental
noise for a behavioral or physiological QTL can often be quite high in a single
measurement, however, and unfortunately, the uniqueness of each F2 animal means
FIGURE 27.1 Recombinant inbred (RI) breeding scheme. Breeding (RI) lines simply
requires sequential intercrosses of offspring, usually starting with inbred parental strains, until
the resulting offsprings are fully inbred (approximately F20).
that observations on a given F2 are limited. This includes genotype observations, so

each F2 population must be separately genotyped for analysis to be possible.
These issues are strong arguments for using recombinant inbred strains (RIs) to
map QTLs. RI strains (Figure 27.1) are generated by repeatedly intercrossing progeny
from an F2 intercross until the offspring are themselves inbred, with genomes made
up of patches of each parental genome. Since these strains breed true, observations
on different animals within the same strain share a common genome (common genetic
influences) and have independent environmental components. Averaging phenotypes
among several animals thus serves to reduce environmental noise. This is especially
important with relatively noisy measurements such as array-based mRNA pheno-
types, because it allows us to pool samples for each array and pool arrays for each
strain, all of which dramatically reduce the noise levels of our phenotypes.
In addition to pooling the same observations, using an RI strain set allows us
to compare different observations. If we measure two phenotypes using different
animals from the same strain set, the common factor between these measurements
is their genetic context. Therefore if the two phenotypes are correlated, we know
that the correlation must represent common genetic underpinnings of the two phe-
notypes. (This is known as a genetic correlation. GN provides several tools for
exploring this sort of correlation, which will be explored later.) The RI strains serve
as an effectively immortal genetic reference population (GRP) about which pheno-
typic data can be accumulated.
27.7 MICROARRAY DATA: THE CHEAPEST, HIGHEST

THROUGHPUT PHENOTYPE
The advent of microarrays gives us a unique opportunity to look at the abundance
of the entire steady-state mRNA populations in a tissue at once. Essentially this
means that we can perform from 10,000 to 1,000,000 measurements of mRNA level
in parallel, depending on the precise platform used. If we estimate that each array
costs $600 and we pool 4 arrays per strain (2 per sex), our cost is still only 6 cents
per phenotype, assuming use of the Affymetrix platform, which measures about
40,000 messages. Array costs continue to decline rapidly even as array quality
improves, making mRNA abundance the cheapest phenotype to acquire.39
27.8 MICROARRAY DATA: CAVEATS FROM COMPLEX

TISSUE
The variation in which we are most interested is that of expression levels of genes
and proteins. However, we cannot estimate per cell expression level. Instead we are
able to estimate the relative expression levels of many transcripts among genetically
different samples and the covariance of many transcripts with each other and with
other phenotypes among this same set of samples. Variation in cell composition will
be one cause of differences between samples. For example, ratios of GABA-ergic
and glutamatergic transcripts (Gad, Psd95, etc) will be under genetic control and
will vary either as a result of genetic cell autonomous expression differences, or as
a result of genetic differences in ratios of excitatory and inhibitory neurons. The
same will be true of neuron:glial ratios.40 The range of variation in expression levels
generally exceeds variation in cell composition, however, so this complication does
not typically invalidate simple observations on gross tissue. Because most of our
microarray observations are performed using genetic reference population (GRPs),
we can always relate QTLs and other relationships back to differences in cell
populations. The Mouse Brain Library, described later in this chapter, is an excellent
tool with which to characterize cell-type composition differences by strain. (This is
not simply a matter of eliminating a nuisance variable. Differences in cell population
ratios are an interesting phenotype with interesting potential correlates; for example,
variation in ratios of old and new neurons in the hippocampus would be of great
interest to researchers studying learning and memory.)
27.9 MRNA ABUNDANCE QTLs

When taken using a GRP or other mapping population, measurements of mRNA
abundance can be used as phenotypes for QTL analysis. This crucial insight into
the value of molecular phenotypes was made for proteins by Damerval and
colleagues41 seven years before Jansen and Nap resuggested it as an approach for
analyzing mRNA abundance using microarrays.42 In the five intervening years,
mRNA abundance has been measured in yeast,43 mice,4446 rats,47 humans,4850 and
even eucalyptus trees.51
While calculating significance thresholds for a single mRNA trait is equivalent
to calculating them for any other complex trait, assigning significance to mRNA
abundance QTLs from an array experiment with 39,000 or more mRNA abundance
traits is not yet well worked out. A False Discovery Rate52 (FDR) approach may be
helpful but may also be too conservative given our assumption that at some level
most transcripts will be genetically modulated by other factors in the genome. In
addition, it may be appropriate to apply different significance criteria to cis-acting
QTLs (those that modulate the transcript maps to the transcripts location on the
genome) and trans-acting QTLs, since the prior expectation that a cis-acting QTL
is real (given that it has mapped to the location of the transcript itself and that such
occurrences are at a rate much higher than chance in the population of transcripts)
is much higher than would be the case for a trans-acting QTL.
27.10 CIS-ACTING QTLs AND TRANS-ACTING QTLs

A useful way to categorize mRNA abundance QTLs is by whether they modulate
mRNA abundance in a cis-acting or trans-acting manner. To put it another way, is
a QTL that modulates a particular gene map near the gene itself? If so, it is likely
that a variation in the genes regulatory region is responsible for the variation in
gene expression and the presence of the QTL.
Two alternative interpretations prevent us from immediately declaring that we
have identified a gene responsible for a QTL when we find apparently cis-acting
QTLs. The first alternate interpretation is that a closely linked gene is exerting a
modulatory influence. The second alternate interpretation is Affymetrix specific. The
probes used in the Affymetrix mouse arrays were designed against public mouse
sequence, meaning their probes were designed to complement C57BL/6J mRNAs.
Single nucleotide polymorphisms (SNPs) and other sequence variants that happen to
occur within the probe sequence reduce the probe binding and may be detected as
cis-acting QTLs. Doss and colleagues53 applied a cis-trans test to apparent cis-acting
QTLs, reasoning that if the QTL were cis-acting the ratio of transcripts in the F1
would differ from 1:1. Using this test they were able to confirm the mode of action
for 64% of significant, apparently cis-acting QTLs.
There are, however, many apparent cis-acting QTLs. Plotting transcript and
QTL positions against each other yields a strong diagonal of cis-acting QTLs and a
number of off-diagonal, potentially trans-acting QTLs (Figure 27.2). In addition to
Position of transcript on Genome (Mb)
2250
1500
750
750 1500 2250

Position of QTL on Genome (Mb)
FIGURE 27.2 cis- and trans-acting QTLs in the BXD brain. This figure plots QTL position
(x-axis) vs. transcript genomic position (y-axis) using QTLs from the INIA M430 Brain data
set (PDNN transform, Jan. 06 version). The diagonal line indicates cis-acting QTLs (the QTL
maps to the same position as the transcript) while dots off of the vertical line indicate trans-
acting QTLs. The precise cause of the light, vertical bands of trans-acting QTLs (often called
trans-bands) is unknown, but caution should be taken interpreting them since they do not
replicate well between data sets.
the diagonal line of cis-acting QTLs, there are several vertical lines where there
appear to be a high concentration of trans-acting QTLs. Exactly what causes
these trans-bands is unclear, but they may relate to the action of one or more
transcription factors with many targets. It is interesting to note, however, that while
many cis-acting QTLs replicate between our RI array data and F2 array data in the
same tissue (whole brain, cerebellum), fewer trans-acting QTLs and only one trans-
band replicates.54
27.11 MRNA IS NOT PROTEIN

Generally speaking, our interest in mRNA abundance is not fascination with mRNA
per se but a desire to learn something about the status of proteins in the cell, with
the mRNA abundance standing in for protein abundance. Of course, it is important
to keep in mind that mRNA is not a protein and that while mRNA and protein levels
will generally co-vary together, the relationship between the two varies from tran-
script to transcript and is often distinctly nonlinear.
27.12 FROM QTL TO GENE

Over the past decade, defining a QTL interval given a reasonably heritable phenotype
has become largely routine. The WebQTL module of GeneNetwork has even reduced
the barrier to entry for routine QTL mapping to the effort required to input the
phenotype, at least for data gathered in RI strains. It is the next stage, determining
which gene or genes in the QTL interval are responsible for the influence of the
QTL that is not yet a routine process.
One approach to proceeding from QTL to gene is fine mapping, possibly fol-
lowed by positional cloning of the polymorphism. There are a number of approaches
to fine mapping, including generation of large intercross populations (F2s) and
congenic animals. This approach could indeed be made routine but involves a great
deal in the way of effort and resources.
Researchers may also aggregate phenotypic information in an attempt to intel-
ligently choose a candidate gene from among the tens to hundreds of possible genes
in the QTL interval. Techniques to further examine the candidate include analysis
of the phenotype of a knockout or transgenic pharmacological interventions to
produce phenocopy, evaluating predictions in another appropriate population, and
other modes of biological confirmation. Generally these methods are expensive in
terms of time and material, so choosing a good candidate is very important. It is
also, unfortunately, difficult.
GN is a valuable tool in the process of taking a phenotype from QTL mapping
to candidate analysis, and is particularly well suited to the pursuit of genes affecting
CNS morphology.
27.13 DATA SETS IN GENENETWORK

There are a wide variety of data sets currently in GenNetwork (GN), and the list is
constantly expanding. Because GN used the BXD RI strains as a GRP from early
on, the deepest data sets, especially where CNS morphology is concerned, are for
these strains. While the crown jewels of the GN data sets are obviously the extensive
microarray data, a valuable tool GN provides is easy access to published phenotype
data in a useful analysis environment for many common GRPs. Again, the BXDs
are the most thoroughly characterized with 830 published phenotypes in the database
at time of writing.
For the BXD strain set there are many CNS-related array data sets (all generated
using Affymetrix arrays, though GN easily supports data from other array platforms)
including data for the forebrain, hippocampus, striatum, cerebellum, and eye. (There
is also a data set for liver mRNA abundance generated using the Agilent array and
an Affymetrix array-based study of hematopoietic stem cells.) Each of these data
sets was generated using at least one array per sex per strain and from 2 to 4 age-
and sex-matched littermates pooled per array. We are currently in the midst of
expanding on this range of CNS-related tissues. Since the BXD strain set has recently
been expanded55 we can also considerably improve the power of our inquiries by
simply expanding our phenotyping effort to the new strains.
For the reader following along on the GN site, these data sets are named
according to six characteristics: the institution or group that funded them, tissue the
array was applied to, array type, array version, date this version of the data was
released, and normalization method. We generally provide pre-normalized and qual-
ity controlled array data using at least three normalization procedures: RMA, PDNN,
and MAS5. Choosing a good normalization to address with your research question
is well beyond the scope of this chapter, but normalization can substantially shift
the results of an array analysis, so care is necessary.
A primary advantage of GRPs is never having to genotype again. Over 10,000
SNP genotypes were assayed in many of our mouse GRPs using the Illumina SNP
genotyping platform, resulting in very high-density genotypes generated from those
SNPs that differed between strains. In the BXD strain set, for instance, there are
now 3,795 combined SNP and traditional simple sequence length polymorphism
(SSLP) markers that differ between the B6 and D2 strains!
While GN is already set up to analyze data for the AKXD, BXH, CXB,
AXB/BXA, and LXS RI strain sets, currently there are only two array data sets
available for non-BXD mouse RI strain sets. The first, of considerably more interest
for brain structure research, is an array data set for the CXB hippocampus. The
second is for mammary tumors in the AKXD strain set. The others are currently
limited to genotypes and published phenotypes. In addition, GN also has genotypes
and published phenotype data for the HXB/BXH rat strain set.
GN is an easily extensible framework, and adding the most simple data sets to
the system is simply a matter of importing phenotypes and genotypes. There are
currently several F2 crosses resident on the site, for instance. While they are not an
effectively immortal inbred population, like the RIs, integrating them into the GN
analytical framework affords us the opportunity to compare and confirm results
between crosses and to apply the GN tool set to their analysis. As we continue to
expand the variety of data sets available in GN, we look forward to the integration
of new tissues, crosses, and treatment conditions into the sites repertoire.
27.14 GENETIC CORRELATION

The use and interpretation of correlations is often regarded with suspicion by biol-
ogists with a strong mechanistic or molecular bent. But correlation is one of the
conditions for establishing causalitycausality without a directional arrow or guar-
antee of no intermediate steps, as it were. Biologists need to exploit and understand
sources of correlationnot shun them. The correlations between traits studied using
genetic reference populations are of a special type called genetic correlations. The
addition of the adjective genetic implies that the correlation has an underlying,
heritable basis. This is the optimistic interpretation of correlations computed in a
GRP for two traits for which we have a reasonably accurate strain, meaning that the
correlation is due to neither poor study design nor noise. Consider as an example,
the high negative correlation between prepulse inhibition (PPI), a phenotype thought
to measure some aspects of schizophrenia in humans,56 and the expression of the
norepinephrine transporter (NET or Slc6a2) in the striatum (Figure 27.3). This
correlation is remarkably tight (r = 0.8), but does it mean anything?
8.0
NET mRNA abundance (Striatum, RMA transform, 04/05)
Pearsons r=0.803 P=5.05E-06

+ + + Spearmans r=0.761 P=3.85E-05
7.9 14 24 32
+
7.8 27 +
+ 06
+ 15
19 + +
13+ 09
+ + 28
7.7 16 DBA/2J +
+
+ 08 11
22 +
31
7.6 +
+ C57BL/6J
+ 01
05
7.5
+
29
7.4
20 10 0.0 10.0 20.0 30.0 40.0 50.0
Prepulse Inhibition (PPI, BXD Published Phenotype 104399)
FIGURE 27.3 Prepulse inhibition and NET mRNA abundance. This scatterplot shows the
genetic correlation between prepulse inhibition (percent inhibition, BXD published phenotype
10399; x-axis) and expression of the norepinephrine transporter mRNA (NET, Slc6a2;
RMA transform 4/05, probe set 1447311; y-axis) in the striatum. Each point in this scatterplot
is the average value for an entire strain (BXD24 is labeled 24, etc.). This covariance is quite
significant. (p = 5 10-6) Relations like these can lead to novel hypotheses or be used to
test existing ones.
Let us examine some possibilities. The correlation between PPI and NET may
be caused by a true molecular and mechanistic interaction. If the transporter gene
has a strong allelic variant (e.g., a premature stop codon in one of the parents of the
BXDs), and if variation in PPI just happens to map near the NET locus on Chr 8
at 91.6 Mb, then we would have a good causal hypothesisthat variation in NET
protein causes correlation between these two traits (there are at least 73 B vs. D
SNPS in NET). In this case, however, there is absolutely no evidence that PPI is
modulated by a QTL on Chr 8.
An alternative explanation is that a cascade of protein intermediaries, some of
which happen to overlap, generates the correlation between these two traits. While
PPI and NET may overlap in this molecular network space, they may not be directly
associated. The correlation may be generated by common gene variants that are far
upstream of both PPI and NET. If this is the case, then NET and PPI are still in
some sense mechanistically coupled, but they do not have a direct causal relationship.
If we understood the molecular cascades and networks controlling both traits, we
could provide a compelling reason for the correlation, but it might not be possible
to induce a change in the PPI simply by changing NET expression as implied by
the graph. Over-expression of NET would not improve PPI scores or cure schizo-
phrenia.
A final alternative is that PPI and NET co-vary tightly but that this covariation
is entirely due to linkage disequilibrium. This source of correlation is a bit harder
to grasp. Linkage disequilibrium is a biological phenomenon, but it usually does not
interest nongeneticists. To understand this source of correlation, consider a pair of
very different QTLs that both just happen to be located on Chr 11 at 98 Mb. One
of these QTLs is actually NeuroD2, a gene apparently essential for the normal
development of the basolateral amygdala.57 Mutations in NeuroD2 can produce
fearless, mean mice. Right next to NeuroD2 is another key developmental gene,
namely growth hormone (GH). Imagine that strain X just happened to have alleles
at NeuroD2 and GH that lead to the production of small and unusually aggressive
mice and that strain Y just happened to have alleles that lead to gentle and large
mice. Imagine also that NeuroD2 and GH had many other independent downstream
molecular targets (which they do). Because these two genes are physically tied
together on Chr 11, these two mechanistically separate downstream networks
will also be tied together in any experimental cross between strains X and Y.
Linkage produces secondary effects that ripple through data sets. Disequilibrium
between modifiers that control separate biochemical networks will cause these
networks to appear to co-vary, leading to a mixture of mechanistic covariance and
linkage covariance.
Fortunately, at least for individual hypotheses where upstream modulators are
known, it is easy to control for linkage disequilibrium effects by stratifying the
correlation analysis by allele or genotype at the upstream modulator. In the example
above, we would stratify our analysis by allele at a marker near NeuroD2 and GH
to disambiguate correlation due to linkage disequilibrium from correlation due to
variation in individual networks.
27.15 CLUSTERING AND NETWORKS

GN includes several tools for examining the clustering and network properties of
groups of transcripts. Since the pathways involved in higher-order phenotypes are
arguably of even greater interest (and applicability to the human condition) than the
particular genes varying between parental strains, understanding the organization of
transcripts is valuable for understanding larger-scale genome organization.
27.16 CLUSTER TREE

The Cluster Tree tool takes selected traits and builds a hierarchical cluster tree
diagram from the distances, measured using 1-r where r is the Pearson prod-
uctmoment correlation, between pairs of traits, then by successively linking groups
of traits to create the hierarchy. The tool then computes the QTL map, using
permuted, genome-wide, adjusted p-values for each transcript and plots it as a
heat map for each transcript, allowing the user to easily compare QTL maps for
related traits.
27.17 NETWORK GRAPH

The Network Graph tool takes selected traits and displays them as nodes, with
correlations displayed as edges, given user-specified cutoffs. The user can choose
from Pearson or Spearman correlation coefficients as well as literature correlations.
Literature correlations are computed using the Semantic Gene Organizer (SGO)
software by Homayouni and colleagues.58 SGO uses latent semantic indexing (LSI)
to extract genegene relationships from MEDLINE titles and abstracts.
LSI represents the semantic structure of each MEDLINE entry as a vector in a
constructed word space and measures the degree of similarity between documents
(where genes are mentioned) as the angle between the vectors. Because a direct link
between genes in a particular document is not required to calculate the necessary
vectors, LSI is a useful tool for proposing novel genegene interactions.
27.18 MORE TO DO WITH A GENE LIST:

WEBGESTALT AND RELATED TOOLS
Not every analysis starts with a QTL. For physiological and molecular phenotypes,
we are frequently as interested in the genes that correlate with the phenotype as in
the regions of the genome containing genes that modulate the phenotype. A first
step in analyzing these lists of related genes is to order them thematically. GN allows
the user to select a list of genes from his or her shopping cart and export the list
automatically to a variety of tools including WebGestalt, a WEB-based GEne SeT
AnaLysis Toolkit. WebGestalt provides tools to organize a human or mouse gene
set by function, tissue expression, chromosome location, and publication co-occur-
rence. All of these tools are intended to layer themes on gene lists, which is a crucial
organizational task for microarray results. For most GN-derived gene lists, however,
the primary interest is functional relatedness.
WebGestalt provides four similar tools for gene set analysis by function. The
most commonly used is the gene ontology tree tool (GO Tree). GO Tree uses gene
ontology categories and compares the observed and expected fraction of genes
represented in the gene set across category, returning the gene set in a tree form
with over-represented areas marked. It is worth exploring broad categories that are
not over-represented because narrower subcategories may still have over-represen-
tation, especially in a large gene list. This can be easily done via the Enriched GO
DAG tool, which visualizes the GO Tree results as a directed acyclic graph. For
convenience this option is directly linked from GN but can also be accessed from
WebGestalt.
WebGestalt also provides three other tools for gene set analysis by function.
Users can fit gene sets to and perform over-representation analyses on KEGG (Kyoto
Encyclopedia of Genes and Genomes; www.genome.jp/kegg) or BioCarta pathways
(www.biocarta.com). In these cases over-representation means more genes (in the
gene set) that participate in a given pathway than would be expected by chance.
These options are particularly good for correlation analyses because genes in the
same pathway would be expected to have correlated expression levels. In addition,
users can look for over-representation of PFAM (www.sanger.ac.uk/Software/Pfam)
protein domains in their gene list.
27.19 BACK TO THE BRAINSMOUSE BRAIN LIBRARY

(MBL)
Behind most morphological measurements of the CNS are collections of high-quality
brain sections. The MBL29 is our way of making such sections accessible to a much
broader audience. MBL is a Web-based tool that contains a variety of mouse atlases
as well as high-resolution images of more than 800 brains from more than 180
genetically characterized strains of mice. These include several RI strain sets: 28
AXB/BXA, 34 BXD, 11 BXH, 13 CXB, and 57 LXS, and a wide variety of other
inbred strains.
GN already contains phenotypes for a variety of CNS morphological traits,
especially in the BXD strain set. These can be found in the BXD Phenotypes database
mentioned earlier, and come from a variety of sources. Well over 100 neuroanatom-
ical phenotypes taken from the MBL have been entered into GeneNetwork: hippoc-
ampal volume, dentate gyrus neuron number, neuron:glial ratio in the dorsal lateral
geniculate nucleus, neuron number in the basolateral amygdala, and volume of the
IGL of the cerebellum. These morphometric and stereological data sets can now be
compared and correlated with large behavioral and neuropharmacological data sets.
Does the cell population or volume of basolateral amygdala co-vary with important
behavioral traits or with the expression of NeuroD2 or GH? These types of complex
multiscale questions can now be addressed very rapidly because the relevant data
have been assembled along with appropriate analysis tools in GN.
If you are interested in collecting your own volumetric data using the MBL data
set, the MBL includes metadata on the age, sex, body, and brain weight for each
animalin addition to strain identificationand each of these terms can be searched
to select animals according to the users requirements. Images are available at a
variety of resolutions, depending on the slide. All slides are available at the base
resolution of 24.5 0.5 m per pixel in the XY plane. Along the z-axis, sections
were taken at a 150 m interval (300 m between sections on each slide, 2 slides
per case). Significantly higher resolution images of single brain sections have been
acquired at 4.5 m per pixel for more than 100 cases (look for the blue, hi-res
button), and 1 m/pixel images for the neocortex, hippocampus, and dorsal lateral
geniculate nucleus are underway.
Images can be used to calculate volumes using several free programs: NIH
Image, Scion Image, or Image/J. (Adobe Photoshop can also be used.) Details on
processing, imaging, and calibration using MBL images can be found on procedures,
pages, or in papers written based on this data, such as our analysis of hippocampal
volume.31
27.20 PUTTING THE PICTURES TOGETHER

While array analysis is useful for analyzing mRNA abundance in brain tissues, it
does not provide any information on the distribution of expression within that tissue
with respect to cell populations. In a homogeneous tissue this is not problematic,
but as discussed above, brain and even brain subregions are anything but homoge-
neous tissues. Fortunately there are two Web-based toolsets that do an excellent job
of filling this gap. The Allen Brain Atlas and GENSAT both provide high-resolution
images of expression in the brain. The ABA provides these for adult C57BL/6J
animals, while GENSAT uses embryonic, neonate, and adult FVB animals; both
plan to provide data for nearly all brain-expressed genes.
There are naturally many uses for these data. On the most basic level, presence
of expression in the same brain region as indicated in an array result suggests the
array is correctly identifying the presence of the gene. While all current array data
in GN was gathered using adults, this sort of comparison can eventually be made
using animals age-matched to the GENSAT samples. This basic comparison is
valuable for analyzing candidate genesif a gene is not expressed in the target
region, it is less likely to be a good candidate.
Putting these expression assays together with MBL images to analyze gene
expression in a cell-type specific manner is another exciting use of these databases.
Since array expression results are based on the ratio of one mRNA species to the
total mRNA population, it is an important caveat that morphological variations
include variation in fractions of a given cell type within a structure. In a GRP we
can untangle this relationship by directly measuring morphological variations in cell
type fraction in the MBL images. Then, if we assume that the cell population type
in which genes are expressed is constant, we can normalize the expression ratios
from the array by the cell type ratios from the morphological data. This will allow
us to make useful statements about cell type populations within morphological
regions without the need to purify single cell types for array analysisan expensive
and difficult proposition.
27.21 ALLEN BRAIN ATLAS (ABA)

The ABA (www.brainatlas.org) is a Web-based application that allows researchers
to visualize the in situ hybridization patterns for genes expressed in the brain. The
ABA is the first project of the Allen Institute for Brain Science (AIBS) and will
soon include each of the 24,000 genes that AIBS researchers believe is expressed
in the brain. The in situ data can be searched by expression in a number of brain
compartments as well as by intensity and coverage of that expression. Keep in mind,
however, that all of these gene expression patterns are specific to the C57BL/6J mice
and do not address variation along the genetic axis.
The ABA is a fantastic tool for determining whether QTL candidate genes are
indeed expressed in the brain region for which they are expected to modulate
expression. While it is possible that the gene responsible for a QTL may exert its
influence from some other brain region, it is likely that genes expressed in a brain
region are better candidates for QTLs that are related to a particular compartment.
The ABA is also valuable as an internal control on quality-of-brain-region
preparations. If you are generating array data, it may be helpful to examine signature
expression profiles for your chosen region. In other words, if a gene is unexpressed
in your tissue of interest but highly expressed in a neighboring tissue, assaying for
the gene expression is a useful way of quality-controlling your dissection protocol.
27.22 GENSAT
One other mouse imaging tool worth mentioning is the GENSAT database59
(www.gensat.org), which contains a mouse gene expression database for embryos
(E15.5), neonates (P7), and adult animals. GENSATs expression model is based on
bacterial artificial chromosomes (BACs) in which coding sequences were replaced
by the enhanced green fluorescent protein (EGFP). Gene expression here is a relative
and somewhat more indirect measure, since the EGFP mRNA and protein may have
different stability than the gene it replaces. GENSAT uses FVB mice to generate its
transgenic embryos, so, like ABA, it is not designed to investigate a genetic axis.
The ability to introduce multiple copies of the BAC enhances the ability to detect
genes at low levels, however, and may be more sensitive than in situ methods. When
adult expression is of primary interest, GENSAT is a useful resource to check, either
to confirm expression observed in the ABA or for when the ABA shows no expres-
sion and there is reason to suspect that the actual expression level may simply be
low. Of course, when embryonic or neonatal expression is crucial, as will often be
the case with morphological characteristics, GENSAT is an extremely important
resource.
27.22.1 EXAMPLE: MAPPING DENTATE GYRUS VOLUME

A sample QTL map from the WebQTL module of GeneNetwork (RecordID/10460)
for dentate gyrus volume in the BXD RI strain set is shown in Figure 27.4. The
trace of association (given as LOD) between genotype and phenotype is given.
The p = 0.05 genome-wide adjusted significance threshold (adjusted for the many
nonindependent tests for association between marker and phenotype) is shown as a
line near the top of the graph. The significance threshold is derived from 1,000
permutations by the method of Churchill and Doerge.60 For each permutation the
relationship between animal ID (strain for RI populations) and phenotype is ran-
domized, the permuted QTL map is generated, and the highest LOD score from
each map is recorded. The ordered list of permuted LOD scores serves as the
reference for genome-wide adjusted p-value thresholds. Note that the trace of
marker-phenotype association only exceeds the genome-wide p = 0.05 threshold
once, on chromosome (Chr) 13. This locus was named DGV13a when it was
discovered32 to indicate its effect on dentate gyrus volume, its location on Chr 13,
and that it was the first such QTL to be identified on that chromosome. Having
identified a QTL affecting a morphological trait, the next logical question is how to
define the interval of the QTLthe range of positions within which we expect to
find the gene responsible for the QTLs presence. We know the QTL is on Chr 13,
and we can use the GN Map Viewer to generate a view of this chromosome only
(Figure 27.5). This expanded view highlights several useful functions of GN. The
bars indicate the frequency that, using a bootstrapped sample, the maximum position
of the QTL was in the area indicated. This is a method for defining the QTL interval.
A more widely used method is to follow the QTL association trace from its highest
point to the nearest point 1 LOD down to the left and right. Dupuis and Siegmond61
discovered, however, that a 1.5 LOD drop is closer to a 95% confidence interval.
This method estimates that the gene causing the effect of DGV13a is likely to lie
between 46 to 64 Mb on Chr 13.
4.2 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 x
2.0
Trait ID: BXDPublish: 10460 LOD
Interval Mapping for Dataset: BXD, mapping on All Chromosome Addititive Effect Frequency of the
Significant LOD = 3.94 Peak LRS
Using Haldane mapping function with no control for other QTLs Suggestive LOD = 2.29
3.0
LOD
2.0
1.0
25
50
75
100
125
150
175
25
50
75
100
125
150
175
25
50
75
100
125
150
25
50
75
100
125
150
25
50
75
100
125
150
25
50
75
100
125
150
25
50
75
100
125
25
50
75
100
125
25
50
75
100
25
50
75
100
125
25
50
75
100
25
50
75
100
25
50
75
100
25
50
75
100
25
50
75
100
25
50
75
25
50
75
25
50
75
25
50
25
50
75
100
125
150
Megabases
FIGURE 27.4 Whole genome QTL map of dentate gyrus volume. This figure shows a typical
QTL map for a morphological phenotype. Dentate gyrus volume has one significant QTL on
Chr 13. The line near the top indicates genome-wide adjusted p = 0.05 threshold.
Click to view the corresponding section of the qenome in an 8x expanded WebOTL map
Click to view the corresponding section of the qenome in the UCSC Genome Browser
Click to view the corresponding section of the qenome in theEnsemnl Genome Browser
Chr 13
4.2
4.0
3.0
LOD
2.0
1.0
0 10 20 30 40 50 60 70 80 90 100 110
Megabases
FIGURE 27.5 Chr 13 QTL map of dentate gyrus volume QTL DGV13a. This is a closeup
of DGV13a. Features are discussed further in the text and include the SNP density map at
the bottom, bars indicating the frequency with which the best QTL position was in that position
in a bootstrapped population, and the light trace indicating the additive effect of the marker
on the phenotype. A B6 allele increases dentate gyrus volume.
27.22.2 EXAMPLE: ROR2 AND MSX2 AS CANDIDATE GENES

