Proteomics: Methods and Protocols

Methods in
Molecular Biology 1550
Lucio Comai · Jonathan E. Katz

Parag Mallick Editors
Proteomics
Methods and Protocols
Methods in Molecular Biology
Series Editor
John M. Walker
School of Life and Medical Sciences
University of Hertfordshire
Hatfield, Hertfordshire, AL10 9AB, UK
For further volumes:

http://www.springer.com/series/7651
Proteomics
Methods and Protocols
Edited by
Lucio Comai
Keck School of Medicine, University of Southern California, Los Angeles, CA, USA
Jonathan E. Katz
Keck School of Medicine, University of Southern California, Los Angeles, CA, USA
Parag Mallick
Stanford University, Palo Alto, CA, USA
Editors
Lucio Comai Jonathan E. Katz
Keck School of Medicine Keck School of Medicine
University of Southern California University of Southern California
Los Angeles, CA, USA Los Angeles, CA, USA
Parag Mallick
Stanford University
Palo Alto, CA, USA
ISSN 1064-3745 ISSN 1940-6029 (electronic)

Methods in Molecular Biology
ISBN 978-1-4939-6745-2 ISBN 978-1-4939-6747-6 (eBook)
DOI 10.1007/978-1-4939-6747-6
Library of Congress Control Number: 2017930810
© Springer Science+Business Media LLC 2017

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is
concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction
on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation,
computer software, or by similar or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not
imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and
regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to
be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty,
express or implied, with respect to the material contained herein or for any errors or omissions that may have been made.
Printed on acid-free paper
This Humana Press imprint is published by Springer Nature

The registered company is Springer Science+Business Media LLC
The registered company address is: 233 Spring Street, New York, NY 10013, U.S.A.
Preface
In the catalog of biochemical techniques, proteomics has barely reached its adolescent stage,
and a very immature adolescent at that. Like any teen, potential still overshadows realized
accomplishment, but the future is still bright with potential. This particular adolescent has
shown quite a level of promise. Indeed, for a number of tasks, proteomics is fully proficient—
determining the identity of a small number of proteins, providing absolute quantitation of a
similar number of proteins. For others, it is still testing its limits: How many proteins? How
many orders of magnitude of sensitivity? And we begin to doubt, but it is not impossible to
imagine the realization of the full parental dream—given a sample, what are the concentra-
tions and identity of every protein and every modification on those proteins. And then there
are the unexpected questions we can answer—our teenager has shown potential in areas we
never first imagined: What is the three-dimensional structure of a protein? How do proteins
interact? What is the turnover rate of various post-translational modifications?
We view proteomics as a pipeline with four discrete components: The isolation of material
from a biological specimen, sample preprocessing, sample analysis, and data interpretation.
Recognizing proteomic analysis almost always is a collaborative effort and that special-
ized analyses will always require domain-specific knowledge, our goals with this book are to
provide step-by-step protocols on a wide range of biochemical methods, analytical
approaches, and bioinformatics tools developed to analyze the proteome. Here are our
specific goals for this book:
1. Accessible. Most scientists in the life sciences will be able to employ the methods described
in this book. Aside from basic mass spectrometers, we have avoided unusual and/or
expensive equipment and reagents. (Specialists do not consult books as a primary
reference.)
2. Practical. The techniques herein described are broadly applicable, commonly employed
protocols.
3. Current. Mature well-established protocols will be referenced and briefly described.
“State-of-the-art” emerging standard protocols will be clearly and completely described—
common wisdom included at no extra charge!
4. One stop. Recognizing that proteomics is often a collaborative effort, this book shall
describe, as we see it, the complete proteomic pipeline, upfront biology through data
analysis. For analysis that has become or is emerging as routine, our hopes are for this to
be the “go to” reference.
Los Angeles, CA Lucio Comai

Los Angeles, CA Jonathan E. Katz
Palo Alto, CA Parag Mallick
v
Contents
Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v
Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix
1 A Robust Protocol for Protein Extraction and Digestion . . . . . . . . . . . . . . . . . 1

Michelle Atallah, Mark R. Flory, and Parag Mallick
2 Improving Proteome Coverage and Sample Recovery with Enhanced
FASP (eFASP) for Quantitative Proteomic Experiments . . . . . . . . . . . . . . . . . . 11
Jonathan Erde, Rachel R. Ogorzalek Loo, and Joseph A. Loo
3 Proteome Characterization of a Chromatin Locus Using the Proteomics
of Isolated Chromatin Segments Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
Sophie L. Kan, Nehmé Saksouk, and Jérome Déjardin
4 Profiling Cell Lines Nuclear Sub-proteome . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
Aline Poersch, Andrea G. Maria, Camila S. Palma, Mariana L. Grassi,
Daniele Albuquerque, Carolina H. Thomé, and Vitor M. Faça
5 Optimized Enrichment of Phosphoproteomes by Fe-IMAC Column
Chromatography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
Benjamin Ruprecht, Heiner Koch, Petra Domasinska, Martin Frejno,
Bernhard Kuster, and Simone Lemeer
6 Full Membrane Protein Coverage Digestion and Quantitative
Bottom-Up Mass Spectrometry Proteomics . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
Joseph Capri and Julian P. Whitelegge
7 Hydrophilic Strong Anion Exchange (hSAX) Chromatography
Enables Deep Fractionation of Tissue Proteomes . . . . . . . . . . . . . . . . . . . . . . . 69
Benjamin Ruprecht, Dongxue Wang, Riccardo Zenezini Chiozzi,
Li-Hua Li, Hannes Hahne, and Bernhard Kuster
8 High pH Reversed-Phase Micro-Columns for Simple, Sensitive,
and Efficient Fractionation of Proteome and (TMT labeled)
Phosphoproteome Digests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
Benjamin Ruprecht, Jana Zecha, Daniel P. Zolg, and Bernhard Kuster
9 Multi-Lectin Affinity Chromatography for Separation,
Identification, and Quantitation of Intact Protein Glycoforms
in Complex Biological Mixtures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
Sarah M. Totten, Majlinda Kullolli, and Sharon J. Pitteri
10 Parallel Exploration of Interaction Space by BioID and Affinity
Purification Coupled to Mass Spectrometry . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
Geoffrey G. Hesketh, Ji-Young Youn, Payman Samavarchi-Tehrani,
Brian Raught, and Anne-Claude Gingras
11 LUMIER: A Discovery Tool for Mammalian Protein Interaction Networks . . . 137
Miriam Barrios-Rodiles, Jonathan D. Ellis, Benjamin J. Blencowe,
and Jeffrey L. Wrana
vii
viii Contents
12 Dual-Color, Multiplex Analysis of Protein Microarrays

for Precision Medicine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
Solomon Yeon, Florian Bell, Michael Shultz, Grace Lawrence,
Michael Harpole, and Virginia Espina
13 Quantitative Proteomics Using SILAC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
Kian Kani
14 Relative Protein Quantification Using Tandem Mass Tag
Mass Spectrometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
Lichao Zhang and Joshua E. Elias
15 Pathway-Informed Discovery and Targeted Proteomic Workflows
Using Mass Spectrometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
Caroline S. Chu, Christine A. Miller, Andy Gieschen, and Steve M. Fischer
16 Generation of High-Quality SWATH® Acquisition Data for Label-free
Quantitative Proteomics Studies Using TripleTOF® Mass Spectrometers . . . . . 223
Birgit Schilling, Bradford W. Gibson, and Christie L. Hunter
17 Annotating Mutational Effects on Proteins and Protein Interactions:
Designing Novel and Revisiting Existing Protocols . . . . . . . . . . . . . . . . . . . . . . 235
Minghui Li, Alexander Goncearenco, and Anna R. Panchenko
18 Protein Micropatterning Assay: Quantitative Analysis
of Protein–Protein Interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
Gerhard J. Schütz, Julian Weghuber, Peter Lanzerstorfer, and Eva Sevcsik
19 Designing Successful Proteomics Experiments . . . . . . . . . . . . . . . . . . . . . . . . . 271
Daniel Ruderman
20 Automated SWATH Data Analysis Using Targeted Extraction
of Ion Chromatograms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289
Hannes L. Röst, Ruedi Aebersold, and Olga T. Schubert
21 Virtualization of Legacy Instrumentation Control Computers
for Improved Reliability, Operational Life, and Management . . . . . . . . . . . . . . 309
Jonathan E. Katz
22 Statistical Assessment of QC Metrics on Raw LC-MS/MS Data . . . . . . . . . . . . 325
Xia Wang
23 Data Conversion with ProteoWizard msConvert . . . . . . . . . . . . . . . . . . . . . . . 339
Ravali Adusumilli and Parag Mallick
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 369
Contributors
Ravali Adusumilli • Department of Radiology, Canary Center at Stanford for Cancer

Early Detection, Stanford University, Stanford, CA, USA
Ruedi Aebersold • Institute of Molecular Systems Biology, ETH Zurich, Zurich,
Switzerland; Faculty of Science, University of Zurich, Zurich, Switzerland
Daniele Albuquerque • Department of Biochemistry and Immunology, Ribeirão Preto
Medical School, University of São Paulo, Ribeirão Preto, SP, Brazil
Michelle Atallah • Canary Center at Stanford for Cancer Early Detection, Stanford
University, Palo Alto, CA, USA
Miriam Barrios-Rodiles • Center for Systems Biology, Lunenfeld-Tanenbaum Research
Institute, Mount Sinai Hospital, Toronto, ON, Canada
Florian Bell • Grace Bio-Labs, Bend, OR, USA
Benjamin J. Blencowe • Donnelly Centre, University of Toronto, Toronto, ON, Canada;
Department of Molecular Genetics, University of Toronto, Toronto, ON, Canada
Joseph Capri • Department of Pharmacology, David Geffen School of Medicine,
Los Angeles, CA, USA
Riccardo Zenezini Chiozzi • Department of Chemistry, Sapienza Università di Roma,
Rome, Italy
Caroline S. Chu • Agilent Technologies, Inc., Santa Clara, CA, USA
Jérome Déjardin • INSERM AVENIR, Institute of Human Genetics CNRS UPR1142,
Montpellier, France
Petra Domasinska • Biomedical Research Center, University Hospital Hradec Kralove,
Hradec Kralove, Czech Republic; Faculty of Chemical Technology, Department of
Biological and Biochemical Sciences, University of Pardubice, Pardubice, Czech Republic
Joshua E. Elias • Department of Chemical & Systems Biology, Stanford University,
Stanford, CA, USA
Jonathan D. Ellis • Donnelly Centre, University of Toronto, Toronto, ON, Canada
Jonathan Erde • Department of Chemistry and Biochemistry, University of California-Los
Angeles, Los Angeles, CA, USA
Virginia Espina • Center for Applied Proteomics and Molecular Medicine, George Mason
University, Manassas, VA, USA
Vitor M. Faça • Department of Biochemistry and Immunology, Ribeirão Preto Medical
School, University of São Paulo, Ribeirão Preto, SP, Brazil; Center for Cell Based
Therapy - Hemotherapy Center of Ribeirão Preto, Ribeirão Preto Medical School,
University of São Paulo, Ribeirão Preto, SP, Brazil
Steve M. Fischer • Agilent Technologies, Inc., Santa Clara, CA, USA
Mark R. Flory • Canary Center at Stanford for Cancer Early Detection, Stanford
Martin Frejno • Chair of Proteomics and Bioanalytics, Technische Universität München,
Freising, Germany; Department of Oncology, University of Oxford, Oxford, UK
Bradford W. Gibson • The Buck Institute for Research on Aging, Redwood City, CA,
USA
ix
x Contributors
Andy Gieschen • Agilent Technologies, Inc., Santa Clara, CA, USA

Anne-Claude Gingras • Lunenfeld-Tanenbaum Research Institute, Mount Sinai Hospital,
Toronto, Canada; Department of Molecular Genetics, University of Toronto, Toronto,
Canada
Alexander Goncearenco • National Center for Biotechnology Information, National
Institutes of Health, Bethesda, MD, USA
Mariana L. Grassi • Department of Biochemistry and Immunology, Ribeirão Preto
Medical School, University of São Paulo, Ribeirão Preto, SP, Brazil; Center for Cell Based
Hannes Hahne • OmicScouts GmbH, Freising, Germany
Michael Harpole • Center for Applied Proteomics and Molecular Medicine, George Mason
Geoffrey G. Hesketh • Lunenfeld-Tanenbaum Research Institute, Mount Sinai Hospital,
Toronto, Canada
Christie L. Hunter • SCIEX, Framingham, MA, USA
Sophie L. Kan • INSERM AVENIR, Institute of Human Genetics CNRS UPR1142,
Montpellier, France
Kian Kani • USC Center for Applied Molecular Medicine, USC Keck School of Medicine,
Los Angeles, CA, USA
Jonathan E. Katz • USC Center for Applied Molecular Medicine, Los Angeles, CA, USA
Heiner Koch • Chair of Proteomics and Bioanalytics, Technische Universität München,
Freising, Germany; German Cancer Consortium (DKTK), Heidelberg, Germany;
German Cancer Research Center (DKFZ), Heidelberg, Germany
Majlinda Kullolli • Department of Radiology, Canary Center at Stanford for Cancer
Early Detection, Stanford University School of Medicine, Palo Alto, CA, USA
Bernhard Kuster • Chair of Proteomics and Bioanalytics, Technische Universität
München, Freising, Germany; Center for Integrated Protein Science Munich (CIPSM),
Freising, Germany; German Cancer Consortium (DKTK), Heidelberg, Germany;
German Cancer Research Center (DKFZ), Heidelberg, Germany; Bavarian
Biomolecular Mass Spectrometry Center, Technische Universität München, Freising,
Germany
Peter Lanzerstorfer • School of Engineering and Environmental Sciences, University
of Applied Sciences Upper Austria, Wels, Austria
Grace Lawrence • Center for Applied Proteomics and Molecular Medicine, George Mason
Simone Lemeer • Chair of Proteomics and Bioanalytics, Technische Universität München,
Freising, Germany; Center for Integrated Protein Science Munich (CIPSM), Freising,
Germany; Biomolecular Mass Spectrometry and Proteomics, Bijvoet Center for
Biomolecular Research and Utrecht Institute of Pharmaceutical Sciences, Utrecht
University, Utrecht, The Netherlands
Li-Hua Li • Department of Pathology and Laboratory Medicine, Taipei Veterans General
Hospital, Taipei, Taiwan, R.O.C.
Minghui Li • National Center for Biotechnology Information, National Institutes
of Health, Bethesda, MD, USA
Joseph A. Loo • Department of Chemistry and Biochemistry, University of California-Los
Angeles, Los Angeles, CA, USA; Department of Biological Chemistry, University of
California-Los Angeles, Los Angeles, CA, USA
Contributors xi
Parag Mallick • Department of Radiology, Canary Center at Stanford for Cancer Early

Detection, Stanford University, Stanford, CA, USA; School of Medicine, Stanford
Andrea G. Maria • Department of Pediatrics, Ribeirão Preto Medical School, University
of São Paulo, Ribeirão Preto, SP, Brazil
Christine A. Miller • Agilent Technologies, Inc., Santa Clara, CA, USA
Rachel R. Ogorzalek Loo • Department of Biological Chemistry, University
of California-Los Angeles, Los Angeles, CA, USA
Camila S. Palma • Department of Biochemistry and Immunology, Ribeirão Preto Medical
Anna R. Panchenko • National Center for Biotechnology Information, National Institutes
of Health, Bethesda, MD, USA
Sharon J. Pitteri • Department of Radiology, Canary Center at Stanford for Cancer
Aline Poersch • Department of Biochemistry and Immunology, Ribeirão Preto Medical
Brian Raught • Princess Margaret Research Institute, Princess Margaret Cancer Centre,
University Health Network, Toronto, Canada; Department of Medical Biophysics,
University of Toronto, Toronto, Canada
Hannes L. Röst • Institute of Molecular Systems Biology, ETH Zurich, Zurich,
Switzerland; Department of Genetics, Stanford University, Stanford, CA, USA
Daniel Ruderman • Lawrence J. Ellison Institute for Transformative Medicine of USC,
USC Keck School of Medicine, Los Angeles, CA, USA
Benjamin Ruprecht • Chair of Proteomics and Bioanalytics, Technische Universität
München, Freising, Germany; Center for Integrated Protein Science Munich (CIPSM),
Freising, Germany
Nehmé Saksouk • INSERM AVENIR, Institute of Human Genetics CNRS UPR1142,
Montpellier, France
Payman Samavarchi-Tehrani • Lunenfeld-Tanenbaum Research Institute, Mount Sinai
Hospital, Toronto, Canada
Birgit Schilling • The Buck Institute for Research on Aging, Redwood City, CA, USA
Olga T. Schubert • Department of Human Genetics, University of California,
Los Angeles, CA, USA; Institute of Molecular Systems Biology, ETH Zurich, Zurich,
Switzerland
Gerhard J. Schütz • Institute of Applied Physics, TU Wien, Vienna, Austria
Eva Sevcsik • Institute of Applied Physics, TU Wien, Vienna, Austria
Michael Shultz • Grace Bio-Labs, Bend, OR, USA
Carolina H. Thomé • Department of Biochemistry and Immunology, Ribeirão Preto
Medical School, University of São Paulo, Ribeirão Preto, SP, Brazil; Center for Cell Based
Sarah M. Totten • Department of Radiology, Canary Center at Stanford for Cancer
xii Contributors
Dongxue Wang • Chair of Proteomics and Bioanalytics, Technische Universität München,

Freising, Germany
Xia Wang • Department of Mathematical Sciences, University of Cincinnati, Cincinnati,
OH, USA
Julian Weghuber • School of Engineering and Environmental Sciences, University
of Applied Sciences Upper Austria, Wels, Austria
Julian P. Whitelegge • The Pasarow Mass Spectrometry Laboratory, The Jane and Terry
Semel Institute for Neuroscience and Human Behavior, David Geffen School of Medicine,
UCLA, Los Angeles, CA, USA
Jeffrey L. Wrana • Department of Molecular Genetics, University of Toronto, Toronto,
ON, Canada; Center for Systems Biology, Lunenfeld-Tanenbaum Research Institute,
Mount Sinai Hospital, Toronto, ON, Canada; Breast Cancer Research, Mary Janigan
Chair in Molecular Cancer Therapeutics, Toronto, ON, Canada
Solomon Yeon • Center for Applied Proteomics and Molecular Medicine, George Mason
Ji-Young Youn • Lunenfeld-Tanenbaum Research Institute, Mount Sinai Hospital,
Toronto, Canada
Jana Zecha • Technische Universität München, Freising, Germany; German Cancer
Consortium (DKTK), Heidelberg, Germany; German Cancer Research Center (DKFZ),
Heidelberg, Germany
Lichao Zhang • Department of Chemical & Systems Biology, Stanford University,
Stanford, CA, USA
Daniel P. Zolg • Technische Universität München, Freising, Germany
Chapter 1
A Robust Protocol for Protein Extraction and Digestion

Michelle Atallah, Mark R. Flory, and Parag Mallick
Abstract
Proteins play a key role in all aspects of cellular homeostasis. Proteomics, the large-scale study of proteins,
provides in-depth data on protein properties, including abundances and post-translational modification
states, and as such provides a rich avenue for the investigation of biological and disease processes. While
proteomic tools such as mass spectrometry have enabled exquisitely sensitive sample analysis, sample prep-
aration remains a critical unstandardized variable that can have a significant impact on downstream data
readouts. Consistency in sample preparation and handling is therefore paramount in the collection and
analysis of proteomic data.
Here we describe methods for performing protein extraction from cell culture or tissues, digesting the
isolated protein into peptides via in-solution enzymatic digest, and peptide cleanup with final preparations
for analysis via liquid chromatography-mass spectrometry. These protocols have been optimized and stan-
dardized for maximum consistency and maintenance of sample integrity.
Key words Proteomics, Protein extraction, Acetone precipitation, Enzymatic solution digest, Liquid
chromatography-mass spectrometry
1 Introduction
Analysis of proteins, the key effectors of most cellular processes, is

critical for understanding biological processes in healthy and dis-
eased states, and liquid chromatography-mass spectrometry (LC-
MS) analysis of peptides has proven to be a workhorse tool to this
end. Over the last 10 years, LC-MS instrumentation for biological
proteomics has undergone a remarkable evolution with significant
gains in analysis speed, precision, consistency, and sensitivity and
has in many cases given rise to commoditized instrumentation
accessible even to non-experts [1]. As a result, a major factor in the
success of proteomics-focused assays has now become the quality
of the input material and the consistency of, and care in perform-
ing, methods to generate proteomic samples for LC-MS assays.
While the relative simplicity of nucleic acid analytes enabled a
reasonably rapid and facile unification of methods in the genomics
field, the huge variation in biophysical properties of proteins, and
Lucio Comai et al. (eds.), Proteomics: Methods and Protocols, Methods in Molecular Biology, vol. 1550,
DOI 10.1007/978-1-4939-6747-6_1, © Springer Science+Business Media LLC 2017
1
2 Michelle Atallah et al.
their several orders of magnitude abundance variation in cells, has

proven a significant challenge to the proteomics field [1]. Variations
in methods for proteomics sample generation across labs, and even
within groups across operators, can significantly affect and even
confound the outputs of MS-based proteomics. This will undoubt-
edly prove to be a challenge going forward for the field in highly
specific experimental contexts and in answering unique biological
questions, especially given the highly varied biophysical properties
exhibited by proteins and protein complexes. However, methods
striving for complete solubilization of proteomes, for example for
global profiling under different states or conditions, are much
more amenable to unification across the field.
However, an operator striving to develop such a sample workup
regime for global proteomics analysis faces what is now a daunting
variety of choices at each step of the workflow. Often reagents for
upstream sample preparation are not compatible with LC-MS analysis
downstream, necessitating additional steps that can potentially cause
sample loss. For example, while robust detergents are generally
required to affect protein solubility, many common detergents used
in this role, such as SDS, are not compatible with downstream pro-
teolytic digestion and/or LC-MS. A wide variety of choices to deal
with this challenge now face the operator including traditional meth-
ods such as protein precipitation for detergent removal [2], commer-
cial innovations such as alternative LC-MS compatible surfactants
(e.g., Rapigest SF Surfactant, Waters Corporation), and development
of relatively new methods such as FASP (Filter-Aided Sample
Preparation), a method facilitating detergent removal and digestion
on a solid phase [3], just to name a few. In developing a composite
workflow, assessing which of these approaches best balances robust-
ness and consistency while avoiding significant sample loss is a difficult
task. Moreover, a similarly daunting set of choices must be addressed
at all other steps of a proteomics workflow, including a host of vari-
ables for proteolytic digestion and peptide cleanup, among others.
Here we provide a simple, robust method for whole-cell pro-
tein extraction from tissue culture cells, including methods for
downstream enzymatic digestion and peptide cleanup prior to
LC-MS. This method employs a classic ionic detergent at a rela-
tively high percentage to efficiently extract and solubilize proteins.
A facile acetone precipitation step is employed to remove this
detergent prior to straightforward and proven in-solution proteo-
lytic digestion. A simple solid-phase extraction method is per-
formed at the end of the workflow to clean and concentrate
peptides in advance of LC-MS. While no one method is perfect,
based on our experiences working in the field we feel the simplicity
and consistency of this method will provide new operators to the
field with an accessible, robust, and consistent starting point, and
in fact we use this routinely in our group for global proteomic
analyses of our valuable primary samples. Specific method steps are
Protein Extraction and Digestion 3
provided, and critical variables for each method are emphasized so

that experimental bias is minimized as much as possible.
2 Materials
1. Cell Lysis and Acetone Precipitation.

Given the sensitivity of mass spectrometry analysis, utmost
care should be taken to ensure clean working conditions free of all
possible contaminants. All reagents should be prepared using
LC-MS grade water (Honeywell brand is preferred). If possible,
sequestering a separate set of pipettes along with the use of nitrile
examination gloves is extremely useful in avoiding contamination
of samples with ubiquitous keratin from the environment.
2.1 Solutions 1. 4 % w/v sodium dodecyl sulfate (SDS).

and Reagents 2. 1 M Tris hydrochloride, pH 7.5.
3. 1 M dithiothreitol (DTT): Dissolve 87 mg in 5 mL 100 %
MeOH, store at −20 °C.
4. Thermo HALT protease and phosphatase inhibitor cocktail
(100×).
5. 1 M (10×) PMSF (phenyl-methyl-sulfonamide): Dissolve
87 mg in 5 mL LC-MS water, store at −20 °C.
6. LC-MS-grade water.
7. 99.5 % + acetone, chilled to −20 °C.
2.2 Equipment 1. Probe tip sonicator (alternatively, a Diagenode or Covaris

water bath sonicator, see Note 1).
2. (for probe tip sonicators) Clear plastic or acrylic box
(6 × 6 × 6 in. works well).
3. (for probe tip sonicators) Foam tube floater.
4. Pierce Micro BCA Protein Assay Kit.
2.3 Peptide 1. 1 M TCEP: Dissolve 287 mg in 1 mL MS water. Store at

Digestion −20 °C in suitable (e.g., 10 μL) aliquots.
2.3.1 Solutions 2. 1 M Tris pH 8, see Note 2.
and Reagents 3. 500 mM iodoacetamide (IAA): Dissolve a single-use pre-
weighed tube of 9.3 mg IAA (Pierce/Thermo Fisher) in
200 μL LC/MS water.
4. Protein resuspension solution: 8 M urea, 100 mM Tris pH 8.
5. Protein dilution solution: 100 mM Tris pH 8.
6. 100 mM CaCl2.
7. Sequence-grade trypsin enzyme.
2.4 C18 Cleanup 1. Honeywell Brand Reverse Phase A (RP-A): LC-MS water with
0.1 % formic acid.
2.4.1 Solutions
and Reagents 2. Honeywell Brand Reverse Phase B (RP-B): acetonitrile with
0.1 % formic acid.
3. Wetting solution (50 % Honeywell RP-B in RP-A).
4. Equilibration/Wash solution (2 % RP-B in Honeywell RP-A),
see Note 3.
5. Elution solution (90 % Honeywell RP-B in RP-A).
2.5 Equipment 1. Thermo Pierce C18 tips (Thermo product 87784).
3 Methods
3.1 Cell Lysis 1. Fill an ice bucket with wet ice and fill up to ice level with water.
and Acetone 2. Prepare 500 μL of cell lysis buffer for each 1e7 cells.
Precipitation
(a) Lysis buffer composition: 3 % SDS, 0.02 M DTT, 0.10 M
Tris–HCl pH 7.5, 1× Thermo Protease/Phosphatase
inhibitor, 1× PMSF. Make fresh and add protease/phos-
phatase inhibitors, especially PMSF that is active in aque-
ous solution for 30 min, just before use.
(b) Preheat lysis buffer at 95 °C. See Note 4.
3. Keep cell pellet tubes on dry ice. Add preheated lysis buffer to
frozen cell pellet; 500 μL for each ten million cells (see Notes
5 and 6). Pipette and/or vortex to mix. Critical point: the
frozen pellet must not be allowed to thaw until covered with
hot SDS lysis solution—thawing then should be rapid and
right into the concentrated surfactant.
4. Put cap lock on tube. Place tube in 95 °C heat block for 3 min.
Vortex every 15–20 s.
5. If necessary, split contents into tubes of ~500 μL each, see
Note 7.
6. Pour chilled ice water from ice bucket into clear plastic/acrylic
box (see Note 8), avoiding ice as much as possible. Wedge an
ice pack to the bottom of the acrylic box to keep the water
cool (see Note 9), and a foam tube holder to float the tubes at
the surface.
7. Vent the tubes by opening them briefly (with the opening cap
oriented away from you). Place them in the foam holder in the
acrylic box.
8. Sonicate samples while the tubes remain in ice water (see
Note 10). There are two critical points in this step, both
facilitated by the use of the clear acrylic box: first, the samples
must be kept cold to avoid degradation of proteins due to the
heat generated by the sonicator. We have found that keeping
the tubes in water is the best way to accomplish this. Secondly,

it is necessary to submerge the sonicator probe tip before
switching on sonicator, and switch it off before removing the
tip. Moving the probe in and out of solution while on will
cause foaming of the solution and may damage proteins.
Additionally, be consistent with tip depth (2/3 into solution
from top is ideal).
(a) Set sonicator amplitude to 40 %.
(b) Press the “Set” button to select “Continuous”.
(c) Sonicate for 3 cycles. 1 cycle = ON for 30 s, OFF on ice for
2 min (see Notes 11 and 12). Avoid contact between
probe tip and tube walls as much as possible to prevent
shedding of polymers into your sample.
9. Centrifuge tubes for 15 min at 15,000 rpm at 20 °C to clarify
the lysate of any particulates or insoluble material (see Note 13).
10. Transfer ~200–250 μL of supernatant to labeled Eppendorf
tubes (see Note 14).
(a) Pull out 10 % of the volume for a separate precipitation to
generate a pellet to bring up in 3 % SDS for BCA quantita-
tion (see Note 15).
(b) Note the exact volume of each sample that goes into each
new tube. This will be important later when calculating
the amount of protein per tube.
11. Using a glass Pasteur pipette (see Note 16), add cold (−20 °C)
100 % acetone to tubes. Fill to 1 mL line (see Note 17). Critical
point: ensure that least 4× volume acetone is added.
12. Invert tubes several times, vortex well, and place tubes at

−20 °C overnight (see Note 18).
13. The next day, spin tubes at 15,000 rpm for 15 min at 4 °C
(see Note 19).
14. Keep tubes on ice. Without disturbing the pellet or pipetting
up and down, remove and discard the acetone supernatant with
a glass Pasteur pipette (see Note 20). Add fresh ice-cold ace-
tone to the 1 mL mark, again without disturbing the pellet.
15. Centrifuge samples at 15,000 rpm for 10 min at 4 °C

(see Note 21).
16. Repeat steps 14 and 15 for a second wash (see Note 22).
17. Remove and discard acetone supernatant with a glass Pasteur
pipette.
18. Air-dry pellets with the lids open (covered with a Kimwipes)
for 1 h while on ice (see Note 23).
19. Store dry pellets at −80 °C until ready for further use.
20. Resuspend the pellet from the 10 % (v/v) aliquot precipitation
from step 10a in 100–300 μL 3 % SDS and perform BCA (see
Note 24) quantification according to the manufacturer’s
instructions. From the results determine how much protein is
in each acetone-precipitated pellet based on the fraction of the
total sample that went into each tube. Critical point: ensure
complete resuspension and solubilization of the pellet. In addi-
tion to visually inspecting the pellet to confirm complete resus-
pension, vortexing the sample and/or heating at 37 °C for up
to 2 min can aid in resolubilization.
3.2 Peptide 1. Bring up dried acetone pellet in 50 μL 8 M urea, 100 mM Tris
Digestion pH 8 for a final concentration of 80 μg in 25 μL (see Note 25).
(a) Use manual pipetting and if necessary heat and/or sonica-
tion to break up pellet. Avoid any heating over 37 °C to
avoid carbamylation by urea.
2. Add TCEP to final concentration of 5 mM and incubate for
RT for 30 min (see Note 26). Mix well by vortexing, and then
knock down by pulse spinning in microfuge.
3. Add IAA to final concentration of 10 mM and incubate for RT
for 30 min in the dark (covered in foil, see Note 27). Mix well
by vortexing, and then knock down by pulse spinning in
microfuge.
4. Bring up to 250 μL with 100 mM Tris pH 8, reducing the
urea concentration to <1 M (see Note 28). Mix well by vortex-
ing, and then knock down by pulse spinning in microfuge.
5. Add 100 mM CaCl2 to a final concentration of 1 mM. Mix
well by vortexing, and then knock down by pulse spinning in
microfuge.
6. Add 8 μg trypsin (1:10 mass:mass) for a final concentration of
30 ng/μL. Mix well by vortexing, and then knock down by
pulse spinning in microfuge.
7. Incubate overnight at 37 °C in static or shaking incubator.
3.3 C18 Cleanup The following is a modified protocol is based on the manufactur-
er’s recommendations for Pierce C18 tips (100 μL bed, Catalog
No. 87784).
Preloading tubes, or well plates, with each solution to be used
can increase throughput speed and minimizes downtime in which
the tip can dry. Each 100 μL aliquot of sample should be loaded
separately, and should be separately aliquoted into wells or tubes.
1. Wet C18 filter tip by aspirating 100 μL of wetting solution and
then discarding solvent (see Note 29).
2. Repeat wash step and discard solvent.
3. Equilibrate tip by aspirating 100 μL of equilibration/wash
solution and discarding the solvent.
4. Repeat equilibration step and discard solvent.