FOR DGV13A
There are many genes within the 18 Mb DGV13a interval, and searching for the
gene or genes responsible for the QTLs effect is more art than science. There are
two major complementary approaches: narrowing the interval and building evidence
for a particular gene. Narrowing the QTL interval reduces the number of potential
candidates and generally involves more genetic approaches, such as building con-
genics or phenotyping additional animals that are recombinant across the QTL
interval. Building evidence for a particular candidate is trickier given the formidable
number of possibilities and the capacity of the human mind for constructing plausible
stories.
That said, our favorite candidates for DGV13a are Ror2 and Msx2. Ror2 is a
receptor tyrosine kinase expressed in the developing nervous system62 as well as
the branchial arches, heart, and limb/tail bud.63 The gene is located from 51.673 to
51.845 Mb, right in the center of the DGV13a interval, and, crucially, in a region
with a considerable number of SNPs. A glance at the SNP track at the bottom of
Figure 27.5 will indicate how unevenly the SNPs between B6 and D2 are distributed.
This is a fairly typical distribution, and if we make the simplifying assumption that
a gene with no SNPs is not a likely candidate, we can immediately eliminate a
substantial fraction of the genes in the interval. In this case, Ror2 has 289 SNPs,
though none of these are non-synonymous (the apparent mis-sense mutation,
mCV23266532, is actually in intron 1).
Ror2 regulates transcription of MSH-like homeo box 2 (Ms 2) by sequestering
Maged1 (Dlxin-1) from the membrane.64 Since Ms 2 is located very near Ror2,
we might expect this to result in what would appear to be a cis-acting QTL. While
Ror2 seems to be expressed in adult tissue according to the ABA, it has extremely
low expression, which is hard to distinguish from noise, in our hippocampus data.
This suggests that the effect of the gene may occur during development. This
7.4
(Hippocampus, RMA transform, 12/05)

Pearsons r=0.537 P=1.81E-03
Spearmans r=0.494 P=4.90E-03
+
7.3 C57BL/6J
Ms2 mRNA abundance

+
40
7.2
+ + +
+38 13 +19
+ 31
39 + 15 + +
7.1 + + + 12 + + 05 22
27 28 36 32 42 +
+ + + +
+ + 33 +24 + + 08
29 23 DBA/2J 14 01
7.0 16 + 09
+ 20
34+
02 +
11
6.9
6.8
3.0 3.5 4.0 4.5 5.0 5.5 6.0
Dentate Gyrus Volume (mm3)
FIGURE 27.6 Correlation of Ms 2 mRNA abundance and dentate gyrus weight. This
scatterplot shows the genetic correlation between Ms 2 (probe set 1449559, BXD hippoc-
ampus data) mRNA abundance and dentate gyrus weight. (BXD published phenotype 10460).
suggestion is bolstered by the expression of Ror2 in the early nervous system.62 The
peri-natal lethality of a Ror2 knockout63 and the association of Ror2 with dwarfism,
cyanosis, and short limbs and tails also underscore its importance in development.
Ms 2 interacts with Ror2, and its expression is strongly correlated with dentate
gyrus volume (Figure 27.6; Pearsons r = 0.53; p = 0.0018), which make it a
compelling candidate in the same pathway. Ms 2 is located from 52.026 to 52.031
Mb on Chr 13 and has an even higher SNP density per Kb than Ror2. Maged1,
which mediates the interaction between Ms 2 and Ror2, is located on Chr X and
has 0 SNPs, so while it interacts with our favorite candidates it is not itself a
candidate.
27.23 A SIDE NOTE ON THE PROBLEMS OF

SPECIFICITY AND POWER
We discussed above a significant QTL called DGV13a, so named because we found
a significant QTL for dentate gyrus volume, but not a significant QTL near that
position for volume of the pyramidal cell layer or other aspects of hippocampal
morphology. However, lack of detection in a QTL analysis does not imply any
definite proof of effect absence.
In addition, it is entirely possible that a QTL may have a large effect on the
dentate gyrus and a small effect in nearby areas, so that in a larger population its
effect size would be sufficient to generate a QTL independently. In this case the
QTL has a real effect on multiple areas yet is likely to be named as affecting only
one. When the areas being analyzed are overlapping rather than neighboring any
QTL affecting a subregion, the situation will seem similar but will have a different
underlying cause. Unless balanced by a corresponding, opposite effect, a QTL
affecting a smaller structure should be picked up by a sufficiently sensitive experi-
ment analyzing a larger structure that overlaps the smaller structure.
Also, at least with the smaller BXD RI strain set, (this is to some extent
ameliorated by the larger, extended BXD strain set) GN experiments are underpow-
ered for detection of QTLs of small to medium effect. This is not to say that one
should not identify a QTL by the designation of the least restrictive population it
effects or by the region in which it has the largest effect size if that is where it
demonstrated significance. This still seems appropriate in the absence of biological
understanding of the QTLs effect. It does, however, suggest liberal interpretation
of candidate-gene evidence regarding the genes effect in larger areasand cautious
interpretation of the importance of having selectively identified a subregion QTL.
27.24 CONCLUSION
The major purpose of understanding biological processes is to be able to make
reliable predictions and to modulate outcome in a logical and controlled way. To
achieve this level of understanding, we must understand complex systems and
networks of phenotypes. We need to understand how combinations of gene variants,
networks of molecules, and neuronal circuits work together in different individuals
and in many different environments to achieve functional equilibrium. We need to
adopt holistic and integrative approaches to experimental design that would have
been completely impractical a decade ago. The advent of sophisticated statistical
and computational methods and the advent of high-throughput methods to acquire
hundreds of thousands of genotypes and phenotypes are rewriting the rules of design.
The scientific and statistical chaos that followed the introduction of microarrays is
just the technical spearhead of this new experimental order. We should not be
surprised if there has been a backlash on the part of those scientists more accustomed
to juggling one ball at a time. We are now living through an awkward transition that
is moving us in the direction of systems biology for good reason: only this approach
will enable us to make reliable predictions of complex outcomes. Animals and
humans are not machines. We are stochastic, fluid, willful, and richly redundant
cybernetic systems.
REFERENCES
1. Adams MD, Soares MB, Kerlavage AR, Fields C, Venter JC: Rapid cDNA sequencing
(expressed sequence tags) from a directionally cloned human infant brain cDNA
library. Nat Genet 1993, 4(4):373380.
2. Sutcliffe JG: mRNA in the mammalian central nervous system. Annu Rev Neurosci
1988, 11:157198.
3. Yeo G, Holste D, Kreiman G, Burge CB: Variation in alternative splicing across
human tissues. Genome Biol 2004, 5(10):R74.
4. Williams RW, Herrup K: The control of neuron number. Annu Rev Neurosci 1988,
11:423453.
5. Gautvik KM, de Lecea L, Gautvik VT, Danielson PE, Tranque P, Dopazo A, Bloom
FE, Sutcliffe JG: Overview of the most prevalent hypothalamus-specific mRNAs, as
identified by directional tag PCR subtraction. Proc Natl Acad Sci USA 1996,
93(16):87338738.
6. Usui H, Falk JD, Dopazo A, de Lecea L, Erlander MG, Sutcliffe JG: Isolation of
clones of rat striatum-specific mRNAs by directional tag PCR subtraction. J Neurosci
1994, 14(8):49154926.
7. Lashley K: Persistent problems in the evolution of mind. Quart Rev Biol 1949,
24:2842.
8. Wimer C, Prater L: Some behavioral differences in mice genetically selected for high
and low brain weight. Psychol Rep 1966, 19(3):675681.
9. Keverne EB, Martel FL, Nevison CM: Primate brain evolution: genetic and functional
considerations. Proc Biol Sci 1996, 263(1371):689696.
10. Aboitiz F: Does bigger mean better? Evolutionary determinants of brain size and
structure. Brain Behav Evol 1996, 47(5):225245.
11. Rensch B: Increase of learning capability with increase of brain size. Amer Natur
1956, 90(8195).
12. Ward BC, Nordeen EJ, Nordeen KW: Individual variation in neuron number predicts
differences in the propensity for avian vocal imitation. Proc Natl Acad Sci U S A
1998, 95(3):12771282.
13. Devoogd TJ, Krebs JR, Healy SD, Purvis A: Relations between song repertoire size
and the volume of brain nuclei related to song: comparative evolutionary analyses
amongst oscine birds. Proc Biol Sci 1993, 254(1340):7582.
14. Crusio WE, Schwegler H, van Abeelen JH: Behavioral responses to novelty and
structural variation of the hippocampus in mice. II. Multivariate genetic analysis.
Behav Brain Res 1989, 32(1):8188.
15. Schwegler H, Lipp HP: Hereditary covariations of neuronal circuitry and behavior:
correlations between the proportions of hippocampal synaptic fields in the regio
inferior and two-way avoidance in mice and rats. Behav Brain Res 1983, 7(1):138.
16. Lipp HP, Schwegler H, Heimrich B, Driscoll P: Infrapyramidal mossy fibers and two-
way avoidance learning: developmental modification of hippocampal circuitry and
adult behavior of rats and mice. J Neurosci 1988, 8(6):19051921.
17. Lipp HP, Collins RL, Hausheer-Zarmakupi Z, Leisinger-Trigona MC, Crusio WE,
Nosten-Bertrand M, Signore P, Schwegler H, Wolfer DP: Paw preference and intra-
/infrapyramidal mossy fibers in the hippocampus of the mouse. Behav Genet 1996,
26(4):379390.
18. Lipp HP, Schwegler H, Crusio WE, Wolfer DP, Leisinger-Trigona MC, Heimrich B,
Driscoll P: Using genetically-defined rodent strains for the identification of hippoc-
ampal traits relevant for two-way avoidance behavior: a non-invasive approach. Expe-
rientia 1989, 45(9):845859.
19. Schwegler H, Crusio WE, Lipp HP, Heimrich B: Water-maze learning in the mouse
correlates with variation in hippocampal morphology. Behav Genet 1988, 18(2):153165.
20. Lipp HP, Schwegler H, Heimrich B, Cerbone A, Sadile AG: Strain-specific correla-
tions between hippocampal structural traits and habituation in a spatial novelty situ-
ation. Behav Brain Res 1987, 24(2):111123.
21. Patel NV, Hitzemann RJ: Detection and mapping of quantitative trait loci for halo-
peridol-induced catalepsy in a C57BL/6J x DBA/2J F2 intercross. Behav Genet 1999,
29(5):303310.
22. Kanes S, Dains K, Cipp L, Gatley J, Hitzemann B, Rasmussen E, Sanderson S,
Silverman M, Hitzemann R: Mapping the genes for haloperidol-induced catalepsy.
J Pharmacol Exp Ther 1996, 277(2):10161025.
23. Bucan M, Abel T: The mouse: genetics meets behaviour. Nat Rev Genet 2002,
3(2):114123.
24. Melo JA, Shendure J, Pociask K, Silver LM: Identification of sex-specific quanti-
tative trait loci controlling alcohol preference in C57BL/ 6 mice. Nat Genet 1996,
13(2):147153.
25. Peirce JL, Derr R, Shendure J, Kolata T, Silver LM: A major influence of sex-specific
loci on alcohol preference in C57Bl/6 and DBA/2 inbred mice. Mamm Genome 1998,
9(12):942948.
26. Takahashi JS, Pinto LH, Vitaterna MH: Forward and reverse genetic approaches to
behavior in the mouse. Science 1994, 264(5166):17241733.
27. Plomin R, McClearn GE, Gora-Maslak G, Neiderhiser JM: Use of recombinant inbred
strains to detect quantitative trait loci associated with behavior. Behav Genet 1991,
21(2):99116.
28. Belknap JK, Phillips TJ, OToole LA: Quantitative trait loci associated with brain
weight in the BXD/Ty recombinant inbred mouse strains. Brain Res Bull 1992,
29(34):337344.
29. Williams RW: Mapping genes that modulate mouse brain development: a quantitative
genetic approach. In: Mouse Brain Development. Edited by Goffinet A, Rakic P.
New York: Springer; 2000: 2149.
30. Williams RW, Strom RC, Goldowitz D: Natural variation in neuron number in mice
is linked to a major quantitative trait locus on Chr 11. J Neurosci 1998, 18(1):138146.
31. Lu L, Airey DC, Williams RW: Complex trait analysis of the hippocampus: mapping
and biometric analysis of two novel gene loci with specific effects on hippocampal
structure in mice. J Neurosci 2001, 21(10):35033514.
32. Peirce JL, Chesler EJ, Williams RW, Lu L: Genetic architecture of the mouse hip-
pocampus: identification of gene loci with selective regional effects. Genes Brain
Behav 2003, 2(4):238252.
33. Williams RW, Airey DC, Kulkarni A, Zhou G, Lu L: Genetic dissection of the
olfactory bulbs of mice: QTLs on four chromosomes modulate bulb size. Behav Genet
2001, 31(1):6177.
34. Rosen GD, Williams RW: Complex trait analysis of the mouse striatum: independent
QTLs modulate volume and neuron number. BMC Neurosci 2001, 2:5.
35. Zhou G, Williams RW: Eye1 and Eye2: gene loci that modulate eye size, lens weight,
and retinal area in the mouse. Invest Ophthalmol Vis Sci 1999, 40(5):817825.
36. Airey DC, Robbins AI, Enzinger KM, Wu F, Collins CE: Variation in the cortical
area map of C57BL/6J and DBA/2J inbred mice predicts strain identity. BMC Neu-
rosci 2005, 6(1):18.
37. Airey DC, Lu L, Williams RW: Genetic control of the mouse cerebellum: identifica-
tion of quantitative trait loci modulating size and architecture. J Neurosci 2001,
21(14):50995109.
38. Li CX, Wei X, Lu L, Peirce JL, Williams RW, Waters RS: Genetic analysis of barrel
field size in the first somatosensory area (SI) in inbred and recombinant inbred strains
of mice. Somatosens Mot Res 2005, 22(3):141150.
39. Williams RW: Expression Genetics and the Phenotype Revolution. Mamm Genome
2006, in press.
40. Seecharan DJ, Kulkarni AL, Lu L, Rosen GD, Williams RW: Genetic control of
interconnected neuronal populations in the mouse primary visual system. J Neurosci
2003, 23(35):1117811188.
41. Damerval C, Maurice A, Josse JM, de Vienne D: Quantitative trait loci underlying
gene product variation: a novel perspective for analyzing regulation of genome expres-
sion. Genetics 1994, 137(1):289301.
42. Jansen RC, Nap JP: Genetical genomics: the added value from segregation. Trends
Genet 2001, 17(7):388391.
43. Brem RB, Yvert G, Clinton R, Kruglyak L: Genetic dissection of transcriptional
regulation in budding yeast. Science 2002, 296(5568):752755.
44. Chesler EJ, Lu L, Shou S, Qu Y, Gu J, Wang J, Hsu HC, Mountz JD, Baldwin NE,
Langston MA et al: Complex trait analysis of gene expression uncovers polygenic
and pleiotropic networks that modulate nervous system function. Nat Genet 2005,
37(3):233242.
45. Bystrykh L, Weersing E, Dontje B, Sutton S, Pletcher MT, Wiltshire T, Su AI, Vellenga
E, Wang J, Manly KF et al: Uncovering regulatory pathways that affect hematopoietic
stem cell function using genetical genomics. Nat Genet 2005, 37(3):225232.
46. Schadt EE, Monks SA, Drake TA, Lusis AJ, Che N, Colinayo V, Ruff TG, Milligan
SB, Lamb JR, Cavet G et al: Genetics of gene expression surveyed in maize, mouse
and man. Nature 2003, 422(6929):297302.
47. Hubner N, Wallace CA, Zimdahl H, Petretto E, Schulz H, Maciver F, Mueller M,
Hummel O, Monti J, Zidek V et al: Integrated transcriptional profiling and linkage
analysis for identification of genes underlying disease. Nat Genet 2005, 37(3):243253.
48. Stranger BE, Forrest MS, Clark AG, Minichiello MJ, Deutsch S, Lyle R, Hunt S,
Kahl B, Antonarakis SE, Tavare S et al: Genome-Wide Associations of Gene Expres-
sion Variation in Humans. PLoS Genet 2005, 1(6):e78.
49. Cheung VG, Spielman RS, Ewens KG, Weber TM, Morley M, Burdick JT: Mapping
determinants of human gene expression by regional and genome-wide association.
Nature 2005, 437(7063):13651369.
50. Morley M, Molony CM, Weber TM, Devlin JL, Ewens KG, Spielman RS, Cheung
VG: Genetic analysis of genome-wide variation in human gene expression. Nature
2004, 430(7001):743747.
51. Kirst M, Basten CJ, Myburg AA, Zeng ZB, Sederoff RR: Genetic architecture of
transcript-level variation in differentiating xylem of a eucalyptus hybrid. Genetics
2005, 169(4):22952303.
52. Benjamini Y, Hochberg Y: Controlling the False Discovery Rate: a practical and
powerful approach to multiple testing. J Royal Statistical Society B 1995, 57:289300.
53. Doss S, Schadt EE, Drake TA, Lusis AJ: Cis-acting expression quantitative trait loci
in mice. Genome Res 2005, 15(5):681-691.
54. Peirce J, Lu L, Hongqiang L, Wang J, Manly KF, Hitzemann B, Belknap JK, Rosen
GD, Williams RW: How replicable are mRNA expression QTLs. Mamm Genome
2006, in press.
55. Peirce JL, Lu L, Gu J, Silver LM, Williams RW: A new set of BXD recombinant
inbred lines from advanced intercross populations in mice. BMC Genet 2004, 5(1):7.
56. McCaughran J, Jr., Bell J, Hitzemann R: On the relationships of high-frequency
hearing loss and cochlear pathology to the acoustic startle response (ASR) and
prepulse inhibition of the ASR in the BXD recombinant inbred series. Behav Genet
1999, 29(1):2130.
57. Lin CH, Hansen S, Wang Z, Storm DR, Tapscott SJ, Olson JM: The dosage of the
neuroD2 transcription factor regulates amygdala development and emotional learning.
Proc Natl Acad Sci U S A 2005, 102(41):1487714882.
58. Homayouni R, Heinrich K, Wei L, Berry MW: Gene clustering by latent semantic
indexing of MEDLINE abstracts. Bioinformatics 2005, 21(1):104115.
59. Heintz N, Hatten M: The Gene Expressio Nervous System Atlas (GENSAT)
Project, NINDS Contract #N01NS02331 to The Rockefeller University, New York.
February 1, 2006. www.gensat.org.
60. Churchill GA, Doerge RW: Empirical threshold values for quantitative trait mapping.
Genetics 1994, 138(3):963971.
61. Dupuis J, Siegmund D: Statistical methods for mapping quantitative trait loci from
a dense set of markers. Genetics 1999, 151(1):373386.
62. Oishi I, Takeuchi S, Hashimoto R, Nagabukuro A, Ueda T, Liu ZJ, Hatta T, Akira
S, Matsuda Y, Yamamura H et al: Spatio-temporally regulated expression of receptor
tyrosine kinases, mRor1, mRor2, during mouse development: implications in devel-
opment and function of the nervous system. Genes Cells 1999, 4(1):4156.
63. Takeuchi S, Takeda K, Oishi I, Nomi M, Ikeya M, Itoh K, Tamura S, Ueda T, Hatta
T, Otani H et al: Mouse Ror2 receptor tyrosine kinase is required for the heart
development and limb formation. Genes Cells 2000, 5(1):7178.
64. Matsuda T, Suzuki H, Oishi I, Kani S, Kuroda Y, Komori T, Sasaki A, Watanabe K,
Minami Y: The receptor tyrosine kinase Ror2 associates with the melanoma-associ-
ated antigen (MAGE) family protein Dlxin-1 and regulates its intracellular distribu-
tion. J Biol Chem 2003, 278(31):2905729064.
28 Synaptic Mechanisms
Involved in Cognitive
Function: Cues from
Mental Retardation Genes
Guntram Borck, Florence Molinari, Birgit Dreier,
Peter Sonderegger, and Laurence Colleaux
CONTENTS
28.1 Introduction .................................................................................................. 436

28.2 At the Beginning Was the X........................................................................ 436
28.2.1 NS-MR Genes Involved in the Regulation of the
Actin Cytoskeleton: Possible Roles for the
Adaptive Response of Postsynaptic Spines..................................... 437
28.2.1.1 Oligophrenin-1.................................................................. 438
28.2.1.2 PAK3 (p21 Activated Kinase 3) ....................................... 438
28.2.1.3 PIX or Cool-2................................................................. 439
28.2.2 Nonsyndromic MR Genes Involved in the Regulation
of Presynaptic Neurotransmitter Release ........................................ 439
28.2.2.1 RabGDI1 (Rab GDP-Dissociation Inhibitor 1)................ 441
28.2.2.2 IL1RAPL1 (Interleukin-1 Receptor Accessory
Protein-Like1) ................................................................... 441
28.2.2.3 FACL4 (Fatty Acid-CoA Ligase 4) .................................. 441
28.3 Lessons from the X...................................................................................... 442
28.4 Neurotrypsin, a Nonsyndromic MR Gene with Synaptic
Localization but Unknown Function ........................................................... 442
28.4.1 Neurotrypsin Mutation in Autosomal Recessive MR ..................... 443
28.5 What Do Nonsyndromic MR Genes Tell Us about the Cellular
Mechanisms That Are Important for Cognitive Functions? ....................... 444
References.............................................................................................................. 445
435
Mental retardation (MR) is likely the result of alterations in molecular pathways

underlying neuronal processes involved in cognitive functions; however, the exact
nature of these pathways remains unknown. Remarkable progress in understanding
the molecular and cellular basis of mental retardation has occurred in recent years.
Genes associated with nonsyndromic mental retardation have been evaluated with
regard to a role in synaptic transmission. Some of them have been linked to presyn-
aptic vesicle release, whereas others are involved in the regulation of actin cytosk-
eleton dynamics via Rho GTPases. Detailed investigations of their molecular and
cellular roles at the synapse may provide cues for synaptic mechanisms that are
essential for cognitive functions.
28.1 INTRODUCTION
Mental retardation (MR) affects about 2% of the general population and is the most
frequent handicap in children and young adults. It is characterized by a broad range
of deficits in higher brain functions that result in significant limitations in adaptive
and cognitive capacities required for competence in daily living, communication,
social interaction and integration, self-direction, and work (DSM-IV). An onset of
the symptoms in the developmental period is an essential diagnostic trait. The severity
of mental retardation is commonly classified on the basis of the intelligence quotient
(IQ) although other criteria have also been used. Based on a population mean of 100
and a standard deviation of 15, MR is usually classified as mild when the IQ ranges
between 50 to 70 and as severe when the IQ value is below 50.1,2
The causes of MR are diverse and include environmental factors, teratogens,
numerical or structural anomalies of chromosomes, gene defects, and metabolic
diseases. An estimated fraction of 25 to 40% of the severe MRs and most mild MRs
remain unexplained.35 They are currently thought to be due to monogenic or mul-
tigenic defects or to a combination of genetic and environmental factors. Only
25 to 50% of the severe forms of MR are estimated to be genetically determined.
They include metabolic diseases impairing neuronal function in a nonspecific man-
ner, conditions which alter the normal patterning of the brain, neuromuscular dis-
orders, as well as MRs that do not show any clinical features besides cognitive
deficits. The last category is termed nonsyndromic MR (NSMR).
Although NSMR conditions are much difficult for geneticist studies, they are
best suited for identifying the molecular basis of cognitive functions, and analysis
of these disorders will very likely give new insight into the neurobiology of human
cognitive processing.
28.2 AT THE BEGINNING WAS THE X

A linkage of a relevant proportion of nonsyndromic MRs to the X chromosome was
suggested already in the 1930s based on the observed predominance of MR in men.
The large number of families in which MR is inherited as an X-linked trait provided
further, more direct evidence for sex linkage of mental retardation.6 Taking advantage
of the ease of gene mapping on the X chromosome and of the large number of
X-linked mental retardation (XLMR) families collected by international consortia,
Synaptic Mechanisms Involved in Cognitive Function 437
NS-XLMRs were first investigated. Since 1996, positional and functional candidate
gene approaches led to the identification of 11 X-linked nonsyndromic MR genes.7,8
Despite the extreme variety of processes in which genes responsible for these
disorders are involved, several common cellular processes have emerged.
28.2.1 NS-MR GENES INVOLVED IN THE REGULATION OF THE ACTIN

CYTOSKELETON: POSSIBLE ROLES FOR THE ADAPTIVE RESPONSE
OF POSTSYNAPTIC SPINES
Three X-linked nonsyndromic MR genes encode components that are directly

involved in signal transduction pathways of the Rho family GTPases (Figure 28.1).
Small GTPases of the Rho family act as transducers of extracellular signals to the
Extracellular signals
(Adhesion molecules, growth factors, electric/synaptic activity)
Rho GDI
Rho-GDP
Rac-GDP
Cdc42-GDP
OPHN1 Rho GAP Rho GEF ARHGEF6

Rho-GTP
Rac-GTP
Cdc42-GTP
Rho effectors PAK3
Actin cytoskeleton
of dendritic spines
FIGURE 28.1 (SEE COLOR INSERT FOLLOWING PAGE 236) One cluster of nonsyn-
dromic MR genes relates to the Rho family of small GTPases. Rho family GTPases transduce
extracellular signals into adaptive responses of the actin cytoskeleton. Actin-dependent pro-
cesses at the synapse include the regulation of the morphology and the dynamics of the
dendritic spines. The Rho family comprises three members, termed Rho, Rac, and Cdc42.
They shuttle between an inactive (red) and an active (green) state under the control of three
types of regulatory proteins (blue). GEFs (guanine nucleotide exchange factors) promote the
release of GDP and its replacement by GTP and, thereby, mediate the transition of Rho, Rac,
and Cdc42 from the inactive into the active state. GAPs (GTPase-activating proteins) activate
the endogenous GTPase function and, thus, the self-inactivation of Rho, Rac, and Cdc42.
GDIs (GDP dissociation inhibitors) bind to the GDP form of the GTPases and prevent their
premature activation as long as the GTPase is not in the correct place and situation for a new
round of activation. Further downstream activators (yellow) eventually act directly or indi-
rectly on regulatory components of the actin cytoskeleton. The genes affected in MRs are
printed red.
cytoskeleton and regulate gene expression. They are involved in neurite outgrowth,
axon guidance, dendrite maturation, synapse formation, and the morphogenesis and
dynamics of dendritic spines.912
The family comprises three members, termed Rho, Rac, and Cdc42. They exert
their diverse cellular functions via distinct effects on the actin cytoskeleton that are
mediated by PAK and ROCK kinases as immediate downstream signaling molecules.
Both PAK and ROCK kinases in turn activate LIM kinase, which acts upon and
regulates the dynamics of the actin cytoskeleton. Three major classes of upstream
regulators control the signaling activity of the Rho family GTPases. First, the guanine
nucleotide exchange factors (RhoGEFs) mediate the release of GDP and its replace-
ment by GTP and, thus, drive the conversion from the inactive into the active form.
Second, the GTPase activating proteins (RhoGAPs) enhance the conversion of bound
GTP into GDP and, thus, drive the conversion from the active into the inactive state.
And third, the GDP dissociation inhibitors (RhoGDIs) bind to the GDP-bound form
of the GTPase and prevent their activation as long as other prerequisites, such as
the correct location for reinitiation of the activation cycle, are not fulfilled.
Over the past years, two regulators and one downstream effector of Rho family
GTPases have been found among the genes causing X-linked nonsyndromic MR,
described as follows.
28.2.1.1 Oligophrenin-1
The OPHN1 gene, encoding oligophrenin-1, maps to Xq12 and is expressed in fetal
brain at high levels. It is disrupted in a mildly mentally retarded female carrier of
a balanced X;12 translocation. Furthermore, it has been shown to be mutated in
affected males of a large X-linked MR family by Billuart et al.13 Oligophrenin
contains a domain typical for Rho-GTPase-activating proteins (RhoGAP). In vitro
assays demonstrated that it can regulate the activity of the Rho GTPases RhoA,
Rac1 and Cdc42Hs.
Recent studies show that oligophrenin-1 is present in neuronal and astroglial
cells and that it colocalizes with actin at the tip of growing neurites.14 In addition,
using siRNAs in organotypic hippocampal slices, Govek and co-workers15 demon-
strated that knocking down oligophrenin-1 significantly decreased dendritic spine
length in CA1 pyramidal neurons. Spine morphological changes of the same mag-
nitude have been reported for a mouse model of fragile X,16 indicating that such
changes can compromise synaptic plasticity and potentially lead to learning and
memory deficits.
28.2.1.2 PAK3 (p21 Activated Kinase 3)
Interestingly, soon after the characterization of OPHN1, a second NS-XLMR gene

implicated in the Rho GTPase pathway was described. Using a candidate gene
approach, Allen et al.17 showed that a mutation in the PAK3 gene was associated
with MR. PAK3 encodes a serine-threonine kinase involved in the Rac/Cdc42-
dependent regulation of actin cytoskeleton dynamics. In neuronal cells, PAK3 acts
downstream of the GTPases in a signaling pathway that drives polarized growth of
the actin cytoskeleton in developing neurites. In agreement with such a function,
PAK3 is highly expressed in fetal human brain. In the newborn mouse brain it
localizes mainly to cortical and hippocampal dendrites and axons. LIM kinase 1, a
downstream effector kinase regulated by PAK kinase, was recently reported to
regulate the morphology of dendritic spines of hippocampal neurons.18
28.2.1.3 PIX or Cool-2
Finally, taking advantage of a balanced X;21 translocation with a breakpoint in

Xq26, Kutsche et al.19 showed that the gene ARHGEF6, encoding a Rac1/Cdc42-
specific guanine nucleotide exchange factor called a PIX, was truncated in a female
patient with severe MR and that a null mutation of ARHGEF6 cosegregated with
the MR phenotype in a X-linked MR family. Due to this regulation of Rac and Cdc42
ARHGEF6 has been implicated in signal transduction pathways involved in cell
migration and axonal outgrowth. Although little is known about the biological
function of ARHGEF6 in neuronal processes, its interaction with the focal adhesion
molecule beta-Parvin (PARVB) suggests that ARHGEF6 is involved in integrin-
mediated signaling leading to the activation of Rac1 and/or Cdc42 GTPases.20 A
related Rho GEF, termed kalirin-7, has recently been reported to be involved in
dendritic spine morphogenesis via activation of Rac1 and its effector PAK.21
Although OPHN1, PAK3 and ARHGEF6 genes affect the signaling of the Rho
family members Rho, Rac, and Cdc42 differently, they all affect actin cytoskeleton
dynamics, as this is the hallmark of Rho family GTPase function. Dendritic plasticity
depends on morphological changes that follow rearrangements of the cytoskeleton
in response to neural activity at the synaptic site.2224 The observation that mutations
in these genes are responsible for NSMR sheds light on the connection between
Rho GTPases and cognition and makes a clear case for the central role played by
these effectors in neuronal signaling. The importance of the regulation of actin
dynamics is further demonstrated by the involvement of downstream effectors of
the Rho GTPase signaling pathway. Not only PAK3, a direct effector kinase of Rac
and Cdc42, but also LIM kinase-1, a kinase downstream of PAK3, have been linked
to MR. If PAK3 mutations are responsible for NS-XLMR, a mutation in LIM kinase-
1 was found in patients with Williams syndrome, a syndromic form of MR. Finally,
the hypothesis that a deficiency in the cytoskeletal dynamics at the synapse may be
responsible for cognitive deficit is also supported by the observation of abnormal
shapes and numbers of dendritic spines.25
28.2.2 NONSYNDROMIC MR GENES INVOLVED IN THE REGULATION

OF PRESYNAPTIC NEUROTRANSMITTER RELEASE
Vesicle release from the presynaptic nerve terminal is an essential process for
synaptic transmission. For a long time it has been known that synaptic vesicle release
is subject to modulations that depend on previous activity patterns of a given synapse.
It is known that this modulatory mechanism, termed synaptic facilitation, depends
on intracellular calcium, but its exact molecular basis is still under investigation.
Neurotransmitter release is essential for basic synaptic function. Its activity-
dependent regulation is of critical importance for the coordinated and dynamic
Neurotransmitter
synthesis/uptake
Rab GEF
Rab-GTP
Rab-GTP Rab GDI1
Rab GDI
Vesicle
IL1RAPL1 recycling
(endocytosis)
Rab-GTP Rab-GTP Rab-GDP
NCS-1
Docked Fused Restricting
FACL4
Rab GAP
vesicle Ca2+ vesicle ilin
bud oph
End
FIGURE 28.2 (SEE COLOR INSERT) A second cluster of nonsyndromic MR genes relates
to the release of neurotransmitters from the presynaptic nerve terminal and the regeneration
of new releasable vesicles by the budding and endocytosis of vesicles from the presynaptic
membrane. The release of neurotransmitters from storage vesicles occurs in several consec-
utive steps. First, the vesicles are docked to the release site, the presynaptic active zone, a
process that involves the interaction of several proteins of the vesicle and the active zone.
Docked vesicles fuse with the presynaptic membrane in a process regulated by transiently
increased intracellular calcium, and release their content into the synaptic cleft. In order to
maintain a constant supply of releasable neurotransmitters, vesicles recycle by budding off
the presynaptic membrane and endocytosis. The budding occurs in a clathrin-dependent
manner. The restriction of the bud neck, a process that prepares the bud for scission, requires
endophilin. Endophilin is a lysophosphatidic acid acyltransferase, which introduces arachi-
donic acid into the cytoplasmic face of the bud neck. It, thereby, facilitates the formation of
a strong negative curvature of the membrane required for the restriction of the bud neck in
order to allow the scission process to begin. It is conceivable that the supply of arachidonoyl-
CoA requires FACL4. The genes affected in MR are printed in red.
function of synaptic circuits. Three nonsyndromic MR genes play essential roles in

the presynaptic vesicle release and recycling machinery, RabGDI1, IL1RAPL1, and
FACL4 (Figure 28.2). Particularly intriguing in the context of MR is the finding that
their function is indispensable for the fine-tuning of the release machinery
during activity-dependent adaptive processes. The deficiency of these genes is
still compatible with basic synaptic function, but adaptive responses of the secretory
machinery, such as those thought to be necessary for learning and memory, may
be deficient.
28.2.2.1 RabGDI1 (Rab GDP-Dissociation Inhibitor 1)
The members of the GDI family (Rab GDP-dissociation inhibitors) play an essential
role in the recycling of Rab GTPases required for vesicular transport. Using a
candidate gene approach, DAdamo et al.26 identified RabGDI1 mutations in two
X-linked MR families linked to the Xq28 region. RabGDI1 binds Rab3a, a small
GTP-binding protein highly enriched in the synapse and involved in neurotransmitter
release. First evidence for a role of RabGDI1 protein in neurite outgrowth and
synaptic function came from in vitro studies showing that RabGDI1-antisense
oligonucleotides impaired neurite extension in cultured rat hippocampal neurons.
Further support for this role came from histological, behavioral and electrophysio-
logical studies of RabGDI1-deficient mice. These mice, which do not have any gross
morphological or neuropathological anomalies, display an altered plasticity of hip-
pocampal neurotransmission.27 In addition, behavioral studies have revealed defects
in short-term memory, a reduced aggression level, and altered social behavior.28 A
recent ultrastructural analysis revealed pronounced presynaptic changes in several
brain regions. Most strikingly, a marked reduction of the number of synaptic vesicles
and a clustering of the residual vesicles were observed. These ultrastructural alter-
ations are in perfect accordance with a role of RabGDI1 in synaptic vesicle recycling.
28.2.2.2 IL1RAPL1 (Interleukin-1 Receptor Accessory

Protein-Like1)
The IL1RAPL1 gene has been identified in the analysis of non-overlapping Xp22
deletions and point mutations in X-linked MR families by Carrie et al.29 The
IL1RAPL1 protein is a member of the IL-1/Toll receptor family. It shows homologies
with the IL-1 receptor accessory proteins (IL1RACPs), and localizes to the plasma
membrane. Recent functional studies showed that IL1RAPL1 is not a receptor for
IL-1 but a specific interaction partner of neuronal calcium sensor 1.30 The NCS-1
protein has been shown to be involved in synaptic facilitation at excitatory hippoc-
ampal synapses.31 Synaptic facilitation is a mechanism of short-term plasticity
that enhances transmitter release from the presynaptic terminal and increases postsyn-
aptic activation as a consequence of recurrent stimulation. Because synaptic facili-
tation has long been known to depend on calcium and because NCS-1 was demon-
strated to be a sensor of presynaptic calcium, NCS-1 is put into a crucial position
for the activity-dependent adaptation of presynaptic transmitter release. By interact-
ing with NCS-1 in the inhibition of exocytosis, IL1RAPL-1 might exert a regulatory
role on the activity-dependent dynamics of presynaptic transmitter release.
28.2.2.3 FACL4 (Fatty Acid-CoA Ligase 4)
By deletion mapping, FACL4, which encodes fatty acid-CoA ligase 4, was identified
as an X-linked nonsyndromic MR gene.32 Acyl-CoA ligases (or synthases) form a
family of enzymes that catalyze the formation of acyl-CoA esters from fatty acids,
ATP and coenzyme A. Both MR-associated mutations lead to a drastic decrease of
enzymatic activity. That fatty acid-CoA ligase 4 has a strong preference for arachi-
donic acid as a substrate raises interesting speculations about its role in the recycling
of synaptic vesicles. Recent reports suggest an essential role of arachidonic acid in
the endocytosis of synaptic vesicles from the presynaptic membrane, a process
required for the regeneration of releasable neurotransmitter vesicles. A crucial step
in this process is the formation of a bud which then constricts and separates from
the presynaptic plasma membrane to form a new vesicle. This process requires the
remodeling of the lipids on the cytosolic side of the bud neck by endophilin A1.33
Endophilin A1 acts as a lysophosphatidic acid acyltransferase, i.e., it binds lyso-
phosphatidic acid and fatty acyl-coenzyme A and condenses them to phosphatidic
acid. The formation of phosphatidic acid promotes the negative curvature required
at the bud neck for bud restriction. Unsaturated fatty acyl-CoAs, such as arachi-
donoyl-CoA, are most effective in promoting negative curvature when inserted into
the cytosolic leaflet. Thus, a deficiency in unsaturated fatty acyl-CoAs, such as
arachidonoyl-CoA, that may result from deficient FACL4 is likely to reduce the
efficiency of bud restriction and thus results in reduced synaptic vesicle recycling.
28.3 LESSONS FROM THE X

Some common features emerge from the studies on NS-XLMR. There are now
several examples of genes initially described as syndromic MR genes that account
for rare cases of NS-XLMR. This is the case for the RSK2, MECP2, and FGD1
genes leading to Coffin-Lowry, Rett, and Aarskog-Scott syndromes, respectively,
and mutated in rare families with NSMR.3436 In addition, the overlap between
syndromic XLMR and NS-XLMR exists in both directions: not only can syndromic
XLMR genes cause NS-XLMR but recently OPHN1 mutations were found to cause
MR and congenital cerebellar hypoplasia.37,38 Another common point is that all NS-
XLMR genes that have been tested are highly expressed in the fetal brain (from as
early as embryonic day E8 on in the mouse) with a high expression in the cortex
and the hippocampus, i.e., structures important for higher cognitive functions such
as memory and learning. The fact that some XLMR genes are part of common
physiological pathways has already been mentioned above. Finally, genetic hetero-
geneity of this pathology is still growing. None of the NS-XLMR genes accounted
for all families linked to a specific region, implying that the number of genes,
previously thought to lie between 8 and 12 based on non-overlapping linkage inter-
vals, is much higher.39
28.4 NEUROTRYPSIN, A NONSYNDROMIC MR GENE

WITH SYNAPTIC LOCALIZATION BUT UNKNOWN
FUNCTION
While an autosomal recessive mode of inheritance may account for nearly one fourth
of individuals with NS-MR, only X-linked genes have been identified until recently.
In fact, the broad genetic heterogeneity of MR and the scarcity of large pedigrees
suitable for linkage analyses have hitherto hampered identification of the genes
responsible for these diseases. The hope that a careful study of individuals with
chromosomal anomalies would lead to the rapid identification of autosomal MR
genes has been largely disappointed. However, linkage studies and chromosomal
breakpoint analyses recently led to the identification of the first autosomal NS-MR
genes.
28.4.1 NEUROTRYPSIN MUTATION IN AUTOSOMAL RECESSIVE MR

The PRSS-12 gene encoding the synaptic serine protease neurotrypsin was recently
identified as the first autosomal-recessive gene involved in nonsyndromic MR. Using
homozygosity mapping in a large consanguineous Algerian family, Molinari et al.
mapped an autosomal recessive nonsyndromic MR gene on chromosome 4q24-q25.40
The sibship comprises four mentally retarded and four healthy children born to first-
cousin Algerian parents. All affected children (3 girls and 1 boy) exhibited a severe
impairment of cognitive functions with an IQ below 50. Analysis of candidate genes
mapping to this part of the genome led to the identification of a homozygous 4-bp
deletion in the neurotrypsin gene, resulting most likely in a null allele due to the
formation of a shortened protein lacking the catalytic domain.40 Neurotrypsin is
predominantly expressed in neurons of the cerebral cortex, the hippocampus and the
amygdala.41 By immuno-electronmicroscopy, neurotrypsin was localized in the pre-
synaptic membrane and the presynaptic active zone of both asymmetrical (excitatory)
and symmetrical (inhibitory) synapses. In vitro studies have demonstrated that it is
a secreted protein which remains associated with the presynaptic membrane after
its secretion. In search for additional mutations in this gene, Molinari et al. screened
18 inbred families and 30 nonconsanguineous families with individuals affected with
nonsyndromic MR.40 They found the same 4-bp deletion in a child born to first-
cousin Algerian parents. The two families appear unrelated, but originate from the
same area of eastern Algeria. In both families, the mutation was carried on the
background of the same haplotype across the neurotrypsin locus, suggesting a
founder effect in the Algerian population.
The pathophysiological phenotype and the age of disease onset in the affected
individuals are consistent with the idea that neurotrypsin might regulate adaptive
synaptic functions, such as synapse reorganization, during later stages of neurode-
velopment and postnatal synaptic plasticity. In all affected children the course of
the disease was similar. They reached the milestones of normal psychomotor devel-
opment in the first 18 months. Signs of MR were first observed by their parents
when they were around 2 years of age. This suggests that neurotrypsin is not involved
critically in the formation of synapses, but rather plays a crucial role for adaptive
synaptic functions, such as those subserving higher cognitive functions.
Altogether, these results provide the first evidence for an association between
cognitive impairment and defects in extracellular proteolytic activity at the synapse,
opening a novel field in the pathophysiology of MR. The generation of animal
models such as mice deficient in the catalytic domain of neurotrypsin in CNS
neurons or mice overexpressing neurotrypsin will provide further insight into the
function of this protein.
28.5 WHAT DO NONSYNDROMIC MR GENES TELL US

ABOUT THE CELLULAR MECHANISMS THAT ARE
IMPORTANT FOR COGNITIVE FUNCTIONS?
The currently available data about monogenic defects resulting in nonsyndromic
MR indeed reveal a relatively high proportion of genes that encode synaptic proteins.
The best-studied among them contribute to presynaptic release of neurotransmitter
and the recycling of transmitter vesicles (IL1RAPL1, RabGDI1, FACL4) or the
regulation of the cytoskeletal dynamics of postsynaptic spines (OPHN1, ARHGEF6,
PAK3). Both the presynaptic release of neurotransmitters and the morphology of the
postsynaptic spines are elements of the adaptive response of synapses that is sum-
marized under the term synaptic plasticity. The first gene involved in autosomal
recessive NS-MR, PRSS-12, encodes the synaptic serine protease neurotrypsin,
whose role in synaptic structure, function, and plasticity has not yet been elucidated.
However, its almost exclusive synaptic localization strongly suggests a regulatory
role of synaptic function and thus makes it an interesting candidate for a gene
involved in cognitive processes.
A common feature that makes nonsyndromic MR genes particularly interesting
is the observation that their deficiency is compatible with life, but results in severe
deficits in cognitive functions while synapse formation and basic synaptic transmis-
sion is not affected. Thus, one can conclude that cognitive functions need more than
basic synaptic activity. They depend on the coordination of extended synaptic cir-
cuits. This in turn depends on synaptic plasticity, the capacity of individual synapses
to adapt their transmission on demand to fit the preconditions of the entire circuit.
Intact mechanisms of synaptic plasticity are thought to be an important prerequisite
for the development of these complex and dynamic circuits in the brain that underlie
cognitive functions. Although at present, none of these genes has been unequivocally
assigned to a specific synaptic function that when compromised would result in a
selective loss of a cognitive function, e.g., memory and learning, their identification
provides the basis for a detailed characterization of cognitive functions at the molec-
ular level.
Many more NS-MR genes remain to be characterized because, for most of the
genes currently known, mutations turn out to be very rare with only few mutations
identified in affected families. The observation of common or closely related path-
ways dysfunction underlying NS-XLMR led to some excitement, but it should not
result in a dangerous oversimplification. Many other molecular pathways are likely
to be involved in NS-MR as well, and in most cases the relationship between these
proteins and cognitive functions remains elusive.4247
This observation is obviously bad news with respect to diagnosis and counseling,
but opens up the possibility that there are still many other genes that play a major
role in NS-MR. A better understanding of the function of MR genes will ultimately
yield not only deeper insight into brain function, but may also provide targets for
future therapies.
REFERENCES
1. Penrose, L.S., A clinical and genetic study of 1280 cases of mental defects (The
Colchester Survey). Special report series, Medical Research Council No. 229, Her
Majestys Stationary Office, London, 1938.
2. Roeleveld, N., Zielhuis, G.A., and Gabreels, F. The prevalence of mental retardation:
a critical review of recent literature, Dev Med Child Neurol, 39, 125, 1997.
3. Turner, G. An etiological study of 1,000 patients with an I.Q. assessment below 51.
Med J Aust, 2, 927, 1975.
4. Lamont, M.A. and Dennis, N.R., Etiology of mild mental retardation. Arch Dis Child,
63, 1032, 1988.
5. Kahler, S.G. and Fahey, M.C., Metabolic disorders and mental retardation. Am J Med
Genet, Part C (Sem Med Genet), 117C, 31, 2003.
6. Stevenson, R.E., Schwartz, C.E., and Schroer, R.J., Emergence of the concept of X-
linked mental retardation, In X-Linked Mental Retardation, Stevenson, R.E.,
Schwartz, C.E., and Schroer, R.J.. New York: Oxford University Press, 2000:2367.
7. Chelly, J. and Mandel, JL. Monogenic causes of X-linked mental retardation. Nature
Rev Genet, 2, 669, 2001.
8. Ramakers, G.J.A. Rho proteins, mental retardation and the cellular basis of cognition.
Trends in Neurosci, 25, 191, 2002.
9. Mackay, D.J., Nobes, C.D., and Hall A. The Rhos progress: A potential role during
neuritogenesis for the Rho family of GTPases. Trends Neurosci, 18, 496, 1995.
10. Luo, L. Rho GTPases in neuronal morphogenesis. Nat Rev Neurosci, 1, 173, 2000.
11. Ishizaki, H. et al. Role of rab GDP dissociation inhibitor alpha in regulating plasticity
of hippocampal neurotransmission. Proc Natl Acad Sci USA, 97, 11587, 2000.
12. Negishi, M. and Katoh, H. Rho family GTPases as key regulators for neuronal network
formation. J Biochem. (Tokyo) 132, 157, 2002.
13. Billuart, P. et al. Oligophrenin-1 encodes a rhoGAP protein involved in X-linked
mental retardation. Nature, 392, 923, 1998.
14. Fauchereau, F. et al. The RhoGAP activity of OPHN1, a new F-actin-binding protein
is negatively controlled by its amino-terminal domain. Mol. Cell. Neurosci., 23, 574,
2003.
15. Govek, E.E. et al. The X-linked mental retardation protein oligophrenin-1 is required
for dendritic spine morphogenesis. Nat Neurosci., 7, 364, 2004.
16. Comery, T.A. et al. Abnormal dendritic spines in fragile X knockout mice: Maturation
and pruning deficits. Proc. Natl. Acad. Sci. USA, 94, 5401, 1997.
17. Allen, K.M. et al. PAK3 mutation in nonsyndromic X-linked mental retardation. Nat
Genet., 20: 2530, 1998
18. Meng, Y. et al. Abnormal spine morphology and enhaced LTP in LIMK-2 knockout
mice. Neuron 35: 121133, 2002.
19. Kutsche, K. et al. Mutations in ARHGEF6, encoding a guanine nucleotide exchange
factor for Rho GTPases, in patients with X-linked mental retardation. Nat Genet. 26,
24750, 2000.
20. Rosenberger, G. et al. Interaction of alphaPIX (ARHGEF6) with beta-parvin
(PARVB) suggests an involvement of alphaPIX in integrin-mediated signaling. Hum
Mol Genet 12: 155167, 2003.
21. Penzes, P. et al. Rapid induction of dendritic spine morphogenesis by trans-synaptic
ephrinB-EphB receptor activation of the Rho-GEF kalirin. Neuron 37: 263274, 2003.
22. Matus, A., Brinkhaus, H., and Wagner, U. Actin dynamics in dendritic spines: a form
of regulated plasticity at excitatory synapses. Hippocampus, 10, 555, 2000.
23. Matus, A. Postsynaptic actin and neuronal plasticity. Curr Opin Neurobiol, 9, 561,
1999.
24. Yuste, R. and Bonhoeffer, T. Morphological changes in dendritic spines associated
with long-term synaptic plasticity. Annu Rev Neurosci, 24, 1071, 2001.
25. Purpura, D.P. Dendritic spine dysgenesis and mental retardation. Science, 186, 1126,
1974.
26. DAdamo, P. et al. Mutations in GDI1 are responsible for X-linked non-specific
mental retardation. Nat Genet, 19,134, 1998.
27. Ishizaki, H. et al. Role of rab GDP dissociation inhibitor alpha in regulating plasticity
of hippocampal neurotransmission. Proc Natl Acad Sci USA, 97, 11587, 2000.
28. DAdamo, P. et al. Deletion of the mental retardation gene Gdi1 impairs associative
memory and alters social behavior in mice. Hum Mol Genet, 11, 2567, 2002.
29. Carrie, A. et al. A new member of the IL-1 receptor family highly expressed in
hippocampus and involved in X-linked mental retardation. Nat Genet 23 , 25, 1999.
30. Bahi, N. et al. IL1 receptor accessory protein like, a protein involved in X-linked
mental retardation, interacts with Neuronal Calcium Sensor-1 and regulates exocyto-
sis. Hum Molec Genet, 12, 1425, 2003.
31. Sippy, T. et al. Acute changes in short-term plasticity at synapses with elevated levels
of neuronal calcium sensor-1. Nature Neurosci, 6, 1031, 2003.
32. Meloni, I. et al. FACL4, encoding fatty acid-CoA ligase 4, is mutated in nonspecific
X-linked mental retardation. Nat Genet, 30, 436, 2002.
33. Huttner, W.B. and Schmidt, A. Lipids, lipid modification and lipid-protein interaction
in membrane budding and fission insights from the roles of endophilin A1 and
synaptophysin in synaptic vesicle endocytosis. Curr Opin Neurobiol, 10, 543, 2000.
34. Merienne, K. et al. A missense mutation in RPS6KA3 (RSK2) responsible for non-
specific mental retardation. Nat Genet, 22, 13, 1999.
35. Couvert, P. et al. MECP2 is highly mutated in X-linked mental retardation. Hum Mol
Genet, 10, 941, 2001.
36. Lebel, R.R. et al. Non-syndromic X-linked mental retardation associated with a
missense mutation (P312L) in the FGD1 gene. Clin Genet., 61, 139, 2002.
37. Philip, N. et al. Mutations in the oligophrenin-1 gene (OPHN1) cause X linked
congenital cerebellar hypoplasia. J Med Genet., 40, 441, 2003.
38. Bergmann, C. et al. Oligophrenin 1 (OPHN1) gene mutation causes syndromic X-
linked mental retardation with epilepsy, rostral ventricular enlargement and cerebellar
hypoplasia. Brain, 126, 1537, 2003.
39. Ropers, HH. et al. Nonsyndromic X-linked mental retardation: where are the missing
mutations? Trends Genet., 19, 316, 2003.
40. Molinari, F. et al. Truncating neurotrypsin mutation in autosomal recessive nonsyn-
dromic mental retardation. Science, 298, 1779, 2002.
41. Gschwend, T.P. et al. Neurotrypsin, a novel multidomain serine protease expressed
in the nervous system. Mol Cell Neurosci, 9, 207, 1997.
42. Fukami, M. et al. A member of a gene family on Xp22.3, VCX-A, is deleted in
patients with X-linked nonspecific mental retardation. Am J Hum Genet, 67, 563,
2000.
43. Zemni, R. et al. A new gene involved in X-linked mental retardation identified by
analysis of an X;2 balanced translocation. Nat Genet, 24, 167, 2000.
44. Freude, K. et al. Mutations in the FTSJ1 Gene Coding for a Novel S-Adenosylme-
thionine-Binding Protein Cause Nonsyndromic X-Linked Mental Retardation. Am J
Hum Genet. 2004.
45. Kleefstra, T. et al. Zinc finger 81 (ZNF81) mutations associated with X-linked mental
retardation. J Med Genet., 41, 394, 2004.
46. Shoichet, S.A. et al. Mutations in the ZNF41 gene are associated with cognitive
deficits: identification of a new candidate for X-linked mental retardation. Am J Hum
Genet., 73, 1341, 2003.
47. Gecz, J. The FMR2 gene, FRAXE and non-specific X-linked mental retardation:
clinical and molecular aspects. Ann Hum Genet., 64, 95106, 2000.
29 Pharmacogenetics
Byron C. Jones
CONTENTS
29.1 Introduction .................................................................................................. 449

29.2 The Two Traditions in Pharmacogenetics ................................................... 449
29.2.1 The Pharmacokinetic Tradition........................................................ 450
29.2.1.1 Inborn Errors of Metabolism............................................ 450
29.2.1.2 Pharmacogenetics of Drug Metabolism ........................... 450
29.2.1.3 The Special Case of the Cytochrome P 450
Superfamily of Phase I Drug Metabolizing Enzymes ..... 451
29.2.2 The Pharmacodynamic Tradition..................................................... 451
29.2.3 Let Us Not Forget the Influence of the Environment..................... 452
29.3 Single Genes and Pharmacogenetics ........................................................... 452
29.4 Pharmacogenetics and Complex Traits Analysis ........................................ 453
29.5 QTL to QTGene........................................................................................... 454
29.6 PharmacogenomicsThe Next Step ........................................................... 454
29.7 Summary ...................................................................................................... 455
References.............................................................................................................. 455
29.1 INTRODUCTION
Pharmacogenetics is the study of individual differences in drug response. The field
has been recognized since the early 1950s, but the biomedical research devoted to
it has been relatively slow in development. Nevertheless, pharmacogenetics has
manifold, important implications for health. For some individuals, for example, what
would otherwise be termed a drugs side effect may in fact be the main drug effect.
Alternatively, an individual may be predisposed to insensitivity to a particular drug,
with implications for dose or to make administration of the drug futile. In this chapter
we will explore the genetic bases for individual differences in drug actions (although
there are environmental effects as well).
29.2 THE TWO TRADITIONS IN PHARMACOGENETICS

There have been two, somewhat distinct paths that the study of pharmacogenetics
has taken over the past half century. The first, historically is the pharmacokinetic
tradition and the second, the pharmacodynamic tradition. In pharmacology, these
449
distinctions are somewhat arbitrary but focus the attention to different aspects of
drug action.
29.2.1 THE PHARMACOKINETIC TRADITION