5. Aspirate 100 μL of sample and, using a clean Eppendorf tube,
slowly aspirate 100 μL up and down 10× (see Note 30).
(a) Repeat as many times as necessary until the entire sample
volume has been absorbed onto the column.
6. Wash tip by aspirating 100 μL of equilibration/wash solution
and discarding the solvent.
7. Repeat the wash step and discard the solvent.
8. Elute the sample by slowly aspirating 100 μL of elution solu-
tion into the tip, and slowly depositing into a new elution tube.
9. Repeat the elution step for a total final sample volume of 200 μL.
Deposit the second elution into the same tube as the first.
10. Dry the sample down to 10 % of its elution volume in a Speed-
Vac (see Note 31).
11. Resuspend peptides in 0.1 % formic acid in 2 % RP-B in

RP-A. Resuspend to a final volume of 1 μg/ul (or other suit-
able concentration and volume depending on your intended
LC-MS injection amount).
12. Spin sample for 3 min at maximum speed in a centrifuge before
transferring the supernatant to an autosampler vial for down-
stream mass spectrometry analysis (see Note 32).
4 Notes
1. The sonicator is used to shear the chromatin in the sample,

decreasing its viscosity to allow for efficient protein recovery.
Branson water baths are not powerful enough for this; only
brands such as Diagenode and Covaris that are typically used
for shearing fixed chromatin (e.g., for Chip-SEQ) will work,
although the sonication efficiencies of water bath type instru-
ments in our experience are much more sensitive to input
amounts (e.g. input overloading) versus probe-tip sonicators.
2. Tris pH 8 is optimal when using trypsin as the digestive
enzyme. If using LysC, substitute Tris pH 9 throughout the
protocol.
3. Including a small amount of acetonitrile in the solution aids in
solubilization of the peptides.
4. When heating solutions to near-boiling temperatures, it is
helpful to add cap locks to the Eppendorf tubes to ensure that
they stay closed. Furthermore, after boiling tubes should be
opened facing away from you to vent and avoid splashing up
of expanded air/liquid in the tube.
5. Scale volume of lysis buffer up for pellets with more cells.
6. Volume of lysis buffer stated is for probe tip sonicator; vol-

umes should be adjusted for different sonicators according to
manufacturer’s recommendations. Additionally, appropriate
tubes should be used if necessary —for example, older
Diagenode models require their own commercial tubes which
have a harder consistency for sufficient sonication efficiency.
7. 500 μL per tube is the ideal volume for sonication on a probe
tip sonicator. Smaller volumes can be used with suitably pow-
erful water bath sonicators.
8. For probe tip sonicators, it is essential to visually monitor the
depth of the probe tip during sonication while keeping the
sample cold. We recommend using a clear enclosure so that
the placement depth of the probe tip can be visually ascer-
tained for as much consistency as possible.
9. It is important to keep the samples as cold as possible to coun-
teract the heating induced by sonication.
10. Invert tubes before and after sonication to check for viscos-
ity—viscosity should be much reduced after sonication.
11. Different brands of sonicators have different strengths, so

optimal conditions (where minimal sonication is applied to
sufficiently disrupt chromatin) should be determined empiri-
cally for each sonicator. Slowly increasing strength and/or
time of sonication and monitoring viscosity of the resulting
lysate with test lysates (e.g., by inverting tubes and observing
fluidity of the solution during a pilot timecourse) is useful for
quickly determining an optimal sonication regime. Over-
sonication should be avoided to minimize heating and/or
physical compromise of proteins in samples.
12. In the case of probe tip sonicators it is absolutely critical is to
submerge the sonicator probe tip before switching on the son-
icator, and switch it off before removing the tip.
13. Orient tubes inside centrifuge so hinge faces outward to con-
sistently position the pellet.
14. Avoid touching pellet with pipette tip—do not hesitate to
leave some supernatant behind if necessary. Split as necessary
as you wan to be able to add at least 4–5 volumes of ice-cold
acetone for a good precipitation.
15. A 10 % precipitation is done to remove DTT and so that the
lysate can be resuspended in a strong denaturant compatible
with the protein quantification BCA assay. Interferences can
significantly bias results of the assay.
16. It is important to always use glass when working with acetone—
the alcohol can potentially solubilize plastic on pipette tips,
causing polymer interference in downstream LC-MS analysis.
17. Again, make sure this is at least a fourfold volume addition of

neat ice-cold acetone to ensure the dielectric constant of the
solution is sufficiently reduced for efficient precipitation. Keep
acetone on ice or put it back in the freezer when not in use.
18. Ideally should see some flecks of particulate matter starting to
precipitate even before putting tubes into freezer, although
visible precipitating material may not be immediately observ-
able with inputs of less than 300 μg.
19. Again, orienting the tubes inside the centrifuge so that the
hinges face outward will ensure you know where the pellet is.
20. Leaving behind a small amount (~10 μL) of acetone is prefer-
able to disturbing the pellet.
21. The pellet will usually appear lightly opaque or white on the
backside of conical bottom.
22. Here it is fine to simply add cold acetone and re-spin. No vor-
texing or release of pellet from tube required, or
recommended.
23. Pellets may shrink during the drying process. Do not over-dry
pellets, as this will make them more difficult to resuspend. Do
not use the Speed-Vac at this step.
24. Run triplicates of each sample. May also run sample undiluted
in addition to 1:5–1:20 dilutions to ensure accurate range of
the BCA assay.
25. If sample amounts permit, reserve a small amount (equivalent
of ~5 μg) for protein gel analysis pre- vs post-digest.
26. Dilute TCEP 1:10 in MS-grade water (e.g., dilute 1:10 and
add 5 μL to each 100 μL digest).
27. Dilute single use tube (9.3 mg per tube) in 200 μL MS-grade
water for 200 mM stock (e.g., add 5 μL to each 100 μL
digest)—use of the pre-weighed tubes minimizes weighing
errors and saves on reagent, which is not reusable once
resuspended.
28. This dilution step is very important—trypsin is only active in
up to 1 M urea. Less trypsin (to 1:50 w/w) can be used to
reduce cost, but care should be taken to avoid incomplete
digestion - trypsin concentration in the final volume should
not drop below 20 ng/μL.
29. Once wet, do not introduce air into the tip resin—this could
dry the column and cause any bound material to become per-
manently attached. This is important throughout the proto-
col, but especially after the methanol wetting step as methanol
will evaporate quickly if delays occur. Likewise, proceed con-
tinuously through the protocol as any stopping or pauses
could cause the tip to dry out.
30. Aspirations up and down into the tip should be done slowly
(each single up/down cycle over 3–5 s, and multiple aspira-
tions up and down into the tip can be done to improve effi-
ciency. Recommended are 10 aspirations for each 100 μL
aliquot of sample loaded, and 5 for each 100 μL elution, and
3 aspirations at every other step. With experience and care,
implementation an 8-channel multichannel pipette and well
plates can improve throughput speed.
31. This requires careful monitoring of the drying, which can be
achieved with a strobe device (e.g. Labconco Centrizap) that
allows visualization of the sample without stopping the
Speed-Vac.
32. Avoid any pelleted material at bottom of tube. This final spin
step helps remove any particulates or debris that may clog the
mass spectrometer.
References
1. Mallick P, Kuster B (2010) Proteomics: a techniques for proteome analysis by mass spec-
pragmatic perspective. Nat Biotechnol 28:
trometry. J Chromatogr A 1418:158–166
695–709 3. Wiśniewski JR, Zougman A, Nagaraj N, Mann
2. Kachuk C, Stephen K, Doucette A (2015) Com M (2009) Universal sample preparation method
parison of sodium dodecyl sulfate depletion for proteome analysis. Nat Methods 6:359–362
Chapter 2
Improving Proteome Coverage and Sample Recovery

with Enhanced FASP (eFASP) for Quantitative
Proteomic Experiments
Jonathan Erde, Rachel R. Ogorzalek Loo, and Joseph A. Loo
Abstract
Enhanced Filter Aided Sample Preparation (eFASP) incorporates plastics passivation and digestion-
enhancing surfactants into the traditional FASP workflow to reduce sample loss and increase hydrophobic
protein representation in qualitative and quantitative proteomics experiments. Resulting protein digests
are free of contaminants and can be analyzed directly by LC-MS.
Key words Enhanced filter aided sample preparation, Quantitative proteomics, Detergent,
Ammonium deoxycholate
1 Introduction
The integrity of proteomic experiments hinges on consistent and

robust protein extraction, solubilization, and digestion. Protocols
using anionic detergents and/or chaotropes to extract and solubi-
lize cellular and matrix proteins efficiently provide samples that
must be purified prior to digestion and analysis. Methods such as
organic precipitation to remove contaminants, denaturants, and
other undesired species (e.g., salts, nucleic acids, lipids, and alkylat-
ing reagents), are subject to poor recoveries, resolubilization
problems, and protein-to-protein variation. Enhanced Filter Aided
Sample Preparation (eFASP) provides efficient protein extraction,
purification, and digestion for a variety of samples [1, 2].
Traditional Filter Aided Sample Preparation (FASP) circum-
vents many protein purification challenges by exchanging buffers
in spin filter units (ultrafiltration assemblies) that can remove
sodium dodecyl sulfate (SDS) and sample contaminants completely
[3, 4]. Proteins are reduced, alkylated, washed, and digested in the
filter unit, releasing product free of detergent, reductant, and alkyl-
ating agent. Nevertheless, when applied to very small sample sizes,
11
12 Jonathan Erde et al.
this method can suffer sample losses near 50 % [5]. Enhanced FASP
(eFASP) addresses this challenge by incorporating deoxycholic
acid and, optionally, TWEEN® 20 into the FASP workflow.
Deoxycholic acid (DCA) is a secondary bile acid that, amongst
many other uses, is employed as a mild detergent for membrane
proteins. It increases the efficiency with which trypsin digests cyto-
solic and membrane proteins and is easily removed by acidification
and phase transfer (PT) to peptide-immiscible ethyl acetate (EA) in
liquid–liquid extraction [1, 6–11]. PT decreases EA-soluble con-
taminants, including SDS, n-octylglucoside, NP-40, and Triton
X-100 [12].
The eFASP protocol optionally uses the surfactant TWEEN®-
20 to passivate surfaces of Microcon® filter units and collection
tubes. TWEEN®-20, as well as SDS, are recognized choices for
minimizing protein binding to surfaces and have been recom-
mended by Amicon Centricon for use in filter units [13]. Passivation
of the filter units and collection tubes used in eFASP can reduce
peptide and protein loss due to nonspecific surface binding, but
care is needed to prevent TWEEN®-related ions from contaminat-
ing mass spectra.
Presented here is the eFASP approach, which utilizes 0.2 %
DCA and (optionally) TWEEN®-20 to quantitatively increase
recovery and proteomic coverage of hydrophilic and hydrophobic
proteins. An express eFASP method variant is also included, which
uses a one-step reduction/alkylation employing tris(2-
carboxyethyl)phosphine (TCEP) and 4-vinylpyridine (4-VP) prior
to deposition on the Microcon® filter, increasing alkylation speci-
ficity and speeding processing [1, 14–17].
2 Materials
Prepare all solutions fresh using ultrapure water and MS-grade

reagents. Follow all waste disposal regulations and chemical safety
guidelines.
2.1 Solutions 1. Passivation Solution: 5 % (v/v) TWEEN® 20.

and Reagents 2. Lysis Buffer: 4 % SDS, 0.2 % deoxycholic acid (DCA, Sigma,
for eFASP D2510), 50 mM TCEP, 100 mM ammonium bicarbonate
(See Subheading 3.1) (ABC), pH 8 (see Notes 1 and 2).
3. Exchange Buffer: 8 M urea, 0.2 % DCA, 100 mM ABC, pH 8
(see Note 1).
4. Alkylation Buffer: 50 mM iodoacetamide, 8 M urea, 0.2 %
DCA, 100 mM ABC, pH 8 (see Notes 1 and 2).
5. Digestion Buffer: 0.2 % DCA, 50 mM ABC, pH 8 (see Note 1).
6. Trypsin Buffer: 0.5 μg/μl trypsin, 50 mM ABC, pH 8.
Enhanced FASP for Proteomic Experiments 13
7. Peptide Recovery Buffer: 50 mM ABC, pH 8.

8. Ethyl acetate.
9. 50 % methanol.
10. Trifluoroacetic acid.
11. MS-grade H2O.
2.2 Solutions 1. Passivation Solution: 5 % (v/v) TWEEN® 20.

and Reagents 2. Lysis Buffer: 4 % SDS, 0.2 % DCA (Sigma, D2510), 50 mM
for Express eFASP TCEP, 100 mM ammonium bicarbonate (ABC), pH 8
(See Subheading 3.2) (see Notes 1 and 2).
3. Exchange Buffer: 8 M urea, 0.2 % DCA, 100 mM ABC, pH 8
(see Note 1).
4. Alkylation Stock: 500 mM 4-vinylpyridine (4-VP) in ethanol
(see Note 3).
5. Quench Buffer: 1 M dithiothreitol (DTT), 100 mM ABC, pH 8.
6. Digestion Buffer: 0.2 % DCA, 50 mM ABC, pH 8 (see Note 1).
7. Trypsin Buffer: 0.5 μg/μl trypsin, 50 mM ABC, pH 8.
8. Peptide Recovery Buffer: 50 mM ABC, pH 8.
9. Ethyl acetate.
10. 50 % methanol.
12. MS-grade H2O.
2.3 Equipment 1. Microcon® ultrafiltration units (YM-30 30 kDa cutoff limit;

Millipore, Billerica, MA).
2. Bench-top centrifuge.
3. Thermo-mixer (initially set to 90 °C).
4. SpeedVac®.
5. Squeeze bottle containing MS-grade H2O.
6. Sonicator/homogenizer for disrupting cells.
7. Eppendorf LoBind® tube, 2 ml.
8. Ultrasonic bath.
3 Methods
Two eFASP protocols are described. The first is the standard

procedure that utilizes in-filter alkylation with iodoacetamide

(see Subheading 3.1). The second is an express procedure that
utilizes in-solution alkylation with 4-vinylpyridine to increase

speed, eliminating some buffer exchange steps (see Subheading 3.2).
Carry out all procedures at room temperature or as specified

and follow instrument and chemical safety guidelines. The passiv-
ation steps may be omitted to save time or if TWEEN®-related
background ions cannot be minimized in mass spectra.
3.1 eFASP: Standard 1. On a shaker, incubate filter units and collection tubes over-
night in Passivation Solution. Small batches of items may be
3.1.1 Surface
incubated in 50 ml Falcon centrifuge tubes.
Passivation (Optional)
2. With clean tweezers, remove each item and rinse its outer and
inner surfaces with MS-grade H2O dispensed from a squeeze
bottle.
3. Transfer items to a clean beaker containing a large volume of
MS-grade H2O; e.g., 250 ml or more. Incubate items for
30 min at room temperature, shaking at low speed.
4. Repeat step 3 two additional times with fresh MS-grade H2O.
5. Reserve the passivated collection tubes for peptide recovery
from ultrafiltration devices.
3.1.2 Sample Lysis 1. Wash and pellet cells according to established guidelines for
cell type.
2. Add sufficient Lysis Buffer to the pelleted cells such that a
25 μl aliquot of lysate will provide the quantity desired for
processing by eFASP (or a maximum protein concentration of
10 μg/μl). Ensure that the selected volume and tube size can
accommodate the sonicator probe.
3. Place the lysate into a 90 °C thermo-mixer and incubate for
10 min, shaking at 600 rpm. Remove lysate and decrease
thermo-mixer temperature to 37 °C for later use.
4. Sonicate lysate (employing sonicator/homogenization probe)
three times for 10 s each.
5. Centrifuge the lysate at 14,000 × g for 10 min.
6. Repeat sonication and centrifugation once (steps 4 and 5).
7. Sonicate the lysate (including any pelleted material), for 10 s.
Cool to 37 °C.
3.1.3 Sample Processing 1. Transfer 25 μl of lysate to a sample tube containing 200 μl of
Exchange Buffer. Vortex briefly to mix.
2. Place a (passivated) filter unit atop a non-passivated collection
tube.
3. Dispense the 225 μl lysate/Exchange Buffer sample to the fil-
ter unit and centrifuge at 14,000 × g for 10 min. Discard
filtrate.
4. Add 200 μl Exchange Buffer to the filter unit and centrifuge at
14,000 × g for 10 min. Discard filtrate.
5. Repeat step 4 two more times.
6. Dispense 100 μl Alkylation Buffer to the filter unit and trans-

fer it to a 37 °C thermo-mixer for 1 h, shaking at 300 rpm.
7. Centrifuge the filter unit at 14,000 × g for 10 min. Discard
filtrate.
8. Add 200 μl Exchange Buffer to the filter unit and centrifuge at
14,000 × g for 10 min. Discard filtrate.
9. Add 200 μl eFASP Digestion Buffer to the filter unit and cen-
trifuge at 14,000 × g for 10 min. Discard filtrate.
11. Transfer the filter unit to a passivated collection tube.
12. Add 100 μl eFASP Digestion Buffer to the filter unit.
13. Calculate the volume of Trypsin Buffer to dispense in order to
achieve the desired enzyme-to-substrate ratio; e.g., 1:50 w:w.
14. Deposit the calculated volume of Trypsin Buffer to the filter
unit, and place in a 37 °C thermo-mixer for 12 h, shaking at
low speed. Secure the filter unit cap to minimize evaporation.
15. Remove the filter unit/collection tube assembly from the

thermo-mixer and centrifuge at 14,000 × g for 10 min. Retain
peptide-containing filtrate.
16. Deposit 50 μl of Peptide Recovery Buffer onto the filter unit
and centrifuge at 14,000 × g for 10 min.
17. Repeat step 16 step once. Retain peptide-containing filtrate.
3.1.4 Phase Transfer 1. To the collection tube with peptide-containing filtrate, add
200 μl of ethyl acetate and transfer to a 2 ml Eppendorf
LoBind® tube.
2. Add 2.5 μl TFA and quickly vortex. A white, thread-like pre-
cipitate may be visible if a large quantity of peptide is present.
3. Add ethyl acetate to nearly fill the tube, leaving only enough
space to agitate without losing liquid.
4. Agitate the mixture for 10 s in an ultrasonic bath and centri-
fuge at 16,000 × g for 10 min.
5. Carefully pipet most of the upper (organic) layer into a tube
for discard. Do not disturb the organic/aqueous boundary
layer.
6. Repeat steps 3–5 two times.
7. Place the uncapped sample tube in a 60 °C thermo-mixer, in a
fume hood, for 5 min to remove residual ethyl acetate.
8. Remove residual organic solvent and volatile salts by vacuum
drying in a SpeedVac®.
9.
Resuspend the dried sample in 50
% methanol and
vacuum-dry.
10. Repeat step 9 two times.
3.2 Express eFASP 1. Wash and pellet cells according to established guidelines for
cell type.
3.2.1 Surface
Passivation 2. Add sufficient Lysis Buffer to the pelleted cells such that a
(Optional) 25 μl aliquot of lysate will provide the quantity desired for
processing by eFASP (or a maximum protein concentration of
3.2.2 Sample Lysis 10 μg/μl). Ensure that the selected volume and tube size can
accommodate the sonicator probe.
3. Place the lysate into a 90 °C thermo-mixer and incubate for
10 min, shaking at 600 rpm. Remove lysate and decrease
thermo-mixer temperature to 37 °C for later use.
4. Sonicate lysate (employing sonicator/homogenization probe)
three times for 10 s each.
5. Centrifuge the lysate at 14,000 × g for 10 min.
6. Repeat sonication and centrifugation once (steps 4 and 5).
7. Sonicate the lysate (including any pelleted material), for 10 s.
Cool to 37 °C.
8. Add Alkylation Stock to the lysate to a final concentration of
25 mM 4-VP, and place the sample tube into a 37 °C thermo-
mixer for 1 h at 300 rpm.
9. Add Quench Buffer to the lysate to a final concentration of
40 mM DTT.
3.2.3 Sample Processing 1. Transfer 25 μl of lysate to a sample tube containing 200 μl of
Exchange Buffer. Vortex briefly to mix.
2. Place a (passivated) filter unit atop a non-passivated collection
tube.
3. Dispense the 225 μl lysate/Exchange Buffer sample to the fil-
ter unit and centrifuge at 14,000 × g for 10 min. Discard
filtrate.
4. Add 200 μl eFASP Digestion Buffer to the filter unit and cen-
trifuge at 14,000 × g for 10 min. Discard filtrate.
6. Detach the filter unit from the non-passivated collection tube,
and place it on top of a passivated tube.
7. Add 100 μl eFASP Digestion Buffer to the filter unit.
8. Calculate the volume of Trypsin Buffer to dispense in order to
achieve the desired enzyme-to-substrate ratio; e.g., 1:50 w:w.
9. Deposit the calculated volume of Trypsin Buffer to the filter
unit, and move the filter/collection tube assembly to a 37 °C
thermo-mixer for 12 h of shaking at low speed. Cap the filter
unit to reduce evaporation.
10. Remove the filter/collection tube assembly from the thermo-
mixer and centrifuge at 14,000 × g for 10 min. Retain the
peptide-containing filtrate.
11. Dispense 50 μl of Peptide Recovery Buffer to the filter unit

and centrifuge at 14,000 × g for 10 min.
12. Repeat step 11 step once.
3.2.4 Phase Transfer 1. To the collection tube with peptide-containing filtrate, add
200 μl of ethyl acetate and transfer to a 2 ml Eppendorf
LoBind® tube.
2. Add 2.5 μl TFA and quickly vortex. A white, thread-like pre-
cipitate may be visible if a large quantity of peptide is present.
3. Add ethyl acetate to nearly fill the tube, leaving only enough
space to agitate without losing liquid.
4. Agitate the mixture for 10 s in an ultrasonic bath and centri-
fuge at 16,000 × g for 10 min.
5. Carefully pipet most of the upper (organic) layer into a tube
for discard. Do not disturb the organic/aqueous boundary
layer.
6. Repeat steps 3–5 two times.
7. Place the uncapped sample tube in a 60 °C thermo-mixer, in a
fume hood, for 5 min to remove residual ethyl acetate.
8. Remove residual organic solvent and volatile salts by vacuum
drying in a SpeedVac®.
9. Resuspend the dried sample in 50 % methanol and
vacuum-dry.
10. Repeat step 9 two times.
4 Notes
1. This protocol employs deoxycholic acid rather than sodium

deoxycholate, in order to minimize analyte exposure to
sodium, which can degrade mass spectrometry analyses. Even
with online liquid chromatography to remove much of the
Na+, acidic peptides may appear sodium-adducted. A trade-off
in substituting deoxycholic acid for sodium deoxycholate is
the former’s low solubility; 0.2 % DCA is about the maximum
solubility achievable in ABC buffer. Dissolving deoxycholic
acid in a small volume of ethanol prior to mixing with buffer
can facilitate dissolution. Alternatively, ABC buffer may be
added to the appropriate quantity of solid DCA immediately
before use and vortexed extensively; even if slightly cloudy, the
freshly prepared DCA solution can be employed in eFASP. DCA
solutions should not be refrigerated; irreversible precipitation
may occur. We have found that 0.1 % DCA performs almost as
well as 0.2 % in eFASP; thus, solubility problems may be eased
by using the lower concentration. Finally, sodium deoxycholate
could substitute for deoxycholic acid, with STAGE tip cleanup

prior to injection onto the LC column.
2. Alternatively, Lysis Buffer can employ 5 mM TCEP and/or
Alkylation Buffer can be formulated with 5 mM iodoacetamide.
3. We assume that neat 4-vinylpyridine is 8.8 M in concentration.
References
1. Erde J, Loo RRO, Loo JA (2014) Enhanced 10. Lin Y, Zhou J, Bi D, Chen P, Wang X, Liang
FASP (eFASP) to increase proteome coverage S (2008) Sodium-deoxycholate-assisted tryp-
and sample recovery for quantitative proteomic tic digestion and identification of proteolyti-
experiments. J Proteome Res 13(4):1885–95 cally resistant proteins. Anal Biochem
2. Erde J (2012) High throughput analysis of 377(2):259–66
proteome perturbations induced by radiation, 11. Lin Y, Liu Y, Li J, Zhao Y, He Q, Han W et al
radiomitigators and chemotherapeutics. (2010) Evaluation and optimization of
University of California, Los Angeles removal of an acid-insoluble surfactant for
3. Manza LL, Stamer SL, Ham A-JL, Codreanu shotgun analysis of membrane proteome.
SG, Liebler DC (2005) Sample preparation Electrophoresis 31(16):2705–13
and digestion for proteomic analyses using 12. Yeung Y-G, Nieves E, Angeletti RH, Stanley
spin filters. Proteomics 5(7):1742–5 ER (2008) Removal of detergents from pro-
4. Wisniewski JR, Mann M (2009) Spin filter– tein digests for mass spectrometry analysis.
based sample preparation for shotgun pro- Anal Biochem 382(2):135–7
teomics. Nat Methods 6(11):785–6 13. Passivation of Amicon Microcon Concentrators
5. Wisniewski JR, Zielinska DF, Mann M (2011) for Improved Recovery (1999). Bedford, MA:
Comparison of ultrafiltration units for pro- Millipore Corporation, Technical Note
teomic and N-glycoproteomic analysis by the PC1001EN00
filter-aided sample preparation method. Anal 14. Sebastiano R, Citterio A, Lapadula M, Righetti
Biochem 410(2):307–9 PG (2003) A new deuterated alkylating agent
6. Masuda T, Tomita M, Ishihama Y (2008 Feb) for quantitative proteomics. Rapid Commun
Phase transfer surfactant-aided trypsin diges- Mass Spectrom 17(21):2380–6
tion for membrane proteome analysis. 15. Bai F, Liu S, Witzmann FA (2005) A “de-
J Proteome Res 7(2):731–40 streaking” method for two-dimensional elec-
7. Masuda T, Sugiyama N, Tomita M, Ishihama trophoresis using the reducing agent
Y (2011) Microscale phosphoproteome analy- tris(2-carboxyethyl)-phosphine hydrochloride
sis of 10,000 cells from human cancer cell and alkylating agent vinylpyridine. Proteomics
lines. Anal Chem 83(20):7698–703 5(8):2043–7
8. Masuda T, Saito N, Tomita M, Ishihama Y 16. Liu S, Bai F, Witzmann F (2006) Destreaking
(2009) Unbiased quantitation of escherichia coli strategies for two-dimensional electrophoresis.
membrane proteome using phase transfer sur- In: Separation methods in proteomics. Eds.:
factants. Mol Cell Proteomics 8(12):2770–7 Smejkal GB, Lazarev A, CRC Press/Taylor &
9. Zhou J, Zhou T, Cao R, Liu Z, Shen J, Chen Francis, Boca Raton, pp. 207–17
P et al (2006) Evaluation of the application of 17. Righetti PG (2006 Sep) Real and imaginary arte-
sodium deoxycholate to proteomic analysis of facts in proteome analysis via two-dimensional
rat hippocampal plasma membrane. maps. J Chromatogr B 841(1–2):14–22
J Proteome Res 5(10):2547–53
Chapter 3
Proteome Characterization of a Chromatin Locus Using

the Proteomics of Isolated Chromatin Segments Approach
Sophie L. Kan, Nehmé Saksouk, and Jérome Déjardin
Abstract
The biological functions of given genomic regions are ruled by the local chromatin composition. The
Proteomics of Isolated Chromatin segments approach (PICh) is a powerful and unbiased method to ana-
lyze the composition of chosen chromatin segments, provided they are abundant (repeated) or that the
organism studied has a small genome. PICh can be used to identify novel and unexpected regulatory fac-
tors, or when combined with quantitative mass spectrometric approaches, to characterize the function of
a defined factor at the chosen locus, by quantifying composition changes at the locus upon removal/addi-
tion of that factor.
Key words Proteomics, Chromatin, Protein Purification, Repeated Sequences, Heterochromatin,