Pharmacokinetics deals with drug administration, absorption, distribution, biotrans-
formation and elimination. The early pharmacokinetic studies focused on the last
two of these, biotransformation and elimination.
29.2.1.1 Inborn Errors of Metabolism
The pharmacokinetic tradition is related to inborn errors of metabolism and is based

largely on work in humans. In 1902, Archibald Garrod1 published an article, The
Incidence of Alkaptonuria: A Study in Chemical Individuality, in The Lancet.
Alkaptonuria is a condition diagnosed when an individuals urine, exposed to air,
turns black. The cause is a problem with the catabolism of tyrosine to produce
homogentisic acid in urine, which is responsible for the color change. Garrod
described the pattern of inheritance as being consistent with an autosomal recessive
allele. In a later work, Flling2 reported on phenylketonuria in a family, a work that
led to phenylketonuria being described as following the same pattern of inheritance
as alkaptonuria. With these pioneering studies, the groundwork was laid to show
that many enzymes that are responsible for biotransformation of autacoidal sub-
stances come in alternate forms (isoenzymes) produced by genes that vary in their
DNA (alleles) and hence in their amino acid sequence.
29.2.1.2 Pharmacogenetics of Drug Metabolism
Not surprisingly, the isoenzymes that catalyze reactions of endogenous hormones,

neurotransmitters, etc., are also involved in the biotransformation of drugs. The usual
consequence is in individual differences in reaction rates based on abundance,
activity or differing affinities among the enzymes for the drug. In the 1950s, Kalow3
described in a family study, individual differences in pseudocholinesterase. This
enzyme is responsible for biotransformation of local anesthetics, such as procaine
and biotransformation of succinylcholine, a muscle paralyzer used as adjunct to
general surgery. Individuals who were slow metabolizers of succinylcholine had
greater difficulty in recovering from its effects on skeletal muscles following general
surgery.
A particularly compelling example of the importance of genetic differences in
pharmacokinetics is in alcohol metabolism. The major metabolic pathway for alcohol
biotransformation is through alcohol dehydrogenase which converts ethyl alcohol
to acetaldehyde (and methyl alcohol to formaldehyde) followed by the action of
acetaldehyde dehydrogenase to produce acetate which goes into intermediate metab-
olism. There are two prominent alleles of acetaldehyde dehydrogenase, ALDH2*1
and ALDH2*2. Individuals who are homozygous for the ALDH2*1 allele are able
to oxidize acetaldehyde much faster than those homozygous for the ALDH2*2 allele
or who are heterozygotes.4 The consequences of consuming alcohol for those car-
rying the ALDH2*2 allele include rapid intoxication, facial flushing, headache and
Pharmacogenetics 451
nauseaall as a result of the toxic effects of accumulated acetaldehyde. About 50%

of East Asians carry the ALDH2*2 allele and flush after consuming alcohol. The
ALDH2*2 allele is rare among Caucasians, Africans, and South Asians. Carrying
the ALDH2*2 allele is thought to protect against alcoholism; however, there are
some carrying the allele and who consume large amounts of alcohol. These individ-
uals incur a many-fold increased risk for esophageal cancer and liver disease com-
pared to those who can metabolize acetaldehyde rapidly.5
29.2.1.3 The Special Case of the Cytochrome P 450 Superfamily

of Phase I Drug Metabolizing Enzymes
Drug metabolizing enzymes are classified into Phase I and Phase II enzymes. Phase
I enzymes by oxidation, reduction or hydrolysis increase the polarity of a drug and
in some cases make the drug more active for example, changing codeine to morphine.
Phase II enzymes act by conjugation, adding for example acetyl group, to further
increase polarity and thus enable elimination by the kidneys. Genetic polymorphisms
may occur in either type, however, some of the most important occur in the Phase
I, especially in the major group of oxidizing enzymes, the Cytochrome P 450
enzymes (CYP). The CYP enzymes consist of more than 150 classified into sub-
families. The CYP2 family is particularly interesting for psychopharmacology,
because many of the drugs used to treat neuropsychiatric illnesses are metabolized
by CYP2D6. Drugs so affected include tricyclic and specific serotonin uptake inhibitor
antidepressants (fluoxetine, paroxetine) and haloperidol, thoridizine, perphenazine
and similar antipsychotics. It is further estimated that in all therapeutic categories,
CYP2D6 is involved in the metabolism of more than 20% of prescribed drugs.6
More than 50 allelic variants of this enzyme are known and are used to categorize
individuals into four classes of metabolizers: poor, intermediate, efficient, and ultra-
rapid. Of particular interest is that the development of Parkinsonian symptoms in
individuals taking antipsychotic medication is associated with the poor metabolizing
phenotype. It is obvious, therefore, that drug dosing recommendations should be
made based on the allelic configuration of the individual and one researcher believes
that the differential distribution of the polymorphisms would make genotyping
worthwhile prior to prescription of medications metabolized by CYP2D6.6
29.2.2 THE PHARMACODYNAMIC TRADITION

The study of individual differences in drug action at target tissues began in the 1950s.
Initially, the almost sole focus of this line of research was on individual differences
in alcohol response and sensitivity and in rats and mice predominantly. More recent
work has focused on other drugs, albeit for the most part on drugs of human misuse.
In 1951, Mardones7 reported research on successful selective breeding for rats that
preferred and for rats that did not prefer alcohol (vs. water). This work showed that
at least some aspects of alcohol consumption were under genetic influence. One of
the most salient demonstrations of genetic influence on drug actions is the work
of G.E. McClearn and colleagues.8 In the 1960s this team screened a large number
of genetically heterogeneous mice for hypnotic sensitivity to ethanol. The animals
were injected intraperitoneally with 3.3 g/kg ethanol in saline and then observed for
loss of righting response by placing the animals in a supine position. The time from
loss of righting response to time of regain was recorded as the dependent variable.
Large individual differences were observed among these animals and mass selection
over several generations produced relatively rapid differentiation between the two
lines. The more sensitive line, as determined by sleep time was named long sleep
(LS) and the less sensitive line was named short sleep (SS). The pharmacogenetics
of alcohol is more completely covered in the next chapter by J.C. Crabbe. The
example of the LS and SS mice is used here to illustrate two important principles
in pharmacogenetics. First, what is the nature of the phenotype? In the example of
the LS and SS mice, does sleep time reflect a pharmacodynamic effect (i.e., target
tissue sensitivity) or is the effect a result of different rates of alcohol disappearance
between the two lines, i.e., do SS mice eliminate alcohol more rapidly than LS?
Tabakoff and Ritzmann addressed this question directly by measuring blood alcohol
concentrations (BEC) at loss and regain of the righting response9 and they reported
that the BEC for the SS was nearly twice that of the LS both at loss and at regain
of righting. Thus, this gives us good evidence that pharmacodynamic processes
account for most of the differential effect of ethanol in these two lines of mice. The
second principle illustrated by the LS-SS example is that differential response pro-
gressed at nearly steady rate across generations, rather than within one or two
generations of selection. This gives good evidence for hypnotic sensitivity to ethanol
as being influenced by multiple genes (alleles) with algebraic additive effects (some
increase effect, some decrease) and that across generations, each line accumulates
more of the alleles that push the phenotype in the direction of selection. In the
pharmacokinetic approach, most of the attention is focused on single genes and their
alleles that alter the function of metabolizing enzymes. In the pharmacodynamic
approach, most of the phenotypes measured show continuous variation, respond to
selection very much like the LS-SS, thus indicating polygenic influence.
29.2.3 LET US NOT FORGET THE INFLUENCE OF THE ENVIRONMENT

Individual differences in drug response are not entirely caused by differences in
genetic makeup. Environment is broadly defined and can include the immediate
external or interior milieu, rearing conditions during infancy, nutrition and even prior
exposure to drugs. The environment may also include hormonal or other humoral
status. Early handling, i.e., removal of pups from the nest for brief periods, can alter
preference for alcohol in some inbred mouse strains but not others,10 individual vs.
group housing can alter ethanol sensitivity11 and iron deficiency early in life can alter
sensitivity to and self administration of cocaine.12,13 It behooves the researcher, there-
fore, to know how environmental conditions can affect the phenotype and how any
environmental perturbation may interact with the genetic makeup of the subject.
29.3 SINGLE GENES AND PHARMACOGENETICS

In Chapter 13, the relationship among the short allele of the 5HT transporter pro-
moter, stressful life events and depression was discussed. A recent study has reported
that individuals who carry at least one long allele of 5-HTTLPR fare better in long-
term antidepressant treatment than those who are homozygous for the short allele.14
In animals, approaches to single-gene effects on drug sensitivity include induced
mutations by irradiation, chemical mutagenesis, and creation of targeted null
mutants. Genes may also be amplified, i.e., by insertion of multiple copies of desired
alleles into the genome. These models can be useful for elucidating the role of
specific proteins, receptors, etc. in drug action. For example, in the serotonin 1b
receptor null mutant mouse, treatment with paroxetine, a selective serotonin reuptake
inhibitor used to treat depression causes a twofold increase in extracellular serotonin
compared with treatment with the same drug in wild-type mice. In humans, allelic
variants of the 5HT1B receptor gene promoter region have been described.15 The
allelic variants are associated with transcriptional activity of the receptor gene,
showing a more than twofold difference between the highest efficiency and lowest
efficiency alleles. Although the allelic configuration may not affect the action of
specific serotonin reuptake inhibitors, the fact that they do exist, in light of the work
in the null mutant mice should give cause for further investigation into possible
ramifications for possible individual differences in drug treatment for depression.
29.4 PHARMACOGENETICS AND COMPLEX

TRAITS ANALYSIS
A rapidly developing approach in pharmacogenetics is complex traits analysis. In
this approach, we recognize the multiple gene influence on most of our phenotypes
of interest and how these genes interact with each other and with the environment.
The major features of complex traits analysis include quantitative trait analysis, gene
network analysis, and gene expression. The object of complex traits analysis is to
identify genes that influence target tissue sensitivity to drugs and to identify related
gene networks. The usual place to start is to specify precisely the phenotype of
interest and perhaps related phenotypes. Quantitative trait analysis is then performed
on data collected from a genetic reference population. Quantitative trait locus (QTL)
analysis gives general locations on chromosomes containing phenotype-related
genes. The next task is to narrow the chromosomal areas, nominate candidate genes
and then verify whether the candidate gene is truly the gene of interest. As an
example of preliminary QTL analysis, Jones et al.16 conducted a QTL analysis of
cocaine-related behaviors and neurochemistry in the BXD/Ty recombinant inbred
mouse panel. The behavioral phenotypes measured were spontaneous activity (loco-
motion, exploration, stereotyped movements and thigmotaxis) under 4 doses of
cocaine, 5, 15, 30 and 45 mg/kg (vs. saline) administered intraperitoneally. In a
separate set of animals, density of dopamine receptors (D1 and D2) and the dopamine
transporter were measured by ligand binding in homogenates derived from the
prefrontal cortex, caudate-putamen, nucleus accumbens, and ventral midbrain. The
purpose of measuring drug response and then measuring putatively related neuro-
biological parameters was to find QTL for the behavioral measures, QTL for the
neurobiological measures and QTL common to both. The last of these would provide
more robust evidence for common genetic mechanisms than would the individual
QTL. Of course the related behavioralneurochemical parameters would have to

make sense in terms of what we know about the neurobiology of the drug action.
Numerous QTL were identified for the behavioral effects of cocaine and neurochem-
ical indices and at least four of these QTLs were common between cocaine-related
behavior and dopamine neurobiology. Of particular interest were markers on chro-
mosome 15 that correlated with both dopamine receptor types in the caudate-
putamen and with locomotion and stereotyped movements. This makes sense in
that the caudate-putamen is involved in movement and is sensitive to psychostimu-
lants and that dopamine actions in this part of the striatum affect movement. Thus,
we have identified QTL that indicate common genetic mechanisms for the expression
of dopamine receptors and for the behavioral actions of cocaine.
29.5 QTL TO QTGENE

QTL analysis is useful for pointing to chromosomal locations containing genes that
modulate drug actions and may also indicate how many genes are involved. The
next step is to move from QTL to actually identifying the genes in question. This
can be a lengthy process, beginning with narrowing the region of the chromosome
to an area that contains dozens rather than hundreds to thousands of genes. QTL
analysis of neurobehavioral traits has a short history of a little more than a dozen
years. In the initial stages, very few genes were identified; however, as analytical
techniques, including high throughput sequencing, have evolved, the pace of
QTGene identification has picked up dramatically. At this writing, more than 30
genes have been identified following QTL mapping and the number is growing as
more refined techniques are brought into play.17 Indeed such an approach has been
undertaken in the laboratory of Buck who identified Mpdz on mouse chromosome
4 as one QTGene underlying susceptibility to alcohol and barbiturate withdrawal-
induced seizures.18
29.6 PHARMACOGENOMICSTHE NEXT STEP

For the most part, pharmacogenetics has been concerned with identifying individual
genes involved with drug safety and sensitivity. Pharmacogenomics casts a wider
net by probing the entire genome for genes and gene products that influence safety,
sensitivity and efficacy of drugs.19 Indeed, QTL analysis has provided one approach
toward developing the methods in multiple-gene identification. Evolving bioinfor-
matics techniques will help us to identify pleiotropic effects of genes, genegene
interactions and gene networks that affect drug actions. The interested reader is
directed to http://www.genenetwork.org for further information. This is a user-
friendly Web site that contains phenotypic data of many kinds, including published
pharmacological data from genetic reference populations. An effective tutorial
guides the user through QTL analysis, genetic correlations, gene networks, epistasis
and more. The site is continually updated and refined.
29.7 SUMMARY
Pharmacogenetics is a small, but important component of the study of individual
differences in response to environmental challenges to homeostatic biological sys-
tems. It is not different in principle from inborn errors of metabolism, dietary
problems such as lactose intolerance, disease resistance, or individual differences in
sensitivity to toxins. What makes pharmacogenetics worthy of study is the great
importance of its impact on human health and the eventual development of drug
treatment strategies based on an individuals genetic constitution.
REFERENCES
1. Garrod, A.E. The incidence of alkaptonuria: a study in chemical individuality. The
Lancet ii:16161620, 1902.
2. Flling, A. Excretion of phenylpyruvic acid in urine as a metabolic. anomaly in
connection with imbecility. Nord Med Tidskr. 8: 10541059, 1934
3. Kalow, W. Familial incidence of low pseudocholinesterase level. The Lancet
2:576577, 1956
4. Crabb, D.W., Edenberg, H. J., Bosron, W. F., Li, T.-K. Genotypes for aldehyde
dehydrogenase deficiency and alcohol sensitivity: the inactive ALDH2*2 allele is
dominant. J. Clin. Invest. 83: 314316, 1989.
5. Yokoyama, A., Kato, H., Yokoyama, T., Tsujinaka, T., Muto, M., Omori, T., Haneda,
T., Kumagai, Y., Igaki, H., Yokoyama, M., Watanabe, H., Fukuda, H., Yoshimizu, H.
and Ingelman-Sundberg, M. Genetic polymorphisms of cytochrome P4502D6
(CYP2D6): clinical consequences, evolutionary aspects and functional diversity. The
Pharmacogenetics Journal 5:613, 2005.
6. Mardones, J. On the relationship between deficiency of B vitamins and alcohol intake
in rats. Q J Stud Alcohol 12:563575, 1951.
7. McClearn, G.E. and Kakihana, R. Selective breeding for ethanol sensitivity: short-
sleep and long-sleep mice. In Development of Animal Models as Pharmacogenetic
Tools G.E. McClearn, R. A. Deitrich and V.G. Erwin, eds., USDHHS-NIAAA
Research Monographs No 6, Washington, pp. 147159, 1981.
8. Tabakoff, B. and Ritzmann, R.F. Acute tolerance in inbred and selected lines of mice
Drug Alcohol Depend 4:8790, 1979.
9. Jones, B., Goldstine, R., Kegel, M., Gurley, M., and Reyes, E. The influence of
infantile handling, age and strain on alcohol selection in mice. Alcohol, 2:327331,
1985.
10. Jones, B. C., Connell, J. M. and Erwin, V. G. Isolate housing affects ethanol sensitivity
in long-sleep and short-sleep mice. Pharmacol Biochem Behav, 35:469472, 1990.
11. Erikson, K.M., Jones, B.C., and Beard, J.L. Iron deficiency alters dopamine trans-
porter functioning in rat striatum. Journal of Nutrition, 130:28312837, 2000.
12. Jones, B.C., Wheeler, D.S., Beard, J.L. and Grigson, P.S. Iron deficiency decreases
acquisition of and suppresses responding for cocaine. Pharmacol Biochem Behav,
73:813819, 2002.
13. De Groote, L., Olivier, B. and Westenberg, G.M. The effects of selective serotonin
uptake inhibitors on extracellular 5-HT levels in the hippocampus of 5-HT1B receptor
knockout mice. Eur J Pharm 439:93100, 2002.
14. Duan, J., Vander Molen, J.E., Martinolich, L., Mowry, B.J., Levinson, D.F., Crowe,
R.R., Silverman, J.M. and Gejman, P.V. Polymorphisms in the 5-untranslated region
of the human serotonin receptor 1B (HTR1B) gene affect gene expression. Mol
Psychiat 8:901910, 2003.
15. Jones, B.C., Tarantino, L.M., Rodriguez, L.A., Reed, C.L., McClearn, G.E., Plomin,
R. and Erwin, V.G. Quantitative trait loci analysis of cocaine-related behaviours and
neurochemistry. Pharmacogenetics 9:607617, 1999.
16. DiPetrillo, K., Wang, X., Stylianou, I.M. and Paigen, B. Bioinformatics toolbox for
narrowing rodent quantitative trait loci. TRENDS in Genetics, 21:68392, 2005.
17. Shirley, R.L., Walter, N.A., Reilly, M.T., Fehr, C. and Buck, K.J. Mpdz is a quanti-
tative trait gene for drug withdrawal seizures. Nat Neurosci 7:699700, 2004.
18. Kalow, W., Meyer, U.A. and Tyndale, R.F., eds. Pharmacogenomics, Marcel Decker,
New York, 1991.
30 Alcohol
Psychopharmacogenetics
John C. Crabbe
CONTENTS
30.1 Introduction .................................................................................................. 457

30.2 Animal Pharmacogenetic Studies ................................................................ 458
30.2.1 Studies with Inbred Strains.............................................................. 458
30.2.2 Studies with Selectively Bred Lines................................................ 459
30.2.3 Gene Identification and Targeting Studies ...................................... 460
30.2.3.1 QTL Mapping Studies ...................................................... 460
30.2.3.2 Transgenic, Null Mutant, Antisense, and
Other Gene-Targeted Expression Studies ........................ 461
30.3 Human Studies ............................................................................................. 462
30.3.1 Twin and Adoption Studies ............................................................. 462
30.3.2 Candidate Gene Studies ................................................................... 463
30.3.3 Identification of Risk Markers and QTLs ....................................... 463
30.4 Gene Expression Analyses........................................................................... 464
30.5 Conclusions .................................................................................................. 464
References.............................................................................................................. 464
30.1 INTRODUCTION
Of all drugs of abuse studied to determine genetic contributions to susceptibility to
their effects, alcohol has been by far the most frequent focus. Both animal model
and human genetic studies will be mentioned here. A historically rich literature with
genetic animal models has explored the general contribution of genetics to alcohol
responsiveness, and has been helpful in elucidating the drugs mechanism of action
on the nervous system. Some studies using genetic animal models have attempted
to identify specific genes that increase or decrease responsiveness for a number of
alcohols effects. Most human studies have been of alcoholics and their relatives,
and have compared relative risks for alcohol dependence disorders in twins, adoptees,
or other relatives. More recently, human studies have also addressed the goal of
identifying individual genes that might contribute to alcoholism risk, or to individual
differences in endophenotypes, which also are associated with alcoholism risk.
457
A comprehensive review of these studies is beyond the scope of this chapter, but a
few examples will be discussed below, and readers will be referred to relevant
reviews.
30.2 ANIMAL PHARMACOGENETIC STUDIES

Two standard pharmacogenetic methods have been employed to investigate the
genetic contribution to alcohol sensitivity in rodents. Some studies have been con-
ducted with standard inbred strains, and others have used lines selectively bred for
enhanced or diminished alcohol response.
30.2.1 STUDIES WITH INBRED STRAINS

Inbred strains have been created through generations of brothersister matings, and
each strain gradually has lost its genetic diversity such that all same-sex members
of a fully inbred strain share two copes of the same allele for each gene. The genotype
is faithfully recreated each generation, with the result that data collected in an inbred
strain many years (and generations) ago can be compared with data collected now.
Since 1959, we have known that inbred mouse strains were characterized by large
and reproducible differences in their drinking preference of weak alcohol solutions.43
C57BL/6 and related strains have a high preference for alcohol, DBA/2 and related
strains are extreme avoiders, and other strains tend to fall somewhere between. This
literature has been systematically reviewed.14,16,18,40,51
Ignoring for the moment the more specialized recombinant inbred (RI) mouse
strains, which will be discussed in a later section, systematic studies with multiple
inbred mouse strains have now been performed for many responses to alcohol. These
include responses representing low-dose stimulant effects,50 moderate-dose effects
such as ataxia and hypothermia, and high-dose effects such as hypnosis and acute
and chronic withdrawal severity.13
One of the features of studies with inbred strains is that pleiotropic gene influ-
ences can be detected. That is, if a gene or genes act to increase sensitivity to one
effect of ethanol in certain strains, and if those strains are also found to be more
sensitive to another effect of ethanol, this provides evidence for the influence of the
same genes on the second response as well.17 In general, the genetic influences on
responses to ethanol appear to be discrete, and there are few striking genetic corre-
lations among different behavioral response domains across inbred strains.13
It is also interesting to note that there are striking strain differences in ethanol
metabolism.19,35,58 However, studies with inbred mice have failed to detect common
genetic influences across strains between ethanol metabolism and behavioral sensi-
tivity.13 This suggests that the strain differences in behavioral sensitivity are mediated
by physiological differences in brain circuits responsive to the alcohol rather than
in the availability of alcohol in the brain.
The stability and availability of inbred mouse strains has led to the recent
development of a systematic database comprising physiological, neurobiological and
behavioral phenotypes of many sorts, including drug and alcohol responses, for
Alcohol Psychopharmacogenetics 459
40 common inbred strains. This effort, the Mouse Phenome Project, can be assessed
to search for evidence of pleiotropic gene effects.48
30.2.2 STUDIES WITH SELECTIVELY BRED LINES

There are more than a dozen sets of mouse and rat lines bidirectionally selected for
their enhanced or diminished response to alcohol.1,7,19,40,51 Selective breeding for
increased intensity of a trait enriches the selected line for alleles that increase the
trait, and when such selected lines are compared, pleiotropic effects of those alleles
can be identified when they produce concomitant increases in a correlated response
to selection.17 The most frequently selected trait is high- or low-alcohol preference
drinking when animals are offered a choice between alcohol and water: at least six
sets of selectively bred rat lines and one set of mouse lines have been selected for
variants of this trait.7,40 The preponderance of studies have been conducted with
preferring (P) and nonpreferring (NP) rat lines, and information characterizing these
lines is extensive.
The similarity of the several selection studies offers a unique opportunity to
identify traits that are reliably differentiated in drinking and nondrinking lines. Many
of the older selection studies were not replicated, which increases the risk that
differences between the selected lines are due to chance fixation of genes unrelated
to ethanol preference.17 However, a few behavioral and neurochemical traits have
been systematically sampled in multiple preference-selected lines, and some appar-
ent commonalities seem likely to exist. P rats show more alcohol tolerance devel-
opment, and it lasts longer, than NP rats. This difference is shared by both pairs of
HAD vs. LAD, and by AA vs. ANA rats as well.40 This could contribute to the
relatively higher drinking of the P, HAD, and AA lines.
The neurotransmitter serotonin (5-HT) and its metabolite 5-HIAA is relatively
depleted in several brain areas in P rats vs. NP rats. This difference is also seen in
HAD vs. LAD rats, but has not been found to characterize the AA vs. ANA rat
lines.40 A possible reason for the discrepant findings with the AA/ANA rats is that,
in contrast to the NP and LAD rats, ANA rats accumulate more acetaldehyde than
AA rats after administration of alcohol, so the toxicity of acetaldehyde may serve
to limit alcohol consumption in these selected lines. Another potentially important
difference between genetically high- and low-drinking lines is in the mesolimbic
dopamine system. P rats will self-administer ethanol directly into the ventral teg-
mental and current studies are exploring the pharmacological control of the circuits
regulating this response.56,57
Several of the preference selected lines were compared in a set of interesting
operant studies that characterized their behavioral patterns of drinking with the goal
of understanding how alcohols rewarding effects were experienced by these rats.29,59
The genotypes selected for high preference tended to self-administer more ethanol
under most conditions studied, suggesting that some genes affect both operant self-
administration and two-bottle preference drinking. However, the high-preference
lines did not always behave similarly, and the overall picture was that operant and
two-bottle preference methods for studying self-administration were influenced
mostly by rather different genes.
Studies with lines selected for low-dose sensitivity to ethanol, measured by

increases in locomotor activity, have been reviewed elsewhere,50 as have those for
alcohols sedative effects, measured as a decrease in core body temperature after
injection.51 A long-term project has generated lines of mice (long- and short-sleep,
or LS and SS) and more recently rats (high and low alcohol sensitive, or HAS and
LAS) for sensitivity to the high-dose effect of ethanol to cause loss of righting reflex.
Studies with these animals number in the hundreds, and have been periodically
reviewed.1,19,51 These and other genetic studies implicate the neurotransmitter GABA
as an important determinant of the genetic differences in sensitivity to high-dose
ethanol.8
Finally, another long-term project has created mice that are withdrawal seizure-
prone or -resistant (WSP and WSR) by exposing them to ethanol vapor for 72 hrs
and measuring handling-induced convulsion severity during withdrawal. Studies
with these animals have also been discussed in detail elsewhere, and have revealed
many interesting features about chronic neuroadaptation to alcohol.7,11,46 Perhaps
most interesting is the finding that a genetic susceptibility to ethanol withdrawal
severity is shared with susceptibility to withdrawal from other depressant drugs such
as barbiturates and benzodiazepines,12 a finding also substantiated in inbred strains.47
30.2.3 GENE IDENTIFICATION AND TARGETING STUDIES

Three general approaches have been taken to explore the role of specific genes in
modulating alcohol sensitivity. The first approach has been to identify novel genes
by using quantitative trait locus (QTL) gene mapping to discover the presence of
chromosomal markers near genes affecting alcohol responses. QTLs have been
reported for ethanol drinking, sensitivity to various acute effects of alcohol, toler-
ance, and dependence liability, which is inferred from studies of withdrawal when
alcohol is discontinued. The second strategy is employed where specific candidate
genes have been identified as potentially important mediators of susceptibility to
alcohol. Classically, such studies employ Northern or Western analyses to study
mRNA or protein levels, respectively, for a gene of interest. The use of antisense
oligodeoxynucleotide treatments to reduce gene translation, or, more recently, RNA
inhibitors to modulate transcription, has been reported in only a few studies. Many
studies have now used gene targeting technologies to create overexpression or null-
mutant (knockout) lines of mice by targeting specific genes of interest. The success
of many recent candidate gene studies with animals has been noted.62
The third, and most recent, method is a hybrid of the first two. Analogously to
QTL mapping, microarray chips can simultaneously study many thousands of genes.
However, where QTL mapping seeks to associate degree of alcohol response with
a specific variation in gene sequence mapped to a specific chromosomal location,
expression microarrays focus on differences in expression in any gene probed.
30.2.3.1 QTL Mapping Studies
Most alcohol QTL studies have been carried out using the BXD RI strains of mice
as a starting point for rough QTL identification, followed by verification tests in
other populations, such as the F2 cross of C57BL/6J and DBA/2J, lines selectively
bred from the F2 cross, or other selected populations. The experiments that have
progressed the furthest have shown that three loci for acute alcohol withdrawal
severity have been definitively mapped with very high LOD scores to regions of
mouse chromosomes 1, 4, and 11.10 Genes encoding the "1, "6, and (2 subunits of
the neuroinhibitory GABAA receptor map near the chromosome 11 QTL. Studies
using inbred strains, lines selectively bred from the B6D2F2 population, and a
number of so-called congenic strains were then used to reduce the area of genome
surrounding the chromosome 4 QTL until it contained only a handful of candidate
genes.28 The remaining candidate genes were gradually eliminated because they did
not differ in sequence, expression, or function in ways that systematically co-varied
with alcohol withdrawal severity. In the end, a single gene remained, Mpdz, which
must be the QT Gene for this QTL. MPDZ is a multiple PDZ domain protein that
affects the coupling of neurotransmitters with their receptors, including the binding
of serotonin with the 5-HT2 receptor. This gene also affects the severity of with-
drawal from pentobarbital, another nervous system depressant that interacts with
GABA receptors. This provides another example of pleiotropic gene effects, and
Mpdz may indeed also play a role in susceptibility to seizures more generally, in
addition to its specific role in alcohol and pentobarbital withdawal.28,63
For sensitivity to high-dose ethanol, five significant QTLs with high LOD scores
have been identified on chromosomes 1, 2, 8, 11, and 15.41 Starting with the LS and SS
selected lines and using RI strains and crosses derived from them, the QTLs were nom-
inated and subsequently verified in a multistage procedure much like that used for acute
alcohol withdrawal.24 Some interesting potential candidate genes near the QTL regions
include Ntsr on chromosome 2, which codes for the high-affinity neurotensin receptor.
Several different research groups have reported mapping QTLs for ethanol
preference drinking and related traits.3,44,45,52,65 The significant QTLs are found on
chromosomes 2, 9, 11 and 15, and several other QTLs have been suggested on
chromosomes 1, 3, 4, 5, 7, and 13. The chromosome 9 QTL is near genes encoding
both the dopamine D2 and serotonin 5-HT1B receptor, Drd2, and Htr1b, respectively.
The results of these studies are not currently in complete agreement, probably
because different crosses, sexes, strains, and mapping strategies were employed, as
well as different measures of preference-related drinking. However, a meta-analysis
revealed a remarkable degree of convergence of findings for several QTLs.6
A number of other alcohol-related traits are also being mapped, although the
projects are in earlier stages. One set of studies mapping the effect of ethanol to
stimulate locomotor activity in mice has used a variety of mapping methods.20,34 Other
traits being mapped include: ethanol-conditioned place preference and taste aversion;
hypothermia; traits related to ataxia; tolerance to some of these effects; and ethanols
anxiolytic effects. Progress in ethanol QTL mapping has been reviewed recently.9,15,30,49
30.2.3.2 Transgenic, Null Mutant, Antisense, and Other

Gene-Targeted Expression Studies
GABAA receptors are known to be important modulators of many behavioral effects

of ethanol. Ethanol enhances GABA stimulated function in vitro, and when antisense
oligodeoxynucleotides to several GABAA receptor subunits were tested in an early

study, ethanol enhancement was prevented only by an antisense probe to receptors
containing the (2L subunit. This subunit differs from (2S only by bearing an additional
8 amino acids which contain a consensus site for phosphorylation by protein kinase
C. Site-directed mutagenesis within this phosphorylation site blocked ethanols
effects.66 As technology developed, subsequent studies showed that null mutant mice
that lack the "6 subunit did not differ from wild types in sensitivity to loss of righting
reflex after ethanol.36 As more subunit null mutants have been studied and more
specific genetic tools have become available, it has become possible to see how
alcohols effects can be understood by comparison with other anesthetic agents with
strong effects on GABA systems.37
Gene targeting strategies have also been employed to study the role of trans-
forming growth factor ", insulin-like growth factor 1, and many other candidate
genes in alcohol responses. Gene targeting studies must be interpreted cautiously to
avoid attributing the influence of background genes to the specific gene that has
been transgenically altered.31 For example, the effect of a gene deletion may only
be detectable on certain genetic backgrounds.53 The strengths and weaknesses of the
various gene targeting approaches applied to alcohol research have been dis-
cussed.14,68 Some studies also have explored the induction of the immediate early
gene, c-fos, as a marker for intracellular changes following ethanol administration.9,69
Initial identification of brain regions activated by ethanol can then be followed by
studies identifying responsive circuits.4
30.3 HUMAN STUDIES

30.3.1 TWIN AND ADOPTION STUDIES
Genetic influence on developing alcoholism has been supported by more than 20
years of controlled studies using twins and adoptees.23,25,32 Early studies used samples
from Scandinavia because the national record keeping systems made the process of
ascertaining subjects and finding their relatives fairly straightforward. Although
these experiments reported a substantial genetic component to risk, the definitions
of alcoholism and alcohol-related problems may have been somewhat idiosyncratic
to Scandinavian populations, possibly because of the strong cultural determinants
of drinking patterns that differ among countries.
Recent studies with populations in the United States have used more modern
diagnostic criteria, and have largely yielded similar estimates of the importance of
inheritance. Such studies have always faced the difficulty of defining the phenotype.
The standard diagnoses of alcoholism are generally based on possession of a number
of symptoms from a list of possible criteria, which range from medical complications
(e.g., compromised liver function) to social (e.g., loss of significant relationship
attributable to drinking) or functional problems (e.g., loss of job). To give an exam-
ple, different diagnoses are available for alcohol abuse vs. alcoholism. Are these
related disorders on a continuum, or are they distinct (the latter would imply that
separate genetic influences should be discernible)? A related complexity is that there
is extensive overlap, or comorbidity, among alcoholism and other individual traits,
such as: other psychiatric disorders (e.g., antisocial personality, depression, attention
deficit hyperactivity disorder); personality characteristics (e.g., impulsivity, aggres-
siveness); and substance abuse.54 Disentangling the chains of cause and effect among
these different diagnostic aspects makes this a very difficult area of research to
interpret. Finally, each typology for conceptualizing alcohol dependence disorders
leads to emphasis on different interactions between heritable risk factors and the
modulating effects of environmnental factors that are also important for turning risk
into diagnosis.33,38
Recently, much attention has turned to analyzing identifiable aspects of alco-
holism and alcohol responses for their separable genetic determinants. By using
other heritable traits as potential indicators, it is hoped that the genetic basis for
such traits will make it easier to determine genebehavior relationships than is trying
to study the diagnoses directly. Such studies are meeting with some success.21,26
30.3.2 CANDIDATE GENE STUDIES

The human alcoholism literature offers a clear example of the influence of a specific
gene on alcoholism susceptibility. Human populations possess several variant iso-
forms of the metabolic enzymes, alcohol dehydrogenase (ADH) and aldehyde dehy-
drogenase (ALDH). Asian populations have a high frequency of specific ALDH2
variants that are associated with slow acetaldehyde metabolism, and consequently,
facial flushing, nausea, and anhedonia after alcohol ingestion. Possession of these
ALDH2 alleles has been shown to reduce risk for alcoholism.67 Other polymorphisms
in ADH genes have a similar but smaller protective effect.21
30.3.3 IDENTIFICATION OF RISK MARKERS AND QTLS

Association and linkage studies can only genotype individuals for a limited subset
of their total DNA sequence, and therefore must compromise between two choices
genome-wide screens using markers spread widely (which may therefore detect most
signals but provide little information about exactly where the responsible gene is)
or focused screens using densely spaced markers, which can localize the signal
better but may miss signals that are not very near the targeted regions. Many bio-
markers of genetic risk have been suggested in studies with human populations, such
as monoamine oxidase and adenylyl cyclase. Generally, these studies remain con-
troversial.2 As studies with genetic animal models increasingly suggest specific
candidate genes, it was to be expected that studies with human populations would
increasingly examine allelic variation in detail for areas surrounding those genes.
Some such studies are now providing evidence for influential genomic regions.21
An interesting risk marker for alcoholism has been revealed by studies of
individuals who have are family history positive (FHP) vs. negative (FHN) for
alcoholism. FHP individuals when tested as young men showed lower subjective
response to alcohol than FHN controls, and a follow-up study showed that low initial
responders (whether of FHP or FHN background) went on to develop alcoholism
at higher rates than high responders.60,61 Another such risk marker that is not iden-
tified with a particular gene is the lower P300 event-related potential that characterize
FHP subjects.5
The National Institute on Alcohol Abuse and Alcoholism is currently sponsoring

a large multisite study called the Collaborative Study on the Genetics of Alcoholism
(COGA). The goal of this project is to perform a large, population-based association
study looking for particular genetic polymorphisms that indicate QTLs associated
with alcoholism diagnosis, P300 event-related potentials, and other traits potentially
associated with alcoholism and alcohol dependence. This study is described else-
where.55 Some genomic regions have appeared from these analyses as having poten-
tial linkages, including those containing GABA receptor subunit genes.22,64
30.4 GENE EXPRESSION ANALYSES

A few studies have now applied microarray technology to explore tissue from human
alcoholics for potential gene expression differences vs. controls. These studies are
quite preliminary, but as the technology matures, have the potential to find important
genes in addition to those that differ by virtue of their sequence polymorphisms.27,39,42
30.5 CONCLUSIONS
The field of alcohol pharmacogenetics offers an interesting viewpoint on the explo-
sive impact modern molecular biological methods have had on genetic analyses. The
rich history of standard genetic animal model research into alcohols effects has
provided many cases where clear genetic influence was known to exist. Thus, the
move to identification of specific candidate genes and markers occurred quite rapidly
after the advent of dense genetic maps and the ability to target specific genes for
deletion and overexpression. Because there are readily available populations of
human alcoholics with well-defined pedigrees, it will be possible for the animal
model and human genetic findings to be related rather directly. This offers the hope
that in addition to improving our understanding of how alcohol works, genetic studies
will be able to provide novel methods for identifying at-risk individuals and for
developing new pharmacotherapies to attack this major disease.
ACKNOWLEDGMENTS
Preparation of this chapter was supported by a grant from the Department of Veterans
Affairs and by NIH Grants from NIAAA and NIDA.
REFERENCES
1. Allan AM, Harris RA. Neurochemical studies of genetic differences in alcohol action,
In: Crabbe JC, Harris RA, editors. The Genetic Basis for Alcohol and Drug Actions.
New York: Plenum; 1991. pp. 10552.
2. Anthenelli RM, Tabakoff B. The search for biological markers. Alcohol Health and
Research World 1996;19:17681.
3. Bachmanov AA, Reed DR, Li X, Li S, Beauchamp GK, Tordoff MG. Voluntary
ethanol consumption by mice: genome-wide analysis of quantitative trait loci and
their interactions in a C57BL/6ByJ x 129P3/J F2 intercross, Genome Res.

2002;12:125768.
4. Bachtell RK, Tsivkovskaia NO, Ryabinin AE. Strain differences in urocortin expres-
sion in the Edinger-Westphal nucleus and its relation to alcohol-induced hypothermia.
Neuroscience 2002;113:42134.
5. Begleiter H, Porjesz B. Neurophysiological phenotypic factors in the development of
alcoholism, In: Begleiter H, Kissin B, editors. Genetic Factors and Alcoholism. New
York: Oxford University Press; 1996. pp. 26993.
6. Belknap JK, Atkins AL. The replicability of QTLs for murine alcohol preference
drinking behavior across eight independent studies. Mamm Genome 2001;12:89399.
7. Browman, K. E., Crabbe, J. C., and Li, T.-K. Genetic strategies in preclinical sub-
stance abuse research. Bloom, F. E. and Kupfer, D. J. Psychopharmacology: A Fourth
Generation of Progress (CD-ROM version 3). 2000. Lippincott, Williams & Wilkins.
Ref Type: Electronic Citation (url = http://www.acnp/org/citations/GN400000077).
8. Buck KJ. Molecular genetic analysis of the role of GABAergic systems in the
behavioral and cellular actions of alcohol. Behav. Genet. 1996;26:31323.
9. Buck KJ, Crabbe JC, Belknap JK. Alcohol and other abused drugs, In: Pfaff DW,
Berrettini WH, Joh TH, Maxson SC, editors. Genetic Influences on Neural and
Behavioral Functions. Boca Raton FL: CRC Press; 2000. pp. 15983.
10. Buck KJ, Metten P, Belknap JK, Crabbe JC. Quantitative trait loci involved in genetic
predisposition to acute alcohol withdrawal in mice. J Neurosci 1997;17:394655.
11. Crabbe JC. A genetic animal model of alcohol withdrawal. Alcohol Clin. Exp. Res.
1996;20(8):96A100A.
12. Crabbe JC, Merrill CD, Belknap JK. Acute dependence on depressant drugs is
determined by common genes in mice. Journal of Pharmacology and Experimental
Therapeutics 1991;257:2:66367.
13. Crabbe, J. C., Metten, P., Cameron, A. J., and Wahlsten, D. An analysis of the genetics
of alcohol intoxication in inbred mice. Neuroscience and Biobehavioral Reviews,
2005;28:785802.
14. Crabbe JC, Phillips TJ. Pharmacogenetic studies of alcohol self-administration and
withdrawal. Psychopharmacology 2004;174:53960.
15. Crabbe JC, Phillips TJ, Buck KJ, Cunningham CL, Belknap JK. Identifying genes
for alcohol and drug sensitivity: recent progress and future directions. Trends Neurosci
1999;22:17379.
16. Crabbe JC, Phillips TJ, Cunningham CL, Belknap JK. Genetic determinants of ethanol
reinforcement. Annals NYAS 1992;654:30210.
17. Crabbe JC, Phillips TJ, Kosobud A, Belknap JK. Estimation of genetic correlation:
interpretation of experiments using selectively bred and inbred animals. Alcohol Clin.
Exp. Res. 1990;14:2:14151.
18. Crawley JN, Belknap JK, Collins A, Crabbe JC, Frankel W, Henderson N et al.
Behavioral phenotypes of inbred mouse strains: implications and recommendations
for molecular studies. Psychopharmacology 1997;132:10724.
19. Deitrich RA, Baker RC. Genetic influences on alcohol metabolism and sensitivity to
alcohol in animals, In: Begleiter H, Kissin B, eds. The Genetics of Alcoholism. New
York: Oxford University Press; 1995. p. 13964.
20. Demarest K, Koyner J, McCaughran JJ, Cipp L, Hitzemann R. Further characteriza-
tion and high-resolution mapping of quantitative trait loci for ethanol-induced loco-
motor activity. Behav. Genet. 2001;31:7991.
21. Dick DM, Foroud T. Candidate genes for alcohol dependence: a review of genetic
evidence from human studies. Alcohol Clin Exp Res 2003;27:86879.
22. Dick DM, Nurnberger J, Jr., Edenberg HJ, Goate A, Crowe R, Rice J et al. Suggestive
linkage on chromosome 1 for a quantitative alcohol-related phenotype. Alcohol Clin.
Exp. Res. 2002;26:145360.
23. Ehringer MA, Thompson J, Conroy O, Goldman D, Smith TL, Schuckit MA et al.
Human alcoholism studies of genes identified through mouse quantitative trait locus
analysis. Addiction Biology 2002;7:36571.
24. Ehringer MA, Thompson J, Conroy O, Yang F, Hink R, Bennett B et al. Fine mapping
of polymorphic alcohol-related quantitative trait loci candidate genes using interval-
specific congenic recombinant mice. Alcohol Clin. Exp. Res. 2002;26:160308.
25. Enoch MA, Goldman D. The genetics of alcoholism and alcohol abuse. Curr
Psychiatry Rep 2001;3:14451.
26. Enoch MA, Schuckit MA, Johnson And BA, Goldman D. Genetics of alcoholism
using intermediate phenotypes. Alcohol Clin. Exp. Res. 2003;27:16976.
27. Fan L, Bellinger F, Ge YL, Wilce P. Genetic study of alcoholism and novel gene
expression in the alcoholic brain. Addict. Biol. 2004;9:1118.
28. Fehr C, Shirley RL, Belknap JK, Crabbe JC, Buck KJ. Congenic mapping of alcohol
and pentobarbital withdrawal liability loci to a <1 centimorgan interval of murine
chromosome 4: Identification of Mpdz as a candidate gene. J Neurosci 2002;22:
373038.
29. Files FJ, Samson HH, Denning CE, Marvin S. Comparison of alcohol-preferring and
nonpreferring selectively bred rat lines. II. Operant self-administration in a continu-
ous-access situation. Alcohol Clin. Exp. Res. 1998;22:214758.
30. Flint J. Analysis of quantitative trait loci that influence animal behavior. J Neurobiol.
2003;54:4677.
31. Gerlai R. Molecular genetic analysis of mammalian behavior and brain processes:
caveats and perspectives. Seminars in the Neurosciences 1996;8:15361.
32. Heath AC. Genetic influences on alcoholism risk: a review of adoption and twin
studies. Alcohol Health and Research World 1995;19:16671.
33. Heath AC, Todorov AA, Nelson EC, Madden PAF, Bucholz KK, Martin NG. Gene-
environment interaction effects on behavioral variation and risk of complex disorders:
The example of alcoholism and other psychiatric disorders. Twin Research 2004;5:3037.
34. Hitzemann R, Demarest K, Koyner J, Cipp L, Patel N, Rasmussen E et al. Effect of
genetic cross on the detection of quantitative trait loci and a novel approach to
mapping QTLs. Pharmacol. Biochem. Behav. 2000;67:76772.
35. Holmes RS. Genetic variants of enzymes of alcohol and aldehyde metabolism.
Alcohol Clin. Exp. Res. 1985;9:53538.
36. Homanics GE, Ferguson C, Quinlan JJ, Daggett J, Snyder K, Lagenaur C et al.
Gene knockout of the alpha6 subunit of the gamma-aminobutyric acid type A recep-
tor: lack of effect on responses to ethanol, pentobarbital, and general anesthetics.
Mol. Pharmacol. 1997;51:58896.
37. Homanics GE, Xu Y, Tang P. Integrated approaches to the action of general anesthetics
and alcohol. Physiol Behav. 2002;77:49599.
38. Johnson EO, van den Bree M, Gupman AE, Pickens RW. Extension of a typology
of alcohol dependence based on relative genetic and environmental loading. Alcohol
Clin. Exp. Res. 1998;22:142129.
39. Lewohl JM, Wang L, Miles MF, Zhang L, Dodd PR, Harris RA. Gene expression in
human alcoholism: microarray analysis of frontal cortex. Alcohol Clin. Exp. Res.
2000;24:187382.
40. Lumeng L, Murphy JM, McBride WJ, Li T-K. Genetic influences on alcohol prefer-
ence in animals, In: Begleiter H, Kissin B, eds., The Genetics of Alcoholism. New
York: Oxford University Press; 1995. p. 165201.
41. Markel PD, Bennett B, Beeson M, Gordon L, Johnson TE. Confirmation of quanti-
tative trait loci for ethanol sensitivity in long-sleep and short-sleep mice. Genome
Res. 1997;7:9299.
42. Mayfield RD, Lewohl JM, Dodd PR, Herlihy A, Liu J, Harris RA. Patterns of gene
expression are altered in the frontal and motor cortices of human alcoholics.
J. Neurochem. 2002;81:80213.
43. McClearn GE, Rodgers DA. Differences in alcohol preference among inbred strains
of mice. Quarterly Journal of Studies on Alcohol 1959;20:69195.
44. McClearn GE, Tarantino LM, Rodriguez LA, Jones BC, Blizard DA, Plomin R.
Genotypic selection provides experimental confirmation for an alcohol consumption
quantitative trait locus in mouse. Mol. Psychiat 1997;2:48689.
45. Melo JA, Shendure J, Pociask K, Silver LM. Identification of sex-specific quantitative
trait loci controlling alcohol preference in C57BL/6 mice. Nat. Genet. 1996;13:14753.
46. Metten P, Crabbe JC. Dependence and withdrawal, In: Deitrich RA, Erwin VG, eds.
Pharmacological Effects of Ethanol on the Nervous System. Boca Raton FL: CRC
Press; 1996. p. 26990.
47. Metten P, Crabbe JC. Genetic determinants of severity of acute withdrawal from
diazepam in mice: commonality with ethanol and pentobarbital. Pharmacol Biochem
Behav 1999;63:47379.
48. Grubb SC, Churchill GA, Boque MA. A collaborative database of infred mouse strain
characteristics. Bioinformatics 2004; 20: 28572859.
49. Palmer AA, Phillips TJ. Quantitative trait locus (QTL) mapping in mice, In: Liu Y,
Lovinger DM, eds. Methods in Alcohol-Related Neuroscience Research. Boca Raton
FL: CRC Press; 2002. p. 130.
50. Phillips TJ. Behavior genetics of drug sensitization. Crit Rev. Neurobiol. 1997;11:2133.
51. Phillips TJ, Crabbe JC. Behavioral studies of genetic differences in alcohol action,
In: Crabbe JC, Harris RA, editors. The Genetic Basis of Alcohol and Drug Actions.
New York: Plenum Press; 1991. p. 25104.
52. Phillips TJ, Crabbe JC, Metten P, Belknap JK. Localization of genes affecting alcohol
drinking in mice. Alcohol Clin. Exp. Res. 1994;18:93141.
53. Phillips TJ, Hen R, Crabbe JC. Complications associated with genetic background
effects in research using knockout mice. Psychopharmacology 1999;147:57.
54. Pickens RW, Svikis DS, McGue M, LaBuda MC. Common genetic mechanisms in
alcohol, drug, and mental disorder comorbidity. Drug and Alcohol Dependence
1995;39:12938.
55. Reich T. A genomic survey of alcohol dependence and related phenotypes: results
from the Collaborative Study on the Genetics of Alcoholism (COGA). Alcohol Clin.
Exp. Res. 1996;20:133A7A.
56. Rodd ZA, Bell RL, Melendez RI, Kuc KA, Lumeng L, Li TK et al. Comparison of
intracranial self-administration of ethanol within the posterior ventral tegmental area
between alcohol-preferring and wistar rats. Alcohol Clin. Exp. Res. 2004;28:121219.
57. Rodd-Henricks ZA, McKinzie DL, Melendez RI, Berry N, Murphy JM, McBride
WJ. Effects of serotonin-3 receptor antagonists on the intracranial self-administration
of ethanol within the ventral tegmental area of Wistar rats. Psychopharmacology
2003;165:25259.
58. Rout UK, Holmes RS. Alcohol dehydrogenases and aldehyde dehydrogenases among
inbred strains of mice: multiplicity, development, genetic studies and metabolic roles.
Addiction Biology 1996;1:34962.
59. Samson HH, Files FJ, Denning C, Marvin S. Comparison of alcohol-preferring and
nonpreferring selectively bred rat lines. I. Ethanol initiation and limited access operant
self-administration. Alcohol Clin. Exp. Res. 1998;22:213346.
60. Schuckit MA. Self-rating of alcohol intoxication by young men with and without
family histories of alcoholism. J. Stud. Alcohol 1980;41:24249.
61. Schuckit MA, Smith TL. An 8-year follow-up of 450 sons of alcoholic and control
subjects. Archives of General Psychiatry 1996;53:20210.
62. Schumann G, Spanagel R, Mann K. Candidate genes for alcohol dependence: animal
studies. Alcohol Clin Exp Res 2003;27:88088.
63. Shirley RL, Walter NA, Reilly MT, Fehr C, Buck KJ. Mpdz is a quantitative trait
gene for drug withdrawal seizures. Nat. Neurosci. 2004;7:699700.
64. Song J, Koller DL, Foroud T, Carr K, Zhao J, Rice J et al. Association of GABAA
receptors and alcohol dependence and the effects of genetic imprinting. Am. J. Med.
Genet. 2003;117B:3945.
65. Vadasz C, Saito M, Balla A, Kiraly I, Gyetvai B, Mikics E et al. Mapping of
quantitative trait loci for ethanol preference in quasi-congenic strains. Alcohol
2000;20:16171.
66. Wafford KA, Burnett D, Harris RA, Whiting PJ. GABAA receptor subunit expression
and sensitivity to ethanol. Alcohol Suppl 1993;2:32730.
67. Wall TL, Ehlers CL. Genetic influences affecting alcohol use among Asians. Alcohol
Health and Research World 1995;19:18489.
68. Wehner JM, Bowers BJ. Use of transgenics, null mutants, and antisense approaches
to study ethanol's actions. Alcohol Clin. Exp. Res. 1995;19:81120.
69. Weitemier AZ, Woerner A, Backstrom P, Hyytia AP, Ryabinin AE. Expression of
c-Fos in Alko Alcohol rats responding for ethanol in an operant paradigm. Alcohol
Clin. Exp. Res. 2001;25:70410.
1903_Index.fm Page 469 Wednesday, July 26, 2006 7:52 PM
Index
A historical research developments, 11
humans, 286287
Aarskog-Scott syndrome, 442 interim solution, 25
ABA (Allen Brain Atlas), 426 mice, 282285, 283, 285286, 287
Abiola studies, 116 phenotypes, 3031
Abkevich studies, 237 rodents, 9
Absence, genetic activity, 23 types, 282
Acceptance of association, 175 voles, 12
Accessory olfactory bulb (AOB), 2223 AILS (advanced intercross lines), 60, 133
Acetaldehyde toxicity, 451, 459 Akaikes information criterion (AIC), 256
Acoustic startle reflex, 295 AKR/J strains, 134
Acoustic stimuli, 320321 Albino strains and albinism, 138, 414
Actin cytoskeleton, 437, 437439 Alcohol, alcohol-related disorders, and abuse,
Activators, regulation of transcription, 81 see also Ethanol and ethanol resistance
Active avoidance paradigms, 295, 298 bipolar disorders, 234
Active genotype-environment correlation, 190 brain dopamine systems, 371372
Addictions, 12, 87, see also specific substance Caenorhabditis elegans, 10
Additive genetic effects and variations, 3334, 34, expression variation sources, 87
40, 184 false positives, 175
Additive model, gene-environment interactions, gene-environment interactions, 35, 191
179 historical research, 7
ADHD (attention-deficit hyperactivity disorder), inbreeding, 136
269 metabolism, 450
Adobe Photoshop, 425 multi-gene disorder, 30
Adoption studies, see also Family and twin pharmacodynamics, 451452
studies; Human studies; Twins studies QTLs, 301
alcohol psychopharmacogenetics, 462463 recurrent unipolar disorders, 236
bipolar disorders, 229 response and sensitivity, 451452
historical research developments, 4, 12 rodents, 9, 11
schizophrenia, 214216 serotonin, 2425
Advanced intercross lines (AILS), 60, 133 Alcohol psychopharmacogenetics
Affected cohorts requirement, 174 adoption studies, 462463
Affective disorders, major animal studies, 458462
bipolar disorders, 228236 candidate gene studies, 463
family studies, 228, 236 fundamentals, 457458, 464
fundamentals, 227, 237238 gene expression analysis, 464
linkage disequilibrium studies, 231236, gene identification, 460462
233 human studies, 462464
molecular linkage studies, 230231, inbred strains studies, 458459
232233, 237 QTLs identification, 463464
recurrent unipolar disorders, 236237 risk marker identification, 463464
twins studies, 228229, 229, 237 selectively bred lines studies, 459460
Affymetrix system, 9091, 381, 416, 418, 420 targeting studies, 460462
Aggression twins studies, 462463
fundamentals, 281282 ALDH2 variants, 450451, 463
gene-environment interactions, 195 Algerian family study, 443
genetic vulnerability, 178 Alkaptonuria, 4, 450
469
Allan studies, 228 Jackson Laboratory, 142143

Allele, W.C., 4 knockout mice, 139
Alleles, gene silencing, 87 laboratory mice strains and substrains, 141,
Allen Brain Atlas (ABA), 425426 142
Allen studies, 438 mixed inbred strains, 138
Allostery, 81 model requirements, 130
Alternative slicing, transcription, 80 multiple inbred strains, 136138
Alternative strategies, high-resolution mapping, outbred strains and colonies, 140
5962, 61 recombinant congenic strains, 137138
Altruism, 268269 recombinant inbred strains, 136137, 137
Alzheimers disease, 65, 138 segregating inbred strains, 138
Ambidirectional dominance, 38, 43 selected strains, 136
Amrein, Bray and, studies, 324 transgenic mice, 138
Amrein studies, 389407 Web sites, 144146
Amygdala hyper-reactivity, 194 wild-derived inbred mice, 140
Analysis of variance (ANOVA) ANOVA (analysis of variance)
effect size, 151152 effect size, 151152
false positives, 153 false positives, 153
gene-gene interactions, 204 gene-gene interactions, 204
inbred strain comparison, 160 inbred strain comparison, 160
one-way design comparison, 161 one-way design comparison, 161
published data estimates, 156 published data estimates, 156
two-way factorial designs, 162163, 165 two-way factorial designs, 162163, 165
Ancient history of behavior genetics, 23, Antidepressant treatment response, 177
see also Historical research development Antisense studies, 461462
Animal Breeders Toolkit, 251, 261 Antisocial behaviors, heritability, 190
Animals, see also specific animal Anttila studies, 204
alcohol psychopharmacogenetics, 458462 Anxiety disorders, 25, 266
gene-environment interactions, 195197, Anxiolytic drugs, 295
196197 AOB (accessory olfactory bulb), 2223
gene-gene interactions, 203204 APP695s we gene, 138
genotype-environment interactions, AR (attributable risk), 170171
195197, 196197 Arginine-vasopressin (AVP), 270271
health quality, JAX rodents, 143 Argonaute protein family, 86
Animals, resources, see also Sample size Aristotle (384-322 B.C.E.), 3, 1718
requirements Asperger syndrome, 178
advanced intercross lines, 133 Association studies
Caenorhabditis elegans, 144 acceptance of association, 175
Charles River Laboratories, 143 advantages, 172173
chemically-induce mutations, mice, 139 case control study, 170, 170171
chromosomal aberration strains, 139140 environment interaction, 178179, 179
chromosome substitution strains, 134 false positives, 174175
co-isogenic strains, 134 family-based study, 171, 171172
Complex Trait Consortium, 143 fundamentals, 169170, 180
congenic strains, 134, 135 gene-environment interaction, 178179,
conplastic strains, 135136 179
consomic strains, 134 genetic heterogeneity, 177178
Drosophila melanogaster, 143144 historical research developments, 13
fundamentals, 130 limitations, 173176
genetically engineered mutant mice, linkage disequilibrium, 173175
138140 neurobehavioral genetic consequences,
genome sequence importance, 131 175176
hybrid crosses, 132133 pedigree analyses, 257258
inbreeding, 131133, 132, 140141, 141 phenotypical heterogeneity, 177
international supply, 142143 power, 172173, 175, 175
Index 471
rejection of association, 175 Behavior Genetics, 6, 8

reproducibility, 176 Behavior genetics, history
Association verification, 393394, 395 1910 to 1952, 45
Attention-deficit hyperactivity disorder (ADHD), 1953 to 1970, 57
269 1971 to 1985, 810
Attributable risk (AR), 170171 1986 to 2005, 1013
Audiogenic seizures, 9 ancient history, 23
Autism, 177178, 271 bees, 11
Autosomal recessive mental retardation, 443 canids, 10
AVP (arginine-vasopressin), 270271 Darwin and Galton, 3
Axel, Buck and, studies, 324 dogs, 7
Drosophila melanogaster, 4, 68, 11
first century, 413
B fundamentals, 2, 1314
humans, 5, 7, 10, 1213
BAC (blood alcohol concentrations), 34, 452, nematodes, 8, 10
see also Alcohol, alcohol-related disorders, preliterate history, 23
and abuse rodents, 45, 79, 1112
Bachner-Melman studies, 263273 Behaviors, C. elegans, see Caenorhabditis
B6;129-Acrv2tm/Zuk strains, 138 elegans
Bacterial artificial chromosomes (BACs), 62, Belay studies, 307314
426 Belknap studies, 11, 153, 371386
Badner and Gershon studies, 231 Belmaker studies, 263273
Bailey, Michael, 12 Benjamini-Hochberg False Discovery Rate
Baker, Laura, 12 (FDR), 91, see also False Discovery
B6.AKR-H2k strains, 134 Rate (FDR)
BALB/cBy strains, 284 Benjamini studies, 153
BALB/c (C) strains, 132, 138 Benjamin studies, 263273
BALB/cJ strains, 195, 373 Bennett, A., studies, 25
BALB/c strains, 116, 135 Bennett, B., studies, 60
Ballabio studies, 65 Bennett, E.L., 7
Baltimore, Meffert and, studies, 386 Ben-Shahar studies, 313
Bank voles, 405 Benzer, Hotta and, studies, 326
Barber studies, 391 Benzer, Seymour, 6, 8
Bargmann, de Bono and, studies, 363 Bergeson, Susan, 111
Bargmann, Wes and, studies, 357 Berrettini studies, 227238
Bargmann and Mori studies, 356357 Bertelsen, Gottesman and, studies, 214
Baron and Richardson studies, 287 Bertelsen studies, 228
Barry, Dave, 201 Bidirectional coactional system, 19
Bartell, Shorey and, studies, 322 BIG (Brain Information Group), 97
B6-15A strain, 124 Bignami, C., 7
Bats, 406 B6-II7r strain, 124
Bausell and Li studies, 157, 165 Billuart studies, 438
BDNF (brain-derived neurotropic factor), 204, BIND database, 104
234 Binding elements, 75
BDs, see Bipolar disorders (BDs) BioCarta pathways, 424
Beagles, 10, see also Dogs Bioinformatics, behavior
Beaman, Betty, 5 applications, 104110
Bees, 11, 14 database design, 9799
Behavioral and Ecological Genetics: A Study in data modeling, 9899
Drosophila, 8 fundamentals, 9697, 110
Behavioral despair, 296 genes, references, 100101
Behavioral genetics, family and twin methods, gene sets, references, 104
183184 genetic reference population, 101104, 102
Behavioral tests standardization, 12 genome integration, 109110
genome location, reference, 100, 101 Brain dopamine systems