Telomeres
1 Introduction
Each cell of an organism harbors the same genetic information. Yet

cells can express distinct genetic programs according to their dif-
ferentiation state or to environmental factors. These features con-
trol the chromatin composition, which in turn regulates the
biological features of a locus of interest. A major tool to study the
composition of chromatin is the chromatin immunoprecipitation
approach (ChIP), which allows to interrogate using various analyti-
cal approaches, whether a DNA sequence is associated with a factor
of interest [1]. This method is widely used because it is sensitive and
it has become relatively easy to implement. However, ChIP is lim-
ited by the knowledge of the factor to test and the availability of
suitable antibodies against that factor. We have developed a method
which we named PICh (Proteomics of Isolated Chromatin seg-
ments) which allows to purify a given genomic region together with
the factors associated in vivo, without the need for any prior knowl-
edge about the identity of bound factors [2]. Briefly PICh is based
on the hybridization capture of a specific chromatin fragment.
19
20 Sophie L. Kan et al.
The PICh method is powerful because it provides a comprehensive

knowledge about the steady state protein composition of the target
locus. However it must be kept in mind that PICh has one impor-
tant technical limitation: the signal to noise ratio. We estimate the
PICh affinity purification enrichment to reach ~10,000 fold, and
thus PICh is not suitable to purify single copy targets from human
cells. A single 1 kb locus of interest represents only ~0.00003 % of
the human genome. To obtain a 50 % pure mixture for a relatively
straightforward downstream mass spectrometric analysis of such
low abundance species, it would require an enrichment factor of
1,500,000 fold, which to our knowledge is challenging with a sin-
gle affinity step. The nonspecific capture of minute amounts of non-
relevant chromatin fragments is also an issue that needs to be
carefully eliminated from the PICh material. For instance, the non-
specific capture of 0.003 % of the starting material would be accept-
able in any ChIP experiment because enrichment of a specific target
is compared to any other similarly abundant “negative control.” Yet
in PICh the entire nonspecific material contributes proteins in the
eluate, 0.003 % of input represents a 100 times more material than
proteins specifically contributed by a unique target. Thus PICh is
applicable only to abundant targets from mammalian genomes and
importantly, entails the use of dedicated reagents. The protocol
described here is optimized for the purification of proteins associ-
ated to abundant sequences such as mammalian telomeres or peri-
centromeric repeats (0.01–3 % of the genomes) (Fig. 1). For regions
with higher complexity such as ribosomal RNA gene promoters,
refer to the end-targeting PICh approach (ePICh) [3]. In PICh,
chromatin is reversibly cross-linked with formaldehyde and sheared
by sonication. The locus of interest is targeted with desthiobiotinyl-
ated oligonucleotide probes containing 50 % of locked nucleic acids
(LNA). LNA–DNA hybrids have high melting temperatures ensur-
ing strong and stable hybridization to the end sequence of the frag-
ment of interest [4]. In our hands, DNA or RNA capture probes do
not hybridize stably enough to allow any specific enrichment in
PICh, presumably because short D-loops or R-loops are notori-
ously highly unstable in aqueous solution [5, 6]. The probe–chro-
matin hybrids are captured by streptavidin coupled magnetic beads.
Desthiobiotin is a biotin analog that has weaker affinity for strepta-
vidin than biotin (Kd ~ 10–12 M) which allows a competitive gentle
elution from streptavidin beads with biotin (Kd ~ 10–14 M) [7].
This gentle condition allows maintaining nonspecific proteins
bound to the beads. Finally the chromatin mixture is precipitated
by trichloroacetic acid, and protein–protein and protein–DNA
cross-links are reversed. The preparation can be used for western-
blot assays or analyzed by mass spectrometry to identify new regula-
tory factors. The quantitative version of the PICh technique
(qPICh) is less suited to identify low abundance factors (the sensi-
tivity is 2-fold lower), but is used to quantify differences in protein
composition at the studied locus in two distinct samples [8]
Proteome Characterization of a Chromatin Locus… 21
Formaldehyde Chromatin Chromatin

Grow cells
cross-linking isolation shearing
Preclearing Telomeric
chromatin
Denaturation
Hybridization
LNA sequence
Hybrids capture
Spacer
Desthibiotin
Streptavidin
dynabead Protein
precipitation
Washes Elution Crosslink reversal
Western-blot
Mass-spectrometry
Fig. 1 Outline of the PICh protocol
provided the abundance of that locus is comparable in the two sam-

ples. qPICh combines SILAC with PICh and is appropriate to
address the function of a given factor at a region of interest. PICh
and qPICh are described in the following protocol.
2 Materials
Prepare all solutions using ultrapure water and analytical grade

reagents. Store all buffers at 2–8 °C. Add phenylmethanesulfonyl-
fluoride (PMSF) just prior to use.
2.1 Special 1. Sonicator with high power sonication probe (Qsonica).

Equipment 2. DynaMag™-2 magnet or equivalent for 1.5 mL microcentri-
fuge tubes (Life Technologies).
3. DynaMag™-15 magnet or equivalent for 15 mL conical tubes
(Life Technologies).
4. 10 mL chromatography columns (Pierce).
2.2 Buffers 1. 100 mg/mL RNaseA.

and Reagents 2. 37 % formaldehyde.
3. High capacity streptavidin agarose resin (Thermo Scientific).
4. Streptavidin C1 MyOne dynabeads (Life Technologies).
5. 50 mM D-Biotin (Invitrogen).
6. 1× phosphate-buffered saline: 8 mM Na2HPO4, 2 mM
KH2PO4, 137 mM NaC, 2.7 mM KCl, 1 mM PMSF.
7. Sucrose buffer: 0.3 M Sucrose, 10 mM HEPES–NaOH,
pH 7.9, 1 % Triton X-100, 3 mM CaCl2, 2 mM MgOAc,
1 mM PMSF.
8. Lysis buffer: 10 mM HEPES-NaOH, pH 7.9, 100 mM NaCl,
2 mM EDTA, pH 8, 1 mM EGTA, pH 8, 0.2 % SDS, 0,1 %
Sarkosyl, protease inhibitors, 1 mM PMSF.
9. Low salt buffer: 10 mM HEPES–NaOH, pH 7.9; 30 mM
NaCl, 2 mM EDTA, pH 8, 1 mM EGTA, pH8, 0.2 % SDS,
0.1 % Sarkosyl, protease inhibitors, 1 mM PMSF.
10. Elution buffer: 12.5 mM D-Biotin, 7.5 mM HEPES-NaOH;
pH 7.9, 75 mM NaCl, 1.5 mM EDTA, pH 8; 0.75 mM
EGTA, pH 8, 0.15 % SDS, 0.075 % Sarkosyl (3/4 lysis buf-
fer + ¼ D-Biotin).
11. Cross-linking reversal solution: 250 mM Tris–HCl, pH 8.8,
2 % SDS, 0.9 M 2-mercaptoethanol.
12. Sephacryl S-400 high resolution gel filtration resin (GE

Healthcare).
13. SilverQuest Staining Kit (Invitrogen).
14. LNA probes: LNA probe synthesis is performed in-house

using an H-6 K&A synthesizer, on a 1 μ-mole scale, following
manufacturer’s recommendations and standard phosphorami-
dite synthesis guidelines. Alternatively LNA probes can be
ordered from Exiqon or any company offering to make LNA
containing probes. Capital letters represent LNA residues
while small letters are DNA residues:
Telomere: Desthiobiotin-TEG-4XC18 spacers-5′TtAgGg
TtAgGgTtAgGgTtAgGgt-3′
Major satellites: Equimolar mixture of 4 distinct capture
probes:
Desthiobiotin-TEG-4XC18 spacers– 5′ CaCtTTaGgaCGTgAa
ATaTGg-3′
Desthiobiotin-TEG-4XC18 spacers– 5′ TgTaGgAcAtGgAaTa
TgGcAaGaAaAc-3′ Desthiobiotin-TEG-4XC18 spacers-
5′CgAgGaAaAcTgAaAaAgGttgAaAaTt-3′ Desthiobiotin-
TEG-4XC18spacers-5′AaTCaCGgAaAaGgAGaaATac-3′
Scramble: Desthiobiotin-TEG-4XC18 spacers –

5′GaTgTgT gGaTgTggAtGtGgAtgTgg-3′
15. Media for stable isotope labeling by amino acids in cell culture:
Arginine and lysine free cell growth medium.
Dialyzed FBS.
Light arginine/lysine.
Heavy arginine/lysine (Cambridge Isotope Labeling).
3 Methods
3.1 Tissue Culture 1. Plate cells on enough 150 mm plates to harvest 109 cells per
condition (see Note 1).
2. Incorporation of isotopically stable amino acids into cellular
proteins for qPICh. Grow the two cell lines (e.g., WT and
mutant for a gene of interest) in culture media with isotopi-
cally distinct amino acids (see Note 2).
3. Count cells of one or two representative plate for each condi-
tion. For qPICh this step is extremely important.
4. FACS analyze the cell cycle profile of the cell populations to
account for potential differences cell cycle distributions among
the distinct cell populations.
3.2 Formaldehyde 1. 1. Add formaldehyde directly to the growth medium to a 3 %

Cross-Linking final concentration for 30 min at room temperature (see Note 3).
Gently shake the plates on an orbital shaker.
3.2.1 Adherent Cells
2. Remove the growth medium and dispose of it appropriately.
3. Wash twice with PBS-PMSF to remove the formaldehyde of
the cross-linking solution (see Note 4).
4. Scrap cells in 3 mL PBS containing 0,05 % Tween-20 per
150 mm plate with a cell lifter and collect them in a 50 mL
conical tube (10–15 plates should fit into one 50 mL conical
tube) (see Note 5).
5. Centrifuge the cross-linked cells at 3200 × g for 10 min at
4 °C.
6. Pool the pellets in one 50 mL conical tube (see Note 6).
7. Wash the pellet three times in PBS-PMSF. Centrifuge each
time at 3200 × g for10 min at 4 °C.
3.2.2 Suspension Cells 1. Cross-link the protein–DNA complexes by incubating the

cells with formaldehyde diluted to a 3 % final concentration for
30 min at room temperature on a nutator (volume of the
cross-linking solution must not be below 50 times the volume
of pelleted cells: for a pellet of 4 mL cells, use at least 200 mL
of cross-linking solution).
2. Centrifuge the cross-linked cells at 3200 × g for 10 min at 4 °C.

3. Remove the growth medium and dispose of it appropriately.
4. Wash the pellet twice in PBS-PMSF (3200 × g, 10 min) (see
Notes 5 and 6).
5. The pellet can be stored at −80 °C up to 1 month (see Note 7).
6. For qPICh mix pelleted cells in a one-to-one ratio (see Note 8).
3.3 Chromatin 1. Resuspend the pellet in 10 mL of sucrose solution by vortex-

Preparation ing and bring volume to 50 mL (see Note 9).
2. Centrifuge at 3200 × g for 10 min at 4 °C.
3. Remove and discard the supernatant, resuspend the pellet in
10 mL of sucrose solution and bring the volume to 20 mL
with sucrose solution.
4. Transfer the solution to a 40 mL dounce homogenizer.
5. Dounce on ice approximately 20 times with a tight pestle.
6. Transfer the lysed cells into a new 50 mL conical tube, wash
the homogenizer twice with 15 mL of sucrose solution and
pool these washes with the sample to minimize material loss.
7. Centrifuge at 3200 × g for 10 min at 4 °C.
8. Discard the supernatant.
9. Resuspend the pellet in the same pelleted volume of PBS.
10. Add Triton-X100 (from a 20 % stock solution) to a final con-
centration of 0.5 % Triton X-100.
11. Add RNaseA to a final concentration of 10 μg/μL (10 μL of
concentrated RNaseA (100 mg/mL) per 1 mL of chromatin).
12. Incubate overnight on a nutator at 4 °C (see Note 10).
13. Transfer the mixture into a 50 mL conical tube and bring the
volume to 50 mL with PBS—PMSF.
14. Centrifuge the sample at 3200 × g for 10 min at 4 °C and dis-
card the supernatant.
15. Wash three times in PBS-PMSF to dilute RNase A.
16. Resuspend the pellet in 50 mL of freshly prepared lysis buffer.
17. Centrifuge the sample at 3200 × g for 10 min and discard the
supernatant.
18. Resuspend the pellet in 2.5 volume of pelleted volume of lysis
buffer by pipetting with 1 mL micropipette and split into
3 mL aliquots in 15 mL conical tubes (see Note 11).
19. Sonicate the sample to shear the chromatin to an average
length of about 1 kb. Sonicate on ice and ensure that the
samples do not foam during the process. Sonication parame-
ters for a Misonix S-4000 sonicator with a high power probe
(see Note 12): power setting amplitude 70 % (36–45 W), 15 s
constant pulse, 45 s pause, 7 min total process time.
3.4 Chromatin 1. Incubate the sonicated chromatin at 58 °C for 5 min

Pre-clearing (see Note 13).
2. Aliquot the chromatin into 1.5 mL microcentrifuge tubes and
pellet insoluble material by centrifugation at 16,000 × g for
10 min at room temperature.
3. Transfer the supernatants into a fresh 15 mL conical tube.
4. Save 20 μL to check the chromatin shearing efficiency (see
Subheading 3.11).
5. In the meantime equilibrate the high capacity streptavidin
agarose resin with lysis buffer for 5 min at room temperature.
Use 200 μL of slurry Ultralink streptavidin resin for every
700 × 106 cells. Centrifuge at 500 × g for 2 min at room tem-
perature. Repeat this step once.
6. Add streptavidin beads to the chromatin sample for the pre-
clearing step to remove any endogenously biotinylated protein
(e.g., carboxylases) that might interfere with the subsequent
capture.
7. Incubate overnight at 4 °C on a nutator (see Note 14).
8. Prepare the desalting columns by placing a 10 mL PIERCE
empty chromatography column into a 50 mL conical tube.
Add the same volume of Sephacryl S-400 HR gel filtration
media as that of the chromatin sample (see Note 15). Do not
add the chromatin at this step.
9. Centrifuge at 750 × g for 3 min to dry the resin in the
column.
10. Place the column in a fresh 50 mL conical tube and add chro-
matin (together with the streptavidin agarose beads used for
pre-clearing) to the dried column.
11. Pre-clear the chromatin by centrifugation at 750 × g for 5 min.
Recover the flow through, it is the pre-cleared chromatin
fraction.
12. Measure the optical density (OD) of the pre-cleared chromatin.
Use the low salt buffer as a blank solution. A high quality PICh
chromatin sample has the following characteristics: OD260:
2–2.5 mg/mL, OD260/OD280: 1.3–1.45 (see Note 16).
13. Bring the pre-cleared chromatin to a final concentration of
0.02 % of SDS (1/100 th of the final volume of 20 % SDS).
14. Spare 10 μL for the input.
3.5 Hybridization 1. Split the chromatin in two equal volumes, one for the experi-
and Chromatin ment and the other for the scrambled control (see Note 17).
Capture 2. Add the LNA probe to a final concentration of 0.25 μM
(see Note 18).
3. Aliquot 150 μL of the chromatin-LNA mixture into PCR
tubes.
4. Hybridize in a standard thermocycler using the following pro-

gram (see Note 19):
(a) 25 °C for 3 min
(b) 71 °C for 9 min.
(c) 38 °C for 1 h.
(d) 60 °C for 3 min.
(e) 38 °C for 30 min.
(f) 60 °C for 3 min.
(g) 38 °C for 30 min.
(h) 25°C final temperature.
5. Pool the samples from the PCR tubes into 1.5 mL microcen-
trifuge tubes and centrifuge at 16,000 × g for 15 min at room
temperature in order to pellet any precipitate that could have
formed during the hybridization step.
6. Pool the supernatants into 15 mL conical tubes.
7. In the meanwhile equilibrate the MyONE C1 magnetic strep-
tavidin beads twice 10 mL of low salt buffer. Use 900 μL of
MyONE C1 magnetic streptavidin beads for every 24 μL of
LNA probe.
8. Immobilize the beads on the magnetic stand and discard the
supernatant.
9. Keep the beads on the stand while adding the chromatin.
10. Still keeping beads on the stand dilute the chromatin in the
same volume of Milli-Q water as the volume of chromatin
(see Note 20).
11. Resuspend the beads very gently and slowly. Incubate for 1 h
and a half on a nutator at room temperature.
12. Bring the volume to 10 mL with lysis buffer.
13. Immobilize on magnetic stand.
3.6 Washes 1. Wash the beads five times with lysis buffer. Resuspend the beads
gently between the washes, avoid vortexing (see Note 21).
Perform the additional wash with low salt buffer.
2. Resuspend the beads in 1.2 mL lysis buffer per tube and trans-
fer to low binding 1.5 mL microtubes (see Note 22).
3. Immobilize the beads on the magnetic stand.
5. Resuspend in 1 mL of lysis buffer.
6. Immobilize the beads on the magnetic stand. A “halo” should
be apparent on the sides of the wall of the microcentrifuge
tube (see Note 23).
8. Resuspend in 1 mL of low salt buffer.

9. Incubate for 5 min at 42 °C in a thermomixer shaking at
500 rpm.
10. Immobilize the beads on the magnetic stand and discard the
supernatant.
3.7 Elution 1. Resuspend the beads in 900 μL of elution buffer.

2. Incubate for 30 min at 42 °C and an additional 10 min at
65 °C in the Thermomixer shaking at 500 rpm.
4. Transfer the eluates to fresh 1.5 mL microcentrifuge tubes.
6. Transfer the eluates again to fresh 1.5 mL microcentrifuge
tubes. This step is important to eliminate all the beads from
the eluates (see Note 24).
7. Verify that the eluates OD260 is around 15–25 ng/μL
(see Note 25).
3.8 Protein 1. Prepare the inputs: add 890 μL of lysis buffer to the 10 μL of
Precipitation input saved at step 37.
2. Precipitate proteins by adding 200 μL of 100 % trichloroacetic
acid to the 900 μL of eluates (the final concentration of tri-
chloroacetic acid should be in the range of 15–20 %).
3. Incubate on ice for 10 min.
4. Centrifuge at 16,000 × g for 10 min at 4 °C.
5. Carefully remove the supernatant with a pipette and leave
about 50 μL (see Note 26).
6. Wash the pellet twice with 1 mL of ice-cold acetone 100 %.
7. Vortex for 10 s.
8. Centrifuge at 16,000 × g for 10 min at 4 °C.
9. Carefully remove the supernatant and leave about 50 μL.
10. Evaporate the acetone by heating the tube at 100 °C for 10 s.
Repeat this step if the acetone is not entirely evaporated.
11. Resuspend in 70 μL of cross-linking reversal solution if the
sample will be analyzed by mass spectrometry or in 100 μL of
reversal cross-linking solution if the sample is used for western-
blot analysis. In both cases, resuspend the input in 400 μL of
cross-linking reversal solution.
12. Incubate for 24 min at 99 °C, after 12 min vortex briefly and
centrifuge to retrieve condensation from the cap back to the
bottom of the tube and incubate another 12 min at 99 °C.
3.9 Control for PICh 1. Separate proteins on a 12 % Bis-tris gel run with MOPS buffer.
Efficiency Load 2 μL of input and 5–10 μL of scrambled and PICh
extracts (see Note 27).
2. Silver stain the gel to reveal the proteins (see Note 28).
If 1/6th of eluted material was used, there should be visible
proteins bands after 3 min of developing. If the bands do
not appear after 5 min it means there are not enough pro-
teins in the remaining PICh extracts for analysis by mass
spectrometry (Fig. 2).
3. The PICh extracts can be used for western-blot analysis
(Fig. 3), mass spectrometry analysis or stored at −80 °C.
3.10 Protein Analysis 1. Resolve proteins (usually up to 40 % of the total, ~30 μL) on a
by Mass 12 % Bis-Tris gel.
Spectrometric 2. Cut each lane into 5–10 bands.
Approaches
3. Process each band using standard mass spectrometric identifi-
cation protocols.
Input Scramble PICh
191 kDa
97 kDa
64 kDa
51 kDa
39 kDa
28 kDa
19 kDa
Fig. 2 Silver staining of a PICh experiment. The protein complexes associated

with mouse ESCs telomeres were specifically enriched in the PICh extract where
a telomere-specific probe was used. Note that the banding pattern in the PICh
lane is different from the input lane. The scrambled lane shows nonspecific pro-
tein interactions eluted from the beads. Proteins were separated by SDS-PAGE
and visualized by silver staining. Molecular weight marker is indicated on the left.
0.0125 % of input and 4 % of scramble and PICh extracts were loaded
Input Scramble PICh
1 2 1 2 1 2
TRF1
H3
Fig. 3 Immuno-blot analyses of a telomere PICh experiment. Telomeric PICh extracts

were separated on a 12 % Bis-Tris gel. 1 and 2 are two independent experiments 2
starting from less material than 1. PICh efficiency is shown by the specific enrich-
ment of the telomere-associated protein TRF1. Note that TRF1 is undetectable in
similar amounts of input chromatin, as this factor is specific to telomeres
4. Protein quantifications in the SILAC experiments are per-

formed using the MaxQuant software using the default param-
eters (see Note 29).
3.11 Chromatin 1. Reverse cross-link 20 μL of chromatin input by adding 200 μL

Shearing Analysis of reverse cross-linking solution and add 10 μL of NaCl 5 M
(final concentration > 200 mM).
2. Incubate overnight at 65 °C by shaking 600 rpm.
3. The next day, add 24 μg of RNase A for at least 1 h at 37 °C.
4. Add 10 μL of Proteinase K (20 μg/μL) and digest proteins for
2 h at 65 °C.
5. Purify DNA with phenol–chloroform and ethanol-precipitate
the DNA.
6. Run 10 μg of DNA on a 0,8 % agarose gel (Fig. 4) (see Note 30).
4 Notes
1. To harvest 109 mouse embryonic stem cells, more or less 50

plates of 150 mm gelatin coated dishes should be foreseen.
Lowering the scale of the experiment is not advised, as this
usually results in higher background contaminations.
2. SILAC is an approach based on the protein incorporation of
stable isotope-labeled amino acids into newly synthetized pro-
teins in vivo. The compared cell populations should be grown
with the “light” isotope of the used amino acids in the control
condition and with the “heavy” version in the experiment.
Use a drop-out medium and also use dialyzed serum. It is
highly recommended to perform the reverse experiment in
which the amino acid isotopes are swapped between the
two conditions. It is also advised to perform an incorporation
on a test plate to make sure there is a negligible conversion of
arginine into proline, which would complicate the analyses.
10 kb
8 kb
6 kb
5 kb
4 kb
3 kb
2 kb
1.5 kb
1 kb
0.5 kb
Fig. 4 Agarose gel electrophoresis of DNA isolated from PICh extracts. Size dis-
tribution of purified mouse embryonic stem cell DNA after sonication. 2 μg of
DNA is analyzed on a 0.8 % agarose gel, post-stained with ethidium bromide
The time needed for complete incorporation into proteins

depends on the cell line. In general six cell cycles are consid-
ered sufficient for full incorporation and replacement of “long”
lived proteins. For stem cells that cycle every 8 h this corre-
sponds to 3 days in culture. Also it is good to check that the
different media formulations are not affecting the cell cycle
distribution by FACS analyses.
3. The cross-linking step usually involves much higher formalde-
hyde concentrations and volumes than in standard chromatin
immunoprecipitation. Follow appropriate working practices
and dispose of solutions properly.
4. Do not quench formaldehyde with glycine solutions as usually
performed during classical chromatin immunoprecipitation
experiments. It may result in nonspecific cross-linking of gly-
cine to proteins and prevent peptide mass attribution during
the mass spectrometry analysis. Instead, dilute and wash out
unreacted formaldehyde with PBS washes.
5. If the samples are used for mass spectrometry analysis, it is
important to avoid any keratin or other source of contamination
that could interfere with the analysis. Therefore it is advisable to
work with powder free gloves, filter tips, disposable pipettes and
clean but non autoclaved tubes. Before starting the SDS PAGE,
extensively wash the migration tank with MilliQ water, use com-
mercial pre-cast gels, premade buffers, loading dye etc., that
have limited chances to be contaminated by keratins or other
unwanted protein. Stain your gel in a brand new plastic dish
(diam 150 mm) and open the cover only to change buffers.
6. A 4–5 mL volume of cell pellet is sufficient for the sequence-
specific pull-down and the scramble pull-down.
7. Flash freeze the cross-linked material in liquid nitrogen and
store it at −80 °C to minimize cross-linking reversal of pro-
teins from the chromatin. It is not recommended to use mate-
rial stored longer than one month as we observed significant
de-cross-linking and much lower protein retrieval. The best
results are obtained when fresh material is used and the stor-
age steps are skipped.
8. It is extremely important that the same number of cells is
mixed for quantitative mass spectrometry analysis.
9. Resuspend pellets by vortexing until the sonication step 22.
To avoid foam vortex very gently and slowly increase to the
maximal speed. Using a pipette to resuspend results in losing
a significant amount of chromatin that sticks to the plastic. Pay
attention to minimize material loss at each step, as the sample
sticks to plastic ware.
10. This RNase A step is to optimize the subsequent LNA probe–
chromatin hybridization step and to avoid with the nonspe-
cific capture of RNA–protein complexes.
11. Do not vortex, as vortexing leads to excessive foaming of the
sample which inhibits sonication efficiency. Resuspend by
pipetting with 1 mL micropipette.
12. At this level of cross-linking, indirect ultrasonication in water
baths does not solubilize the chromatin. We have recently
developed an alternative solubilization method involving
restriction enzyme digestion and high pressure solubilization
with a French high pressure system [3].
13. Heating at 58 °C favors the unmasking of endogenously bio-
tinylated proteins which have to be cleared from the sample.
14. Pre-clearing is necessary to remove most of endogenously bio-
tinylated proteins that might compete with desthiobiotinyl-
ated probe during the streptavidin binding step. If the signal is
weak on silver-stained gel, reduce the pre-clear to 2 h at room
temperature instead of overnight.
15. This step is very important for the quality of the purification.
The ultra-fast gel filtration reduces the salt concentration of
the sample by about two-thirds leaving roughly 30 mM of
NaCl. Low salt concentrations make the hybridization step
more stringent. Also, it will prevent nonspecific precipitation
of chromatin to the dynabeads during the capture. The use of
DNA denaturing agents in buffers to increase hybrid specific-

ity (formamide or urea) is not advised at any step, as this will
result in non specific precipitation on the magnetic beads dur-
ing the capture step.
16. Higher values of 260/280 ratio would mean that RNaseA
treatment was not effective and that chromatin is contami-
nated with RNA–protein complexes. This will strongly affect
the outcome of the PICh procedure as this usually results in
the nonspecific capture of those RNP complexes. In such a
case both scrambled and telomere pull downs contain many
hnRNPs ribosomal proteins.
17. The scrambled probes are nonspecific probes used as a nega-
tive control.
18. An LNA containing probe offers increased affinity for its com-
plementary strand and the LNA–DNA hybrids have high
melting temperatures ensuring strong and stable interactions
compared to traditional DNA or RNA oligonucleotides.
19. The hybridization program was optimized so that LNA probe
can invade DNA sequence into chromatin without observable
protein de-cross-linking. Denaturation temperatures higher
than 75 °C should be avoided as they result in significant
cross-linking reversal and proteins loss from the target chro-
matin. Denaturation temperature below 68 °C should also be
avoided because no capture was observed.
20. The addition of water prevents chromatin precipitation on the
beads.
21. Do not vortex but resuspend the beads by gently inverting the
tubes until all the beads are totally resuspended.
22. To avoid material loss in the pipette tip add 400 μL of lysis
buffer with one pipette and resuspend the beads and transfer
the mixture with another pipette (keep the tip on this pipet).
Repeat this step 2 additional times (3 × 400 μL = 1.2 mL).
23. The “halo” indicates that the PICh is efficient. The chromatin
bound dynabeads need more space when immobilized on the
magnetic stand thus forming a “halo” on the wall of the
microcentrifuge tube.
24. Any bead remaining in the eluate will contribute, together
with nonspecifically bound proteins, to the nonspecific back-
ground. As the target is usually of low abundance, the nonspe-
cific background can significantly alter the outcome of the
experiment as discussed in the introduction.
25. This value is a good indication that the LNA probe has been
eluted from the beads. Usually at the same concentration, the
scramble LNA probe absorbs more than the telomere LNA
probe in the eluates. The reason for this discrepancy is
unknown.
26. Put the 1.5 mL microcentrifuge tube with the hinge outwards.
The protein pellet will end up all along on this side of the wall
of the tube which explains why it is sometimes difficult to see.
27. If the PICh assay is very clean the scrambled extract should be
devoid of any proteins. Most frequent contaminants are his-
tones. The PICh extract should show different banding pat-
terns that input meaning the probe used specifically purified
proteins bound to the genomic region of interest.
28. Silver staining is about 50–100 times more sensitive than