high-throughput gene set analysis, catalepsy phenotype, 373375, 374376
108109 DRD2 expression, 379386, 380381,
integration, 109110 383385
LIMS, 97 D2 receptor density, 379383, 380381,
microarray gene set analysis, 108109 383385
natural keys, 99104 fundamentals, 371372, 372, 386
phenome integration, 109110 transporter and receptor density, 376,
phylogeny, 105106 377379, 378379
proteomic gene set analysis, 108109 Brain Information Group (BIG), 97
QTL candidate gene selection, 107108, Brain lesions, 4344
108 Brain reward center, 269, see also Reward
sequence analysis scoring matrices, dependence
105106 Brain structure, see Expression and brain structure
text mining, 109 Brain weights, 9, 150151
Biological rhythms, 10, see also Circadian clock Brand & Perimon method, 324
and rhythms Bray and Amrein studies, 324
Biology Student Workbench, 106 Breeding
Biology Workbench, 104 alcohol psychopharmacogenetics, 458460
Biometrical Genetics, 6 selective, IIPMF, 392393, 393
Biosynthetic genes, 321324 variance theoretical background, 4546
Bipolar disorders (BDs) Brennan, Hancock and Keverne studies, 22
family studies, 228 Brenner, Sydney, 355
fundamentals, 227 Broadhurst, Peter, 7
linkage disequilibrium studies, 231236, Brodsky and Lombroso studies, 23
233 Bromodomains, 82
molecular linkage studies, 230231, Brood parasitism, cuckoos, 3
232233 Bruell, J., studies, 7, 46
phenotypical heterogeneity, 177 Brush, Bob, 9
twins studies, 228229, 229 B6129SF2/J strains, 133
Birds, 132, 413 B6 strains
B6-I110 strain, 123 consomic strains, 119
Bivariate analysis, 4546 knockout/congenic strains, 120, 122124
Black and white box, 295 mapping QTLs, 123124
BLAST, 64 B6-Tcrd strain, 123
B6-Lipc strain, 123 Buck and Axel studies, 324
Blizard, David, 9 Buckle and Van Tol, Wong and, studies, 24
Blood alcohol concentrations (BAC), 34, 452, Buck studies, 1112, 454
see also Alcohol, alcohol-related disorders, BXD/BXD RI strains
and abuse behavioral phenotypes, 376
BN salt-sensitive strains, 60 catalepsy phenotype, 373374
Bocchetta studies, 235 complex trait analysis, 453
Bolivar and Flaherty studies, 123124 fundamentals, 136137
Bolivar studies, 115126 GeneNetwork data sets, 419420
Bonferroni correction, 153, 176 genetic reference populations, 102103
Boomsma, Dorett, 10, 12, 48 Mouse Brain Library, 424
Boot strappig strategy, 211 bZIP, 83
Borck studies, 435444
Borenstein, Cohen and, studies, 151
Borenstein studies, 157, 160, 164 C
Bouchard, Thomas, Jr., 10
Brain activity imaging, 339340 Cadoret studies, 229
Brain-behavior relations, 403406 Caenorhabditis elegans
Brain-derived neurotropic factor (BDNF), 204, 1971 to 1985, 8
234 1986 to 2005, 10
Index 473
behaviors, 355367 Causation, phylogenetic aspects

chemosensory behavior, 356, 356359 examples, 43
Drosophila melanogaster comparison, 6 natural selection/genetic architecture, 3839
egg laying, 359360, 360 theoretical background, 3943
ethanol resistance, 362 CBA/J strain, 373
food response, 361 C57BL/A/J strain, 449
fundamentals, 144, 353, 367 C57BL/6 (B6) strain, 116, 132
locomotion response regulation, 361 C57BL/6By strain, 284
memory, 364367, 365366 C57BL/6 C2J strain, 134
as model organism, 131, 353355 C57BL/6J-Chr 19 SPR strain, 134
mutations impact, 355367 C57BL/6J strain
nervous system, 353355, 354 advanced intercross lines, 60
plasticity, 356, 356359 bioinformatics, 96
social feeding behavior, 363364 co-isogenic strains, 134
suitability for studies, 355 D2 dopamine receptor density, 381
supplier Web sites, 145146 mixed inbred strains, 138
Cairns studies, 9, 195 QTLs, 418, 461
Calico cats, 31, see also Cats recombinant inbred strains, 136
cAMP response element binding (CREB) C57BL/10J strain, 195
expression variation sources, 87 C57BL6K strain, 373
recurrent unipolar disorders, 237 C57BL/6L strain, 449
signal transduction, 8385 C57BL/6-mtBALB/c strain, 135
Canastar studies, 281287 C57Bl strain, 141, 142
Candidate gene studies, 463 C57BL/6 strain
Canids, 10, see also Dogs alcohol sensitivity, 458
Cannibis, 179 congenic strains, 134
Capping, 79 conplastic strains, 135
Capsi studies, 193 linkage solution, 396
Carey studies, 44 C57BL/10 strain, 4
Carlier studies, 9, 211, 213 C57BL/6xSJL strain, 138
Carr, Lucinda, 301 CcS1 strain, 138
Carrie studies, 441 CcS2 strain, 138
Carter, H.D., 5 CcSX strains, 138
Case control study approach, 170, 170171, cDNA, 6466
176 Cellular mechanisms, 444
Caspi studies, 25 Center for Advanced Studies in the Behavioral
CAST/Ei strains, 122 Sciences, 6
Catalepsy phenotype, 373375, 374376 Central Dogma, 23, 73
Catechol O-methyltransferase (COMT), Central nervous system (CNS), see also Nervous
203 system
Cats, 14, 31 bipolar disorders, 234
Cattle, 14, 131 complex trait analysis, 413414
Causation, phenogenetic aspects gene-environment interactions, 335
brain lesions, 4344 gene expression, 74
correlation approach, 4344 CFA (concern for appropriateness) scale, 271
examples, 4849 Chalfie studies, 22
genetic correlations, 44 Chargraff, Erwin, 23
IIPMF, 48 Charles River Laboratories, 143
IQ, 4849 Chemically-induce mutations, mice, 139
locality assumption, 4344 Chemosensory behavior, 356, 356359
nerve conduction velocity, 4849 Chemotaxis, 10
QTLs, 48 Cherny, Stacy, 12
rearing behavior, 48 Chertemps-Thomas studies, 323
SIP-IQ, 4849 Chesler studies, 95110, 374
theoretical background, 4548 Chiang studies, 341
Chickens, 14, 131 Collaborative Study on the Genetics of

Child Behavior Check List, 287 Alcoholism (COGA), 464
Childhood maltreatment, 287 Colleaux studies, 435444
Chimpanzees (Pan troglodytes), 14, Collier studies, 235
see also Pedigree analyses; Collins, Francis, 65
Rhesus monkeys Comas studies, 333347
Chinchilla strains, 138 Combinatorial control, 81
Chi-square tests, 256 Comparison of cohorts requirement, 174
Chlorembucil, 139 Complete penetrance, 50
Choice tests, 295, 296 Complexity, expression and brain structure, 413
Chromatin, 74, 81 Complex Trait Consortium (CTC), v, 6162, 143
Chromodomain-containing proteins, 82 Complex traits analysis
Chromosomal aberration strains, mice, 139140 additive effects, 3334
Chromosomal location, schizophrenia, 219220 pharmacogenetics, 453454
Chromosomal substitution strains (CSSs), 60, QTLs mapping, 413415
134, 302 COMT (catechol O-methyltransferase), 203
Chumakov studies, 234 Concern for appropriateness (CFA) scale, 271
Churchill and Doerge studies, 427 Conditional logistic approach, 205
Cichlidae fish species, 39 Conditioned place preference, 381
Circadian clock and rhythms, 1112, Conditioning experiments, 295296
see also Biological rhythms; Clock mutants Conflicting data, 401403
Cis-acting QTLs, 88, 418, 418419 Congenic strains and substrains, see also
Classical genetics, 31 Consomic strains
Clock mutants, 12, 64, see also Circadian clock animal resources, 134, 135
and rhythms fundamentals, 115119, 117118, 133
Cloningers temperament and character inventory high-resolution mapping, 5659, 5758
(TCI), 202 historical research developments, 9
Cloningers Tridimensional Personality knockout mice, 123126
Questionnaire (TPQ), 265, 267268 knockout strains, 118, 119122, 121122
Cloninger studies, 12, 35, 191, 272 mapping QTLs, 123126, 124125
Clustering, 423 Connecticut Department of Children and Families
Cluster Tree tool, 423 (DCF), 193
Cn3D tool, structure analysis, 106 Connolly studies, 344
Coaction, 20 Conplastic strains, 133, 135136
Coat color, rodents, 9 Consequences, genetic activity, 20, 22
Coates and de Bono studies, 364 Consomic strains
Cocaine and cocaine addiction alternative strategies, 60
behavioral phenotypes, 376 animal resources, 134
brain dopamine systems, 372 fundamentals, 115116, 133
Caenorhabditis elegans, 10 genetic reference populations, 103
catalepsy phenotype, 374375 Constitutive alternative splicing, 80
expression variation sources, 87 Constitutive transcription factors, 81
Coding mutations, rodents, 12 Contigs, 6263, 65
Coding region, 74 Contributions to Behavior-Genetic Analysis:
Coffin-Lowry syndrome, 442 The Mouse as a Prototype, 8
COGA (Collaborative Study on the Genetics of Control cohorts requirement, 174
Alcoholism), 464 Cool-2, 439
Cognition, human research development, 12 Cooperation, gene-environment interactions,
Cohen and Borenstein studies, 151 190197
Cohen and Wahlsten studies, 157 Cooperative binding, 81
Cohen studies, 151152, 156157, 160164 Cordell studies, 205
Co-isogenic strains, 133134 Core promoter, transcription, 76, 78
Cold spots, recombination, 174 Correlations, see also Genetic correlations
Coleman, D.L., 9 expression and brain structure, 421,
Collaborative Cross, 6162, 67 421422
Index 475
gene-environment interactions, 190 Dains studies, 382

genetic, 44 Damerval studies, 417
genetic reference populations, 102 Daniels, Plomin and, studies, 184
IIPMF, 396, 397398, 398 Darvasi, A., 60
misapplication, quantitative genetics, 50 Darwin, Charles, 24, 13
phenogenetic aspects, causation, 4344 Database design, 9799
trait association verification, 394 Data modeling, 9899
Correns studies, 4 DAT (dopamine transporter) gene, 204, 376,
Costa studies, 310 378
Courtship behavior, Drosophila melanogaster DAVID, 104, 109
(fruit flies) Davies studies, 362
acoustic stimuli, 320321 Davis, Ronald, 341
biosynthetic genes, 321324 DBA/2 (D2) strains, 132, 396
fruitless gene, 326328 DBA/2J strains, 60, 373, 461
fundamentals, 319 DBA/2 strains
gustatory receptors, 324325 alcohol sensitivity, 458
olfactory receptors, 324325 congenic strains, 116
pheromones, 321324 knockout/congenic strains, 122
sex determination cascade, 326328 recombinant inbred strains, 136
visual stimuli, 320321 D4DR genotype, 202203
Covalent modifications, 81 de Bono, Coates and, studies, 364
Cows, 14, 131 de Bono and Bargmann studies, 363
Coyle studies, 236 De Bono studies, 363
Coyotes, 10, see also Dogs Defecation, 144, 297
CpG islands, 6364 DeFries, John, 78, 10, 12
Crabbe studies, 9, 1112, 452, 457464 DeFries, Plomin, Loehlin and, studies, 190
Crawley studies, 11 Deletion mutations, 139
CREB, see cAMP response element binding Demant and Hart studies, 59
(CREB) Dementia praecox, 210, see also Schizophrenia
Crick, Francis, 5, 73, 75 Demir and Dickson studies, 327
Crick, Watson and, studies, 75, see also Watson- Denate gyrus volume, mapping, 427,
Crick base pairing 427428
Criminal behavior, heritability, 287 Denmarks National Twin Register, 214
cRNA (complementary RNA), 90 Denny studies, 62
Crook, M.N., 5 Denzler, Brenda, 26
Crossbreeding, 4243 Deoxyribose nucleic acid (DNA)
Crusio and van Abeelen studies, 43 complementary nature, 76
Crusio studies, 12, 3751, 396, 401 congenic strains and substrains, 56
Cruz studies, 297 dogma invalidity, 23
CSSs (chromosomal substitution strains), 60, 134, historical behavioral genetics, 45
302 regulation of transcription, 8081
CTC, see Complex Trait Consortium (CTC) scoring matrices, 106
C-terminal tail, 7879 transcriptional activity, 73
Cuckoo brood parasitism, 3 Department of Children and Families (DCF),
Cunningham studies, 381 Connecticut, 193
Curtis studies, 272 Dephosphorylation, transcription, 78
CYP (Cytochrome P 450) enzymes, 451 Depression
Cystic fibrosis, 30 gene-environment interactions, 193194
Cytochrome P 450 (CYP) enzymes, 451 historical behavioral genetics, 4
interim solution, 25
multi-gene disorder, 30
D phenotypical heterogeneity, 177
serotonin, 2425
D Adamo studies, 441 The Descent of Man, 3
Dahl salt-sensitive strains, 60 Detection failure, 154155
Developmental neurobehavioral genetics D2 receptor density, 379383, 380381,

development-as-explanation, 1920, 2021, 383385
2223 Dreier studies, 435444
epigenesis, 1719 Drosha endonuclease, 86
fundamentals, 17, 26 Drosophila melanogaster (fruit flies)
interim developmental neurobehavioral 1910 to 1952, 4
genetics, 21, 2425 1953 to 1970, 67
model, 20, 21 1971 to 1985, 8
preformation, 1719 1986 to 2005, 11
Developmental-relational casuality, 25 circadian rhythms, 12
De Vries studies, 4 fundamentals, 143144
DFREML program, 257 inbreeding, 132
DGV13A candidate genes, 428429, 428429 mutagenesis, 139
Diabetes, 56, 183 supplier Web sites, 145
Diabetes line, 9 Drosophila melanogaster (fruit flies), courtship
Diallel cross behavior
paradise fish, 43 acoustic stimuli, 320321
quantitative genetics, 3738 biosynthetic genes, 321324
rearing behavior, 48 fruitless gene, 326328
theoretical background, 42, 47 fundamentals, 319
Diallel design, 299 gustatory receptors, 324325
Dicer endonuclease, 86 olfactory receptors, 324325
Dickson, Demir and, studies, 327 pheromones, 321324
Differential display, 90 sex determination cascade, 326328
DiLalla, Liz, 12 visual stimuli, 320321
Ding studies, 269 Drosophila melanogaster (fruit flies), food search
DIP database, 104 behavior
Directional selection, 39 food search behavior, 311312
Direct sequencing, 64 foraging gene role comparisons, 312314
Disruptive selection, 39 fundamentals, 307308
Dissimilarity, life experiences, 24 as model organism, 308
Dizygotic (DZ) twins, see Twins studies polygenic analysis, 308311
DNA, see Deoxyribose nucleic acid (DNA) single-gene analysis, 308311
DNA methyltransferase, 86 Drosophila melanogaster (fruit flies), learning
Dobzhansky studies, 4, 6 and memory studies
Doerge, Churchill and, studies, 427 brain activity imaging, 339340
Dogs, 7, 10, 14, 131 fundamentals, 333334, 335337
Dominance localization, olfactory memory, 338,
heritability estimate, 256 343345
intralocus interaction, 202 mutants, 340341
natural selection, 38 neuronal circuits, 338339, 338339
theoretical background, 40, 45 olfactory memory, 343347
Dominant traits, 33 phases, olfactory memory, 345347,
Dopamine, see Brain dopamine systems 346347
Dopamine D4 receptor (DRD4) temporal control, gene expression, 341343,
antisocial behaviors, 269 342
fibromyalgia, 267 tools, anatomo-functional maps, 343
molecular genetics, 266 Web resources, 335337
Dopamine output area, 87 Drug abuse, multi-gene disorder, 30
Dopamine transporter (DAT) gene, 204, 376, Drugs, 9, 11, 453, see also specific drug
378 D8S499 marker, 272
Doss studies, 418 Dubnau studies, 341
Downs Syndrome, 7, 12 Dudek, Bruce, 9, 12
DRD4 (dopamine D4 receptor), 266267, 269 Dupuis and Siegmond studies, 427
DRD2 expression, 379386, 380381, 383385 DZ twins, see Twins studies
Index 477
E shared vs. nonshared, 184

theoretical background, 4647
Early-onset familial Alzheimers disease, 65, transcription, 74
see also Alzheimers disease Environment sensing and response, C. elegans,
EASE, 104, 109 144
Eaves, Lindon, 10, 12 Epigenesis, 1719
Ebert, Patricia, 9 Epigenetic regulation, gene silencing, 87
Ebstein studies, 202, 263273 Epigenetics, 30
Economic/ethnic group disparities, 187 Epilepsy, see Hypertension epilepsy
Edwards, Johnston and, studies, 20 Epistasis, 35, 103, 202203
Effect size, 150152, 150152, 155156 EPS (extrapyramidal symptoms), 373
Egan studies, 234 Erlenmeyer-Kimling, Niki, 6
EGFP (enhanced green fluorescent protein), 426 Estimation, sample size requirements, 155158
Egg laying, C. elegans, 144, 359360, 360 ESTs, scoring matrices, 106
Ehrman, Lee, 6, 8 Ethanol and ethanol resistance, see also Alcohol,
Eleftheriou, Basil, 9 alcohol-related disorders, and abuse
Elevated plus maze, 295, 297298, 301, alcohol metabolism comparison, 458459
see also Mazes Caenorhabditis elegans, 362
Ellison Foundation, 62 complex trait analysis, 34
Elongation phase, 79 gene-environmental interactions, 3435
Emergence tests, 295 sensitivity, 451452
Emerging hypotheses, IIPMF, 402403 Ethnic/economic groups disparities, 187
Emotional behaviors Ethylnitrosourea (ENU), 139
choice tests, 295, 296 Eukaryotic cells and genes, 76, 79
conditioning experiments, 295296 Everyday speech personality approach, 265
forced exposure, 293, 294, 295 Evidence, gene-environment interactions
fundamentals, 291292 animals, 195197, 196197
gene targeting, 303 humans, 191192, 191194
genetic analysis, 298303 Evolutionary conservation, 63
measurements, 292298 Ewens, Spielman and, studies, 171
multivariate approach, 296, 296298 Exons, 64, 79
QTLs, 300303, 302 Experiments, see Sample size requirements
quantitative genetic analysis, 299 Exploratory activity, 293
rodents, 11 Expression and brain structure, see also Gene
selection experiments, 298299 expression; Transcription
trait links, 299300 Allen Brain Atlas, 426
transgenesis, 303 behavior, 412413
unconditioned stress responses, 293, 294, brains, 412413
295, 296 cis-acting QTLs, 418, 418419
Emotionality, 292 clustering, 423
Endocrine systems, rodents, 9 cluster tree, 423
Enhanced green fluorescent protein (EGFP), 426 complexity, 413
Enhancers, 75 complex trait analysis, 413415
Enns, R. Mark, 258 denate gyrus volume, mapping, 427,
Ensembl browser and database, 100, 386 427428
EntrezGene, 101 development, 412
ENU (ethylnitrosourea), 139 DGV13A candidate genes, 428429,
Environmental association studies, 257258 428429
Environmental influences and interactions, see fundamentals, 412, 425, 430
also Gene-environment interactions genes, 412
economic/ethnic group disparities, 187 genetic correlation, 421, 421422
genetic vulnerability, 178179, 179 GenNetwork, data sets, 419420
life experience component, 24 GENSAT database, 426429
pharmacogenetics, 452 microarray data, 416417
potential danger of ignoring, 184 Mouse Brain Library, 424425
mRNA abundance, 417, 419 F1 hybrids

Msx2, DGV13A candidate genes, 428429, aggression, 283
429 ambidirectional dominance, 38
network graph, 423 congenic strains, 116
networks, 423 historical research developments, 9
power, 429430 inbred strains, 132133
QTLs, 413415, 417419 theoretical background, 40, 42
reference population value, 415416, 416 F2 hybrids
Ror2, DGV13A candidate genes, 428429 advanced intercross lines, 60
specificity, 429430 aggression, 283
trans-acting QTLs, 418, 418419 genetic correlations, 44
WebGestalt, 423424 historical research developments, 9
The Expression of Emotions in Man and inbred strains, 132133
Animals, 3 interval-specific congenic strains, 60
Expression profiling, 66, 90 misapplication, quantitative genetics, 50
Extended TDT (transmission disequilibrium test), reference populations value, 415416
171 theoretical background, 4142, 46
Externalizing behavior problems, 271 F3-Fn hybrids, 283
Extrapyramidal symptoms (EPS), 373 Fibromyalgia, 267
Eysenck studies, 272, 298 Figueredo, A.J., 258
File studies, 297
Fink studies, 373, 375
F Fiorentini studies, 386
Fire studies, 355
Facilitation, behavior development, 22 Fischer studies, 214
FACL4 (fatty acid-CoA ligase 4), 441442 Fisher rats (FISH), 132
Falconer, D.S., 6 Fisher studies, 4
Falconer and Mackay studies, 309 Fish species, 14, 39, 43
False Discovery Rate (FDR), 91, 417 Fitzpatrick and Sokolowski studies, 314
False positives, 152154, 174176 Five-factor model (FFM), 249, 253, 265
Familial Alzheimers disease, 65, 138 Fixed genotype, fundamentals, 30
Familial fatal insomnia, 33 Flaherty, Bolivar and, studies, 123124
Family and twin studies, see also Adoption Flaherty studies, 115126
studies; Human studies; Twins studies Flat file format, 98
Algerian families, 443 Flint studies, 11, 301
association studies, 171, 171172 Floripa H and L rat lines, 300
autosomal recessive MR, 443 Fluoxetine, 451
behavioral genetics, 183184 Flling studies, 450
bipolar disorders, 228 Food response, C. elegans, 361
fundamentals, 183 Food search behavior, Drosophila melanogaster
health, 183184 (fruit flies)
historical research developments, 3, 10, food search behavior, 311312
1213 foraging gene role comparisons,
integration of approaches, 186188 312314
methods, 184186, 186 fundamentals, 307308
recurrent unipolar disorders, 236 as model organism, 308
schizophrenia, 211213, 212 polygenic analysis, 308311
Farah studies, 43 single-gene analysis, 308311
Fatty acid-CoA ligase 4 (FACL4), 441442 Foraging gene role comparisons, 11, 312314
FDR (False Discovery Rate), 91, 417 Forced exposure, emotional behaviors, 293, 294,
Feeding, 9, 1112, 144 295
Feldker, D.E., 12 Forced swimming test, 296
Fertilization, gene silencing, 87 Foundation of Behavior Genetics, 8
Festing studies, 132 Fowler studies, 373
FFM (five-factor model), 249, 253, 265 Fragile X syndrome, 23, 438
Index 479
Fraternal twins, see Twins studies striatal, 382384

Freeman, F.N., 5 transcription, 7683, 7778
Free University of Amsterdam (Netherlands), Gene expression profiling, 90
48 Gene-gene interactions
Freud, Sigmund, 264 animal studies, 203204
Friedman, J.M., 12 epistasis, 202203
Frog, 131 fundamentals, 35, 201202, 205
Fruit flies, see Drosophila melanogaster (fruit human diseases, common, 204
flies) statistical methods, 204205
Fruitless gene, 326328 Gene identification
Fulker, David, 10, 12 alcohol psychopharmacogenetics, 460462
Fuller, John L., 57, 9 alternative strategies, 5962, 61
Fuller and Thompson studies, 8 congenic strains and substrains, 5659,
Fullerton studies, 272 5758
Fully bidirectional coactional system, 19 fundamentals, 5556, 67
Furlong studies, 236 high-resolution mapping, 5662
Future directions, pharmacogenetics, 454 positional candidate approach, 6567
positional cloning, 6267, 63
pure positional cloning, 6364
G schizophrenia, 220221
GeneKeyDB, 101
GABA, 460462 GeneNetwork (GN), see also GenNetwork;
GAD (generalized anxiety disorder), 177 WebQTL
-Galactosidase, 83 data sets, 419420
Galen (129-200 C.E.), 3 dentate gyrus volume, 427
Gall, Franz Josef, 44 genetic correlation exploration, 416
Galton, Francis, 24, 13 QTL mapping tools, 415
Gametogenesis, 32 Gene Onotology Tree Machine, 104
Garipy studies, 195 Gene ontology, 104, 105
Garrod, A., 5, 178, 450 Generalized anxiety disorder (GAD), 177
Gauderman studies, 205 Generalized single locus (GSL) model, 217218
GeneChip microarray, 90 General transcription factors, 77
Gene-environment interactions Gene Reference Into Function (GRIF), 104, 109
association studies, 178179, 179 Genes
considerations and discussion, 197198 expression and brain structure, 412
cooperation, 190197 fundamentals, 7476, 75
family and twin methods, 186188 natural keys, biological databases, 100101
fundamentals, 3435, 189190 targeting, 303, 461462
human research development, 13 Genes, Brain and Behavior, 10
importance, 189190 Genes as a Tool in the Study of Behavior, 6
interim solution, 2425 Genetic activity, 20, 2223
pedigree analyses, 257258 Genetically engineered mutant mice, 138140
Gene expression, see also Expression profiling; Genetic analysis, emotional behaviors, 298303
Transcription Genetic architecture, 3839
alcohol psychopharmacogenetics, 464 Genetic association studies, 257258
fundamentals, 7374 The Genetic Basis of Alcohol and Drug Actions,
gene basics, 7476, 75 11
measurement, 8891 Genetic correlations, see also Correlations
MRNA biogenesis, 7680, 7778 bioinformatics, 110
receptor density mismatch, 380 expression and brain structure, 421,
regulation, transcription, 8085, 84 421422
RNA interference, 8586 phenogenetic aspects, causation, 44
signal transduction, 8385, 84 Genetic effects impact, schizophrenia, 216219
silencing genes, 8687 Genetic heterogeneity, 177178
sources of variation, 8788 Genetic power calculator program, 173, 175
Genetic quality, JAX rodents, 143 Hall, Calvin S., 46, 46, 292
Genetic reference populations (GRPs), 101104, Hall, Jeff, 11
102 Haloperidol, 374375, 451
The Genetics and Social Behavior of the Dog, 7 Hamer, Dean, 12
The Genetics of Behavior, 8 Hamilton Station for Behavioral Research, 4
Genetic variation, schizophrenia, 211216 Hamster prion protein, 138
GenMapp, 109 Hancock and Keverne, Brennan studies, 22
GenNetwork, 419420, see also GeneNetwork Haplotype and haplotype mapping
(GN) alternative strategies, 60
Genome Institute of Novartis, 107 bioinformatics, 96
Genome integration, 109110 family-based association studies, 171
Genome location, 100, 101 linkage, 32
Genome sequence importance, 131 positional cloning, 6667
Genome size, human, 169 schizophrenia, 221
Genome-tagged mice, 122 Haplotype Trend Regression, 272
Genotype-environment interactions HAPMAP, 173
animal studies, 195197, 196197 Hariri studies, 194
human studies, 191192, 191194 Harm avoidance
Genotype-phenotype relationships, 31 epistasis, 202
GENSAT, 412, 425429 fibromyalgia, 267
Geotaxes, 6 linkage studies, 271
Gerlai studies, 43 phenotypical heterogeneity, 177
Gershenfeld studies, 49 theoretical personality approach, 265
Gershon, Badner and, studies, 231 Harris, R.Adron, 11
Ginsburg, Benson, 47, 910, 14 Hart, Demant and, studies, 59
Giot studies, 334 Hayman studies, 46
Gomez studies, 366 HDL (high density lipoprotein) cholesterol, 24
Gorwood studies, 169180 Health, 183184, 187
Gottesman and Bertelsen studies, 214 Heath, Andrew, 13
Gottesman and McGue studies, 219 Hedgecock and Russel studies, 364
Gottesman and Shields studies, 217 Hendersons mixed model equation, 254255
Gottesman studies, 12, 209222 Henderson studies, 9, 12, 195
Gottliebs approach, 22 Hendley studies, 299
Gottlieb studies, 1726 Hereditary Genius, 3
Govek studies, 438 Heritability
GRAIL, 64 additive-genetic effects, 41
Greenspan, Ralph, 11 aggression, 287
GRIF (Gene Reference Into Function), 104, 109 alcoholism typologies, 191
Grilli studies, 386 altruism, 268
Grossfield, K., 8 Alzheimers disease, 65, 138
Gross morphological changes, 20 antisocial behaviors, 190
GSL (generalized single locus) model, 217218 dominance, 256
Guenet studies, 119 fundamentals, 185
Gurling studies, 219 genetic reference populations, 102
Gustatory receptors, 324325 misapplication, quantitative genetics, 5051
Gutierrez studies, 235 phenotypes, 266
quantitative genetic analysis, 299
rearing behavior, 48
H recurrent unipolar disorders, 237
schizophrenia, 219
HAB rats, 300 SIP-IQ relationship, 49
Hahn, Martin, 9, 12 theoretical background, 41
Haldane, J.S., 18 Heroin addiction, 287, 372
Haldane studies, 4 Heston studies, 7, 215
Half-life, association, 174 Heterochromatin, gene silencing, 86
Index 481
Heterogeneous stocks, 60 5-HT1B (serotonin 1B receptor), 204, 287

Hewitt, John, 9 5HTTLPR gene, see also Serotonin transporter
5-HIAA, alcohol sensitivity, 459 (5-HTT) gene
Hierarchical format, 98 COMT, 203
High density lipoprotein (HDL) cholesterol, 24 epistasis, 202
High-resolution mapping fibromyalgia, 267
alternative strategies, 5962, 61 gene-environment interactions, 191,
congenic strains and substrains, 5659, 193194
5758 molecular genetics, 266
High-throughput gene set analysis, 108109 Huberlein, U., 11
Hippocampal mossy fibers, 12, 43, see also Intra- Hukema studies, 358
and infrapyramidal mossy fiber (IIPMF) Human Genome Project, 187
Hirsch, Jerry, 6, 11 Human genome size, 169
Hirsch studies, 46 Human studies, see also Adoption studies; Family
Histone acetylases, 82 and twin studies; Twins studies
Histone deacetylases, 82 1910 to 1952, 5
Histone methylases, 82 1953 to 1970, 7
Histones 1971 to 1985, 10
fundamentals, 74 1986 to 2005, 1213
gene silencing, 87 aggression, 286287
regulation of transcription, 8182 alcohol psychopharmacogenetics, 462464
transcription, 78 diseases, gene-gene interactions, 204
Histone tails, 82 gene-environment interactions, 191192,
Historical research development 191194
1910 to 1952, 45 pedigree analyses, 249250
1953 to 1970, 57 Huntingtons chorea, 184
1971 to 1985, 810 Huntingtons disease
1986 to 2005, 1013 dominant traits, 33
ancient history, 23 genotype-phenotype relationships, 31
bees, 11 single-gene disorder, 30, 210, 414415
canids, 10 Hwang studies, 385
Darwin and Galton, 3 Hybrid crosses, inbred strains, 132133
dogs, 7 Hybridization method, 90
Drosophila melanogaster, 4, 68, 11 Hyde, Janet, 9
first century, 413 Hypertension, prenatal environment, 183
fundamentals, 2, 1314 Hypertension epilepsy, 56
humans, 5, 7, 10, 1213 Hyponeophagia, 293
nematodes, 8, 10 Hypothalamo-pituitary-adrenal axis, 196
preliterate history, 23 Hypothesis of interaction, 164
rodents, 45, 79, 1112
Hitzemann studies, 371386
Ho, M.-W., 18 I
Hoefgen studies, 193
Holmans studies, 205, 237 Iakoubova studies, 122
Holzinger, K.J., 5 IBANGS (International Behavior and Neural
Homayouni studies, 423 Genetics Association), 10
Homogentisic acid, 450 ICR mice, 140
Honeybees, see Bees IEGs (immediate early genes), 85, 87
Honzig, M., 7 IIPMF, see Intra- and infrapyramidal mossy fiber
Hood, Kathryn, 9 (IIPMF)
Horvitz, Bob, 355 Illumina SMP genotyping platform, 420
Hot spots, recombination, 174 IL1RAPL1 (interleukin-1 receptor accessory
Hotta and Benzer studies, 326 protein-like 1), 441
Housekeeping genes, 8990 Image/J, 425
hSPT5 factor, 79 Image program (NIH), 425
Immediate early genes (IEGs), 85, 87 Introduction to Quantitative Genetics, 6

Imprinting, gene silencing, 87 Introns, transcription, 79
Inbred strains IQ (intelligence quotient), 4849
alcohol psychopharmacogenetics, 458459 Isabel studies, 333347
origins, 140141, 141 ISCS (interval-specific congenic strains), 60
sample size, 167 IS/Kyo rat, 141
theoretical background, 40 ISPG (International Society for Psychiatric
Inbreeding, animal resources, 131133, 132 Genetics), 10
The Incidence of Alkaptonuria: A Study in IT15, 33, see also Huntingtons disease
Chemical Individuality, 450 Ito studies, 339
Incomplete penetrance, 174175, 216
Independent Assortment, Law of (Mendel), 32
Inducible transcription factors, 81 J
Induction, behavior development, 22
Infant locomotion, 20, see also Locomotion Jackson Laboratory
Inquiry into Human Faculties and Its BXD RI strains available, 136
Development, 3 consomic strains, 119
In situ hybridization method, 90 historical research development, 4, 7, 12
Institute for Behavioral Genetics, 8 mutagenesis facility, 139
Insulin resistance, 66 research model supply, 142143
Integration, approaches, 186188 Jallon, Venard and, studies, 321
Integration, behavior bioinformatics, 109110 Jallon studies, 11, 319328
Intelligence quotient (IQ), 4849 Jamain studies, 177178
Interactions of genes, 3435, see also specific Janowsky studies, 376
interaction Jansen and Nap studies, 417
Interactome, 104 Jansen studies, 353367
Interim developmental neurobehavioral genetics, JAX rodents, 143
21, 2425 Jinks, Mather and, studies, 309
Interleukin-1 receptor accessory protein-like 1 Johnson, Thomas, 9, 11
(IL1RAPL1), 441 Johnston and Edwards studies, 20, 22
Intermediate phenotype, 177 Jones, B.C., studies
International Behavior and Neural Genetics alcohol and drug effects, 11
Association (IBANGS), 10 brain dopamine systems, 372, 374,
International Society for Psychiatric Genetics 376379
(ISPG), 10 cocaine-related behaviors, 453
International supply, animal resources, 142143 gene-environment interactions, 189198
Internet, see Web sites pharmacogenetics, 449455
Interval-specific congenic strains (ISCS), 60, 118 QTL analysis, 11, 453
Interventionist studies, 43 trait fundamentals, 2935
Intra- and infrapyramidal mossy fiber (IIPMF) Jones, L.C., studies, 189198
association verification, 393394, 395
conflicting data, 401403
correlations, 396, 397398, 398 K
emerging hypotheses, 402403
fundamentals, 389391, 406407 Kakihana, McClearn and, studies, 34
hippocampal mossy fiber projection, 391, Kallman, F.J., 5
392 Kalow studies, 450
linkage problem solutions, 394, 395, 396 Kaltschmidt, ONeill and, studies, 386
non-hippocampal correlations, 399400, Kandel, Mayford and, studies, 367
400 Kanes studies, 373
outlook, 406407 Kaufman studies, 193
phenogenetic aspects, causation, 48 Kcnj9 polymorphism, 88
real world testing, 403406, 404405 KEGG database, 104, 424
selective breeding, 392393, 393 Kendler studies, 12, 228
Introduction to Behavioral Genetics, 8 Kety studies, 7, 215216
Index 483
Keverne, Brennan, Hancock and, studies, 22 LD, see Linkage and linkage disequilibrium (LD)
KID (kinase inducible domain), 83 Learned helplessness, 295
Kimura studies, 328 Learning and memory
Kinase inducible domain (KID), 83 Caenorhabditis elegans, 10, 364367,
King studies, 247258 365366
Kirov studies, 235 gene expression, 74
Kitamoto studies, 343 historical research developments, 9
Klinefelters condition, calico cats, 31 rodents, 11
Klinefelters Syndrome, 12 Learning and memory, Drosophila melanogaster
Knockout mice (fruit flies)
congenic strains and substrains, 118, brain activity imaging, 339340
119126, 121122 fundamentals, 333334, 335337
fundamentals, 139 historical research development, 11
gene-gene interactions, 35 localization, olfactory memory, 338,
historical research developments, 1112 343345
Konopka, Ronald, 8 mutagensis, 139
Korstanje studies, 56 mutants, 340341
Kraemer and Thiemann studies, 156157 neuronal circuits, 338339, 338339
Kraepelin, Emil, 210211 olfactory memory, 343347
Kraus studies, 386 phases, olfactory memory, 345347,
Krech, D., 7 346347
Kruglyak, Lander and, studies, 153 temporal control, gene expression, 341343,
Kubiak studies, 363 342
Kunugi studies, 236 tools, anatomo-functional maps, 343
Kutsche studies, 439 Web resources, 335337
Kyoto University, 141 Least squares estimation, 253
Kyriacou, Charalambos P., 11 Lee studies, 385
Lennox and Wolfe studies, 270
Lesch studies, 191193, 202
L Levinson studies, 231
Lewis rats (LEW), 132, 300301
Laboratory animals, see Animals; Sample size Li, Bausell and, studies, 157, 165
requirements Limbic-system lesions, 43
Laboratory Information Management Systems LIMS (Laboratory Information Management
(LIMS), 97 Systems), 97
Laboratory mice strains and substrains, 141, 142, Lindzey, Gardner, 8
see also specific strain or substrain Linkage and linkage disequilibrium (LD)
LAB rats, 300 association studies, 172175
Lack, genetic activity, 23 bipolar disorders, 230236, 232233
Lac repressor, 83 fundamentals, 32
Lactation, aggression, 282 genetic correlations, 44
Lagerspetz, K., 7 IIPMF, 394, 395, 396
LAL (long attack latency), 12 law of independent assortment, 32
Landau, Virginia, 258 phenotypes, 271272
Lander, Eric, 13 recurrent unipolar disorders, 233, 237
Lander and Kruglyak studies, 153 schizophrenia, 219221
Lans studies, 357 theoretical background, 4647
Lassalle, Roullet and, studies, 401 Lipp and Schwegler studies, 403
Lassalle studies, 129146 Lipp and Wolfer studies, 392
Latent semantic indexing (LSI), 423 Lipp studies, 12, 44, 389407
Latent semantic relationships, 109 Lipsky studies, 386
Latent variable modelo, 251, 252 Lissencephaly, 23
Laumonnier studies, 178 Locality assumption, 4344
Law of Independent Assortment (Mendel), 32 Localization, olfactory memory, 338,
Law of Segregation (Mendel), 3132 343345
Locomotion MAS (marker-assisted selection), 57

Caenorhabditis elegans, 144, 361 MAS5 normalization procedures, 420
development as explanation, 20 MAST, 106, 108
emotional behaviors, 297 Maternal factors, animal studies, 9, 195
response regulation, 361 Mather and Jinks studies, 309
LOD (log of odds) score, 56, 116 Mather studies, 6, 39
Loehlin, John, 10, 12 Mathis studies, 49
Loehlin and DeFries, Plomin, studies, 190 Mating, 9, 11, 144, see also Courtship behavior,
Log of odds (LOD) score, 56, 116 Drosophila melanogaster (fruit flies)
Lombroso, Brodsky and, studies, 23 Maudsley Emotional/Non-emotional strains, 7, 9,
Long attack latency (LAL), 12 298
Long-Evans rats, 140 Maximum likelihood binomial (MLB), 272
Long sleep lines, 9, 34, 136, 452 Maxson studies, 114, 281287
Lousiville Twin Study, 7, 10, see also Twins Mayford and Kandel studies, 367
studies Mazes
LSI (latent semantic indexing), 423 elevated plus maze, 295, 297298,
Luxenburger studies, 213 301
Lynch, Carol, 9 maze bright/dull rats, 4, 7
Morris water maze, 398
multiple T-maze learning, 396
M radial, 396, 401
water maze, 163, 403404
Macaque monkeys (rhesus), 2425, 131 MBL, see Mouse Brain Library (MBL)
Mackay, Falconer and, studies, 309 McCartney, Scarr and, studies, 190
Mackay studies, 309 McClearn and Kakihana studies, 34
Macropodus operculatus (paradise fish), McClearn studies, 78, 30
43 McCrae, R.R., 258
Madrigal studies, 386 McEwen, R.S., 4
Maintenance, behavior development, 22 McGill, Thomas, 9
Maintenance methylase, 87 McGue, Gottesman and, studies, 219
Major affective disorders McGue, Matt, 13
bipolar disorders, 228236 McGuffin, Peter, 12
family studies, 228, 236 McGuire studies, 344
fundamentals, 227, 237238 McWeeney studies, 371386
linkage disequilibrium studies, 231236, MDD (mood depressive disorder), 177
233 Measurements
molecular linkage studies, 230231, emotional behaviors, 292298
232233, 237 gene expression, 8891
recurrent unipolar disorders, 236237 Medical College of Wisconsin, 60
twins studies, 228229, 229, 237 Medline database, 109
Major depressive disorder, 177 Meffert and Baltimore studies, 386
Malaria, 32 Meiosis, 32, 87
MANCOVA (multiple analysis of covariance MEME, 106, 108
analyses), 204 Memory, C. elegans, 364367, 365366,
Manning, Aubrey, 6 see also Learning and memory
MAO, see Monoamine oxidase (MAO) Mendelian traits and rules
MAOA, see Monoamine oxidase A (MAOA) dominant traits, 33
Mapping candidate genes, 6566 historical developments, 4
Mapping QTLs, 103104, see also Quantitative quantitative genetics, 37
trait loci (QTLs) recessive traits, 3233
Mardones studies, 451 theoretical background, 38, 42
Marijuana, see Cannibis Mendel studies, 2, 201, 299
Markel studies, 119 Mendes de Oliveira studies, 236
Marker-assisted selection (MAS), 57, 118 Mendlewicz and Rainer studies, 228
Martin, Nicholas, 10, 12 Mendlewicz studies, 236
Index 485
Mental retardation (MR) Mismatch (MM) pairs, mRNA, 91

actin cytoskeleton, 437, 437439 Mitosis, gene silencing, 87
autosomal recessive MR, 443 Mixed inbred strains (MIs), 138
cellular mechanisms, 444 MLB (maximum likelihood binomial), 272
Cool-2, 439 Model requirements, animal resources, 130
FACL4, 441442 Moisan studies, 5567, 301
fundamentals, 436 Molecular genetics, 266
IL1RAPL1, 441 Molecular linkage studies, 230231, 232233
neurotrypsin, 442443 Molecular Modeling Database, 106
oligophrenin-1, 438 Moles, 405
PAK3, 438439 Molinari studies, 435444
PIX, 439 Momentary-effective genotype, fundamentals,
postsynaptic spines, 437, 437439 30
presynaptic neurotransmitter release, Monkeys, 14, see also Pedigree analyses; Rhesus
439442, 440 monkeys
RabGDI1, 441 Monoamine oxidase A (MAOA)
X chromosome, 436442 aggression, 287
MERLIN, 272 bipolar disorders, 234
Merlin-Regress, 272 candidate genes, 204
Messenger RNA (mRNA) environmental interactions, 178
brain structure, 417, 419 interim solution, 24
expression, 7374, 417, 419 Monoamine oxidase (MAO), 203
gene presence demonstration, 23 Monogamy, voles, 12
mapping gene expression, 88 Monozygotic (MZ) twins, see Twins studies
measuring gene expression, 8889 Monte Carlo procedures, 254
regulation of transcription, 8082 Mood depressive disorder (MDD), 177
in situ hybridization method, 90 Moore studies, 204
transcription, 7680, 7778, 80 Moral behavior, chimpanzees, 249
variation sources, 87 Morgan, T.H., 4
Meta-analysis, published data estimation, 156 Morgan and Sedensky studies, 362
Metabolism, rodents, 9, 12 Mori, Bargmann and, studies, 356357
Methamphetamine addiction, 372 Mori studies, 364
Method of contrasts, 162 Mormde studies, 291303
Methylation, 82, 87 Morris water maze, 398, see also Mazes; Water
MFT (multifactorial threshold) model, 217219 maze
MGI, genes as references, 101 Moscow laboratory, 404
Mice, see also Animals; Rodents MOTIF, 109
aggression, 282285, 283, 285286, 287 Motif search and alignment, 106
causation phenogenetic aspects, 43 Motivational conflicts, 293
genome-tagged mice, 122 Motor skills, rodents, 11
resources, 145 Mouse, see Mice; Rodents
Microarray chips, gene identification, 460 Mouse Brain Library (MBL), 417, 424425
Microarray data and analysis Mouse Genome Informatics database, 132
behavior bioinformatics, 108109 Mouse Phenome Database (MPD)
expression and brain structure, 416417 alcohol sensitivity, 459
QTL mapping and expression profiling, 66, genetic reference populations, 102
88 inbred strain comparison, 159
Microphrenology, 44 systems-level bioinformatics, 109
MicroRNA (miRNA) mechanisms, 8586, 96 Mouse Phenome Project, 159
Minnesota Multiphasic Personality Inventory, 264 Mouse SNP Selector, 100
Minowa studies, 385 MPD, see Mouse Phenome Database (MPD)
MINT database, 104 MR, see Mental retardation (MR)
MiRNA, see MicroRNA (miRNA) mechanisms mRNA, see Messenger RNA (mRNA)
MIs, see Mixed inbred strains (MIs) Msx2, DGV13A candidate genes, 428429, 429
Misapplications, quantitative genetics, 4951
MTDFREMLA (multitrait derivative fee National Twin Register (Denmark), 214

restricted maximum likelihood analysis), Natural history, pedigree analyses, 249
254258, 262 Natural keys, biological databases
mtDNA mutations, 136 correlations, 102
Multifactorial disorders and complex traits, fundamentals, 99
174175 genes, 100101
Multifactorial threshold (MFT) model, genetic reference populations, 101104,
217219 102
Multigene-multilife experience, 26 genome location, 100, 101
Multiple analysis of covariance (MANCOVA) mapping QTLs, 103104
analyses, 204 Natural language processing, 109
Multiple gene disorder comparison, 30 Natural selection, 3839
Multiple genes, interactions, 202203, Nature-nuture dichotomy, 1819
see also Gene-gene interactions NCBI (National Center for Biotechnology
Multiple inbred strains, 136138, 159161, Information), 63, 106
160161 NC100 line, 195196
Multiple regression NC900 line, 195196
alternative strategies, 60 NDG (Neuroscience Database Gateway), 97
gene-gene interactions, 204205 Neale studies, 272
pedigree analyses, 252 Nectar-eating bats, 406
Multiple test situations, 297 Negative transcription factors, 81
Multiple T-maze learning, 396, see also Mazes Negative variance components, 253
Multitrait derivative fee restricted maximum Nelson, Randy, 11
likelihood analysis (MDFREMLA), 254258, Nelson, Tracy L., studies, 183188
262 Nematodes, see Caenorhabditis elegans
Multivariate analysis and approach, 46, 296, Neoepinephrine transporter (NET), 204, 421
296298 Nerve conduction velocity, 4849
Mundo studies, 235 Nervous system, see also Central nervous system
Munn, Elizabeth, 167 (CNS)
Murphy, Possidente and, studies, 325 Caenorhabditis elegans, 144, 353355, 354
Murphy and Myors studies, 156, 165 Johnston-Edwards model, 22
Murphy studies, 204 Nested model comparisons, 256
Mus musculus spp., 140141 NET (neoepinephrine transporter), 204, 421
Mus spretus, 119 Network Graph tool, 423
Mutagenesis, 144 Networks, expression and brain structure, 423
Mutants, Drosophila melanogaster, 340341, Neuman, H.H., 5
see also specific mutants Neuroanatomy, rodents, 9
Mutations impact, C. elegans, 355367 Neurobehavioral genetic consequences, 175176
Mynett-Johnson studies, 235 Neuronal circuits, Drosophila spp., 338339,
Myors, Murphy and, studies, 156, 165 338339
MZ twins, see Twins studies Neuropsychiatric Genetics, 10
Neuroscience Database Gateway (NDG), 97
Neuroticism, linkage studies, 272
N Neurotrypsin, 442443
N2 generation, congenic strains, 116
Nadeau, Joseph, 119 NHGRI (National Human Genome Research
Nap, Jansen and, studies, 417 Institute), 183, 186
National Center for Biotechnology Information Niche-picking, 190
(NCBI), 63, 106 Nicotine, C. elegans, 10
National Human Genome Research Institute NIH Image program, 425
(NHGRI), 183 NMRI rodents, outbred strains, 140
National Institute on Alcohol Abuse and Non-hippocampal correlations, IIPMF, 399400,
Alcoholism, 464 400
National Institutes of Health, 183, 186 Nonpathological variation, 38
National Library of Medicine, 109 Nonshared environmental effects, 184
Index 487
Nonsyndromic MR genes, 436437, P

see also Mental retardation (MR)
Normotenisve but hyperactive (WKHA) rats, pACC (perigenual anterior cingulate cortex), 194
299 PACs (P1-derived artificial chromosomes), 62
Northern blot, 89, 460 p21 activated kinase 3 (PAK3), 438
Northwestern University, 12 PAGE (polyacrylamide gel electrophoresis), 89
Novel environment, 293 Pak, W., 8
Novelty-seeking, 267 PAK3 (p21 activated kinase 3), 438439
Nucleosomes, 74, 81 Panic disorder, 236
Null hypothesis, 152154, 164 Pan troglodytes (chimpanzees), 14, see also
Null mutant studies, alcohol Pedigree analyses; Rhesus monkeys
psychopharmacogenetics, 461462 Paradise fish (Macropodus operculatus), 43
Parental strains, 51, 58
Parenting, voles, 12
O Paroxtine, 451
Parsons, Peter, 8
Obese line, 9, 12 Passive genotype-environment correlation, 190
Obesity, rodents, 9 Patel studies, 373
Odds ratio (OR), association studies, 170171 PathwayAssist, 109
Offense aggression, 282284 Paw preferences, 400, 402
Ogawa, Sonoko, 11 P1-derived artificial chromosomes (PACs), 62
OGOD (one-gene, one-disease) scenarios, 184, PDNN normalization procedures, 420
201 PDT (pedigree disequilibrium test), 171
Olfactory effects Pearson product-moment correlation, 423
historical research developments, 10 Peas, Mendel study, 3132, 201
localization, 338, 343344 Pedersen, Nancy, 10
phases, 345347, 346347 Pedigree analyses
receptors, 324325 Akaikes information criterion, 256
Oligogenic models, 217, 219 association studies, 257258
Oligophrenin-1, 438 behavior, 249
ONeill and Kaltschmidt studies, 386 bipolar disorders, 230
129P3/J strains, 138 chi-square tests, 256
129 strains, 123124, 138 environmental association studies, 257258
One-way design, sample size requirements, fundamentals, 248
161162, 162 gene-environment interactions, 257258
Ontogeny, 17 genetic association studies, 257258
Open data base connectivity (PDBC), 97 Hendersons mixed model equation,
Open field activity 254255
brain-behavior relations, 404 historical behavioral genetics, 4
emotional behaviors, 297 human comparisons, 249250
historical research developments, 11 latent variable modelo, 251, 252
QTLs, 301 least squares estimation, 253
rearing behavior, 48 linkage studies, 272
selected breeding, 298, 396 Monte Carlo procedures, 254
test, 293 MTDFREMLA, 254258
oPOSSUM algorithm, 385 multiple regression, 252
The Origin of Species, 3 natural history, 249
OR (odds ratio), association studies, 170171 negative variance components, 253
Osborne studies, 312 nested model comparisons, 256
Osher studies, 201205 pedigree analysis methods, 250251
Ospina-Duque studies, 236 personality comparisons, 249
Outbred strains and colonies, mice, 140 recurrent unipolar disorders, 237
Outlook, IIPMF, 406407 squared phenotypic differences, 251252
Ovists, 1718 study goals, 250, 255256
Oxytocin, vole behavior, 12 subjective well-being comparisons, 250
symmetric differences squared, 251254 fibromyalgia, 267

univariate case foundation, 254255 fundamentals, 3031
variances, 251252 heritability, 266
Pedigree analysis methods, 250251 heterogeneity, association studies, 177
Pedigree disequilibrium test (PDT), 171 historical research developments, 12
Peirce, Jeremy, 111 importance, 273
Penetrance, incomplete, 174175, 216 linkage studies, 271272
Penrose, L., 5 molecular genetics, 266
Pentobarbital withdrawal, 461 personality, 263265
Perfect match (PM) pairs, mRNA, 91 schizophrenia, 210211
Perigenual anterior cingulate cortex (pACC), 194 social personality, 270271
Period mutants, 12 theoretical background, 3940
Perphenazine, 451 variability, 191
Persistence, fibromyalgia, 267 Phenylketonuria (PKU), 4, 10, 32, 184, 450
Personalities, see also Pedigree analyses; Pheromones, 23, 321324
Phenotypes and phenotypic selection Phillips studies, 1112, 376, 379
comparisons, 249 Phosphorylation, 78, 8385
historical behavioral genetics, 4 Phototaxis, 6
human research development, 12 Phylogenetic aspects, causation
phenotypes, 263265 examples, 43
Peypelut, Lionel, 314 natural selection/genetic architecture,
Pezawas studies, 194 3839
PFAM overrepresentation, 424 theoretical background, 3943
Pharmacodynamic tradition, 451452 Phylogeny, 105106
Pharmacogenetics Physgen project, 60, 67
complex traits analysis, 453454 Physical mapping, 62
environmental influences, 452 Pierce-Shimomura studies, 357
fundamentals, 449, 455 Pierce studies, 411430
future directions, 454 Pigs, 14
pharmacodynamic tradition, 451452 PIX, 439
pharmacokinetic tradition, 450451 PKA (protein kinase A), 8485
QTL/QTLGene, 453454 PKU, see Phenylketonuria (PKU)
single genes, 452453 PLAD, 107
Pharmacokinetic tradition, 450451 Plasticity
Pharmacological interventions, 43 Caenorhabditis elegans, 356, 356359
Pharmacology, historical research developments, dendritic, 439
9 hippocampal mossy fiber projection, 391
Phases, olfactory memory, 345347, 346347 hippocampal neurotransmission, 441
Phenogenetic aspects, causation nonsyndromic MR genes, 444
brain lesions, 4344 short-term, 441
correlation approach, 4344 Plato (427-346 B.C.E.), 3
examples, 4849 Platt, J.R., 99
genetic correlations, 44 Pleiotropic gene influences, 458
IIPMF, 48 Plomin, Loehlin and DeFries studies, 190
IQ, 4849 Plomin and Daniels studies, 184
locality assumption, 4344 Plomin studies, 10, 1213, 185, 202
nerve conduction velocity, 4849 Plus maze, see Elevated plus maze
QTLs, 48 Pogue-Geile studies, 209222
rearing behavior, 48 Point mutations, 139
SIP-IQ, 4849 Pol II activity, 78, 81
theoretical background, 4548 Polyacrylamide gel electrophoresis (PAGE), 89
Phenome integration, 109110 Polyadenylation, 79
Phenotypes and phenotypic selection Poly-A polymerase, 79
aggression, 284285 Poly-A signal sequence, 79
altruism, 268269 Polycistronic miRNA genes, 85
Index 489
Polygenic analysis, 308311 Prokaryotes, 83