Coomassie blue staining. Although some Silver staining meth-
ods are compatible with protein identification by mass spec-
trometry, we found that the analysis from Silver stained bands
strongly decreased the sensitivity of the analysis. Thus, while
the identification of proteins from such silver stained gels is
technically feasible, we do not recommend this.
29. The direct mass spectrometric analysis of liquid samples is
doable. However, it reproducibly resulted in a much
smaller and less complex proteome and therefore is not
recommended.
30. The DNA smear ranges from below 0.5 to more than 3 kb.
This is higher than the average fragment size distribution
obtained after the typical lower cross-linking conditions used
in ChIP (1 % HCHO for 10 min). Do not pre-stain, but post-
stain the gel for accurate size estimation.
Acknowledgments
We would like to thank Titia de Lange for the kind gift of the
TRF1 antibody.
References
1. Carey MF, Peterson CL, Smale ST (2009) 6. Landgraf R, Chen CH, Sigman DS (1995)
Chromatin immunoprecipitation (ChIP). Cold R-loop stability as a function of RNA structure
Spring Harb Protoc 9:pdb prot5279 and size. Nucleic Acids Res 23:3516–3523
2. Dejardin J, Kingston RE (2009) Purification of 7. Hirsch JD, Eslamizar L, Filanoski BJ,
proteins associated with specific genomic Loci. Malekzadeh N, Haugland RP, Beechem JM,
Cell 136:175–186 Haugland RP (2002) Easily reversible desthio-
3. Ide S, Dejardin J (2015) End-targeting pro- biotin binding to streptavidin, avidin, and other
teomics of isolated chromatin segments of a biotin-binding proteins: uses for protein label-
mammalian ribosomal RNA gene promoter. ing, detection, and isolation. Anal Biochem
Nat Commun 6:6674 308:343–357
4. Braasch DA, Corey DR (2001) Locked nucleic 8. Saksouk N, Barth TK, Ziegler-Birling C, Olava
acid (LNA): fine-tuning the recognition of N, Nowak A, Rey E, Mateos-Langerak J,
DNA and RNA. Chem Biol 8:1–7 Urbach S, Reik W, Torres-Pedilla ME, Imhof
5. Belotserkovskii BP, Reddy G, Zarling DA A, Dejardin J, Simboeck E (2014) Redundant
(1999) DNA hybrids stabilized by heterologies. mechanisms to form silent chromatin at peri-
Biochemistry 38:10785–10792 centromeric regions rely on BEND3 and DNA
methylation. Mol Cell 56:580–594
Chapter 4
Profiling Cell Lines Nuclear Sub-proteome

Aline Poersch, Andrea G. Maria, Camila S. Palma, Mariana L. Grassi,
Daniele Albuquerque, Carolina H. Thomé, and Vitor M. Faça
Abstract
Proteins are very dynamic within the cell and their localization and trafficking between subcellular com-
partments are critical for their correct function. Indeed, the abnormal localization of a protein might lead
to the pathogenesis of several diseases. The association of cell fractionation methods and mass spectrom-
etry based proteomic methods allow both the localization and quantification of proteins in different sub-
compartments. Here we present a detailed protocol for enrichment, identification, and quantitation of the
nuclear proteome in cell lines combining nuclear subproteome enrichment by differential centrifugation
and high-throughput proteomics.
Key words Nuclear fractionation, Subcellular proteomics, Protein localization, Cell Line, Mass
spectrometry
1 Introduction
Eukaryotic cells are arranged in compartments characterized by

distinct biochemical process and particular sets of proteins [1]. As
each compartment provides specific physiological conditions, the
presence in a subcellular compartment and the traffic into different
organelles are crucial to determine the function of a protein as well
as the function of the compartment [2]. Moreover, the subcellular
localization of a protein defines whether it will go through post-
translational modifications, interact with other molecules and inte-
grate with different biological networks [1]. Because proteins are
very dynamic within the cell, the localization of a protein into a
specific cellular compartment defines its correct function [3] and,
therefore, the abnormal localization of a protein might lead to
pathologies that include cardiovascular, neurodegenerative, cancer,
and metabolic diseases [1]. In addition, modulation of some pro-
tein subcellular localization has been proposed as promising thera-
peutic strategies [4]. Altogether, these factors lead to an increased
interest in performing studies that enable the identification and
quantification of proteins in specific subcellular localization.
35
36 Aline Poersch et al.
The nucleus has important roles in several cellular functions

such as gene expression, cell cycle control, cell growth and signal
transduction [5]. The shuttling of specific proteins out of the
nucleus is essential for the regulation of important functions in the
cell [6, 7]. A variety of tumor suppressors demand to localize in the
nucleus for their correct function and aberrant protein localization
or failure on the dynamics of spatio-temporal cell signaling are
involved in tumor cancer cell survival, tumorigenesis, tumor pro-
gression, and drug resistance [1, 8, 9]. Consequently, for particular
sets of proteins, cytoplasmic localization can be explored as a marker
of cell growth disorder and tumor development risk [6, 10, 11].
For instance, nuclear export machinery through XPO1 (Exportin-1/
Chromosome Region Maintenance 1/CRM1) is upregulated in
different cancer types and may be used as a prognostic indicator
[7]. Abnormal XPO1 intracellular localization might affect the cor-
rect function of important suppressors and oncogenic proteins such
as p53, Rb, FOXO, p21, and others, contributing to cancer devel-
opment and progression [12–16]. Therefore, identification of the
correct location of proteins is a potential strategy for therapeutic
intervention in many types of cancer.
A subcellular proteome may be identified by applying specific
enrichment strategies that concentrate organelles and cellular com-
partments [17]. These strategies are usually based on differences in
sedimentation coefficients and organelles density [18]. Thus, the
subcellular compartments can be separated by centrifugation, tak-
ing advantage of the density gradient and buffers that maintain the
compartment integrity. Moreover, detergents might be used to
enrich the subcellular fractions, depending on the compartment
solubility [19, 20]. Once the subcellular fraction of interest is iso-
lated and proteins are enriched, analytical methods can be applied
to identify and quantify the proteins present in each compartment.
Several studies have applied mass spectrometry for this purpose
[17] and large-scale identification of proteins with proteomic
methods is showing to be an important approach to elucidate pro-
tein function and regulation [21]. Here, we present a protocol to
study the nuclear proteome using centrifugation to enrich for cel-
lular nucleus, combined to in-depth proteomics based on intact
protein fractionation and highthroughput LC-MS/MS protein
identification. Further, we show data obtained by our group apply-
ing the protocol described here, illustrating the quality and impor-
tance to analyze the protein content in the nucleus.
2 Materials
2.1 Cell Culture 1. Cell culture media: Mammary Epithelial Cell Growth
Medium (MEBM) supplemented with 100 ng/mL cholera
toxin (Sigma), MEGM SingleQuots kit (bovine pituitary
extract—BPE, human epidermal growth factor—hEGF,
Profiling Nuclear Sub-proteome 37
insulin, hydrocortisone, and gentamicin amphotericin-b—

GA-1000; Lonza), 10–0 % (v/v) Fetal Bovine Serum (FBS)
and 1 % (v/v) penicillin/streptomycin (see Note 1).
2. Pre-warmed (37 °C) phosphate buffered saline (PBS), pH 7.2.
3. 100 mm treated culture dishes for adherent cell culture.
2.2 Cell Lysis 1. Lysis buffer: 50 mM HEPES pH 7.4, 10 mM NaCl, 5 mM
Components MgCl2, 0,1 mM EDTA, 1 mM Na3VO4 (fosfatase inhibitor),
1 mM NaF and 1 mM Na4P2O7.10dH2O, with protease inhib-
itor cocktail 5 % (v/v) (Sigma-Aldrich) (see Note 2).
2. Cell scraper.
3. Ice-cold phosphate buffered saline (PBS), pH 7.2.
4. 3 mL syringe with needle
5. Tissue homogenizer model D-130 (Biosystems) or equivalent
model.
6. 2 mL microcentrifuge tubes.
7. Refrigerated microcentrifuge.
2.3 Nuclear 1. Nuclear protein extraction buffer: 8 M urea and 2 % (v/v)
Fractionation CHAPS containing 5 % (v/v) protease inhibitor cocktail and
Components 1 mM Na3VO4 (see Note 3).
2. Vortex mixer.
3. Refrigerated microcentrifuge.
2.4 Sample 1. Bradford quantification kit.

Preparation 2. 10 % Mini-PROTEAN® TGX™ Precast Gel (Bio-Rad) or
for Proteomic Analysis equivalent, and gel electrophoresis equipment (Bio-Rad).
2.4.1 Protein 3. 2× Laemmli sample buffer (Bio-Rad).
Fractionation by SDS-PAGE 4. Reduction solution: Dissolve 10 mg of dithiothreitol (DTT)
in 1 mL of ultrapure water (10 μg/μL) (see Note 4).
5. Alkylation solution: Dissolve 5 mg of iodoacetamide (IAM) in
0.1 mL of ultrapure water (50 μg/μL) (see Note 5).
6. Fixing solution: 40 % ethanol, 10 % acetic acid, and 50 % ultra-
pure water.
7. Staining solution: 0.1 g of Coomassie Brilliant Blue G-250,
1.175 mL of phosphoric acid, 10 g of ammonium sulfate, and
q.s.p. 100 mL of ultrapure water (see Note 6).
8. Staining solution 20 % methanol: Dilute 40 mL of staining
solution in 10 mL of methanol.
9. Washing solution: 25 % methanol–ultrapure water.
2.4.2 In-Gel Trypsin 1. Clean glass plate and sterile scalpel (see Note 7).
Digestion 2. Ammonium bicarbonate 100 mM pH 8 solution: Dissolve
1.58 g of ammonium bicarbonate in 200 mL of ultrapure
water (see Note 8).
3. Destain solution: 50 mL of ammonium bicarbonate 100 mM
pH 8 with 50 mL of 100 % acetonitrile.
4. Trypsin solution: Resuspend one vial containing 20 μg of
sequencing-grade modified trypsin (Promega) in 100 μL of
ammonium bicarbonate 100 mM pH 8 solution (see Note 9).
5. Thermomixer.
2.5 Peptide 1. Peptide extraction buffer I (50 % acetonitrile–50 % ultrapure

Extraction and Sample water–0.1 % formic acid solution).
Desalting Components 2. Peptide extraction buffer II (70 % acetonitrile–30 % ultrapure
water–0.1 % formic acid solution).
3. Peptide extraction buffer III (100 % acetonitrile–0.1 % formic
acid solution).
4. ZipTip® C18 pipette tips (Merck Millipore).
5. Mass spectrometry compatible injection vials.
6. Thermomixer.
7. SpeedVac concentrator.
2.6 High-Throughput 1. Equipment: LTQ-Orbitrap Velos mass spectrometer (Thermo

Mass Spectrometry Fisher) coupled to a nanoflow chromatography system
Components (Eksigent).
2. Chromatographic column: 25 cm long column (Picofrit
75 mm ID, New Objectives, packed in-house with MagicC18
resin).
3. Solvents for reversed-phase chromatography: aqueous solvent
(A)—5 % acetonitrile–95 % water–0.1 % formic acid; organic
solvent (B)—95 % acetonitrile–5 % water–0.1 % formic acid.
Bottled water, acetonitrile, and formic acid are obtained from
Fisher Scientific.
3 Methods
3.1 Cell Lysis 1. Cultivate MCF10A cell line in 75 cm2 flasks in described cul-
ture media. For the fractionation, plate 5 × 106–1 × 107 cells in
a 100 mm culture dish.
2. After 24 h and/or cellular confluence >80 %, remove the
media, wash the cells twice with pre-warmed PBS, substitute
with new culture media and proceed with the treatments of
interest.
3.2 Preparation 1. After treatments of interest, remove the cell media and wash
of Cell Lysate the cells twice with ice-cold PBS. Remove any solution excess
that remains on the dish. Add 150 μL of lysis buffer contain-
ing protease inhibitors in each dish and scrape cells using a cell
scraper. Collect cell lysate and transfer to 2 mL microtubes.
Keep samples on ice.
2. Pass the cell lysate 20 times through a thin 25-gauge needle
using a 3 mL syringe and homogenize in a tissue homogenizer
(dounce homogenizer) for 1 min. Keep samples on ice.
3. Centrifuge samples at 500 × g for 20 min at 4 °C to pellet
nuclei (no brake applied during the deceleration of the centri-
fuge). Carefully transfer all of the supernatant to a 1.5 mL
microtube (see Note 10).
3.3 Nuclear 1. Add to the dry pellet obtained previously (see Subheading 3.2)
Fractionation 100 μL of nuclear extraction buffer. Nuclear proteins extrac-
tion are obtained by sonication of samples using an ultrasound
probe for 5 min followed by cycles of vortexing for 20 s and
ice bath for 5 min. Repeat this cycle three times.
2. Centrifuge samples at 20,000 × g for 30 min at 4 °C and trans-
fer the supernatant enriched in nuclear proteins to a 1.5 mL
microtube. Store samples at −80 °C until use.
3.4 Sample 1. Quantify samples containing nuclear proteins using Bradford

Preparation for Assay kit according to the manufacturer’s instructions.
Proteomic Analysis 2. Aliquot 50 μg of nuclear proteins sample, mix with 2× Laemmli
3.4.1 SDS- sample buffer and perform protein reduction by adding 5 μL of
Polyacrylamide Gel reduction solution and maintaining the reaction at 95 °C for 5 min.
Electrophoresis Centrifuge samples for 5 min at 8500 × g at room temperature.
3. After cooling, perform protein alkylation by adding 3 μL of
alkylating solution and maintain the reaction at room tem-
perature for 20 min protected from light.
4. Load each sample onto a 10 % 1.0 mm precast gel and perform
SDS-polyacrylamide gel electrophoresis according to the
manufacturer’s instructions.
5. Fix the gel with 25 mL of fixing solution, shaking for 30 min
at room temperature.
6. Remove fixing solution and add 25 mL of staining solution.
Incubate overnight shaking at room temperature.
7. Remove staining solution and wash several times with 25 %
methanol washing solution (see Note 11).
3.4.2 In-Gel Trypsin 1. Place the gel on a clean glass plate and excise each lane con-
Digestion taining proteins with a clean scalpel. Cut each lane into
approximately 1 cm square pieces and transfer to a clean
1.5 mL microtube pre-rinsed with methanol.
2. Add 200 μL of destaining solution and incubate for 10 min,

shaking at room temperature. Repeat this step four times.
Finally, add 200 μL of destaining solution and incubate over-
night at 4 °C (see Note 12). Add 200 μL of acetonitrile for
10 min and dry with SpeedVac concentrator.
3. Perform in-gel trypsin digestion of gel slices adding 20 μL
(0.5 μg) of trypsin solution and agitate vigorously. Carry out
digestion for 30 min at 37 °C constantly shaking in a thermo-
mixer (450 rpm) (see Note 13).
4. Add 200 μL of ammonium bicarbonate 100 mM pH 8 and
incubate overnight at 37 °C constantly shaking in a thermo-
mixer (450 rpm).
3.5 Peptide 1. After digestion, transfer to a 1.5 mL microtube the trypsin/

Extraction and Sample ammonium bicarbonate 100 mM pH 8 solution and dry in
Desalting SpeedVac.
2. Add 200 μL of peptide extraction buffer to the gel pieces and
incubate for 90 min at room temperature, constantly shaking
in a thermomixer (450 rpm).
3. Combine the supernatant containing the peptide extract I
with the initial supernatant in the same 1.5 mL microtube and
dry in SpeedVac.
4. Again, add 200 μL of peptide extraction buffer II to the gel
pieces and incubate for 90 min at room temperature, con-
stantly shaking in a thermomixer (450 rpm).
5. Combine the supernatant containing the peptide extract II
supernatant with previous peptide extracts and dry in
SpeedVac.
6. Add 200 μL of peptide extraction buffer to the gel slices and
incubate for 90 min at room temperature, constantly shaking
in a thermomixer (450 rpm).
7. Finally, combine supernatant containing the peptide extract
III and dry in SpeedVac.
8. Dissolve peptide mixture in 10 μL 5 % acetonitrile–0.1 % for-
mic acid.
9. Desalt samples using ZipTip® C18 pipette tips as follows:
10. Condition tips with 20 μL of 100 % acetonitrile–0.1 % formic
acid. Repeat three times.
11. Equilibrate tips with 20 μL 5 % acetonitrile–0.1 % formic acid.
Repeat three times.
12. Apply sample through the tip.
13. Wash with 20 μL 5 % acetonitrile–0.1 % formic; repeat three
times.
14. Elute peptides with 20 μL of 50 % acetonitrile–water in 0.1 %

formic acid. Repeat three times.
15. Dry eluted peptides in SpeedVac. Samples can be stored dried
at −80 °C until ready for LC-MS analysis.
16. Reconstitue samples in 15 μL 5 % acetonitrile–0.1 % formic
acid. Centrifugate at 12,000 × g for 15 min and transfer to
mass spectrometry compatible injection vials.
3.6 High-Throughput 1. Carry out the high throughput LC-MS/MS data collection
Mass Spectrometry for each individual fraction. Inject 10 μL of peptide extract
and analyze samples over a 90 min linear gradient from 5 to
35 % of organic solvent at 350 nL/min in the system described
in section 2.6 (see Note 14).
2. Process LC-MS/MS files through data bank search, protein
inference and quantitative analysis (see Note 15).
3. Match the lists of proteins identified in the nuclear enriched
fraction with proteins identified in the total cell extract, cyto-
plasmic or membrane enriched fractions (see Note 10). Select
for nuclear proteins based on the higher value of enrichment
obtained from the ratio of spectral counts observed in enriched
nuclear fractions/total cell extract cytoplasmic or membrane
enriched fraction profile (see Note 16). A Sample dataset
obtained for MCF10A cell line is presented in Table 1 and
Fig. 1 (see Note 17).
4 Notes
1. Cultivate cell lineage according to the culture method speci-

fied in ATCC. Cultivate cells in 75 cm2 flasks or 100 mm
dishes and maintain at 37 °C in a 5 % CO2 humidified incuba-
tor. Cultivate cells in SILAC media, using standard SILAC
protocols, if relative quantification is required [22].
2. Prepare a fresh lysis buffer solution and add protease inhibitor
cocktail (P8340, Sigma Aldrich) and Na3VO4 (phosphatase
inhibitor) right before proceeding with the cell lysis.
3. Prepare a fresh nuclear protein extraction buffer solution and
add protease inhibitor cocktail and Na3VO4 right before per-
forming the protein extraction and solubilization.
4. Dissolve DTT in water right before proceeding protein
reduction.
5. Dissolve IAM in water right before proceeding protein alkyla-
tion and protect from light. Protein alkylation before SDS-
PAGE run improves peptide recovery after in situ protein
digestion.
Table 1
Top 20 proteins enriched in the nuclear fraction of MCF10A breast epithelial cells.
42
MCF10A
Cytoplasm Nucleus “Enrichment“

Uniprot entry name Gene name Description counts counts nucleus/cytoplasm
1 FBRL_HUMAN FBL rRNA 2′-O-methyltransferase fibrillarin 1 54 54
2 IF16_HUMAN IFI16 Gamma-interferon-inducible protein 16 N/D 45 45
3 H2AX_HUMAN H2AFX Histone H2AX 1 44 44
Aline Poersch et al.
4 HMGA1_HUMAN HMGA1 High mobility group protein HMG-I/HMG-Y N/D 35 35

5 DKC1_HUMAN DKC1 H/ACA ribonucleoprotein complex subunit 4 N/D 35 35
6 CSK21_HUMAN CSNK2A1 Casein kinase II subunit alpha N/D 34 34
7 HBA_HUMAN HBA1 Hemoglobin subunit alpha N/D 34 34
8 NEP1_HUMAN EMG1 Ribosomal RNA small subunit methyltransferase NEP1 N/D 31 31
9 FUS_HUMAN FUS RNA-binding protein FUS N/D 31 31
10 SNUT1_HUMAN SART1 U4/U6.U5 tri-snRNP-associated protein 1 N/D 27 27
11 HNRDL_HUMAN HNRPDL Heterogeneous nuclear ribonucleoprotein D-like 1 25 25
12 SRSF2_HUMAN SRSF2 Serine/arginine-rich splicing factor 2 N/D 25 25
13 TMOD3_HUMAN TMOD3 Tropomodulin-3 1 24 24
14 DDX21_HUMAN DDX21 Nucleolar RNA helicase 2 1 24 24
15 HNRPM_HUMAN HNRNPM Heterogeneous nuclear ribonucleoprotein M 2 48 24
16 SP16H_HUMAN SUPT16H FACT complex subunit SPT16 N/D 24 24
17 NOP56_HUMAN NOP56 Nucleolar protein 56 N/D 24 24
18 TOP1_HUMAN TOP1 DNA topoisomerase 1 N/D 22 22
19 CBX5_HUMAN CBX5 Chromobox protein homolog 5 N/D 21 21
20 SRRT_HUMAN SRRT Serrate RNA effector molecule homolog N/D 21 21
Breast epithelial cell line MCF10A was profiled by high-throughput proteomics according the method here described. The ratio of spectral counts of proteins detected in the nuclear
enriched fraction over spectral counts detected in the cytoplasmic enriched fraction was used as an “enrichment factor” for nuclear proteins. The list presents the top 20 proteins of
MCF10A cell line detected in the nuclear fraction. For calculation purposes, proteins not detected (N/D) in the cytoplasmic fraction were considered as having one spectral count.
Total Cytoplasm and Nucleus

2% 2% 1% 1%
2% 2% 3%
3%
3%
4%
4%
6%
8%
7%
nucleus (GO:0005634)
52%
10%
Cytoskeleton (GO:0005856)
67%
23% chromosome (GO:0005694)
mitochondrion (GO:0005739)
Cytoplasm Nucleus Cytoplasmic membrane-bounded vesicle (GO:0016023)
2% 1%
2% 2% 2% 1% endosome (GO:0005768)
2%
4% Golgi apparatus (GO:0005794)
5%
6%
vacuole (GO:0005773)
16% 12% 46% endoplasmic reticulum (GO:0005783)
53%
18% 28%
Fig. 1 Protein dataset for nuclear profile of MCF10A breast epithelial cell line. (a) Venn diagram of proteins
identified by LC-MS/MS in the nuclear and cytoplasmic enriched fractions of MCF10A cells. The fractionation
methodology allowed confident overall protein identification of 2220 proteins with less than 1 % FDR. The
overlap between nuclear enriched and cytoplasmic enriched fraction (55 %) as well as the number of proteins
identified in only one of those fractions (45 %) shows the complementarity of both subcellular proteomic pro-
files. (b) Gene Ontology classification of proteins enriched in the nuclear and cytoplasmic compartments also
supports the complementarity of the proteomic profiles
6. Staining solution can be prepared before use and stored in a

clean amber or brown glass bottle at room temperature.
Prepare a fresh staining solution in 20 % methanol right before
proceeding gel staining.
7. To avoid sample keratin contamination wash glass plate with

70 % ethanol and always wear gloves.
8. After preparing ammonium bicarbonate 100 mM pH 8 solu-
tion, pass it thru a 0.22 μm filter.
9. Before starting in-gel digestion, resuspend trypsin vial in
100 μL of ammonium bicarbonate 100 mM pH 8 solution
and incubate for 15 min at 37 °C.
10. To obtain additional subcellular enriched fractions, such as
cytoplasm and membrane, follow the next steps: (1) centrifuge
the supernatant obtained (Subheading 3.2) at 16,000 × g for
20 min at 4 °C to pellet membranes. The supernatant is
denoted enriched cytosol fraction; (2) resuspend pellet in
50 μL of membrane extraction buffer (25 mM MES
(2-(N-morpholino)-ethanesulfonic acid), pH 6.5, 150 mM
NaCl, 2 % Triton X-100) containing 5 % (v/v) protease inhibi-
tor cocktail and 1 mM Na3VO4 and incubate on ice for 60 min
with agitation every 5 min; (3) centrifuge sample at 16,000 × g
for 20 min at 4 °C. The supernatant is denoted enriched mem-
brane fraction. Store samples at −80 °C until use.
11. Stained gel can be stored in 25 % methanol washing solution at
4–6 °C for a few days.
12. Ensure that all gel slices are completely destained. If necessary,
repeat destaining step adding 200 μL of destaining solution.
13. Incubate the gel pieces for few minutes (30 min) until the
trypsin solution is completely absorbed. Only after that com-
plete with ammonium bicarbonate solution to keep the gel
piece completely immersed in solution.

14. Using the described conditions and instrumentation, it is
expected that each run will yield 200–500 good protein iden-
tifications. When all nuclear enriched fractions protein identifi-
cations are combined, it is expected to obtain a list of confident
identifications containing more than 2000 protein hits.
15. Acquired data can be automatically processed by the open
source Labkey Server (www.labkey.org) platform, which
employs the TransProteomic Pipeline, developed at the
Institute of Systems Biology [23]. Search data against the
most recent version of the human proteome database (Uniprot)
or other appropriate human protein database of your choice.
A fixed modification of 57.021464 is added to cysteine resi-
dues and a variable modification of 15.994915 is added to
methionine residues for database search. Optionally, when
SILAC strategy is used, account for incorporation of the light
and heavy amino acid isotopes. To estimate the significance of
peptide and protein matches, we apply the tools PeptideProphet
[24] and ProteinProphet [25]. Identifications with a
PeptideProphet probability score greater than 0.9 are selected
and submitted to ProteinProphet to account for the protein

inference problem. Overall, false discovery rates for this proce-
dure are less than 1 %. Protein quantification can also be per-
formed automatically with the Q3 tool available in the
distribution of labkey server [23].
16. The spectral counting method can be used to estimate nuclear
protein enrichment as previously described [26]. Briefly, the nor-
malized spectral counts for each protein group output by
ProteinProphet for nuclear protein profile divided by the normal-
ized spectral counts output for the total cell extract or cytoplas-
mic/membrane profile can be considered as a semiquantitative
enrichment analysis. The total number of counts in the entire
experiment is used as a normalization parameter for each profile.
17. Several proteins were identified only in the nuclear enriched
fraction dataset. As expected, protein markers for nucleus were
detected among the top enriched protein in our analysis, such
as histones and DNA binding proteins (see Table 1). Gene
ontology analysis supports the enrichment of nuclear protein
based on annotation for cellular compartments (see Fig. 1). A
comparison for annotations observed in a cytoplasmic profile
for the same cell line highlights the enrichments obtainded by
the methodology described here.
Acknowledgments
This research was supported by FAPESP (Young Scientist Grant—

Proc.No. 2011/0947-1), CNPq, Center for Cell Based Thereapy—
CTC-CEPID (Proc.FAPESP 2013/08135-2) and CISBi-NAP.
A.G.M. C.S.P, M.L.G., C.H.T., and D.A. received fellowships
from FAPESP Proc. No., 2014/16839-2, 2012/09682-4,
2013/08755-0, 2013/07675-3, and 2012/02518-4, respec-
tively. A.P receives a PNPD fellowship from CAPES. V.M.F.
receives a fellowship from CNPq, Proc.No. (308561/2014-7).
We thank Profs. Emanuel Carrilho and Daniel Cardoso for allow-
ing our data collection with the LTQ-Orbitrap Velos at the
Analytical Central – Chemistry Institute of São Carlos—University
of São Paulo.
References
1. Hung MC, Link W (2011) Protein localiza- generative model. Bioinformatics 31(12):i365–
tion in disease and therapy. J Cell Sci 124(Pt i374. doi:10.1093/bioinformatics/btv264
20):3381–3392. doi:10.1242/jcs.089110 3. Butler GS, Overall CM (2009) Proteomic identifi-
2. Simha R, Briesemeister S, Kohlbacher O, cation of multitasking proteins in unexpected loca-
Shatkay H (2015) Protein (multi-)location tions complicates drug targeting. Nat Rev Drug
prediction: utilizing interdependencies via a Discov 8(12):935–948. doi:10.1038/nrd2945
4. Tomas A, Futter CE, Eden ER (2014) EGF export. J Biol Chem 277(10):8517–8523.
receptor trafficking: consequences for signal- doi:10.1074/jbc.M108867200
ing and cancer. Trends Cell Biol 24(1):26–34. 17. Drissi R, Dubois ML, Boisvert FM (2013)
doi:10.1016/j.tcb.2013.11.002 Proteomics methods for subcellular pro-
5. Chahine MN, Pierce GN (2009) Therapeutic teome analysis. FEBS J 280(22):5626–5634.
targeting of nuclear protein import in patho- doi:10.1111/febs.12502
logical cell conditions. Pharmacol Rev 18. Lee YH, Tan HT, Chung MC (2010)
61(3):358–372. doi:10.1124/pr.108.000620 Subcellular fractionation methods and strate-
6. Turner JG, Sullivan DM (2008) CRM1- gies for proteomics. Proteomics 10(22):3935–
mediated nuclear export of proteins and drug 3956. doi:10.1002/pmic.201000289
resistance in cancer. Curr Med Chem 19. Ramsby ML, Makowski GS, Khairallah EA
15(26):2648–2655 (1994) Differential detergent fractionation of iso-
7. Takeda A, Yaseen NR (2014) Nucleoporins lated hepatocytes: biochemical, immunochemical
and nucleocytoplasmic transport in hemato- and two-dimensional gel electrophoresis charac-
logic malignancies. Semin Cancer Biol 27:3– terization of cytoskeletal and noncytoskeletal
10. doi:10.1016/j.semcancer.2014.02.009 compartments. Electrophoresis 15(2):265–277
8. Kau TR, Way JC, Silver PA (2004) Nuclear 20. Sawhney S, Stubbs R, Hood K (2009)
transport and cancer: from mechanism to Reproducibility, sensitivity and compatibility
intervention. Nat Rev Cancer 4(2):106–117. of the ProteoExtract subcellular fractionation
doi:10.1038/nrc1274 kit with saturation labeling of laser microdis-
9. Wang SC, Hung MC (2005) Cytoplasmic/ sected tissues. Proteomics 9(16):4087–4092.
nuclear shuttling and tumor progression. Ann doi:10.1002/pmic.200800949
N Y Acad Sci 1059:11–15. doi:10.1196/ 21. Walther TC, Mann M (2010) Mass
annals.1339.002 spectrometry-based proteomics in cell biology.
10. Fabbro M, Henderson BR (2003) Regulation J Cell Biol 190(4):491–500. doi:10.1083/
of tumor suppressors by nuclear-cytoplasmic jcb.201004052
shuttling. Exp Cell Res 282(2):59–69 22. Ong SE, Mann M (2006) A practical recipe for
11. Salmena L, Pandolfi PP (2007) Changing ven- stable isotope labeling by amino acids in cell
ues for tumour suppression: balancing destruc- culture (SILAC). Nat Protoc 1(6):2650–
tion and localization by monoubiquitylation. 2660. doi:10.1038/nprot.2006.427
Nat Rev Cancer 7(6):409–413. doi:10.1038/ 23. Rauch A, Bellew M, Eng J, Fitzgibbon M,
nrc2145 Holzman T, Hussey P, Igra M, Maclean B, Lin
12. Santiago A, Li D, Zhao LY, Godsey A, Liao D CW, Detter A, Fang R, Faca V, Gafken P, Zhang
(2013) p53 SUMOylation promotes its H, Whiteaker J, States D, Hanash S, Paulovich
nuclear export by facilitating its release from A, McIntosh MW (2006) Computational pro-
the nuclear export receptor CRM1. Mol Biol teomics analysis system (CPAS): an extensible,
Cell 24(17):2739–2752. doi:10.1091/mbc. open-source analytic system for evaluating and
E12-10-0771 publishing proteomic data and high through-
13. Ohtani N, Brennan P, Gaubatz S, Sanij E, put biological experiments. J Proteome Res
Hertzog P, Wolvetang E, Ghysdael J, Rowe M, 5(1):112–121. doi:10.1021/pr0503533
Hara E (2003) Epstein-Barr virus LMP1 blocks 24. Keller A, Nesvizhskii AI, Kolker E, Aebersold
p16INK4a-RB pathway by promoting nuclear R (2002) Empirical statistical model to esti-
export of E2F4/5. J Cell Biol 162(2):173– mate the accuracy of peptide identifications
183. doi:10.1083/jcb.200302085 made by MS/MS and database search. Anal
14. Yang H, Zhao R, Yang HY, Lee MH (2005) Chem 74(20):5383–5392
Constitutively active FOXO4 inhibits Akt 25. Nesvizhskii AI, Keller A, Kolker E, Aebersold
activity, regulates p27 Kip1 stability, and sup- R (2003) A statistical model for identifying
presses HER2-mediated tumorigenicity. proteins by tandem mass spectrometry. Anal
Oncogene 24(11):1924–1935. doi:10.1038/ Chem 75(17):4646–4658
sj.onc.1208352 26. Faca VM, Ventura AP, Fitzgibbon MP, Pereira-
15. Hu MC, Lee DF, Xia W, Golfman LS, Ou-Yang Faca SR, Pitteri SJ, Green AE, Ireton RC,
F, Yang JY, Zou Y, Bao S, Hanada N, Saso H, Zhang Q, Wang H, O’Briant KC, Drescher
Kobayashi R, Hung MC (2004) IkappaB kinase CW, Schummer M, McIntosh MW, Knudsen
promotes tumorigenesis through inhibition of BS, Hanash SM (2008) Proteomic analysis of
forkhead FOXO3a. Cell 117(2):225–237 ovarian cancer cells reveals dynamic processes
16. Alt JR, Gladden AB, Diehl JA (2002) of protein secretion and shedding of extra-
p21(Cip1) Promotes cyclin D1 nuclear accu- cellular domains. PLoS One 3(6):e2425.
mulation via direct inhibition of nuclear doi:10.1371/journal.pone.0002425
Chapter 5
Optimized Enrichment of Phosphoproteomes

by Fe-IMAC Column Chromatography
Benjamin Ruprecht, Heiner Koch, Petra Domasinska, Martin Frejno,
Bernhard Kuster, and Simone Lemeer
Abstract
Phosphorylation is among the most important post-translational modifications of proteins and has numer-
ous regulatory functions across all domains of life. However, phosphorylation is often substoichiometric,
requiring selective and sensitive methods to enrich phosphorylated peptides from complex cellular digests.
Various methods have been devised for this purpose and we have recently described a Fe-IMAC HPLC
column chromatography setup which is capable of comprehensive, reproducible, and selective enrichment
of phosphopeptides out of complex peptide mixtures. In contrast to other formats such as StageTips or
batch incubations using TiO2 or Ti-IMAC beads, Fe-IMAC HPLC columns do not suffer from issues
regarding incomplete phosphopeptide binding or elution and enrichment efficiency scales linearly with the
amount of starting material. Here, we provide a step-by-step protocol for the entire phosphopeptide
enrichment procedure including sample preparation (lysis, digestion, desalting), Fe-IMAC column chro-
matography (column setup, operation, charging), measurement by LC-MS/MS (nHPLC gradient, MS
parameters) and data analysis (MaxQuant). To increase throughput, we have optimized several key steps
such as the gradient time of the Fe-IMAC separation (15 min per enrichment), the number of consecutive
enrichments possible between two chargings (>20) and the column recharging itself (<1 h). We show that
the application of this protocol enables the selective (>90 %) identification of more than 10,000 unique
phosphopeptides from 1 mg of HeLa digest within 2 h of measurement time (Q Exactive Plus).
Key words Phosphorylation, Proteomics, Phosphocapture, LC-MS
Abbreviations
ACN Acetonitrile
AGC Acquisition gain control
CAA Chloroacetamide
DTT Dithiothreitol
FA Formic acid
FCS Fetal calf serum
HCD Higher energy collision induced dissociation
HCl Hydrochloride
HPLC High-performance liquid chromatography
47
48 Benjamin Ruprecht et al.
I.D. Inner diameter

IMAC Immobilized metal ion affinity chromatography
MeOH Methanol
MS Mass spectrometry
MS/MS Tandem mass spectrometry
PBS Phosphate buffered saline
Ppm parts per million
PSM Peptide spectrum match
pY/pS/Pt Phosphotyrosine, -serine, -threonine
TFA Trifluoro acetic acid
TiO2 Titanium dioxide
Tris Tris(hydroxymethyl)aminomethane
v/v Volume/volume
w/w Weight/weight
ZrO2 Zirconium dioxide
1 Introduction
Reversible protein phosphorylation is a posttranslational modifica-

tion that plays a key role in signal transduction and aberrant regu-
lation has been implicated in a number of diseases [1]. As a
consequence, mass spectrometry-based large-scale identification of
phosphorylation events has received considerable attention over
the last years. Due to the low abundance and sub stoichiometric
levels of the phosphorylation events, enrichment of phosphopep-
tides or phosphoproteins is required prior to mass spectrometric
detection [2, 3]. In the field of phosphoproteomics, various large-
scale enrichment strategies have been developed over the last years,
all with their own strengths and weaknesses. The most widely used
enrichment strategies utilize the affinity of phosphate groups to
metal ions (immobilized on a solid support). Examples of such
strategies are metal oxide affinity chromatography (TiO2, ZrO2)
[4, 5], immobilized metal ion affinity chromatography (IMAC)
with different metal ions (Fe3+, Ga3+, or Zr4+) [6–8] and Ti-IMAC
[9]. It is thought that there is a high degree of complementarity
between these different enrichment materials [9–12]; however, we
have recently shown that such complementarity can be attributed
to the format of the enrichment, the inefficient elution from the
material, as well as the insufficient acquisition speed of the mass
spectrometer, rather than the material [13]. To date, most phos-
phopeptide enrichments are still performed in either batch mode
or in micro-column format with the material packed in gel-loader
tips. These formats suffer from a high degree of variability due to
the multitude of manual handling steps in these protocols. In addi-
tion, variability is further increased by the use of, for example, dif-
ferent loading solvents, different washing procedures and
incubation times [14, 15]. More importantly, the enrichment effi-
ciency and selectivity is largely dependent on the so called
Phosphoproteome Enrichment by Fe-IMAC 49
bead-to-sample ratio [16, 17]. Consequently, batch- and tip-based

enrichment strategies require optimization for each sample. Here,
we describe the workflow for the reproducible and comprehensive
enrichment of phosphopeptides using Fe-IMAC HPLC columns
that overcomes most of these issues. As reported previously, the
Fe-IMAC column does not suffer from bead-to-sample ratio issues
and allows for the comprehensive depletion of phosphopeptides
from digests without showing any bias in the type of phosphopep-
tides that are enriched [13]. The protocol describes the entire
workflow starting from sample preparation to data analysis.
Moreover, it includes several improvements over the published
method such as shortened gradient length and improved column
recharging, ultimately resulting in considerably increased through-
put. We provide a detailed description of the column setup and
operation (including charging of the columns and gradients) and
describe guidelines for monitoring column performance. Finally,
we apply the protocol to the enrichment phosphopeptides from
1 mg cell line digest which led to the identification of >10,000
unique phosphopeptides in 2 h of measurement time.
2 Material
Unless stated otherwise, all solvents and solutions are prepared

fresh, using ultrapure water and analytical grade reagents. Devices
such as centrifuges, vacuum centrifuges/lyophilizer, thermoshaker,
or refrigerators (−20 °C/−80 °C) are not explicitly listed.
2.1 Preparation 1. Cell culture: RPMI 1640 medium supplemented with 10 %
of Proteome Digests fetal calve serum (FCS). Add 55 ml FCS to 500 ml RPMI
for Phosphopeptide 1640 medium. Sterile phosphate buffered saline (PBS) with-
Analysis out calcium and magnesium. 150 × 20 mm cell culture dishes.
Cell scraper. HeLa S3 cervix carcinoma cells (DMSZ,
Braunschweig, Germany).
2. Lysis buffer: Prepare a 40 mM Tris(hydroxymethyl)amino-
methane (Tris)–HCl, pH = 7.6 solution, containing 8 M urea,
protease and phosphatase inhibitors. Prepare a stock solution of
2 M Tris–HCl by dissolving 2.42 g Tris in 5 ml water, adjust the
pH to 7.6 using a 5 M HCl solution and fill up to 10 ml with
water. A 100 fold stock solution of phosphatase inhibitor cock-
tail 1, 2, and 3 is commercially available (Sigma Aldrich, Munich,
Germany). Add 4.8 g of urea to a 15 ml falcon tube, add 200 μl
Tris–HCl stock solution, add one protease inhibitor tablet com-
plete mini EDTA-free (Roche, Mannheim, Germany) and
100 μl of each phosphatase inhibitor stock solution and fill up
to 10 ml with water. Store the lysis buffer on ice.
3. Reducing agent: 1 M stock solution of dithiothreitol (DTT) in
water. Dissolve 1.54 g DTT in a falcon tube and fill up to 10 ml
with water. Prepare 200 μl aliquots and store the reducing agent
at −20 °C.
4. Alkylating agent: 550 mM stock solution of CAA (CAA) in
water. Dissolve 514 mg CAA in a falcon tube and fill up to
10 ml with water. Prepare 200 μl aliquots and store the alkyl-
ating agent at −20 °C.
5. 40 mM Tris–HCl solution, pH 7.6: Prepare a stock solution
of 2 M Tris–HCl by dissolving 2.42 g Tris in 5 ml water.
Adjust the pH to 7.6 using a 5 M HCl solution and fill up to
10 ml with water. Take 200 μl of the 2 M stock solution and
fill up to 10 ml with water.
6. Trypsin stock solution: Prepare a stock solution of 1 μg/μl
trypsin (sequencing grade modified trypsin, Promega) in
50 mM acetic acid, store at −80 °C.
7. Sep-Pak C-18 peptide purification: 50 mg Sep-Pak cartridges
(Waters Corp., Eschborn, Germany). Solvent A: 0.07 % (v/v)
TFA in water. Dilute 70 μl of 100 % TFA in 99.93 ml water.
Solvent B: 50 % (v/v) ACN, 0.07 % (v/v) TFA in water.
Prepare 50 ml ACN and 70 μl of TFA and dilute in 49.93 ml
water. Store at 4 °C.
8. Vacuum manifold for Sep-Pak desalting.
2.2 Phosphopeptide To avoid column clogging, solvents should be degassed and

Enrichment vacuum-filtered prior to use.
by Fe-IMAC Column
1. Formic acid solvent: 100 ml of 0.1 % FA (v/v) in water.
Chromatography
2. IMAC charging solvent: 250 ml of 25 mM FeCl3 (reagent
grade, Sigma-Aldrich, Product No. 157740) in 100 mM ace-
tic acid. Put 200 ml water into a cylinder, add 1.43 ml acetic
acid and fill up to 250 ml with water. Add 1.014 g FeCl3 and
use a magnetic stirrer to dissolve FeCl3. Leave the solution
stirring for 30 min and vacuum filter it afterwards to remove
insoluble FeCl3 remnants. Store the filtered solution at 4 °C.
3. IMAC stripping solvent: 250 ml of 50 mM EDTA in water,
pH 8. Use a magnetic stirrer to dissolve 3.653 g EDTA in
200 ml water. Add 5 M NaOH solution until the EDTA is
dissolved, adjust the pH to 8 and fill up to 250 ml with water.
Vacuum filter the solvent to remove insoluble EDTA remnants
and store at 4 °C.
4. IMAC loading solvent: 1 l of 0.07 % (v/v) TFA in 30 % (v/v)
ACN. Always prepare fresh.
5. IMAC elution solvent: 100 ml of 0.3 % (v/v) NH4OH in
water (caution, see Note 1).
6. ProPack IMAC-10 column: 10 μm, nonporous, polymeric
beads; 4 mm inner diameter ×
50 mm length (Thermo
Scientific, Product No. 063276).
7. HPLC system with the following requirements: flow rates

ranging from 0.1 ml/min to 4 ml/min; 0.5–1 ml sample loop;
UV detector set to read fixed wavelengths of 214 nm and
280 nm; stable at pH 2–12.
2.3 Desalting 1. Preparation of the C-18 StageTips: A detailed instruction on

of the Fe-IMAC Eluate how to construct micro-column tips is provided in Chapter 8.
Prepare a StageTip containing five C-18 disks (Empore
Octadecyl C-18 47 mm Solid Phase Extraction Disks #2215,
3 M Purification, Eagan, MN, USA).
2. Desalting solvents: Solvent A: prepare 5 ml of 0.07 % (v/v)
TFA in water. Solvent B: prepare 5 ml of 0.07 % (v/v) TFA
and 60 % (v/v) ACN in water.
2.4 LC-MS/MS 1. 50 mM citric acid and 1 % (v/v) FA in water. Dissolve 2.1 g of
and Data Analysis citric acid in 9.9 ml of water and add 100 μl of FA.
2. LC-MS/MS: nano-HPLC setup coupled to a high resolution
mass spectrometer. Here, we use an Eksigent NanoLC-Ultra
1D+ (Eksigent, Dublin, CA) coupled to a Q Exactive Plus
mass spectrometer (Thermo Scientific, Bremen). LC-trap col-
umn: 75 μm × 2 cm, packed with 5 μm Reprosil-Pur ODS-3
C-18 material (Dr. Maisch, Ammerbuch, Germany). Analytical
column: 75 μm × 42 cm, packed with 3 μm Reprosil-Gold
C-18 material (Dr. Maisch, Ammerbuch, Germany).
3. Nano-HPLC solvents: Loading solvent: 0.1 % (v/v) FA in
water. Solvent A: 0.1 % (v/v) FA and 5 % (v/v) DMSO [18] in
water. Solvent B: 0.1 % (v/v) FA and 5 % (v/v) DMSO in ACN.
4. Data analysis: Freely available MaxQuant [19] software pack-
age (e.g., version 1.5.2.8) with the integrated search engine
Andromeda [20]. Protein sequence database in FASTA format
(e.g., UniprotKB).
5. Spreadsheet editor or the freely available Perseus software
package.
3 Methods
3.1 Preparation A schematic overview of the experimental steps covered in this pro-
of Proteome Digests tocol is provided in Fig. 1.
for Phosphopeptide
1. Seed HeLa cells under sterile conditions in RPMI 1640
Analysis medium supplemented with 10 % FCS. Use 30 ml medium for
150 mm cell culture dishes. Grow the cells to 80 % confluency
under humidified atmosphere, 5 % CO2 at 37 °C. For lysis,
place the cell culture dishes on ice or work at 4 °C. Wash cells
two times with cold PBS. Use a pipette to aspirate residual
PBS from cell culture plates after the final washing step (see
Note 2).
Fig. 1 Experimental workflow for comprehensive phosphopeptide enrichment depicting cell culture and lysis,
phosphopeptide enrichment using the Fe-IMAC column, desalting of enriched phosphopeptides, and LC-MS/
MS analysis
2. Add 550 μl of precooled lysis buffer to cell culture dishes.

Carefully pan the dish to distribute the lysis buffer evenly over
all cells and incubate the dishes for 10 min on ice. Use a cell
scraper to mix cell lysate in the cell culture plates. Transfer cell
lysates to 1.5 ml reaction vessels and spin down insoluble
debris for 20 min at 21,000 × g and 4 °C. Transfer supernatant
to a new tube. Use a Bradford assay (or similar photometric
assay) to determine the protein concentration. Store lysates at
−80 °C or continue directly.
3. Use 1 mg of protein lysate. To reduce disulfide bonds, 1 M
DTT stock solution is added to a final concentration of 10 mM
(1:100 dilution). Incubate in a thermoshaker for 40 min at
37 °C and 700 rpm.
4. For alkylation of cysteine residues, add 550 mM CAA to a final
concentration of 50 mM (1 : 10 dilution). Carefully invert the
sample once and incubate for 30 min at room temperature in
the dark.
5. Dilute sample with four volumes of 40 mM Tris–HCl (pH 7.6)
to decrease urea concentration to 1.2 M (see Note 3). Add
trypsin in a protease-to-protein ratio of 1:100 (w/w) and pre-
digest 4 h in a thermo-shaker at 37 °C and 700 rpm. Add
another 1:100 (w/w) trypsin and incubate the digestion mix-
ture over night in a thermo-shaker at 37 °C and 700 rpm.
6. Cool samples down to room temperature and acidify the sam-
ple by addition of 0.5 % (v/v) TFA. Centrifuge acidified pep-
tides at 5000 × g to precipitate insoluble matter. Use 50 mg
Sep-Pak columns and place them into a vacuum manifold
(see Note 4). Prime Sep-Pak columns by adding 1 ml of sol-
vent B. Equilibrate column by adding 2 × 1 ml of solvent
A. Transfer the supernatant of the acidified sample to the col-
umn and load slowly (see Note 5). Reapply flow-through a
second time and discard it afterwards. Wash the column with
3 × 1 ml solvent A. Elute peptides with 2 × 150 μl solvent B
into a 1.5 ml reaction vessel. Adjust the volume to 0.5 ml by
addition of solvent A (see Note 6). The Sep-Pak eluate has a

final concentration of 30 % ACN and can thus be directly
applied to phosphopeptide enrichment (see Note 7) or alter-
natively stored at −80 °C.
3.2 Phosphopeptide For first time use, the column can be directly charged with FeCl3
Enrichment solvent (it does not have to be stripped) (see Note 8). The column
by Fe-IMAC Column is usually operated below 1000 psi. Column stripping and charging
Chromatography should be repeated after 20 enrichments or in case the column has
not been used for more than one week (see Note 9).
1. Column stripping: Connect the IMAC column to your HPLC
system and rinse it with ultrapure water (1 ml/min, 10 ml). Inject
1 ml of IMAC stripping solvent into the sample loop and let it
run through the column (1 ml/min). After 1 min, inject another
1 ml IMAC stripping solvent. Repeat this step eight more times.
Make sure that the sample loop is flushed with water afterwards.
Rinse the column with ultrapure water (2 ml/min, 5 ml).
2. Column charging: Inject 1 ml of IMAC charging solvent into
the sample loop and let it run over the column (0.2 ml/min).
After 5 min, inject another 1 ml of IMAC charging solvent.
Repeat this step four more times. Wash the sample loop with
3 × 1 ml EDTA to scavenge remaining Fe3+ ions (from both
syringe and sample loop) and 10 × 1 ml water to get rid of
residual EDTA. Rinse the column with 50 ml FA solvent to
wash away unbound Fe3+ ions (2 ml/min).
3. Fe-IMAC enrichment: Connect the IMAC loading solvent and
IMAC elution solvent to the HPLC system. Flush the column
with 5 ml of 50 % IMAC elution solvent (3 ml/min). Re-
equilibrate the column with 20 ml IMAC loading solvent (3 ml/
min) (see Note 10). Perform a standard enrichment to ensure
proper charging (15 min gradient, see Table 1, see Note 10).
Table 1
Settings for a 15 min Fe-IMAC column enrichment, displaying the programmed
time, the flow (in ml/min), and percentages IMAC elution solvent used
Time [min] Flow [ml/min] IMAC elution solvent [%]

0–0.1 1 0
0.1–5.1 0.2 0
5.1–6.72 3 0 - 16
6.72–11.68 0.55 16 - 26.25
11.68–12.35 3 26.25 - 50
12.35 0 50 - 0
12.35–15.02 3 0
Fig. 2 Typical chromatogram of a 15 min Fe-IMAC column enrichment using

1 mg of HeLa digest. The first peak (retention time between 2 and 5 min.) con-
tains the non-phosphorylated peptides. The second peak (retention time around
9 min) contains the phosphorylated peptides
4. Inject the desalted sample (see Note 11) dissolved in 0.5 ml

IMAC loading solvent. Prepare two 1.5 ml reaction vessels for
collecting flow-through and phosphopeptide eluate. Start the
15 min gradient (the gradient setup is shown in Table 1).
Monitor the absorption at 280 nm (see Fig. 2) and collect
1.3 ml of the flow-through in a 1.5 ml eppendorf tube. The
IMAC eluate peak containing the phosphopeptides is col-
lected in another 1.5 ml eppendorf tube (total volume of 1 ml,
see Fig. 2 and Table 1). Freeze both eppendorf tubes at −80 °C
and subsequently dry the samples down using a vacuum cen-
trifuge or a lyophilizer. A typical 15 min Fe-IMAC enrichment
chromatogram is depicted in Fig. 2.
5. There is no need to run blanks in between consecutive enrich-
ments. The carryover is minimal. You can reinject the Fe-IMAC
column flow-through in order to monitor (absorption at 214
and 280 nm) if all the phosphopeptides were properly depleted.
3.3 Desalting Although most of the ammonia will evaporate during the vacuum
of the Fe-IMAC Eluate centrifugation/lyophilization step, residual ammonia salts might
remain. Hence, it is recommended to desalt Fe-IMAC eluates
using C-18 StageTips [21]. Pass all liquids through the tips by
centrifugation (~800 × g, room temperature; see Note 12).
1. Dissolve the dried sample in 250 μl of solvent A and keep the
sample on ice while the StageTips are prepared. Check the pH
of the dissolved peptide solution and, if required, adjust it to
pH 2 using FA.
2. Sequentially activate the tips using 250 μl of MeOH, 250 μl of
solvent B and equilibrate with 250 μl of solvent A. Empty the
eppendorf tube in between.
3. Load the dissolved sample and reapply the flow-through.

Discard the flow-through afterwards and wash the column
with 250 μl solvent A.
4. Use 40 μl of solvent B to elute the peptides of the C-18 mate-
rial. Transfer the eluate into a 96-well plate and dry the sample
down using a vacuum centrifuge/lyophilizer. At this point,
the plate can be stored at −20 °C.
3.4 LC-MS/MS 1. Reconstitute desalted IMAC eluate in 20 μl of 1 % FA in

and Data Analysis 50 mM citrate (see Note 13).
2. Perform LC-MS/MS measurements by coupling an Eksigent
NanoLC-Ultra 1D+ to a Q Exactive Plus instrument. 5 μl of
IMAC enriched phosphopeptides corresponding to the
enrichment from 250 μg peptide digest are delivered to the
trap column at a flow rate of 5 μl/min in loading solvent (0.1 %
FA in water). During 10 min of sample loading chelated iron
is washed out while phosphopeptides are retained.
3. Transfer peptides to the analytical column and separate at a
flow rate of 300 nl/min using a 110 min gradient from 0 to
27 % solvent B (0–2 min: 0 % B; 2–100 min: 0–27 % C; 100–
101 min 27–80 % B, 101–105 min: 80 % B, 105–106 min:
80–0 % B, 106–110 min: 0 % B) (see Note 14).
4. Operate the Q Exactive Plus in data-dependent mode, auto-
matically switching between MS1 and MS2. Acquire full-scan
MS spectra at 360–1300 m/z, 70,000 resolution with acquisi-
tion gain control (AGC) target value of 3 × 106 charges and
maximum injection time of 100 ms for MS1. Allow up to 20
precursor ions for HCD fragmentation in tandem mass spectra.
Acquire MS2 spectra at 17,500 resolution, AGC target value of
1 × 105 charges and max injection time of 50 ms (see Note 15).
Set precursor ion isolation width to 1.7 Th and dynamic exclu-
sion to 20 s. Figure 3 shows an expected MS1 base peak inten-
sity chromatogram of the LC-MS/MS measurement.
Fig. 3 Typical base peak intensity chromatogram of a desalted Fe-IMAC phospho-

peptide eluate (1/4 of a 1 mg HeLa enrichment) measured on a Q Exactive Plus
5. Analyze data using a proteomics software capable of label-free

quantification. All results shown in this chapter are based on
peptide identifications by search of raw data against the
UniProtKB human database, version July 2013 (88,354
sequences) using the freely available MaxQuant version 1.5.2.8
and its built-in Andromeda search engine. Parameters used are
specified in Table 2.
6. To facilitate any kind of data analysis, filter the MaxQuant evi-
dence.txt or the phospho(STY)sites.txt output file to remove
reverse sequences and potential contaminants. To determine
the selectivity of the phosphopeptide enrichment, the reported
number of peptides annotated with a phosphorylation sites is
divided by the total number of identified sequences. The
intensity-based selectivity is acquired similarly by dividing the
summed intensity of phosphorylated peptides by the total
intensity. Filter the “Modified Sequence” column for dupli-
cates to remove redundancies and obtain the number of
Table 2
Group-specific and global parameters for data analysis using MaxQuant version 1.5.2.8
Group-specific parameters
Type Standard
Label No
Variable modifications Acetyl (Protein N-term), Oxidation (M), Phospho (STY)
Digestion mode Specific (Trypsin/P)
Max. missed cleavages 2
Main search peptide tolerance 5 parts per million (ppm)
Max. number of modifications per peptide 5
Global parameters
Database UniProtKB
Fixed modifications Carbamidomethyl
PSM FDR 0.01
Protein FDR 0.01
Site decoy fraction 0.01
Min. peptide length 7
Min. score for (un)modified peptides 0
Min. delta score for (un)modified peptides 0
MS/MS match tolerance 20 ppm
Second peptide search Enabled
Table 3
Overview of results typically expected from a single Fe-IMAC enrichment,
measured on a 2 h LC-MS/MS gradient on a Q Exactive Plus. The Fe-IMAC
column eluate was reconstituted in 20 μl of 50 mM Citrate, 1 % FA and
5 μl were injected
Phosphopeptides (MaxQuant—evidence.txt)
Identified unique phosphopeptides 10089
Quantified unique phosphopeptides 9392
Mono phosphorylated 8392 (83 %)
Multiply phosphorylated 1697 (17 %)
Identification-based phosphopeptide selectivity 81 %
Intensity-based phosphopeptide selectivity 94 %
Phosphorylation sites (MaxQuant—Phospho(STY)sites.txt)
Identified phosphorylation sites 8973
Quantified phosphorylation sites 7451
Class I sites (Loc. prob. > 0.75) 6566
pS sites (class I) 5674 (86 %)
pT sites (class I) 727 (11 %)
pY sites (class I) 165 (3 %)
unique phosphopeptides. Similarly, the phospho (STY)sites.