Polymorphisms, expression variation sources, Promoter-proximal elements, 76
87 Promoter region, 75
Population parameters symbols, 151 Protein Data Bank, 106
Portenier, L., 5 Protein kinase A (PKA), 8485
Positional candidate approach, 6567, 221 Protein-protein interactions, 81
Positional cloning Proteomic gene set analysis, 108109
expression profiling, 66 Psychopathology
fundamentals, 6263, 63 altruism, 268
haplotype mapping, 6667 human research development, 12
mapping candidate genes, 6566 personality approach, 264265
positional candidate approach, 6567 P-TEFb factor, 79
pure type, 6364 Published data, estimation, 155156
Positive transcription factors, 81 PubMed, 104, 109
Possidente and Murphey studies, 325 Pure positional cloning, 6364
Post-alcohol injection sleep lines, 9
Post-partum psychotic episodes, 236
Postsynaptic spines, 437, 437439 Q
Potash studies, 228
Power Q2/CAD (constitutive active domain), 83
association studies, 172173, 175, 175 Qian studies, 375, 379
detection failure, 154155 QTDT (quantitative traits disequilibrium test),
expression and brain structure, 429430 171
genetic power calculator program, 173, 175 QTGene, 454
5HTTLPR polymorphism, 193 Quantitative genetics
molecular linkage studies, 230231 brain lesions, 4344
sample size requirements, 156, 158159, correlation approach, 4344
158159 emotional behaviors, 299
theoretical background, 46 examples, 43, 4849
two-way factorial designs, 165 fundamentals, 3738, 51
Power and Precision computer program, 157 genetic architecture, 3839
PPI (prepulse inhibition), 421 genetic correlations, 44
Prader-Willi syndrome, 23 locality assumption, 4344
Prat studies, 333347 misapplications, 4951
Predetermined epigenesis, 18 natural selection, 3839
Preformation, 1719 phenogenetic aspects, causation, 4349
Pregnancy, aggression, 282 phylogenetic aspects, causation, 3843
Pre-initiation complex, 77 theoretical background, 3943, 4548
Preliterate history, 23 Quantitative trait loci (QTLs)
Pre-miRNA, signal transduction, 86 aggression, 283284
Pre-mRNA, transcription, 79 alcohol psychopharmacogenetics, 460461,
Prenatal environment, disease impacts, 183 463464
Prepulse inhibition (PPI), 421 candidate gene selection, 107108, 108
Presynaptic neurotransmitter release, 439442, chromosome substitution strains, 134
440 complex trait analysis, 453
Pri:B6,D2-G# strain, 133 congenic strains and substrains, 116,
Prima facie evidence, 43 123126, 124125, 134
Primary transcript, 79 consomic strains, 134
Primates, maternal contact, 195, see also specific emotional behaviors, 300303, 302
primate expression and brain structure, 413415,
Prion protein (hamster), 138 417419
Prior studies, 401 expression variation sources, 88
Probabilistic epigenesis, 19 gene targeting, 460
Probe sets, 91 genetic reference populations, 102104
Progeny testing, congenic strains, 116 hierarchical format, 98
knockout/congenic strains, 121 Recombinant inbred strains (RIs/RISs)

localization, 49 advanced intercross lines, 60
misapplication, quantitative genetics, 5051 behavioral phenotypes, 378379
pharmacogenetics, 453454 D2 dopamine receptor density, 379
phenogenetic aspects, causation, 48 genetic correlations, 44
recombinant inbred strains, 136137 genetic reference populations, 103
rodents, 11 historical research developments, 11
sample size, 167 multiple inbred strains, 136137, 137
theoretical background, 48 QTLs, 49, 56, 416
Quantitative trait loci (QTLs), gene identification theoretical background, 48
alternative strategies, 5962, 61 Recruitment, regulation of transcription, 81
congenic strains and substrains, 5659, Recurrent unipolar (RUP) disorders, 227
5758 Reductionist approaches, 413
fundamentals, 5556, 67 Rees studies, 235
high-resolution mapping, 5662 Reference population value, 415416, 416
positional candidate approach, 6567 Ref Viz, 109
positional cloning, 6267, 63 Regulated alternative splicing, 80
pure positional cloning, 6364 Regulation, transcription, 8085, 84
Quantitative traits disequilibrium test (QTDT), Regulator binding sites, 75
171 Regulator of G protein signaling 4, chr 1 (RGS4)
Quasi-congenic strains, 122 gene, 221
Regulatory region, 7475
Reichard, Werner, 4
R Rejection of association, 175
Relatedness, 251, 261262
RabGDI1 (Rab GDP-dissociation inhibitor 1), Relational causality, interim solution, 25
441 Relational database management system
Radcliffe studies, 7391 (RDBMS), 9899, 99
Radial maze, 396, 401, see also Mazes Relative risk (RR)
Radiation hybrid mapping, 65 association studies, 170171
Radioisotopes, 90 gene-environment interactions, 179
Rainer, Mendlewicz and, studies, 228 schizophrenia, 212
Ramos studies, 291303 Replicability, multigene-multilife experience, 26
Ramoz studies, 169180 Repressors, regulation of transcription, 8182
Rampon studies, 19 Reproducibility, lack of, 176
Ranganathan studies, 361 Research stage, power and sample size, 158159
Rats, 66, 145, see also Animals; Rodents Rett syndrome, 442
Rattus norvegicus spp., 140 Reverse genetics, C. elegans, 144
Raymond studies, 195 Reverse-transcriptase polymerase reaction (RT-
Rays exposure, 139 PCR), 8990
RCs, see Recombinant congenic strains Revised self-monitoring scale (RSMS), 270271
(RCs/RCSs) Reward dependence, 265, 267, see also Brain
RDBMS (relational database management reward center
system), 9899, 99 RGS4 (Regulator of G protein signaling 4, chr 1)
RD (risk difference), association studies, 170171 gene, 221
Reactive genotype-environment correlation, 190 Rhesus monkeys, 2425, 131, see also
Real world testing, IIPMF, 403406, 404405 Chimpanzees (Pan troglodytes)
Rearing behavior, 48, 396 Ribonucleases, 88
Receptor density, brain dopamine systems, 376, Richardson, Baron and, studies, 287
377379, 378379 Riddle studies, 353, 355
Recessive traits, 3233 Righting reflex, 460
Recombinant congenic strains (RCs/RCSs) Rijsdijk studies, 49
alternative strategies, 59 RIs, see Recombinant inbred strains (RIs/RISs)
genetic correlations, 44 Risch studies, 219
multiple inbred strains, 137138 RISC (RNA-induced silencing complex), 86
Index 491
Risk difference (RD), 170171 one-way design, 161162, 162

Risk marker identification, 463464 power, 158159, 158159
Rissman, Emilie, 11 published data, estimation, 155156
Ritzmann, Tabakoff and, studies, 452 sample size estimation, 156158
RMA normalization procedures, 420 specialized experiments, 167
RNA-editing, 23, 80 specific group comparison, 161162, 162
RNA-induced silencing complex (RISC), 86 strain treatment experiments, 162165,
RNA interference (RNAi), 8586 163, 165
RNA polymerase, 76 two independent groups, 158159, 158159
Robertsonian chromosomes, 139140 Samuel studies, 365
Robinson, Gene, 11 Sasco-Sprague-Dawley rats, 140
Rodents, see also Animals; specific rodent Satiety, rodents, 9, 12
1910 to 1952, 45 Savarit studies, 322
1953 to 1970, 7 Sawin studies, 361
1971 to 1985, 89 Scarr, Sandra, 12
1986 to 2005, 1112 Scarr and McCartney studies, 190
supplier Web sites, 145 Scheiner studies, 313
Roderick, Thomas, 7 Schinka studies, 266
Rogers studies, 363 Schizophrenia
Roman High and Low Avoidance strains, 7 bipolar disorders, 228
Root voles, 405 brain dopamine systems, 371
Ror2, DGV13A candidate genes, 428429 cannabis role, 179
Rose, Richard, 12 fundamentals, 209210, 222
Rosen, Glen, 386 gene chromosomal location, 219220
Rosenthal studies, 7, 215 gene-environment interactions, 189, 198
Rosenzweig, M.R., 7 gene-gene interactions, 204
Roubertoux, Pierre, 9, 11 gene identification, 220221
Roullet and Lassalle studies, 401 genetic effects impact, 216219
Rowe, David, 13 genetic variation, 211216
RR (relative risk), 170171 historical behavioral genetics, 4
RSMS (revised self-monitoring scale), 270271 linkage studies, 272
RT-PCR (reverse-transcriptase polymerase phenotype definition, 210211
reaction), 8990 social skills, 271
Rdin studies, 210211 Schlesinger, Kurt, 9
Rumbaugh studies, 249 Schwegler, Lipp and, studies, 403
RUP, see Recurrent unipolar (RUP) disorders Schwegler studies, 401
Russel, Hedgecock and, studies, 364 Scion Image, 425
Scoring matrices, 105106
Scott, John Paul, 45, 7
S SDS (symmetric differences squared), 251254
Sedensky, Morgan and, studies, 362
SAGE (serial analysis of gene expression), 90 Sgalat studies, 353367
SAL (short attack latency), 12 Segal studies, 10, 184
SamplePower computer program, 157158, 164 Segregating inbred strains (SIs), 138
Sample size, 5HTTLPR polymorphism, 193 Segregation, Law of (Mendel), 3132
Sample size estimation, 156158 Segurado studies, 231
Sample size requirements, see also Animals Selected strains, 136, see also Animals
22 design, 165166, 166167 Selection experiments, emotional behaviors,
detection failure, 154155 298299
effect size, 150152, 150152, 155156 Selective breeding
estimation, 155158 Drosophila melanogaster, 6
false positive probability, 152154 humans, 3
fundamentals, 149150 IIPMF, 392393, 393
multiple inbred strain comparison, 159161, lines studies, 459460
160161 rodents, 89
Semantic Gene Organize (SGO), 109, 423 haplotype mapping, 66

SEM (structural equation modeling), 185 linkage disequilibrium, 173
Sen studies, 266 QTLs, 418
Septal intervention, 43 Single test situations, 297
Sequence analysis scoring matrices, 105106 SIP-IQ (speed-of-information processing), 4849
Sequence comparison, scoring matrices, 106 Sir proteins, 86
The Sequence of the Human Genome, 23 SIs, see Segregating inbred strains (SIs)
Sequence tag sites (STS), 62, 65 Size requirements, see Sample size requirements
Ser-133, 8485 Skeels, H.M., 7
Serial analysis of gene expression (SAGE), 90 Skodak, M., 7
Serotonin, interim solution, 2425 Skolnick, Trullas and, studies, 297
Serotonin 1B receptor (5-HT1B), 204 Slater, E., 5
Serotonin system (5HT) SLC6A4 gene, 191
alcohol sensitivity, 459 SLE (stressful life events), 193194
antidepressant treatment, 453 Sluyter studies, 9, 400
gene-environment interactions, 191192 Smolen, Andrew, 9
Serotonin transporter (5-HTT) gene, Smolen, Toni, 9
see also 5HTTLPR gene Snell, George, 115
bipolar disorders, 235236 SNP, see Single nucleotide polymorphisms
gene-environment interactions, 191 (SNPs)
interim solution, 25 SNPview, 100
phenotypical heterogeneity, 177 Snyder studies, 270
Serotonin transporter (SERT), 204 Social behavior, C. elegans, 10
Serotonin uptake inhibitor antidepressants, 451 Social feeding behavior, C. elegans, 363364
SERT (serotonin transporter), 204 Social interaction test, 293
Sex determination cascade, 326328 Social personality, phenotypes, 270271
Sexual orientation, 12 Society for Neuroscience, 97
Seyfried, Thomas, 9 Sokolowski, Fitzpatrick and, studies, 314
SGO (Semantic Gene Organize), 109, 423 Sokolowski studies, 11, 307314
Sham, Pak, 13 Sonderegger studies, 435444
Shared environmental effects, 184 Songbirds, 413
Sheilds, J., 7 Sources of variation, gene expression, 8788
Shields, Gottesman and, studies, 217 Southern blot, 63
Shorey and Bartell studies, 322 Spearman correlation, 423
Short attack latency (SAL), 12 Specialized experiments, sample size
Short sleep lines, 9, 34, 136, 452 requirements, 167
Sibling Relationship Questionnaire (SRQ), Specific group comparison, 161162, 162
270271 Specificity, expression and brain structure,
Sib-transmission disequilibrium test (S-TDT), 429430
171 Speed congenic strains, 118119
Sickle cell anemia, 32 Speed-of-information processing (SIP) IQ, 4849
Siegmond, Dupuis and, studies, 427 Speiss studies, 6
Signal transduction, 8385, 84 Spermists, 1718
Signal transduction pathways, 83 Spielman and Ewens studies, 171
Significance, 152 Spieth, H.T., 4
Silencers, 75 Spliceosomes, 7980
Silencing genes, 8687 Splicing, transcription, 79
Simpson studies, 386 Spontaneous Hypertensive Rats (SHR), 132,
Simulations, confirmation, 230 299301
Single-gene disorders, 30 Squared phenotypic differences, 251252
Single genes, 308311, 452453 SRQ (Sibling Relationship Questionnaire),
Single nucleotide polymorphisms (SNPs) 270271
bipolar disorders, 231232 Stabilizing selection, 39
DGV13A candidate genes, 428429 State of Connecticut Department of Children and
genetic reference populations, 103 Families (DCF), 193
Index 493
Statistical corrections, 175176 Targeting studies, alcohol

Statistical methods, gene-gene interactions, psychopharmacogenetics, 460462
204205 TAT-SF1 factor, 79
S-TDT (sib-transmission disequilibrium test), Taylor studies, 99, 374
171 TCI (temperament and character inventory), 202
Stem-loop structure, 86 TDT (transmission disequilibrium test), 171
Stockinger studies, 328 Tellegen studies, 46
Strain crossers, 11, see also specific strains and Telomere regions, 86, 174
substrains Temperament and character inventory (TCI), 202
Strain treatment experiments, 162165, 163, Temporal control, gene expression, 341343, 342
165 Terracciano, Antonio, 258
Stress, analytical view, 292 Testosterone-dependent chemosignals, 284
Stressful life events (SLE), 193194 Text mining, 109
Striatal gene expression, 382384 TFIID complex, 82
Structural equation modeling (SEM), 185 TFIIS factor, 79
Structure analysis, behavior bioinformatics, Tg(huAPP695.K670N-M671L)2576 transgenic
106 mice, 138
STS (sequence tag sites), 62, 65 The Descent of Man, 3
Studbooks, 250251 The Expression of Emotions in Man and
Study goals, pedigree analyses, 250, 255256 Animals, 3
Sturtevant, A., 4 The Genetic Basis of Alcohol and Drug Actions,
Suarez studies, 230 11
Subjective well-being (SWB), 248, 250, 256, The Genetics and Social Behavior of the Dog, 7
see also Pedigree analyses The Genetics of Behavior, 8
Sub-Saharan African descent, 32 The Incidence of Alkaptonuria: A Study in
Substance abuse, 371, see also specific substance Chemical Individuality, 450
Succinycholine metabolism, 450 Theissen, Delbert, 89
Suicidal behavior, 177 Thelen studies, 20
Suitability for studies, C. elegans, 355 Theognis (6th century, B.C.E.), 23
Sulston, John, 355 Theoretical background
Suomi studies, 2425 phenogenetic aspects, causation, 4548
Super-control cohorts, 174 phylogenetic aspects, causation, 3943
Svare, Bruce, 9 Theoretical personality approach, 265
SWB (subjective well-being), 248, 250, 256, The Origin of Species, 3
see also Pedigree analyses Theory of mind, chimpanzees, 249
Swiss Webster mice, 140, 284 Thermal nest size, 9
SymAtlas, 382 Thermotaxis, 10
Symbols, population parameters, 151 The Sequence of the Human Genome, 23
Symmetric differences squared (SDS), 251254 Thiemann, Kraemer and, studies, 156157
Synaptic mechanisms, see Mental retardation Thompson, Fuller and, studies, 8
(MR) Thompson, R., 6
-Synuclein, 66 Thoridizine, 451
Syracuse high/low avoidance lines, 9 Threshold cycle, 89
Systems-level bioinformatics, 109110, see also Thyroid function, 9
Bioinformatics, behavior Tienari studies, 198
Szekely studies, 202 Time, Love, and Memory, 8
Timeless mutants, 12
Toma studies, 309
T Tools, anatomo-functional maps, 343
Touch sensory systems, 10, 22
Tabakoff and Ritzmann studies, 452 TPQ (Tridimensional Personality Questionnaire),
Takahashi, Joseph, 12 265, 267268
Talbot studies, 60 Trait links, emotional behaviors, 299300
Taqman-based quantitative reverse-transcriptase Trans-acting QTLs, 88, 418, 418419
polymerase reaction (RT-PCR), 8990 Transcript, mRNA biogenesis, 76
Transcription, see also Expression profiling; University of Tennessee Health Sciences Center,
Gene expression 12, 136137
factors, 76 Unpacked model, 20, 21
fundamentals, 7680, 7778 3 untranslated region (3UTR), 82
regulation, 8085, 84
signal transduction, 8385, 84
Transgenesis, 144, 303 V
Transgenic rodents and studies
alcohol psychopharmacogenetics, 461462 Vale, Jack, 9
fundamentals, 138 Van Abeelen, Crusio and, studies, 43
historical research developments, 1112 van Abeelen studies, 8, 38, 52
Translocations, 139 Vandenberg, S., 8
Transmission disequilibrium test (TDT), 171 van Oortmerssen, Geert, 7, 9
TRANSMIT program, 172 Van Tol, Wong, Buckle and, studies, 24
Transporter and receptor density, 376, 377379, Variable number of tandem repeats (VNTRs),
378379 235
Tridimensional Personality Questionnaire (TPQ), Variances
265, 267268 Hendersons mixed model equation,
Trullas and Skolnick studies, 297 254255
Tryon effect, 46 latent variable modelo, 251, 252
Tryon studies, 4, 46 least squares estimation, 253
Tryptopham hydroxylase, aggression, 287 multiple regression, 252
Tsien, J.Z., 11 negative variance components, 253
Tully, Tim, 11 pedigree analyses, 254255
Turners Syndrome, 7, 12 proportions estimation, 252
Turri studies, 301 theoretical background, 4546
Twins studies, see also Adoption studies; Family Variations, genetic, 8788, 211216
and twin studies; Human studies Vasopressin systems, voles, 12
alcohol psychopharmacogenetics, 462463 Vaughn, C.H., 11
bipolar disorders, 228229, 229 Venard and Jallon studies, 321
general behavior genetic methods, 184185 Venard studies, 324
historical research developments, 12 Vincent studies, 235
Lousiville Twin Study, 7, 10 Visual stimuli, 320321
nerve conduction velocity, 49 VNTRs (variable number of tandem repeats), 235
recurrent unipolar disorders, 237 Volavka studies, 287
schizophrenia, 212, 213214 Voles, 12
22 design, 165166, 166167 Volkow studies, 372
Two independent groups, 158159, 158159 Von Knorring studies, 229
Two-way avoidance learning, 394, 403 von Teschermak studies, 4
Two-way factorial designs, 162165
Tyrosine, 450
W
U Waddington, Conrad H., 30, 34
Wahlsten, Cohen and, studies, 157
Ultrasonic pup vocalization, 295, 298 Wahlsten studies, 12, 149167
Unconditioned stress responses, 293, 294, 295, Wald, George, 4
296 Waldman, Irwin, 12
Unique environmental effects, 184 Walker studies, 382
Univariate analysis, 45, 185186 Ward studies, 357
Univariate case foundation, 254255 Water maze, 163, 403404, see also Mazes;
University of California San Diegos Biology Morris water maze
Workbench, 105 Watson, James, 5, 75
University of Illinois Biology Student Watson-Crick base pairing, 76, 86, 88
Workbench, 106 Watson-Crick complementarity, 77
Index 495
WebGestalt, 101, 423424 Wes and Bargmann studies, 357

WebQTL, see also GeneNetwork (GN) Western analysis, 460
catalepsy phenotype, 373 Wet laboratory procedures, 91
D2 dopamine receptor density, 380382 Whalsten, Douglas, 9
dentate gyrus volume, 427 White studies, 334
genetic reference populations, 102103 Whitfield studies, 183188
interval map view, 107 Wicker-Thomas studies, 322
QTL mapping, 419 Wicks studies, 357
systems-level bioinformatics, 109 Wild-derived mice, 140, 453
Web sites Wild voles, 405
Allen Brain Atlas, 426 Williams studies, 11, 386387, 411430
animal resources, 144146 Williams syndrome, 23, 439
BioCarta, 424 Wilson, Ron, 10
Biology Student Workbench, 106 Wistar Kyoto (WKY) rats, 132, 299, 301
Biology Workbench, 105 Wistar rats, 140, 297
Caenorhabditis elegans, 145146 WKHA (normotenisve but hyperactive) rats,
Chilibot, 109 299
comparative mapping, 60 Wolf, F.E., 11
Complex Trait Consortium, 62 Wolfe, Lennox and, studies, 270
Drosophila melanogaster, 145, 335337 Wolfer, Lipp and, studies, 392
Ensembl browser, 100 Wolfer studies, 389407
EntrezGene, 101 Wolves, 10, see also Dogs
GeneKeyDB, 101 Wong, Buckle and Van Tol studies, 24
GeneNetwork, 415 Wood mice, 405
gene Onotology Tree Machine, 104 Wood studies, 353
genetic power calculator program, 173, Worker bees, see Bees
175 Wrights coefficient of relatedness, 251
GENSAT, 426 Wright studies, 4
HAPMAP, 173
KEGG, 424
MEME, 106 X
meta-analysis, 156
model organisms virtual library, 144 X chromosome
MOTIF, 109 actin cytoskeleton, 437, 437439
mouse, 145 consomic strains, 119
Mouse Phenome Project, 159 Cool-2, 439
Mouse SNP Selector, 100 FACL4, 441442
National Center for Biotechnology fundamentals, 436437, 442
Information, 106 IL1RAPL1, 441
Neuroscience Database Gateway, 97 oligophrenin-1, 438
PFAM, 424 PAK3, 438439
PLAD, 107 PIX, 439
Protein Data Bank, 106 postsynaptic spines, 437, 437439
rat, 145 presynaptic neurotransmitter release,
rodents suppliers, 145 439442, 440
SNPview, 100 RabGDI1, 441
WebGestalt, 101
WebQTL, 102
Wehner, Jeanne, 11 Y
Weimer, Cynthia, 9
Weimer, Richard, 9 Yang studies, 338, 385
Weiner Jonathan, 8 Y chromosomes, consomic strains, 119
Weiss studies, 247258 Yeast artificial chromosomes (YACs), 62, 65
Well-bred parents and children, 3 Yin studies, 345
Wender studies, 229 Young, Larry J., 12
Z Zoo effects, 251, 253, 255

Zubenko studies, 237
Zars studies, 344 Zucker rats, 140
Zebrafish, 14 Zrich laboratory, 404
1903_Color Insert.fm Page 497 Thursday, July 27, 2006 6:39 PM
497
FIGURE 16.2 Confirmed linkage loci for bipolar disorder (B), schizophrenia (S), and both
disorders (*) on a diagram of the human genome.
KC KC
Ca Ca
Ca
LH
' '
' LH
AGT AGT

' '
AL AL
FIGURE 23.1 Drosophila olfactory memory system. Olfactory sensory neurons project to
the antennal lobes (ALs). From there, projection neurons project through the antenno-glom-
erular tract (AGT) and connect mushroom body (MB) dendrites localized in the calyx (Ca),
as well as the lateral horn (LH). Each MB is composed of about 2,500 neurons, the Kenyon
cells (KC). Three types of KC project in five lobes: /, / and .
498 Neurobehavioral Genetics: Methods and Applications, Second Edition
Mushroom body + Enhancer-trap transposable element Mushroom body-expressed

enhancer 5IR 3IR gene
GAL4 w+ ampR
AAAAA
GAL4-mRNA
GFP expressed
within MB
AAAAA
GFP-mRNA
GAL4
+
UAS
UAS
UAS
UAS
UAS
GFP
5IR 3IR
FIGURE 23.2 The GAL4/UAS system. When an enhancer-trap transposable element is

inserted near transcription enhancers that control expression in a given structure, the GAL4
gene that is contained in the P-element is expressed in the same structure. If this fly contains
also a reporter gene downstream of the UAS sequences, GAL4 will transcribe the reporter.
FIGURE 24.3 DIC photograph of the head of a worm. The outline of the pharynx (p) can
be seen and the beginning of the gut (g). The approximate positions of the amphid sensory
neuron cell bodies are indicated with colored circles. The dendrite (d) and an axon (a) of one
of the cells are drawn. Below, the names of the corresponding cells are given, and the
compounds perceived by these cells.
499
Extracellular signals
(Adhesion molecules, growth factors, electric/synaptic activity)
Rho GDI
Rho-GDP
Rac-GDP
Cdc42-GDP
OPHN1 Rho GAP Rho GEF ARHGEF6

Rho-GTP
Rac-GTP
Cdc42-GTP
Rho effectors PAK3
Actin cytoskeleton
of dendritic spines
FIGURE 28.1 One cluster of nonsyndromic MR genes relates to the Rho family of small
GTPases. Rho family GTPases transduce extracellular signals into adaptive responses of the
actin cytoskeleton. Actin-dependent processes at the synapse include the regulation of the
morphology and the dynamics of the dendritic spines. The Rho family comprises three
members, termed Rho, Rac, and Cdc42. They shuttle between an inactive (red) and an active
(green) state under the control of three types of regulatory proteins (blue). GEFs (guanine
nucleotide exchange factors) promote the release of GDP and its replacement by GTP and,
thereby, mediate the transition of Rho, Rac, and Cdc42 from the inactive into the active state.
GAPs (GTPase-activating proteins) activate the endogenous GTPase function and, thus, the
self-inactivation of Rho, Rac, and Cdc42. GDIs (GDP dissociation inhibitors) bind to the
GDP form of the GTPases and prevent their premature activation as long as the GTPase is
not in the correct place and situation for a new round of activation. Further downstream
activators (yellow) eventually act directly or indirectly on regulatory components of the actin
cytoskeleton. The genes affected in MRs are printed red.
500 Neurobehavioral Genetics: Methods and Applications, Second Edition
Neurotransmitter
synthesis/uptake
Rab GEF
Rab-GTP
Rab-GDP Rab GDI1
Rab GDI
Vesicle
IL1RAPL1 recycling
(endocytosis)
Rab-GTP
Rab-GTP Rab-GDP
NCS-1
Docked Fused
FACL4
Rab GAP Restricting
vesicle Ca2+ vesicle ilin
oph
V
bud End
FIGURE 28.2 A second cluster of nonsyndromic MR genes relates to the release of neu-
rotransmitters from the presynaptic nerve terminal and the regeneration of new releasable
vesicles by the budding and endocytosis of vesicles from the presynaptic membrane. The
release of neurotransmitters from storage vesicles occurs in several consecutive steps. First,
the vesicles are docked to the release site, the presynaptic active zone, a process that involves
the interaction of several proteins of the vesicle and the active zone. Docked vesicles fuse
with the presynaptic membrane in a process regulated by transiently increased intracellular
calcium, and release their content into the synaptic cleft. In order to maintain a constant
supply of releasable neurotransmitters, vesicles recycle by budding off the presynaptic mem-
brane and endocytosis. The budding occurs in a clathrin-dependent manner. The restriction
of the bud neck, a process that prepares the bud for scission, requires endophilin. Endophilin
is a lysophosphatidic acid acyltransferase, which introduces arachidonic acid into the cyto-
plasmic face of the bud neck. It, thereby, facilitates the formation of a strong negative curvature
of the membrane required for the restriction of the bud neck in order to allow the scission
process to begin. It is conceivable that the supply of arachidonoyl-CoA requires FACL4. The
genes affected in MR are printed in red.

Neurobehavioral Genetics - Methods and Applications 2nd Ed - B. Jones, P. Mormede (CRC, 2007) WW

Uploaded by

Copyright:

Available Formats

Neurobehavioral Genetics - Methods and Applications 2nd Ed - B. Jones, P. Mormede (CRC, 2007) WW

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Neurobehavioral Genetics - Methods and Applications 2nd Ed - B. Jones, P. Mormede (CRC, 2007) WW

Uploaded by

Copyright:

Available Formats

Second Edition

Pierre Mormde, D.V.M., Ph.D.

Boca Raton London New York

CRC is an imprint of the Taylor & Francis Group,

No claim to original U.S. Government works

International Standard Book Number-10: 0-8493-1903-X (Hardcover)

Donald J. Nash Lorraine Flaherty

B.C.J & P.M.

Byron C. Jones, Ph.D., is a professor of biobehavioral health and pharmacology at

Pierre Mormde, D.V.M., Ph.D., is director of research at the French National

John Belknap Elissa J. Chesler

Wim E. Crusio Guillaume Isabel

Stephen C. Maxson Michael F. Pogue-Geile

Shannon McWeeney Thomas Prat

Marie-Pierre Moisan Richard A. Radcliffe

Peter Sonderegger Keith E. Whitfield

Douglas Wahlsten Robert W. Williams

1.2 PRELITERATE AND ANCIENT HISTORY

A History of Behavior Genetics 3

1.3 TWO VICTORIAN COUSINS

1.4 THE FIRST CENTURY OF BEHAVIOR GENETICS

1.4.1 1910 TO 195212,17

1.4.1.1 Fruit Flies12

A History of Behavior Genetics 5

1.4.2 1953 TO 197023,25,27

1.4.2.1 Fruit Flies

A History of Behavior Genetics 7

1.4.3 1971 TO 19854,9,14

1.4.3.2 Fruit Flies18,39

In 1973, Peter A. Parsons published Behavioral and Ecological Genetics: A Study

In 1970, Gardner Lindzey and Delbert D. Thiessen edited Contributions to Behavior-

A History of Behavior Genetics 9

Research continued on cognition, personality, and psychopatholgy with family, twin,

1.4.4 1986 TO 20058,31,32,37

A History of Behavior Genetics 11

1.4.4.2 Fruit Flies21,35,37

A History of Behavior Genetics 13

1.5 JANUS: THE PAST AND FUTURE OF BEHAVIOR

1.6 A HIGHLY PERSONAL NOTE

A History of Behavior Genetics 15

2.1 PREFORMATION AND EPIGENESIS

Developmental Neurobehavioral Genetics 19

agents of development thus brought into question, the idea of a naturenurture

2.2 DEVELOPMENT AS EXPLANATION

reductionistic in the sense that psychological understanding or explanation does not

Developmental Neurobehavioral Genetics 21

FIGURE 2.2 Completely unpacked model of developmental neurobehavioral genetics, show-

Developmental Neurobehavioral Genetics 23

to be involved in changes in female olfactory responsiveness to male pheromones

2.3 A CURRENTLY ACCEPTABLE DEVELOPMENTAL

Developmental Neurobehavioral Genetics 25

2.4 SUMMARY AND CONCLUSIONS

Developmental Neurobehavioral Genetics 27

3 Some Basics, Mendelian

3.1 Introduction and Levels of Investigation....................................................... 29

3.1 INTRODUCTION AND LEVELS OF

and development of behavior dominated the discipline of psychology in America.

3.2 FUNDAMENTAL QUESTIONS. WHAT

3.3 PHENOTYPES AND PHENOTYPIC SELECTION

Some Basics, Mendelian Traits, Polygenic Traits, Complex Traits 31