txt is used to determine the number of unique and quantifi-
able sites. Filter for “Localization probability” ≥ 0.75 to obtain
the number of class I sites [22].
7. Table 3 shows expected results in terms of unique phospho-
peptides and phosphorylation sites obtained from processing
1 mg of HeLa digest (1/4 of the enrichment was subjected to
MS measurement) according to the procedures described in
this protocol.
4 Notes
1. We noticed that ammonia is evaporating if the NH4OH bot-

tle is not tightly sealed or if it has been opened and closed
repeatedly. Thus, the 0.3 % (v/v) in the IMAC elution solvent
refers to a freshly opened bottle of NH4OH and might have
to be adjusted upon prolonged use. This can be delayed by
ensuring proper sealing of the bottle or alternatively by work-
ing at 4 °C.
2. Aspiration of remaining PBS is important to avoid dilution of

the lysis buffer. Low concentrations of the chaotropic reagent
urea might result in insufficient protein denaturation.
3. Considering the additional volume of the lysed cells and the
remaining PBS, the concentration of urea is reduced from 8 M
to roughly 6 M. Hence, a 1:4 dilution is sufficient to reduce
the urea concentration down to 1.2 M.
4. Sep-Pak sorbent weight has to be chosen according to the
amount of digest you intend to load. As a rule of thumb, the
capacity of Sep-Pack cartridges equals 5 % of the sorbent
weight (e.g., 2.5 mg for the 50 mg sorbent weight cartridges
and 10 mg for the 200 mg sorbent weight cartridges).
5. Load the sample slowly onto the Sep-Pak column. Lower the
flow rate by adjusting the vacuum at the vacuum manifold.
Loading should take at least 10 min to ensure proper binding
of the phosphopeptides. Reapplying the flow-through
increases recovery.
6. Avoid letting the columns run dry.
7. Adjusting the volume to 0.5 ml using solvent A enables direct
sample loading onto the IMAC column without the need to
intermediately dry the sample down.
8. Upon first time use, the column should be thoroughly flushed
with water followed by 0.1 % (v/v) FA solvent. It is advisable
to note down the column backpressure for different types of
solvents. This facilitates to monitor column performance over
time. If the pressure increase is too severe, the column should
be exchanged.
9. From our experience, 20 enrichments can be performed with-
out any performance decrease. However, if the column is not
used for a longer period of time (conservatively more than one
week), column performance seems to be decreasing. If you
want to verify or monitor column performance, run a standard
before and after your enrichment set.
10. Make sure to properly equilibrate the column after charging.
The absorption at 280 nm has to have reached a stable base-
line. We recommend to always run a standard or a blank run
before you enrich your first sample.
11. The 4 mm I.D. column can be applied to sample amounts
ranging between 0.5 and 3 mg. Please be aware that the
enrichment efficiency is also cell line dependent as the degree
of cellular phosphorylation is highly dynamic and may there-
fore vary considerably.
12. Using volumes of 250 μl prevents columns from running dry
even upon prolonged centrifugation. This is especially benefi-
cial when parallelized fractionation is intended as not all col-
umns run at the same speed. If only one sample is intended to

be desalted, the procedure can be accelerated by manually
pushing the liquids through the tips using a 5 ml Eppendorf
CombiTip. The volumes can be scaled down accordingly
(~40 μl for each step).
13. Citric acid acts as a chelating agent for residual Fe3+ ions that
might co-elute from the Fe-IMAC column. Remaining Fe3+
ions can stick to the trap/analytical nano-HPLC columns and
deplete phosphopeptides. Ever since we use citrate we have
not detected any iron contamination [23]. If you are in doubt,
specify iron as a variable modification during data processing
and check if any iron-bound peptides are identified.
14. Phosphopeptides are generally more hydrophilic than non-
phosphopeptides. Compared to full proteome separations, we
use a shallow gradient which leads to a more efficient use of
gradient and MS time.
15. Phosphopeptides often show a pronounced loss of the phos-
phate group upon fragmentation. Because this neutral loss
peak may constitute a big part of the fragment ion intensity,
backbone fragment ions might get lost. Therefore you may
want to evaluate if increased MS2 injection time and or
increased MS2 target values (by a factor of 2) leads to an
increase in phosphopeptide identifications. The substantially
higher identification rate compensates for decreased scan
numbers. Moreover, we found that this AGC/injection time
increase is beneficial for phosphosite localization.
References
1. Lu Z, Jiang G, Blume-Jensen P et al (2001) 6. Andersson L, Porath J (1986) Isolation of
Epidermal growth factor-induced tumor cell phosphoproteins by immobilized metal (Fe3+)
invasion and metastasis initiated by dephos- affinity chromatography. Anal Biochem 154:
phorylation and downregulation of focal adhe- 250–254
sion kinase. Mol Cell Biol 21:4016–4031 7. Posewitz MC, Tempst P (1999) Immobilized
2. Ruprecht B, Lemeer S (2014) Proteomic anal- gallium(III) affinity chromatography of phos-
ysis of phosphorylation in cancer. Expert Rev phopeptides. Anal Chem 71:2883–2892
Proteomics 11:259–267 8. Zhou H, Xu S, Ye M et al (2006) Zirconium
3. Lemeer S, Heck AJ (2009) The phosphopro- phosphonate-modified porous silicon for
teomics data explosion. Curr Opin Chem Biol highly specific capture of phosphopeptides and
13:414–420 MALDI-TOF MS analysis. J Proteome Res
4. Pinkse MWH, Uitto PM, Hilhorst MJ et al 5:2431–2437
(2004) Selective Isolation at the femtomole 9. Zhou H, Low TY, Hennrich ML et al (2011)
level of phosphopeptides from proteolytic Enhancing the identification of phosphopep-
digests using 2D-nanoLC-ESI-MS/MS and tides from putative basophilic kinase substrates
titanium oxide precolumns. Anal Chem 76: using Ti (IV) based IMAC enrichment. Mol
3935–3943 Cell Proteomics 10:M110.006452
5. Kweon HK, Håkansson K (2006) Selective zir- 10. Bodenmiller B, Mueller LN, Mueller M et al
conium dioxide-based enrichment of phos- (2007) Reproducible isolation of distinct,
phorylated peptides for mass spectrometric overlapping segments of the phosphopro-
analysis. Anal Chem 78:1743–1749 teome. Nat Methods 4:231–237
11. Tsai C-F, Hsu C-C, Hung J-N et al (2014) 17. Zhou H, Di Palma S, Preisinger C et al (2013)
Sequential phosphoproteomic enrichment Toward a comprehensive characterization of a
through complementary metal-directed immo- human cancer cell phosphoproteome.
bilized metal ion affinity chromatography. Anal J Proteome Res 12:260–271
Chem 86:685–693 18. Hahne H, Pachl F, Ruprecht B et al (2013)
12. Thingholm TE, Jensen ON, Robinson PJ et al DMSO enhances electrospray response, boost-
(2008) SIMAC (sequential elution from ing sensitivity of proteomic experiments. Nat
IMAC), a phosphoproteomics strategy for the Methods 10:989–991
rapid separation of monophosphorylated from 19. Cox J, Mann M (2008) MaxQuant enables
multiply phosphorylated peptides. Mol Cell high peptide identification rates, individual-
Proteomics 7:661–671 ized p.p.b.-range mass accuracies and
13. Ruprecht B, Koch H, Medard G et al (2015) proteome- wide protein quantification. Nat
Comprehensive and reproducible phospho- Biotechnol 26:1367–1372
peptide enrichment using iron immobilized 20. Cox J, Neuhauser N, Michalski A et al (2011)
metal ion affinity chromatography (Fe-IMAC) Andromeda: a peptide search engine inte-
columns. Mol Cell Proteomics 14:205–215 grated into the MaxQuant environment.
14. Larsen MR, Thingholm TE, Jensen ON et al J Proteome Res 10:1794–1805
(2005) Highly selective enrichment of phos- 21. Rappsilber J, Mann M, Ishihama Y (2007)
phorylated peptides from peptide mixtures Protocol for micro-purification, enrichment,
using titanium dioxide microcolumns. Mol pre-fractionation and storage of peptides for
Cell Proteomics 4:873–886 proteomics using StageTips. Nat Protoc
15. Kettenbach AN, Gerber SA (2011) Rapid and 2:1896–1906
reproducible single-stage phosphopeptide 22. Olsen JV, Blagoev B, Gnad F et al (2006)
enrichment of complex peptide mixtures: Global, in vivo, and site-specific phosphoryla-
application to general and phosphotyrosine- tion dynamics in signaling networks. Cell
specific phosphoproteomics experiments. Anal 127:635–648
Chem 83:7635–7644 23. Winter D, Seidler J, Ziv Y et al (2009) Citrate
16. Li Q, Ning Z, Tang J et al (2009) Effect of boosts the performance of phosphopeptide
peptide-to-TiO2 beads ratio on phosphopep- analysis by UPLC-ESI-MS/MS. J Proteome
tide enrichment selectivity. J Proteome Res Res 8:418–424
8:5375–5381
Chapter 6
Full Membrane Protein Coverage Digestion

and Quantitative Bottom-Up Mass Spectrometry
Proteomics
Joseph Capri and Julian P. Whitelegge
Abstract
A true and accurate bottom-up global proteomic measurement will only be achieved when all proteins in
a sample can be digested efficiently and at least some peptides recovered on which to base an estimate of
abundance. Integral membrane proteins make up around one-third of the proteome and require special-
ized protocols if they are to be successfully solubilized for efficient digestion by the enzymes used in bot-
tom-up proteomics. The protocol described relies upon solubilization using the detergents sodium
deoxycholate and lauryl sarcosine with heating to 95 °C. A subset of peptides is purified by reverse-phase
solid-phase extraction and fractionated by strong-cation exchange prior to nano-liquid chromatography
with data-dependent tandem mass spectrometry. For quantitative proteomics experiments a protocol is
described for stable-isotope coding of peptides using dimethylation of primary amines allowing for three-
way sample multiplexing.
Key words Trypsin, Electrospray ionization, Proteome, StageTip, Dimethylation, Phase transfer
1 Introduction
Integral membrane proteins make up around one-third of the

global proteome and a larger proportion of drug targets because of
their wide-ranging critical functions in cell biology. A global pro-
teomics method for measuring changes in protein expression level
in different conditions must fully cover the membrane proteome.
Researchers have battled this challenge with a variety of approaches
over the years and detergent solubilization prior to proteolytic
digestion has emerged as the most efficacious. Central to the suc-
cess of the method lies efficient digestion of integral membrane
proteins enabling recovery of loop-region peptides for quantitative
bottom-up proteomics with transmembrane domains typically
ignored since they are lost during peptide work up prior to analysis
as a result of their hydrophobic properties. Analysis of membrane
protein posttranslational modifications (PTMs) is accommodated
61
62 Joseph Capri and Julian P. Whitelegge
when they are localized to peptides recovered from loop regions.

Modifications in transmembrane domains require alternative
approaches such as top-down mass spectrometry [1, 2].
Two approaches to detergent solubilization have yielded the
most complete coverage of the integral membrane proteome;
filter-
assisted sample preparation (FASP) was described by the
Liebler group [3] and refined by Mann’s group [4] relying upon
sodium dodecylsulfate (SDS) for solubilization with subsequent
removal of detergent through a filter using urea followed by enzy-
matic digestion. In-solution digestion (ISD) relies upon detergent
solubilization using detergents that are tolerated by the proteolytic
enzyme used for digestion with subsequent removal of detergent
via acid precipitation or phase-transfer after digestion [5]. Masuda’s
work first described the use of sodium deoxycholate–lauryl sarco-
sine mixtures with heating to achieve efficient digestion of mem-
brane proteins [5]. Loo’s group combined FASP with ISD
replacing urea with deoxycholate to remove SDS [6]. It is gener-
ally agreed that use of FASP results in some selective sample loss on
the filter used and the ISD technique is gaining prevalence [7, 8].
The ISD protocol described herein is based upon the efficiency
of membrane protein solubilization with 0.5 % sodium deoxycho-
late and 12 mM lauryl sarcosine, with heating to 95 °C [5, 8]. This
mixture is highly denaturing at high-temperature but becomes
non-denaturing at 37 °C for trypsin digestion.
Occasionally a protein’s amino acid sequence results in failure
to detect a membrane protein. Acidic proteins may lack suitable
basic residues for trypsin cleavage, while basic proteins may have
too many. In such cases a second proteolytic enzyme with different
specificity must be used for complete coverage proteomics.
2 Materials
Prepare all solutions using ultrapure water (prepared by purifying

deionized water to attain a conductivity of 18 MΩ cm at 25 °C)
and analytical grade reagents. All reagents should be stored at
room temperature unless otherwise noted.
1. Cell lysis buffer: 0.5 % sodium deoxycholate, 12 mM
N-laurylsarcosinate sodium, and 50 mM ammonium bicar-
bonate pH 8.5 with 89 μg/mL of Sigma Protease Inhibitor
Cocktail powder (catalog number: P2714) Store at 4 °C. Only
store for a maximum of 2 weeks. Lysis buffer that does not
contain protease inhibitor and can be stored at room tempera-
ture for several months.
2. Reduction buffer stock solution: 1 M tris(2-carboxyethyl)
phosphine in water.
3. Alkylation stock solution: 1 mM iodoacetamide in water.
Full Membrane Protein Coverage 63
4. Bicinchoninic acid protein assay (Pierce).

5. Digestion buffer: 50 mM ammonium bicarbonate, pH 8.5.
6. Sequencing grade trypsin.
8. 200 mg tC18 Sep-Pak cartridges (Waters).
9. HPLC grade methanol.
10. Sep-Pak elution buffer: 80 % acetonitrile with 0.1 % trifluoro-
acetic acid.
11. Sep-Pak loading buffer: 2 % acetonitrile with 0.1 % trifluoro-
acetic acid.
12. Dimethyl labeling solutions:
Prewash: 250 mM 2-(N-morpholino)ethanesulfonic acid
pH 5.5
Light: 60 mM sodium cyanoborohydride, 0.4 % formaldehyde,
and 250 mM 2-(N-morpholino)ethanesulfonic acid pH 5.5
Intermediate: 60 mM sodium cyanoborohydride, 0.4 % form-
aldehyde (CD2O), and 250 mM 2-(N-morpholino)ethanesul-
fonic acid pH 5.5
Heavy: 60 mM sodium cyanoborodeuteride, 0.4 % formalde-
hyde (13CD2O), and 250 mM 2-(N-morpholino)ethanesul-
fonic acid pH 5.5.
13. 15 mL Conical vials.
14. Pierce Quantitative Colorimetric Peptide Assay (Thermo, cat-
alog #23275).
15. StageTip loading buffer: 2 % acetonitrile with 0.5 % acetic acid.
16. StageTip elution buffer: 80 % acetonitrile with 0.5 % acetic

acid.
17. Strong cation exchange (SCX) elution buffers: 30 % acetoni-
trile with 0.5 % acetic acid and increasing amounts of ammo-
nium acetate (NH4AcO): (1) 25 mM NH4AcO, (2) 35 mM
NH4AcO, (3) 50 mM NH4AcO, (4) 70 mM NH4AcO, (5)
100 mM NH4AcO, (6) 150 mM NH4AcO, (7) 350 mM
NH4AcO, and (8) 750 mM NH4AcO.
18. C18-SCX StageTips are made according to Rappsilber et al
[9]. Briefly, a Hamilton 16 gauge blunt-ended needle is used
to puncture chromatographic filters and seated into a P200
pipet tip. Two SCX frits (Empore™ Cation 47 mm Extraction
Disc, Model 2251, Millipore) are seated first followed by two
C18 frits (Empore™ C18 47 mm Extraction Disc, Model
2215, Millipore) seated above.
19. For every C18-SCX StageTip, eight C18 StageTips need to be
made in order to desalt the SCX fractions. These C18 StageTips
are made similarily to step 18, except only one C18 frit is
seated in the P200 pipet tip.
20. DDA capable mass spectrometer with associated nano flow

HPLC, we use a Thermo Orbitrap XL with an Eksigent 2D
nanoLC and Spark autosampler.
21. Nano capillary columns, 25 cm × 75 μm, packed with C18
(300Å, 3 μm particle size) resin.
22. NanoLC, mobile phase A: 3 % acetonitrile, 3 % dimethylsulfox-
ide, and 0.1 % formic acid.
23. NanoLC, mobile phase B: 97 % acetonitrile, 3 % dimethylsulf-
oxide, and 0.1 % formic acid.
24. MaxQuant analysis software.
3 Methods
3.1 Cell Lysis 1. Adherent cells are washed twice directly on plate with ice-cold
and Proteolytic Digest PBS pH 7.6 (see Note 1).
2. 0.5 mL per 1 × 107 cells of cell lysis buffer is added directly to
plate, cells are scraped with a cell scraper, and lysates are tritu-
rated with P1000 pipettor.
3. Cell lysates are transferred to 1.5 mL Eppendorf lo-bind
microcentrifuge tubes, water bath sonicated at RT for 5 min,
and heated at 95 °C for 5 min.
4. Bicinchoninic acid protein assay (Pierce) is performed to
determine protein concentration.
5. Disulfide bridges are reduced with 5 mM tris(2-carboxyethyl)
phosphine (final concentration) at RT for 30 min and subse-
quently alkylated with 10 mM iodoacetamide (final concentra-
tion) at RT in the dark for 30 min.
6. Cell lysates are transferred to 15 mL Falcon tubes and diluted
1:5 (v:v) with 50 mM ammonium bicarbonate pH 8.5.
7. Proteins are digested with sequencing grade trypsin 1:100
(enzyme:protein by mass) for 4 h at 37 °C under gentle agita-
tion followed by a second aliquot of trypsin 1:100
(enzyme:protein) overnight at 37 °C under gentle agitation.
8. Samples are acidified with 0.5 % trifluoroacetic acid (final con-
centration), vortexed rapidly for 5 min, and centrifuged at
16,000 × g for 5 min at RT to pellet sodium deoxycholate.
9. Transfer supernatant to a new tube and proceed to peptide
desalting. If needing to store for up to a week, keep peptide
samples at 4 °C, otherwise freeze at −80 °C.
3.2 Peptide 1. 200 mg tC18 Sep-Pak cartridges (Waters) are wetted with
Desalting 2 mL of 100 % methanol, with solvent pulled through the car-
and Reductive tridge using a vacuum manifold. It is critical to stop the flow
Dimethylation before all solvent has passed through to prevent any air from
entering the packing material (this applies for all subsequent

steps for Sep-Paks). Leaving ~100 μL of solvent above the
packing material is ideal.
2. 1 mL of Sep-Pak elution buffer is passed under vacuum and
repeated 1×.
3. 1 mL of Sep-Pak loading buffer is passed under vacuum and
repeated 2×.
4. Peptide digests are loaded onto Sep-Paks via gravity.
5. 1 mL of 250 mM 2-(N-morpholino)ethanesulfonic acid
pH 5.5 is passed under vacuum [10].
6. 3 mL of the respective dimethyl labeling solution is passed
through. This process needs to take at least 10 min to ensure
complete labeling. This can be accomplished by passing solu-
tion by gravity.
7. 1 mL of Sep-Pak loading buffer is passed under vacuum and
repeated 1×.
8. 1 mL of Sep-Pak elution buffer is passed under gravity, col-
lected in 15 mL conical vial, and repeated 1×.
9. Dimethyl-labeled peptide samples are lyophilized to dryness.
3.3 Strong Cation For each of the following steps, C18-SCX StageTips will be
Exchange denoted (S) and C18 StageTips will be denoted (C) if the follow-
Fractionation ing step is to be performed on that particular StageTip. Unless
otherwise noted, all solvent can be discarded properly (see Note 2).
1. Pierce Quantitative Colorimetric Peptide Assay is performed.
2. Peptide samples are reconstituted in StageTip loading buffer
at a concentration of 0.2 mg/mL.
3. Light, medium, and heavy labeled peptides are mixed 1:1:1.
4. (S,C) StageTips are wetted with 20 μL of 100 % methanol,
pushing solvent through by applying pressure with hand
syringe. It is critical to prevent air from entering frits, leaving
~1–2 μL above frits.
5. (S,C) 20 μL of StageTip elution buffer is passed through using
pressure from hand syringe.
6. (S,C) 20 μL of StageTip loading buffer is passed through
using pressure from hand syringe.
7. (C) 100 μL of 0.5 % acetic acid is deposited into C18 StageTip
and set aside for later use.
8. (S) 20 μL of SCX elution buffer 8 is passed through using
9. (S) 20 μL of StageTip loading buffer is passed through using
10. (S) 32 μg of differential-labeled peptides are loaded to C18-
SCX StageTip using pressure from hand syringe.
11. (S) 20 μL of StageTip loading buffer are passed through using
12. (S) 20 μL of StageTip elution buffer is passed through using
13. (S) 20 μL of 30 % acetonitrile with 0.5 % acetic acid is passed
through using pressure from hand syringe.
14. (S) 20 μL of SCX elution buffer 1 is passed through using
pressure from hand syringe and collected in 100 μL of pre-
deposited 0.5 % acetic acid above pre-conditioned C18
StageTip from step 7. This is repeated for SCX elution buffers
2–8 and collected into separate pre-conditioned C18 StageTips
from step 7.
15. (C) SCX fractions are pipetted up and down to mix with pre-
deposited 100 μL of 0.5 % acetic acid and then passed through
C18 using pressure from hand syringe.
16. (C) 20 μL of StageTip loading buffer is passed through using
17. (C) 20 μL of StageTip elution buffer is passed through using
pressure from hand syringe and collected in 1.5 mL microcen-
trifuge tubes.
18. Peptide fractions are concentrated in vacuum centrifuge to
~2 μL, typically 4 min.
19. Concentrated peptide fractions are reconstituted with 10 uL
of 2 % acetonitrile with 0.1 % formic acid and transferred to
autosampler injection vials.
3.4 NanoLC- 5 μL of each peptide fraction is analyzed using 180 min data-

RP-MS/ MS dependent reverse-phase nLC-MS/MS on Thermo Orbitrap XL
equipped with Eksigent Spark autosampler, Eksigent 2D nanoLC,
and Thermo nano-ESI source (see Note 3).
1. Samples are loaded onto a laser-pulled reverse-phase nanocap-
illary (75 μm I.D., 360 μm O.D. × 25 cm length) with C18
(300Å, 3 μm particle size) for 30 min with mobile phase A
(3 % a cetonitrile, 3 % dimethylsulfoxide, and 0.1 % formic acid)
at 600 nL/min.
2. Peptides are analyzed over 180 min linear gradient to 100 %
mobile phase B (97 % acetonitrile, 3 % dimethylsulfoxide, and
0.1 % formic acid) at 300 nL/min.
3. Electrospray ionization and source parameters are as follows:
spray voltage of 2.2 kV, capillary temperature of 200 °C, capil-
lary voltage at 35 V, and tube lens at 90 V.
4. Data-dependent MS/MS is operated using the following
parameters: full MS from 400 to 1700 m/z with 60,000 reso-
lution at 400 m/z and target ion count of 3 × 105 or fill time
of 700 ms with lock-mass at 401.922718 m/z, and 12 MS/
MS with charge-state screening excluding +1 and unassigned
charge states for ions surpassing 6000 counts, target ion count
of 5000 or fill time of 50 ms, CID collision energy of 35, and
dynamic exclusion of 30 s.
5. Raw data is searched against respective species Uniprot fasta
database using MaxQuant 1.5.3.30 with standard preset
search parameters. The search parameters are as follows: 3-plex
dimethyl labeling to lysine and peptide N-terminus, trypsin
cleavage allowing up to two missed cleavages, fixed modifica-
tion of carbamidomethyl to cysteines, variable modifications
of acetylation to protein N-terminus and methionine oxida-
tion, 10 ppm and 0.5 Da mass errors for Full MS and MS/
MS, respectively, 1 % false-discovery rate on peptide and pro-
tein identifications, and peptide match between run feature
with 1.5 min time window.
4 Notes
1. If proteomic samples are derived from tissue organs or organ-

isms with cell walls, a bead beater should be utilized to homog-
enize the sample in cell lysis buffer. After homogenization,
proceed with Subheading 3.1, step 3 and centrifuge at 16,000 × g
for 5 min at room temperature to pellet insoluble material.
2. Depending on sample complexity, more or less strong cation
exchange fractions can be performed.
3. Online reverse-phase chromatography, in terms of time and
gradient, will need to be optimized depending on sample
being analyzed and type of mass spectrometer acquiring data.
References
1. Ryan CM, Souda P et al (2010) Post- 6. Erde J, Loo RR, Loo JA (2014) Enhanced
translational modifications of integral mem- FASP (eFASP) to increase proteome coverage
brane proteins resolved by top-down Fourier and sample recovery for quantitative proteomic
transform mass spectrometry with collisionally experiments. J Proteome Res 13:1885–1895
activated dissociation. Mol Cell Proteomics 7. Kulak NA, Pichler G et al (2014) Minimal,
9:791–803 encapsulated proteomic-sample processing
2. Whitelegge JP (2013) Integral membrane pro- applied to copy-number estimation in eukary-
teins and bilayer proteomics. Anal Chem otic cells. Nat Methods 11:319–324
85:2558–2568 8. León IR, Schwämmle V et al (2013)
3. Manza LL, Stamer SL et al (2005) Sample Quantitative assessment of in-solution diges-
preparation and digestion for proteomic tion efficiency identifies optimal protocols for
analyses using spin filters. Proteomics 5:
unbiased protein analysis. Mol Cell Proteomics
1742–1745 12:2992–3005
4. Wiśniewski JR, Zougman A, Mann M (2009) 9. Rappsilber J, Mann M, Ishihama Y (2007)
Combination of FASP and StageTip-based Protocol for micro-purification, enrichment,
fractionation allows in-depth analysis of the pre-fractionation and storage of peptides for
hippocampal membrane proteome. J Proteome proteomics using StageTips. Nat Protoc
Res 8:5674–5678 2:1896–1906
5. Masuda T, Tomita M, Ishihama Y (2008)
10. Wilson-Grady JT, Haas W, Gygi SP (2013)
Phase transfer surfactant-aided trypsin diges- Quantitative comparison of the fasted and re-fed
tion for membrane proteome analysis. mouse liver phosphoproteomes using lower pH
J Proteome Res 7:731–740 reductive dimethylation. Methods 61:277–286
Chapter 7
Hydrophilic Strong Anion Exchange (hSAX)

Chromatography Enables Deep Fractionation
of Tissue Proteomes
Benjamin Ruprecht, Dongxue Wang, Riccardo Zenezini Chiozzi,
Li-Hua Li, Hannes Hahne, and Bernhard Kuster
Abstract
The bottom-up proteomic analysis of cell line and tissue samples to a depth > 10,000 proteins still repre-
sents a considerable challenge because of the sheer number of peptides generated by proteolytic digestions
and the high dynamic range of protein expression. As a result, comprehensive protein coverage requires
multidimensional peptide separation. Recently, off-line hydrophilic strong cation exchange (hSAX) chro-
matography has proven its merits for high resolution separation of peptides due to its high degree of
orthogonality to reversed-phase liquid chromatography. Here we describe the use of hSAX for the deep
analysis of tissue proteomes. The protocol includes optimized sample preparation steps (lysis with the aid
of mechanical disruption, one-step disulfide bridge reduction and alkylation), setup and operation of hSAX
columns and gradients, desalting of hSAX fractions prior to LC-MS/MS analysis, and suggestions for the
choice of data acquisition parameters and data analysis using MaxQuant. Application of the protocol to the
fractionation of 300 μg human brain tissue digest led to the identification of more than 100,000 unique
peptide sequences representing over 10,195 proteins and 9,500 genes in 3 days of measurement time on
a Q Exactive Plus mass spectrometer.
Key words Proteomics, Deep fractionation, Chromatography, Strong anion exchange, Tissue
proteomics
Abbreviations
ACN Acetonitrile
CAA Chloroacetamide
DTT Dithiothreitol
FA Formic acid
FDR False discovery rate
HCl Hydrochloric acid
hSAX Hydrophilic strong anion exchange
69

IT Injection time
MeOH Methanol
MS Mass spectrometer
PBS Phosphate buffered saline
RP Reversed-phase
SAX Strong anion exchange
SCX Strong cation exchange StageTip stop and go extraction tip
TCEP Tris-(2-carboxyethyl)-phosphin
TFA Trifluoroacetic acid
Tris Tris(hydroxymethyl)aminomethane v/v volume/volume
w/w Weight/weight
ZIC-HILIC Zwitterionic hydrophilic interaction liquid chromatography
1 Introduction
Despite the fact that tremendous progress has been made recently
in mapping out the human proteome [1, 2] and breathtaking
advances at all levels of proteomic sample preparation and mass
spectrometric instrumentation have been realized [3, 4], identifi-
cation and quantification of a single proteome to a depth > 10,000
proteins is still a considerable challenge. This has initially been
accomplished by two independent research groups in 2011 using
the cell lines HeLa [5] and U2OS [6], respectively. At the time,
this effort comprised the use of multiple enzymes and the analysis
of 72 fractions in 288 h of LC-MS/MS measurement time [5].
Such in-depth proteomic analysis was subsequently extended to a
greater number of cell lines [7, 8] but also tissue proteomes [9,
10], the latter of which represents an even greater challenge
because protein expression in tissues tends to span a broader
dynamic range than cell lines and the analysis is often complicated
by the presence of blood, fat or connective tissue. High-resolution
two-dimensional peptide separation/fractionation is an efficient
means to boost proteome coverage, sequence coverage and quan-
tification performance in bottom-up proteomics experiments.
Given that the stationary phase used as the second dimension sepa-
ration in nano-LC-MS/MS setups is almost exclusively comprised
of reversed-phase (RP) material, the first peptide separation dimen-
sion should ideally be orthogonal to RP and offer high chromato-
graphic resolution. Many different techniques have been employed
for this purpose such as HILIC [11], ZIC-HILIC [12], ERLIC
[13], WAX [14], high-pH reversed-phase [15], SAX [16], or SCX
[17]. Recently Ritorto et al. have demonstrated the merits of hSAX
chromatography [18], which separates peptides primarily based on
the number of acidic residues and the stationary phase is character-
ized by ultralow hydrophobicity. This combination enables
Hydrophilic Strong Anion Exchange (hSAX) Chromatography Enables Deep… 71
orthogonal and robust peptide fractionation with high resolution.

We have subsequently adapted the approach for the analysis of
phosphoproteomes, taking advantage of the very high retention of
negatively charged (phosphate) groups [19].
Here we describe the use of hSAX for the deep characterisation
of tissue proteomes. Importantly, with some minor modifications,
the procedure is equally applicable to the analysis of cell line digests
and phosphoproteome samples. The protocol includes sample
preparation steps, hSAX column setup and operation, StageTip
desalting of hSAX fractions, data acquisition parameters and
instructions on data analysis using MaxQuant [20]. We have opti-
mized lysis conditions (using a bead-beater), disulfide bond reduc-
tion (using Tris-(2-carboxyethyl)-phosphin (TCEP), [21]) and
alkylation (using chloroacetamide (CAA)) for improved protein
extraction and peptide coverage. Applied to the analysis of human
brain tissue, the described procedure enabled the identification
and quantification of more than 10,000 proteins in 3 days of mea-
surement time on a Q Exactive Plus mass spectrometer starting
from only a few mg of tissue.
2 Materials
Unless stated otherwise, all solvents and buffers should be pre-

pared fresh, using ultrapure water and analytical grade reagents.
Devices such as centrifuges, vacuum centrifuges/lyophilizer, ther-
moshaker, or refrigerators (−20/−80 °C) are not explicitly listed.
To avoid hSAX column clogging, solvents should be degassed and
vacuum-filtered prior to use.
2.1 Preparation 1.
Tissue preparation: Precellys 24 Homogenizer (Bertin
of Tissue Technologies, France), Precellys ceramic kit (1.4 mm “small”,
and Proteome Digest 0.5 ml tubes).
2. 550 mM CAA stock solution in water. Dissolve 514 mg CAA
in a falcon tube and fill up to 10 ml with water. Prepare 1 ml
aliquots and store at −20 °C.
3. 1 M TCEP-HCl stock solution in water: Dissolve 2.87 g of
TCEP-HCl in 10 ml of water. Prepare 25 μl aliquots and store
at −20 °C.
4. Tissue lysis solution (see Notes 1 and 2): 50 mM Tris–HCl,
pH = 7.6, containing 8 M urea, 10 mM TCEP-HCl, 40 mM
CAA, protease and phosphatase inhibitors. A 100 fold stock
solution of phosphatase inhibitor cocktail 1, 2, and 3 is com-
mercially available (Sigma Aldrich, Munich, Germany). Prepare
a stock solution of 2 M Tris–HCl by dissolving 2.42 g Tris in
5 ml water. Adjust the pH to 7.6 using HCl and fill up to
10 ml with water. Transfer 4.8 g of urea to a 15 ml falcon tube.
Add 250 μl Tris–HCl stock solutions, one protease inhibitor
tablet complete mini EDTA-free (Roche, Mannheim,

Germany), 100 μl of each phosphatase inhibitor stock solu-
tion, 726 μl of the 550 mM CAA stock solution and 10 μl of
the 1 M TCEP-HCl stock solution. Fill up to 10 ml with water.
Store the tissue lysis solution on ice.
5. Trypsin stock solution: Prepare a stock solution of 1 μg/μl
trypsin (sequencing grade modified trypsin, Promega) in
50 mM acetic acid. The trypsin stock can be reused several
times and is stored at −80 °C.
6. 50 mM Tris–HCl solution, pH 7.6: Prepare a stock solution of
2 M Tris–HCl by dissolving 2.42 g Tris in 5 ml water. Adjust
the pH to 7.6 using HCl and fill up to 10 ml with water. To
obtain a 50 mM Tris–HCl solution, transfer 200 μl of the 2 M
stock solution to a new 15 ml falcon tube and fill up to 10 ml
with water.
7. Sep-Pak C18 peptide purification: 50 mg Sep-Pak cartridges
(Waters Corp., Eschborn, Germany). Solvent A: 0.1 % (v/v)
FA in water. Solvent B: 60 % (v/v) ACN, 0.1 % (v/v) TFA in
water. Store at 4 °C.
2.2 hSAX 1. hSAX solvent A: 5 mM Tris, pH 8.5. Fill 900 ml water in a
Chromatography graduated 1 l cylinder. Use a magnetic stirrer to dissolve
0.606 g of Tris(hydroxymethyl)aminomethane (Tris) and
adjust the pH to 8.5 with 1 M HCl. Fill up to 1 l with water.
2. hSAX solvent B: 5 mM Tris, 1 M NaCl, pH 8.5. Fill 400 ml
water in a graduated 1 l cylinder. Use a magnetic stirrer to dis-
solve 0.303 g of Tris and 29.221 of sodium chloride and adjust
the pH to 8.5 with 1 M HCl. Fill up to 500 ml with water.
3. hSAX analytical column: Dionex IonPac AS24, hydroxide-
selective anion-exchange analytical column (2 × 250 mm, Thermo
Fisher Scientific, Waltham, USA, Product No. 064153).
4. hSAX guard column: Dionex IonPac AG24, hydroxide-selective
anion-exchange guard column (2 × 50 mm, Thermo Fisher
Scientific, Waltham, USA, Product No. 064151) (see Note 3).
5. HPLC system with the following requirements: flow rates
ranging from 0.1 ml/min to 1 ml/min; 0.1–1 ml sample loop;
UV detector set to read fixed wavelengths of 214 nm and
280 nm (here we used a Dionex Ultimate 3000 system with a
flow rate of 0.25 ml/min and a 100 μl sample loop).
2.3 StageTip 1. StageTip construction: Small, round punch to cut out C-18
Desalting of hSAX disks. 200 μl plastic pipette tip, 1.5 ml reaction vessel, 5 ml
Fractions Eppendorf CombiTip.
2. Empore Octadecyl C18 47 mm Solid Phase Extraction Disks
#2215 (3 M Purification, Eagan, MN, USA).
3. Desalting solvents: Solvent A: 0.1 % (v/v) FA in water. Solvent

B: 0.1 % (v/v) FA and 60 % (v/v) ACN in water.
2.4 LC-MS/MS 1. 0.1 % (v/v) FA in water.
and Data Analysis 2. LC-MS/MS: nano-HPLC setup coupled to a high resolution
mass spectrometer. Here, we used an Eksigent NanoLC-Ultra
1D+ (Eksigent, Dublin, CA) coupled to a Q Exactive Plus mass
spectrometer (Thermo Scientific, Bremen, Germany). LC-trap
column: 75 μm × 2 cm, packed with 5 μm Reprosil-Pur ODS-3
C-18 material (Dr. Maisch, Ammerbuch, Germany). Analytical
column: 75 μm × 42 cm, packed with 3 μm Reprosil-Gold C-18
material (Dr. Maisch, Ammerbuch, Germany).
water. Solvent A: 0.1 % (v/v) FA and 5 % (v/v) DMSO [22] in
water. Solvent B: 0.1 % (v/v) FA and 5 % (v/v) DMSO in ACN.
4. Data analysis: Freely available MaxQuant [20] software pack-
age (e.g., version 1.5.1.0) with the integrated search engine
Andromeda [23]. Protein sequence database in FASTA format
(e.g., UniprotKB).
package.
3 Methods
3.1 Preparation 1. Add 250 μl of precooled tissue lysis solution to 5–20 mg of wet
of Tissue Proteome tissue and transfer into Precellys tubes containing ceramic
Digest beads. Mount Precellys tubes in the Precellys 24 bead-milling
device and perform tissue lysis and homogenization (5500 rpm,
1 × 25 s, 5 s pause).
2. Use a Bradford assay (or a similar photometric assay) to deter-
mine the protein concentration. Store lysates at −80 °C or
continue directly. Continue with a lysate volume correspond-
ing to 200–300 μg of total protein (see Notes 4 and 5).
●● Dilute sample with four volumes of 50 mM Tris–HCl,
pH 7.6 to decrease urea concentration to 1.6 M. Add tryp-
sin in a protease-to-protein ratio of 1:100 (w/w) and pre-
digest 4 h in a thermoshaker at 37 °C and 700 rpm. Add
another 1:100 (w/w) trypsin and incubate the digestion
mixture over night in a thermo-shaker at 37 °C and 700 rpm.
●● Cool samples down to room temperature and acidify the
sample to a pH of ~2 by addition of 1 % (v/v) FA (check the
pH afterwards). Centrifuge acidified peptides at 14,000 × g
to p
recipitate insoluble matter. Use 50 mg Sep-Pak columns
and place them into a vacuum manifold (see Note 6). Prime
Sep-Pak columns by adding 1 ml of solvent B. Equilibrate
column by adding 2 × 1 ml of solvent A. Transfer the acidi-
fied supernatant to the column and slowly load the sample

(see Note 7). Reapply the flow-through a second time and
discard it afterwards. Wash the column with 1 ml solvent A
and repeat this step two more times. Elute the peptides with
1 ml solvent B into a 1.5 ml reaction vessel. Use a vacuum
centrifuge to dry the desalted digest down. At this point the
sample can be stored at −80 °C.
3.2 hSAX Connect the analytical hSAX column and the hSAX guard column
Chromatography to you HPLC system. Upon first time use the column has to be
properly flushed with hSAX solvent A until the pressure is stable (see
Note 8). Setup a gradient following the specifications in Table 1.
Monitor the UV absorption at fixed wavelengths of 280 and 214 nm
and use a flow rate of 0.25 ml/min throughout the gradient.
1. Run a standard digest: A Standard ensures column integrity
and proper column equilibration (see Note 9). Inject ~100 μg
in 100 μl hSAX solvent A.
2. Run a blank: Inject 100 μl hSAX solvent A to clean and equili-
brate the column and to avoid carry-over from the previous
sample (see Note 10).
3. Sample fractionation (see Note 11): Dissolve the desalted
digest in your 1.5 ml reaction vessel in 105 μl solvent A (see
Note 12, for sonication). Centrifuge the sample at 20,000 × g
for 10 min to pellet insoluble debris which might lead to
column clogging. Inject 100 μl of the dissolved sample. Use
a 96-well plate to collect the eluting fractions in 1 min inter-
vals (0.25 ml/fraction; see Note 13) starting 2 min into the
gradient. Collect a total of 38 fractions (see Fig. 1). Freeze
the fractions at −80 °C and dry them down using a vacuum
centrifuge.
Table 1
Settings for a 50 min hSAX column gradient, including the programmed time, the solvent flow
(in ml/min) and percentage of hSAX solvents used
Retention time
[min] hSAX solvent A [%] hSAX solvent B [%] Flow [ml/min]
0 100 0 0.25
3 100 0 0.25
27 75 25 0.25
40 0 100 0.25
44 0 100 0.25
45 100 0 0.25
50 100 0 0.25
Fig. 1 Typical UV (216 nm) chromatogram of a 50 min hSAX separation using 300 μg of tissue digest and the
specified gradient composition (in % of hSAX solvent B). The inset below the chromatogram illustrates the
suggested fractionation and pooling scheme applied for this example
4. Run a blank: Inject 100 μl hSAX solvent A to clean the column

and to avoid carry-over from the previous sample (see Note 10).
5. Detach the column which now contains hSAX solvent A and
store at 4 °C.
3.3 StageTip Given the high salt concentration, it is necessary to desalt the hSAX
Desalting of hSAX fractions using C-18 StageTips [24]. Pass all liquids through the
Fractions tips by centrifugation (~800 × g, room temperature, see Note 14).
1. Resuspend all dried fractions, except for fraction 6 and 38 in
250 μl of solvent A (see Note 15). Pool fractions according to
the scheme depicted in Fig. 1 by transferring the dissolved
fraction 5 into the well containing fraction 6 and the dissolved
fraction 37 into the well containing fraction 38. This results in
a total of 36 fractions.
2. Preparation of C-18 StageTips [24]: Use the small, round
punch to cut out five C18 extraction disks from Empore mate-
rial. Construct a micro-column by packing the disks into a
200 μl pipette tip. Use a sharp scalpel to cut the lid of a 1.5 ml
reaction vessel. The reaction vessel will serve as container for
the flow-through. Push the micro-column through the cut lid
into the 1.5 ml reaction tube. Prepare one tip containing five
C-18 disks (Empore Octadecyl C-18 47 mm Solid Phase
Extraction Disks #2215, 3 M Purification, Eagan, MN, USA)
for each of the 36 hSAX fractions (see Note 16).
3. Sequentially activate the tips using 250 μl of MeOH, 250 μl of

solvent B and 250 μl of solvent A (see Note 17). Empty the
reaction vessels in between.
4. Load one hSAX fraction onto each equilibrated StageTip and
reapply the flow-through. Discard the flow-throughs after-
wards and wash the columns with 250 μl of solvent A. Empty
the reaction vessel.
5. Use 100 μl of solvent B to elute the peptides of the C18 mate-
rial and into the 1.5 ml reaction vessel. Transfer the eluates
into a 96-well plate and dry the sample down using a vacuum
centrifuge/lyophilizer. At this point, the plate can be stored at
−20 °C.
3.4 LC-MS/MS 1. Reconstitute the desalted hSAX fractions in 50 μl of 0.1 % FA.
and Data Analysis 2. Inject 5 μl per fraction (see Note 18) and wash peptides bound
to the trap column for 10 min using loading solvent (0.1 % FA
in water) at a flow rate of 5 μl/min. Then, transfer peptides to
the analytical column and separate them at a flow rate of
300 nl/min using the following gradient: elute peptides using
a linear gradient from 4 to 32 % solvent B for the first 100 min
followed by a 10 min wash out and re-equilibration phase
(increase to 80 % solvent B within 1 min, hold at 80 % solvent
B for 4 min, decrease to 2 % solvent B within 1 min, hold at 2 %
solvent B for 4 min).
3. During peptide elution, directly inject peptides into the mass
spectrometer via electrospray ionization in positive ionization
mode. Suggested parameters for data dependent acquisition
on a Q Exactive Plus are specified in Table 2 (see Note 19).
4. Analyze the data using a proteomics software capable of label-
free quantification. All results shown in this chapter are based
on peptide identifications by search of raw data against the
UniProtKB human database, version July 2013 (88,354
sequences) using the freely available MaxQuant version 1.5.1.0
and its built-in Andromeda search engine. Parameters applied
are specified in Table 3.
5. To obtain the number of protein groups and unique genes
open the proteinGroups.txt and exclude reverse and contami-
nant hits. Count the unique entries in the ”Protein IDs” col-
umn and the “Gene names” column. Average and report the
sequence coverage in percent for the unique proteins. Use the
peptides.txt, remove reverse and contaminant hits and subse-
quently remove the duplicates from the sequence column.
Count and report the number of unique sequences. The num-
ber of acquired PSMs can be extracted from the “MS/MS
identified” column in the summary.txt (see Table 4).
Table 2
Suggested parameters for the measurement of hSAX fractions on a Q
Exactive Plus
Full MS
Resolution 70,000
AGC target 3e6
Maximum IT 100 ms
Scan range 360–1300 m/z
MS2
Resolution 17,500
AGC target 1e5
Maximum IT 50 ms
TopN 20
Isolation window 1.7 m/z
Fixed first mass NCE 25
Additional settings
Underfill ratio 1.0 %
Charge exclusion unassigned, 1, 7, 8, >8
Peptide match Preferred
Exclude isotopes On
Dynamic exclusion 35.0 s
6. To create an orthogonality plot for hSAX separation (see

Fig. 2a), filter out modified sequences from the evidence.txt
and split the data according to the fraction they are reported
in. Separately filter out modified sequence duplicates and plot
the number of nonredundant peptides identified per retention
time bin for each fraction.
7. To determine the hSAX separation power (see Fig. 2b), extract
the unique modified sequences per fraction from the evidence.
txt. Count the number of fractions each peptide was identified
in and plot the percentage against the number of fractions.
8. To obtain the unique peptides per hSAX fraction (see Fig. 2c),
extract the unique modified sequences per fraction from the
evidence.txt. Plot the number of unique modified sequences
for each hSAX fraction.
9. Expected results: Table 4 displays the expected results for the
analysis of 300 μg brain tissue digest, which was separated into
36 fractions using hSAX chromatography.
Table 3
Search parameters used for data analysis with MaxQuant version 1.5.1.0.
In case nothing is specified, default parameters were used
Type Standard
Label No
Variable modifications Acetyl (Protein N-term),
Oxidation (M)
Main search peptide tolerance 4.5 ppm
Max. number of modifications per peptide 5
Global parameters
Database UniProtKB
PSM FDR 0.01
Protein FDR 0.05
Site decoy fraction 0.01
Min. score for unmodified peptides 0
Min. score for modified peptides 40
Min. delta score for unmodified peptides 0
Min. delta score for modified peptides 6
Second peptide search Enabled
Table 4
Overview of results typically expected from a 36 fraction hSAX separation
of human brain tissue digests, where each fraction was measured using a
2 h LC-MS/MS gradient on a Q Exactive Plus
PSMs, proteins and peptides (summary.txt, proteinGroups.txt and peptides.txt)
PSMs 473,882
Unique peptides 111,840
Proteins 10,195
Genes 9,516
Average sequence coverage [%] 30.6
Fig. 2 Two-dimensional peptide separation characteristics. (a) Unique peptide sequences per hSAX fraction
across the 110 min LC-MS gradient (10 min LC-MS retention time bins). The size of the dots scales with the
number of identified peptides. This clearly shows that hSAX separation is highly orthogonal to RP chromatog-
raphy. (b) Separation efficiency of the hSAX fractionation shown as the percentage of peptides found in one or
more fractions (the numbers above the bars indicate percentages). (c) Number of peptide sequences identified
per hSAX fraction
4 Notes
1. The described lysis procedure is not applicable to the analysis

of phosphoproteomes because TCEP-HCl interferes with
phosphopeptide enrichment. If phosphoproteome enrichment
is intended, please refer to Chapter 5 for suitable lysis and
digestion conditions. The dried down Fe-IMAC column flow-
through and the desalted Fe-IMAC eluate can be reconstituted
in hSAX solvent A and is subsequently ready for hSAX separa-
tion (see step 3, Subheading 3.2).
2. The outlined protocol can be easily adapted for the analysis of
cell line proteomes. Simply skip the bead beater step and put
the lysis solution directly on top of the cells after the culture
vessel has been thoroughly rinsed with PBS. Use a cell scraper
to remove the cells from the culture vessel and transfer the
suspension into a 1.5 ml reaction vessel. Determine the protein
concentration and continue with 300 μg. Follow the described
procedure starting with step 3 of Subheading 3.1.
3. The hSAX analytical column can be operated without a guard
column. However, a guard column holds back particles, dirt or
insoluble debris and thus protects the analytical column. The

exchange of a guard column is considerably cheaper than the
exchange of an analytical column.
4. The chemical compatibility of TCEP-HCl and CAA allows the
combined reduction and alkylation in one step, which makes
for a time saving alternative and does not negatively affect the
results [21].
5. Standard proteomic workflows usually include clarification of
the lysate prior to protein digestion. Omission of this step
actually results in a higher protein/peptide recovery and there-
fore the identification of a larger number of membrane and
nuclear proteins. This makes the use of detergents such as SDS
during lysis largely dispensable [21].
amount of digest intended to be desalted. As a rule of thumb,
the capacity of Sep-Pack cartridges equals 5 % of the sorbent
weight (e.g., 2.5 mg peptide for the 50 mg sorbent weight car-
tridges and 10 mg for the 200 mg sorbent weight cartridges).
7. Avoid letting the columns run dry. Load the sample slowly
onto the Sep-Pak column. Lower the flow rate by adjusting the
vacuum at the vacuum manifold. Loading should take at least
10 min to ensure proper binding of the peptides. Reapplying
the flow-through increases recovery.
8. Always keep track of the hSAX column pressure at your given
flow rate. Pressure increases are early indicators of column
clogging or deteriorating column performance. If the increase
is too severe consider changing the guard column and/or the
analytical column.
9. Prepare a standard digest stock solution in hSAX solvent A
according to the procedure described in this protocol. This can
be a tissue or a cell line digest. Run a standard prior to each
sample batch and compare the chromatograms of the standards
to each other in order to spot column deterioration early on.
10. In our experience, it is normal to observe three distinct peaks
in every blank run. As the hSAX column was originally designed
for the separation of small organic molecules and halo acetic
acids [18], we assume that these peaks originate from such
components which were not completely removed (e.g., FA).
11. Keep the sample amount below 500 μg in order to be within
the limits of the column capacity and avoid overloading. This
ensures high resolution peptide separation.
12. In case a visual inspection suggests insufficient dissolution of
the sample in hSAX solvent A, a sonicator bath might support
sample solubilisation. Dip the 1.5 ml vessel containing the
sample into the sonicator bath for 3 × 1 min with 30 s incuba-
tions on ice in between the sonication steps.
13. The fraction volume and number is roughly adjusted to the sepa-
ration power of the hSAX column. In our experience collecting
narrower fractionations does not considerably improve protein
identification. However, depending on the sample complexity,
the available MS machine time and the MS performance, it is also
possible to pool the 38 collected fractions into 24 or 12 fractions
prior to measurement. Although the protein identification is not
severely decreased, the achieved sequence coverage and the num-
ber of identified peptides might be considerably lower.
14. Make sure that each tip equilibration step, the washing step,
each sample loading step (sample and flow-through application)
and each elution step takes approximately 5 min. Since parame-
ters such as C18 material packing density might vary in between
experiments, we suggest to separately adjust the centrifugation
speed to fit the specified time scale for each experiment.
even upon prolonged centrifugation. This is especially benefi-
cial for parallelized desalting because not all columns run at the
same speed.
16. Although only 36 StageTips are necessary for desalting of
hSAX fractions, we recommend preparing and equilibrating 40
StageTips. This way you can save yourself the trouble of start-
ing from the beginning should one or more tips be of insuffi-
cient quality.
17. To ensure that the C18 material is not running dry, make sure
that there is no air trapped between the applied liquid and the
packed C18 material.
18. Although we inject 5 μl per fraction (which is a good starting
point) we advise to initially test different injection volumes in
order to choose the right amount of loading. This will depend
on the hSAX input amount, the capacity of the LC-MS trap
and analytical columns and ultimately also on the sensitivity of
your mass spectrometer.
19. For label-free experiments, a Top N method should be chosen
such that an adequate sampling of the chromatographic peak is
assured (~10 MS1 scans per peak). Given the high input
amount and dynamic range of full proteome digests, some
peptides are highly abundant. Thus the cycle time of the instru-
ment has to be adjusted to the chromatographic peak width
which should be determined beforehand. Likewise, the MS
dynamic exclusion time should be adjusted to the chromato-
graphic peak width. Since every LC-MS system has somewhat
different separation and dead volume characteristics, we sug-
gest adjusting this value to the median peak width at base.
References
1. Wilhelm M, Schlegl J, Hahne H et al (2014) 14. Hennrich ML, Groenewold V, Kops GJPL et al
Mass-spectrometry-based draft of the human (2011) Improving depth in phosphopro-
proteome. Nature 509:582–587 teomics by using a strong cation exchange-
2. Kim M-S, Pinto SM, Getnet D et al (2014) A weak anion exchange-reversed phase
draft map of the human proteome. Nature multidimensional separation approach. Anal
509:575–581 Chem 83:7137–7143
3. Richards AL, Merrill AE, Coon JJ (2015) 15. Gilar M, Olivova P, Daly AE et al (2005) Two-
Proteome sequencing goes deep. Curr Opin dimensional separation of peptides using
Chem Biol 24:11–17 RP-RP-HPLC system with different pH in first
4. Mann M, Kulak NA, Nagaraj N et al (2013) and second separation dimensions. J Sep Sci
The coming age of complete accurate, and 28:1694–1703
ubiquitous proteomes. Mol Cell 49:583–590 16. Zhou F, Sikorski TW, Ficarro SB et al (2011)
5. Nagaraj N, Wisniewski JR, Geiger T et al Online nanoflow reversed phase-strong anion
(2011) Deep proteome and transcriptome exchange-reversed phase liquid
mapping of a human cancer cell line. Mol Syst chromatography- tandem mass spectrometry
Biol 7:548 platform for efficient and in-depth proteome
sequence analysis of complex organisms. Anal
6. Beck M, Schmidt A, Malmstroem J et al (2011) Chem 83:6996–7005
The quantitative proteome of a human cell
line. Mol Syst Biol 7:549 17. Wolters DA, Washburn MP, Yates JR (2001)
An automated multidimensional protein iden-
7. Geiger T, Wehner A, Schaab C et al (2012) tification technology for shotgun proteomics.
Comparative proteomic analysis of eleven com- Anal Chem 73:5683–5690
mon cell lines reveals ubiquitous but varying
expression of most proteins. Mol Cell 18. Ritorto MS, Cook K, Tyagi K et al (2013)
Proteomics 11:M111.014050 Hydrophilic strong anion exchange (hSAX)
chromatography for highly orthogonal peptide
8. Azimifar SB, Nagaraj N, Cox J et al (2014) separation of complex proteomes. J Proteome
Cell-type-resolved quantitative proteomics of Res 12:2449–2457
murine liver. Cell Metab 20:1076–1087
19. Ruprecht B, Koch H, Medard G et al (2015)
9. Deshmukh AS, Murgia M, Nagaraj N et al Comprehensive and reproducible phosphopep-
(2015) Deep proteomics of mouse skeletal tide enrichment using iron immobilized metal
muscle enables quantitation of protein iso- ion affinity chromatography (Fe-IMAC) col-
forms, metabolic pathways, and transcription umns. Mol Cell Proteomics 14:205–215
factors. Mol Cell Proteomics 14:841–853
20. Cox J, Mann M (2008) MaxQuant enables
10. Wiśniewski R, Dus-Szachniewicz K, high peptide identification rates, individualized
Ostasiewicz P et al (2015) Absolute proteome p.p.b.-range mass accuracies and proteome-
analysis of colorectal mucosa, adenoma and wide protein quantification. Nat Biotechnol
cancer reveals drastic changes in fatty acid 26:1367–1372
metabolism and plasma membrane transport-
ers. J Proteome Res 14:4005–4018 21. Kulak NA, Pichler G, Paron I et al (2014)
Minimal, encapsulated proteomic-sample pro-
11. Alpert AJ (1990) Hydrophilic-interaction cessing applied to copy-number estimation in
chromatography for the separation of peptides, eukaryotic cells. Nat Methods 11:319–324
nucleic acids and other polar compounds.
J Chromatogr 499:177–196 22. Hahne H, Pachl F, Ruprecht B et al (2013)
DMSO enhances electrospray response, boost-
12. Boersema PJ, Divecha N, Heck AJR et al ing sensitivity of proteomic experiments. Nat
(2007) Evaluation and optimization of ZIC- Methods 10:989–991
HILIC-RP as an alternative MudPIT strategy.
J Proteome Res 6:937–946 23. Cox J, Neuhauser N, Michalski A et al (2011)
Andromeda: a peptide search engine integrated
13. Hao P, Guo T, Li X et al (2010) Novel applica- into the MaxQuant environment. J Proteome
tion of electrostatic repulsion-hydrophilic Res 10:1794–1805
interaction chromatography (ERLIC) in shot-
gun proteomics: comprehensive profiling of rat 24. Rappsilber J, Mann M, Ishihama Y (2007)
kidney proteome. J Proteome Res Protocol for micro-purification, enrichment,
9:3520–3526 pre-fractionation and storage of peptides for
proteomics using StageTips. Nat Protoc
2:1896–1906
Chapter 8
High pH Reversed-Phase Micro-Columns for Simple,

Sensitive, and Efficient Fractionation of Proteome
and (TMT labeled) Phosphoproteome Digests
Benjamin Ruprecht, Jana Zecha, Daniel P. Zolg, and Bernhard Kuster
Abstract
Despite recent advances in mass spectrometric sequencing speed and improved sensitivity, the in-depth
analysis of proteomes still widely relies on off-line peptide separation and fractionation to deal with the
enormous molecular complexity of shotgun digested proteomes. While a multitude of methods has been
established for off-line peptide separation using HPLC columns, their use can be limited particularly when
sample quantities are scarce. In this protocol, we describe an approach which combines high pH reversed-
phase peptide separation into few fractions in StageTip micro-columns. This miniaturized sample prepara-
tion method enhances peptide recovery and hence improves sensitivity. This is particularly useful when
working with limited sample amounts obtained from e.g., phosphopeptide enrichments or tissue biopsies.
Essentially the same approach can also be applied for multiplexed analysis using tandem mass tags (TMT)
and can be parallelized in order to deliver the required throughput. Here, we provide a step-by-step pro-
tocol for TMT6plex labeling of peptides, the construction of StageTips, sample fractionation and pooling
schemes adjusted to different types of analytes, mass spectrometric sample measurement, and downstream
data processing using MaxQuant. To illustrate the expected results using this protocol, we provide results
from an unlabeled and a TMT6plex labeled phosphopeptide sample leading to the identification of
>17,000 phosphopeptides in 8 h (Q Exactive HF) and >23,000 TMT6plex labeled phosphopeptides (Q
Exactive Plus) in 12 h of measurement time. Importantly, this protocol is equally applicable to the frac-
tionation of full proteome digests.
Key words Fractionation, Proteomics, Sample preparation, Mass spectrometry, Phosphorylation,

Isotope labeling
Abbreviations
CAN Acetonitrile
FA Formic acid
HCD Higher energy collision induced dissociation
hSAX Hydrophilic strong anion exchange
83
IT Injection time
MS Mass spectrometer
NH4FA Ammonium formate
RP Reversed-phase
SAX Strong anion exchange
SCX Strong cation exchange
StageTip Stop and go extraction tip
TEAB Triethylammonium bicarbonate
TFA Trifluoroacetic acid
TiO2 Titanium dioxide
TMT Tandem mass tag
Ppm Parts per million
pY/pS/pT Phosphotyrosine, -serine, -threonine
ZIC-HILIC Zwitterionic hydrophilic interaction liquid chromatography
1 Introduction
Recent advances in mass spectrometric instrumentation enable the

identification and quantification of peptides and phosphopeptides
at an unprecedented depth [1]. Despite the ever increasing
sequencing speed of modern mass spectrometers, single measure-
ments are often not sufficient to fully resolve the (phospho)pro-
teome complexity and thus coverage must generally be improved
by some form of two-dimensional peptide fractionation. In shot-
gun proteomics, the most widely applied second dimension of pep-
tide separation is reversed-phase C18 chromatography directly
coupled to tandem mass spectrometry. Many different choices for
orthogonal first dimension separations exist: especially strong cat-
ion exchange (SCX) [2, 3], which separates peptides by charge, has
been widely applied for phosphopeptide (pre)fractionation. But
also hydrophilic interaction liquid chromatography (HILIC) [4],
hydrophilic strong anion exchange (hSAX) [5] and high pH
reversed-phase columns [6] have proven their merit for phospho-
proteomic studies. Although off-line 2D approaches using stan-
dard inner diameter columns (1–10 mm) are highly robust, efficient
and deliver very good analytical depth, fractionation can be time-
consuming and parallelization capabilities are limited. In addition,
we [5] and others [7] have shown that phosphopeptide fraction-
ation after enrichment requires an increased amount of starting
material, due to sample losses that occur during the fractionation
process. Here, we outline a simple and robust alternative which
combines high pH reversed-phase separation and self-packed
StageTip micro-columns [8, 9]. Such high pH reversed-phase col-
umns as well as SAX and SCX materials have recently been used to
separate digests of full proteomes [8, 9, 10] and membrane protein
preparations [11, 12].
High ph Micro-Columns for Simple Efficient Proteome Fractionation 85
Importantly, the enhanced sensitivity and recovery provided

by the high pH reversed-phase micro-column separations is par-
ticularly attractive when working with limited sample amounts
including but not limited to phosphopeptide enrichments. The
described fractionation procedure, which does not require special-
ized equipment, can be accomplished in a time efficient and highly
parallelizable fashion and can easily be combined with TMT label-
ing [13]. Since TMT labeled (phospho)peptides are generally
more hydrophobic than the respective unlabeled peptides, we pro-
vide detailed fractionation and pooling schemes for both sample
types. This protocol further describes recommended parameters
for the mass spectrometric measurement on a Q-Exactive mass
spectrometer and outlines how to analyze the obtained data using
MaxQuant [14, 15]. Finally, we illustrate the expected results
using a phosphoproteome and a TMT6plex labeled phosphopro-
teome obtained by IMAC enrichment of a cancer cell line digest
(see Chapter 5). This protocol is equally applicable to the fraction-
ation of full proteome digests, following the exact same fraction-
ation and pooling scheme which is described for TMT labeled
phosphopeptides.
2 Materials
Unless stated otherwise, all solvents and buffers are prepared

freshly using ultrapure water and analytical grade reagents. Devices
such as centrifuges, vacuum centrifuges/lyophilizer, thermoshaker,
or refrigerators (−20 °C/−80 °C) are not explicitly listed.
2.1 Sample For materials related to lysis, digestion, and phosphopeptide

Preparation enrichment we refer to the Chapter 5.
2.2 TMT6plex 1. Sep-Pak C18 peptide purification before TMT6plex labeling:

Labeling of Whole 50 mg Sep-Pak cartridges (Waters Corp., Eschborn, Germany).
Proteome Digests Prior Solvent A: 0.1 % (v/v) FA in water. Solvent B: 60 % (v/v)
to Phosphopeptide ACN, 0.1 % (v/v) FA in water.
Enrichment 2. TMT6plex labeling reagent: TMT6plex isobaric label reagent
set (Thermo Fisher Scientific Inc., Bremen, Germany). This
protocol uses 1 mg of each TMT6plex labeling reagent to label
250 μg protein digest per channel (see Note 1). Allow the vials
containing the six different labeling reagents to warm up to
room temperature (see Note 2). Prepare TMT6plex stock
solutions by adding 41 μl of the anhydrous ACN (see Note 3)
to each reagent vial (0.8 mg TMT). Vortex vials for 30–60 s
and briefly centrifuge. The TMT6plex stock solution should
be used immediately or can alternatively be stored at −80 °C
for up to 1 week. To increase storage time, the remaining
TMT6plex reagents should be dried down using a vacuum
centrifuge.
3. 50 mM triethylammonium bicarbonate (TEAB), pH 8.5: Mix

50 μl of 1 M TEAB (Sigma-Aldrich, St. Louis, MO), pH 8.5
with 950 μl of water. Make sure that the pH is not below 8.
4. 5 % hydroxylamine in water.
5. 10 % (v/v) FA and 10 % (v/v) ACN in water.
6. 10 % (v/v) FA in water.
7. Sep-Pak C18 peptide purification after TMT6plex labeling:
50 mg Sep-Pak cartridges (Waters Corp., Eschborn, Germany).
Solvent A: 0.07 % (v/v) TFA in water. Solvent B: 50 % (v/v)
ACN, 0.07 % (v/v) TFA in water.
8. Vacuum manifold for Sep-Pak desalting.
2.3 High pH 1. Empore Octadecyl C18 47 mm Solid Phase Extraction Disks
Reversed-Phase #2215 (3 M Purification, Eagan, MN, USA).
Micro-Column 2. Small, round punch to cut out C-18 disks (Fig. 1b).
Fractionation
3. 200 μl plastic pipette tip.
4. 1.5 ml reaction vessel.
5. High pH reversed-phase stock solution: 50 mM ammonium
formate (NH4FA, Sigma-Aldrich, St. Louis, MO, Product
number 156264, reagent grade 97 %), pH 10. Weigh in
315.3 mg of NH4FA and transfer it into a glass beaker or cyl-
inder. Add water to a volume of 90 ml. Mix with a magnetic
stirrer and adjust the pH to 10 using ammonium hydroxide.
Fill up to 100 ml with water.
Fig. 1 Construction of a micro-column for high pH reversed-phase fractionation. (a) The lid of a 1.5 ml reaction
tube is cut crosswise with a small scalpel. (b) A punching device, assembled from a syringe and a piece of wire
is used to punch out 5 disks of Empore C18 material. (c) The punching device is used to push the disks into
the 200 μl pipette tip. (d) The constructed column is placed in the reaction tube by pushing the tip through the
cut lid
6. High pH reversed-phase solvents: Solvent A: Dilute 50 mM

high pH reversed-phase stock solution 1:1 with water to obtain
25 mM NH4FA, pH 10. Solvent B: Dilute high pH reversed-
phase stock 1:1 with ACN to obtain 25 mM NH4FA, 50 %
ACN, pH 10.
7. Mix high pH reversed-phase solvents A and B to obtain elution
solvents (see Note 4). In case you want to separate an unla-
beled phosphoproteome, prepare elution solvents containing
2.5 %, 7.5 %, 12.5 % ACN in 25 mM NH4FA. For the TMT6plex
labeled phospho sample, prepare elution buffers with 5 %,
7.5 %, 10 %, 12.5 %, 15 %, and 17.5 % ACN in 25 mM NH4FA
(see Note 5). Table 1 displays the mixing scheme for the differ-
ent elution buffers, sufficient for 10 micro-column elution
steps each. The solvents are prepared sequentially; as an exam-
ple, prepare the 7.5 % solvent by diluting 0.6 ml of the 10 %
ACN elution solvent with 0.2 ml solvent A. For the fraction-
ation of unlabeled phosphopeptides, start by diluting the 50 %
ACN elution solvent 1:4 with solvent A and continue accord-
ing to the scheme shown in Table 1.
8. 5 ml Eppendorf CombiTip.
9. Desalting of the high pH reversed-phase micro-column flow-
through: Desalting solvent A: 0.07 % (v/v) TFA in water.
Desalting solvent B: 0.07 % (v/v) TFA, 60 % (v/v) ACN in water.
2.4 LC-MS/MS 1. 50 mM citric acid and 1 % (v/v) FA in water. Dissolve 105 mg
and Data Analysis of citric acid monohydrate (VWR, Product No. 20278.298) in
9.90 ml of water and add 100 μl of 100 % FA.
2. LC-MS/MS: nano-HPLC setup coupled to a high resolution
mass spectrometer. Here, we used an Eksigent NanoLC-Ultra
1D+ (Eksigent, Dublin, USA; to measure fractions of
TMT6plex labeled phosphopeptides) or an Thermo Ultimate
Table 1
Mixing scheme for high pH reversed-phase elution solvents
Elution solvent Take Add solvent A (ml) Total [ml]

17.5 % ACN 0.7 ml Solvent B (50 % ACN) 1.3 2.0
15.0 % ACN 1.2 ml Solvent 17.5 % ACN 0.2 1.4
3000 (Thermo Scientific, Bremen, Germany; to measure frac-

tions of unlabeled phosphopeptides) coupled to an Orbitrap Q
Exactive type mass spectrometer (Thermo Scientific, Bremen,
Germany). LC-trap column: 75 μm × 2 cm, packed with 5 μm
Reprosil-Pur ODS-3 C18 material (Dr. Maisch, Ammerbuch,
Germany). Analytical column: 75 μm × 42 cm, packed with
3 μm Reprosil-Gold C18 material (Dr. Maisch, Ammerbuch,
Germany).
water. Solvent A: 0.1 % (v/v) FA and 5 % (v/v) DMSO in water.
Solvent B: 0.1 % (v/v) FA and 5 % (v/v) DMSO in ACN.
4. Data analysis: Freely available MaxQuant software package
(e.g., version 1.5.2.8) with the integrated search engine
Andromeda. Protein sequence database in FASTA format
(e.g., UniprotKB).
package.
3 Methods
3.1 Sample For methods related to lysis, digestion, and peptide desalting using
Preparation Sep-Pak columns, we refer to Chapter 5. The Fe-IMAC protocol
also describes the purification of phosphopeptides using Fe-IMAC
columns. The procedure can be readily applied to combined,
TMT6plex labeled digests. We recommend using at least 1.5 mg of
protein digest for phosphopeptide enrichment (Fe-IMAC input
amounts ranging between 1.5 and 3 mg are required to obtain
optimal results). For a TMT6plex experiment this translates into
250 μg of protein per channel (see Note 6).
3.2 TMT6plex 1. Desalting of samples prior to TMT6plex labeling (see Note 7):
Labeling of Whole Acidify digested samples to ~ pH 2 by addition of 100 % FA to
Proteome Digests Prior a final concentration of 1 % (v/v) FA (check the pH). Centrifuge
to Phosphopeptide peptides at 5000 × g to precipitate insoluble matter. Place six
Enrichment 50 mg Sep-Pak columns into a vacuum manifold (see Note 8).
Prime Sep-Pak columns by adding 1 ml of solvent B. Equilibrate
column by adding 2 × 1 ml of solvent A. Load the supernatant
of the acidified sample slowly onto the column, reapply the
flow-through and discard the second flow-through afterwards.
Wash the column with 2 × 1 ml of solvent A. Elute peptides of
each column with 2 × 150 μl of solvent B into a 1.5 ml reaction
vessel. Freeze the samples at −80 °C and make sure that they
are still frozen when they are placed in the vacuum centrifuge
to dry them down.
2. Labeling reaction (see Note 9): Reconstitute the six desalted sam-
ples in 200 μl of 50 mM TEAB. To start the labeling reaction, add
50 μl of the respective TMT6plex stock solution to each sample,

mix by repeatedly pipetting up and down, briefly centrifuge the
samples and incubate them for 1 h at 20 °C and 400 rpm. Stop
the labeling reaction by adding 20 μl of 5 % hydroxylamine to
each sample (final concentration of ~0.4 % hydroxylamine).
Incubate the samples for 15 min at room temperature, shaking at
400 rpm and briefly centrifuge them afterwards.
3. Sample combination (see Note 10): Combine the six samples by
transferring half of each one into two separate 1.5 ml reaction
vessels. To acidify the sample to a pH of 2, add 50 μl of 10 % FA
to each vessel (check the pH). Add 100 μl of 10 % FA in 10 %
ACN to each of the six original vessels, incubate for 5 min and
combine the solvents with the pooled sample. Briefly centrifuge
both reaction vessels, freeze them at −80 °C, and vacuum-cen-
trifuge to dryness. The two vials can now be stored at -80 °C.
4. Sep-Pak desalting of the combined, labeled sample: Reconstitute
peptides in both vessels in 0.5 ml of 0.07 % TFA and pool the
sample. Except for the applied solvents, desalting is essentially
performed as described in step 1 of this section. Here, the sam-
ple is loaded in 0.07 % (v/v) TFA in water and eluted off the
Sep-Pak column (50 mg sorbent weight) using 2 × 150 μl of
0.07 % TFA, 50 % ACN in water. The eluate is subsequently
filled up to 0.5 ml with 0.07 % TFA in water and can thus
directly be applied to Fe-IMAC column based phosphopeptide
enrichment.
5. Fe-IMAC enrichment (see Note 11): see Chapter 5. Dry the
Fe-IMAC eluate down using a vacuum centrifuge. The StageTip
desalting of the phosphopeptide containing Fe-IMAC eluate
can be omitted. Store the sample at −80 °C or continue directly
with high pH reversed-phase micro-column fractionation.
3.3 High pH The following steps for fractionation and desalting are performed
Reversed-Phase by centrifugation of the micro-column at ~800 × g (see Note 12).
Micro-Column Pre-cool and store all solvents on ice. Unless stated otherwise,
Fractionation avoid letting the columns run dry.
1. High pH reversed-phase micro-column construction: Use the
small, round punch to cut out five C18 extraction disks from
Empore material (see Note 13 and Fig. 1). Construct a micro-
column by packing the disks into a 200 μl pipette tip. Use a
sharp scalpel to cut the lid of a 1.5 ml reaction vessel. The
reaction vessel will serve as container for the flow-through.
Push the micro-column through the cut lid into the 1.5 ml
reaction tube.
2. Dissolve your dried sample in 250 μl solvent A (see Note 14).
Vortex, spin down and store on ice while the column is equili-
brated. Check if the pH of the dissolved sample is ~10 using
pH indicator strips.
3. Column equilibration (see Note 15): Add 250 μl of ACN to

the top of the micro-column to soak the extraction material
and remove air bubbles. Centrifuge the tip and rinse residual
ACN using a 5 ml Eppendorf CombiTip. Wash column with
250 μl solvent B followed by 250 μl solvent A and discard the
flow-through.
4. Sample loading: Slowly load the sample onto the column.
Reapply the flow-through a second time. Transfer the flow-
through in a new 1.5 ml vessel and dry it down using a vacuum
centrifuge (for flow-through desalting, continue to step 6, see
Note 16). Wash the micro-column with 250 μl solvent A and
discard the washing fraction.
5. Sample fractionation: The peptides bound to the extraction
material are sequentially eluted with increasing concentration
of ACN. Use 40 μl of each elution solvent. Let the column run
dry in between each step. After each elution step, the eluate is
transferred into a 96-well plate. For separation of unlabeled
phosphopeptides, sequentially elute peptides from the C18
material with four solvents containing 2.5 %, 7.5 %, 12.5 % and
50 % ACN in 25 mM NH4FA. The desalted sample flow-
through fraction is pooled with the 50 % ACN fraction (see
step 6). For TMT6plex labeled phosphopeptide separation use
seven solvents containing 5 %, 7.5 %, 10 %, 12.5 %, 15 %, 17.5 %
and 50 % ACN in 25 mM NH4FA. Combine the desalted sam-
ple flow-through fraction (see step 6) with the 17.5 % ACN
fraction and the 5 % ACN fraction with the 50 % ACN
fraction.
6. Desalting of the high pH reversed-phase micro-column flow-
through (see Note 16): Dissolve the dried sample in 250 μl of
desalting solvent A and keep it on ice while the StageTip is
prepared. Check the pH of the dissolved peptide solution and,
if required, adjust it to pH 2 using 100 % TFA. Sequentially
activate the tip using 250 μl of ACN, 250 μl of desalting sol-
vent B, and 250 μl of desalting solvent A. Empty the 1.5 ml
vessel in between each step. Load the dissolved sample and
reapply the flow-through. Discard the second flow-through
and wash the column with 250 μl desalting solvent A. Use
40 μl of desalting solvent B to elute the peptides from the C18
material. Transfer the eluate into the 96-well plate containing
the other fractions (see step 6) and dry the fractions down
using a vacuum centrifuge/lyophilizer. At this point, the sealed
plate can be stored at −20 °C.
3.4 LC-MS/MS 1. Reconstitute the micro-column fractions in 10 μl of 1 % FA in

and Data Analysis 50 mM citrate (see Note 17).
2. Inject 10 μl per TMT6plex fraction and 5 μl for fractions con-
taining unlabeled phosphopeptides (see Note 18). Wash pep-
tides bound to the trap column for 10 min using loading
solvent (0.1 % FA in water) at a flow rate of 5 μl/min. Then,

transfer peptides to the analytical column and separate them at
a flow rate of 300 nl/min using gradients as follows (see Note
19): elute unlabeled phosphopeptides using a linear gradient
from 2 to 15 % solvent B for the first 70 min followed by a
linear increase to 27 % solvent B within the next 30 min. In
contrast, separation of TMT6plex labeled phosphopeptides is
performed by an initial increase from 2 to 4 % solvent B within
the first 2 min followed by a 98 min linear gradient from 4 to
32 % solvent B. In both cases the gradient ends with a 10 min
wash out phase (increase to 80 % solvent B within 2 min, hold
at 80 % solvent B for 2 min, decrease to 2 % solvent B within
2 min, hold at 2 % solvent B for 4 min) (Fig. 2).
3. During peptide elution, directly inject peptides into the mass
spectrometer via electrospray ionization in positive ionization
mode. Suggested parameters for data dependent acquisition
and HCD fragmentation on a Q Exactive Plus/HF are speci-
fied in Table 2 (see Note 20).
Fig. 2 Total ion current chromatograms displaying peptide elution patterns across the different high pH
reversed-phase fractions for unlabeled phosphopeptides (blue, Fig. 2a) and TMT6plex labeled phosphopep-
tides (red, Fig. 2b). The number in the corner of each plot indicates the absolute ion current intensity and the
respective fraction number
4. Analyze data using a proteomics software capable of label-free

and TMT6plex quantification, respectively. All results shown in
this chapter are based on peptide identifications by search of raw
data against the UniProtKB human database, version July 2013
(88,354 sequences) using the freely available MaxQuant version
1.5.2.8 and its built-in Andromeda search engine. Parameters
applied are specified in Table 3.
5. Filter the MaxQuant evidence.txt or the phospho(STY) sites.
txt output file for reverse sequences and potential contami-
nants. To determine the selectivity of the phospho enrichment,
the reported number of peptides annotated with a phosphory-
lation site is divided by the total number of identified sequences.
The intensity based selectivity is acquired similarly by dividing
the summed intensity of phosphorylated peptides by the total
intensity. Filter the “Modified Sequence” column for dupli-
cates to remove redundancies and obtain the number of unique
phosphopeptides. Similarly, the phospho (STY)sites.txt is used
to determine the number of unique and quantifiable sites.
Filter for “Localization probability” ≥0.75 to obtain the num-
ber of class I sites [16].
6. To ensure proper TMT6plex labeling, count the number of
TMT6plex labeled peptides and divide them by the sum of
TMT6plex labeled and non TMT6plex labeled peptides (the
labeling efficiency is usually >99 %).
7. To obtain the orthogonality plot for phosphopeptides (Fig. 3),
filter out non-phospho sequences from the evidence.txt and
split the data according to the fraction they are reported in.
Filter out modified sequence duplicates in every fraction sepa-
rately. Plot the number of nonredundant phosphopeptides
identified per retention time bin for each fraction.
8. To determine the separation power of the high pH reversed-
phase micro-column, extract the unique phosphorylated
sequences per fraction from the evidence.txt (use the modified
sequence column). Count the number of fractions each phos-
phorylated sequence was identified in and plot the percentage
against the number of fractions.
9. Expected Results: Table 4 displays the identified phosphopeptides
and phosphorylation sites from both experimental approaches.
4 Notes
1. Although the amount of labeling reagent used in this protocol

is below that specified by the manufacturer, we found it to be
sufficient for complete labeling of peptides.
Table 2
Suggested parameters for the measurement of high pH reversed-phase phosphopeptide fractions on
a Q Exactive HF (unlabeled phosphopeptides) and a Q Exactive Plus (TMT6plex labeled
phosphopeptides) mass spectrometer
Unlabeled (QE HF) TMT6plex labeled (QE Plus)

Full MS
Resolution 60,000 70,000
AGC target 3e6 3e6
Maximum IT 20 ms 20 ms
Scan range 360–1300 m/z 360–1300 m/z
MS2
Resolution 30,000 17,500
AGC target 2e5 2e5
Maximum IT 100 ms 100 ms
TopN 15 20
Isolation window 1.7 m/z 1.3 m/z
Isolation offset 0.0 m/z 0.0 m/z
Fixed first mass 100 m/z 100 m/z
NCE 25 33
Additional settings
Underfill ratio 1.0 % 1.0 %
Charge exclusion unassigned, 1, 7, 8, >8 unassigned, 1, 7, 8, >8
Peptide match Preferred Preferred
Exclude Isotopes On On
Dynamic exclusion 25.0 s 35.0 s
2. Equilibration of TMT reagents to room temperature prior to

opening avoids water condensation and hydrolysis of the mois-
ture-sensitive labeling reagents.
3. Make sure the anhydrous ACN bottle is kept water-free at all
times. Use a dry syringe pre-filled with Argon to transfer ali-
quots of ACN to a dry Eppendorf tube pre-filled with argon.
4. Due to the attached TMT6 label, peptides become more
hydrophobic. Therefore, it is necessary to adjust the ACN con-
centration of the high pH reversed-phase elution solvents
accordingly.
5. High pH reversed-phase micro-column fractionation can also
be applied to the fractionation of full proteome digests using
Table 3
Search parameters used for data analysis with MaxQuant version 1.5.2.8. In case default parameters
were used, nothing is specified
Unlabeled TMT6plex labeled

Type Standard Reporter ion MS2
Label No 6plex TMT
Variable modifications Acetyl (Protein N-term), Oxidation (M), Phospho (STY)
Main search peptide tolerance 5 ppm
Max. number of modifications per 5
peptide
Global parameters
Database UniProtKB
PSM FDR 0.01
Min. score for (un)modified peptides 0
Min- delta score for (un)modified 0
peptides
Match between runs Enabled
Match time window 1 min
Alignment time window 20 min
the same fractionation/elution scheme as described for

TMT6plex labeled phosphopeptides. We refer to Chapter 5 for
the preparation of proteome digests and advise to use input
amounts between 20 and 50 μg as well as five C18 disks per
high pH reversed-phase tip. You should expect the identifica-
tion of ~6000–6500 proteins in 12 h of measurement time on
a Q Exactive type mass spectrometer.
6. To increase the multiplexing capacity, TMT10plex labeling
reagents can be readily used and following the same proce-
dure. As ten instead of six samples are combined, less input
material per channel is required (≥150 μg digest).
7. Avoid letting the columns run dry. Load the sample slowly
onto the Sep-Pak column. Lower the flow rate by adjusting the
Fig. 3 Orthogonality and peptide separation characteristics of high pH reversed-phase micro-column fraction-
ation. (a) Unique phosphopeptide sequences per micro-column fraction across the 110 min LC-MS gradient
(blue: unlabeled phosphopeptides; red: TMT6plex labeled phosphopeptides; grouped into 10 min LC-MS reten-
tion time bins). The size of the dots scales with the number of identified phosphopeptides. (b) Separation
efficiency of the high pH reversed-phase micro-column fractionation shown as the percentage of peptides
found in one or more fractions (blue: unlabeled phosphopeptides; red: TMT6plex labeled phosphopeptides; the
numbers above the bars indicate percentages)
vacuum at the vacuum manifold. Loading should take at least

10 min to ensure proper binding of the peptides. Re applying
the flow-through increases recovery. Drying down frozen sam-
ples improves solubility afterwards.
amount of digest you intend to load. As a rule of thumb, the
capacity of Sep-Pack cartridges equals 5 % of the sorbent weight
(e.g., 2.5 mg for the 50 mg sorbent weight cartridges and
10 mg for the 200 mg sorbent weight cartridges).
9. TMT reagents are amine-reactive, thus all amine-containing buf-
fers and additives must be removed before labeling. TEAB con-
centrations >100 mM reduce labeling efficiency. Keep the final
ACN concentration during the labeling reaction below 40 %.
10. In case of using higher amounts of input digest, the combined
sample might have to be split into more than two 1.5 ml reac-
tion vessels.
11. High pH reversed-phase micro-column fractionation is also
compatible with other phosphopeptide enrichment methods
Table 4
Summary of expected results for fractionation of TMT6plex and unlabeled phosphopeptides prepared
from 1.5 mg and 2 mg of digest, respectively
Unlabeled TMT6plex labeled

Phosphopeptides (MaxQuant—evidence.txt)
Identified unique phosphopeptides 17,140 23,939
Quantified unique phosphopeptides 16,391 22,418
Mono phosphorylated 14,320 (87 %) 18,820 (84 %)
Multiply phosphorylated 2,071 (13 %) 3,598 (16 %)
ID based phosphopeptide selectivity 96 % 94 %
Intensity based phosphopeptide selectivity 93 % 99 %
Phosphorylation sites (MaxQuant—Phospho(STY)sites.txt)
Identified phosphorylation sites 12,971 19,359
Quantified phosphorylation sites 11,221 17,681
Class I sites (Loc prob > 0.75) 9,634 14,534
pS sites (class I) 8,208 (85 %) 12,352 (85 %)
pT sites (class I) 1,368 (14 %) 1,463 (10 %)
pY sites (class I) 67 (<1 %) 728 (5 %)
such as TiO2, IMAC and Ti-IMAC beads in batch or tip

format.
12. Make sure that each tip equilibration step, the washing step,
each sample loading step (sample and flow-through application)
and each elution step takes approximately 5 min. Since parame-
ters such as C18 material packing density might vary in between
experiments, we suggest to separately adjust the centrifugation
speed to fit the specified time scale for each experiment.
13. In our experience, every C18 disk has a capacity for 5–10 μg of
protein digest. We suggest using 5–6 disks per micro-column.
even upon prolonged centrifugation. This is especially
beneficial when parallelized fractionation is intended as not all
columns run at the same speed.
15. If only one sample is intended to be fractionated, the proce-
dure can be accelerated by manually pushing the liquids
through the tips using a 5 ml Eppendorf CombiTip. The vol-
umes can then be scaled down accordingly (~40 μl per step).
16. The high pH reversed-phase micro-column flow-through can
be dried down while fractionation is carried out. Once the frac-
tionation is finished, the 96-well plate containing the eluate

fractions is sealed with adhesive foil and stored at −20 °C until
the desalting step is finished. Afterwards the plate is thawed
and the desalted peptides can be directly eluted into the respec-
tive well of the 96-well plate.
17. Citric acid acts as a chelating agent for residual Fe3+ ions that
might bleed from the Fe-IMAC column. Remaining Fe3+ ions
can stick to the trap/analytical nano-HPLC columns and
deplete phosphopeptides. Ever since we used citrate we have
not detected any iron contamination [17]. If in doubt, specify
iron as a variable modification during data processing and
check if any iron bound peptides are identified.
18. Since TMT6plex labeled phosphopeptide fractionation result
in six instead of four fractions, we opted to inject the complete
fraction. But it should be sufficient to inject only half of the
10 μl also for the TMT6plex labeled fractions.
19. Unlabeled and TMT6plex labeled phosphopeptides require
different nano-HPLC gradients owing to differences in hydro-
phobicity imparted by the TMT label.
20. For label-free experiments, a Top N method should be chosen
such that an adequate sampling of the chromatographic peak is
assured (~ten MS1 scans per peak). Thus, the cycle time of the
instrument has to be adjusted to the chromatographic peak
width. For TMT based experiments, the quantification data is
collected in the tandem mass spectrum which generally allows
using a higher number of MS2 scans per cycle (Top N). For
TMT labeled samples, we further recommend to narrow the
isolation window of the quadrupole to reduce ratio compres-
sion due to co-eluting and co-isolated features [18]. The colli-
sion energy has to be optimized for unlabeled and TMT labeled
phosphopeptides separately and is usually higher for TMT
labeled samples. Likewise, dynamic exclusion should be opti-
mized for every LC system separately. Since every LC-MS sys-
tem has somewhat different separation and dead volume
characteristics, we suggest adjusting this value to the median
peak width at base. For TMT10plex labeled phosphopeptides,
the MS2 resolution has to be adjusted to 60,000 [19] in order
to resolve the TMT reporter ion isotopologues.
References
1. Wilhelm M, Schlegl J, Hahne H et al (2014) 3. Gauci S, Helbig AO, Slijper M et al (2009)
Mass-spectrometry-based draft of the human Lys-N and trypsin cover complementary parts
proteome. Nature 509:582–587 of the phosphoproteome in a refined SCX-
2. Villén J, Gygi SP (2008) The SCX/IMAC based approach. Anal Chem 81:4493–4501
enrichment approach for global phosphoryla- 4. McNulty DE, Annan RS (2008) Hydrophilic
tion analysis by mass spectrometry. Nat Protoc interaction chromatography reduces the com-
3:1630–1638 plexity of the phosphoproteome and improves
global phosphopeptide isolation and detection. 12. Wiśniewski JR, Zougman A, Mann M (2009)
Mol Cell Proteomics 7:971–980 Combination of FASP and StageTip-based
5. Ruprecht B, Koch H, Medard G et al (2015) fractionation allows in-depth analysis of the
Comprehensive and reproducible phosphopep- hippocampal membrane proteome. J Proteome
tide enrichment using iron immobilized metal Res 8:5674–5678
ion affinity chromatography (Fe-IMAC) col- 13. Thompson A, Schäfer J, Kuhn K et al (2003)
umns. Mol Cell Proteomics 14:205–215 Tandem mass tags: a novel quantification strat-
6. Batth TS, Francavilla C, Olsen JV (2014) Off- egy for comparative analysis of complex protein
line high-pH reversed-phase fractionation for mixtures by MS/MS. Anal Chem
in-depth phosphoproteomics. J Proteome Res 75:1895–1904
13:6176–6186 14. Cox J, Mann M (2008) MaxQuant enables
7. Kettenbach AN, Gerber SA (2011) Rapid and high peptide identification rates, individualized
reproducible single-stage phosphopeptide p.p.b.-range mass accuracies and proteome-
enrichment of complex peptide mixtures: wide protein quantification. Nat Biotechnol
application to general and phosphotyrosine- 26:1367–1372
specific phosphoproteomics experiments. Anal 15. Cox J, Neuhauser N, Michalski A et al (2011)
Chem 83:7635–7644 Andromeda: a peptide search engine integrated
8. Ishihama Y, Rappsilber J, Mann M (2006) into the MaxQuant environment. J Proteome
Modular stop and go extraction tips with Res 10:1794–1805
stacked disks for parallel and multidimensional 16. Olsen JV, Blagoev B, Gnad F et al (2006)
peptide fractionation in proteomics. Global, in vivo, and site-specific phosphoryla-
J Proteome Res 5:988–994 tion dynamics in signaling networks. Cell
9. Rappsilber J, Mann M, Ishihama Y (2007) Protocol 127:635–648
for micro-purification, enrichment, pre-fraction- 17. Winter D, Seidler J, Ziv Y et al (2009) Citrate
ation and storage of peptides for proteomics using boosts the performance of phosphopeptide
StageTips. Nat Protoc 2:1896–1906 analysis by UPLC-ESI-MS/MS. J Proteome
10. Lawrence RT, Perez EM, Hernández D et al Res 8:418–424
(2015) The proteomic landscape of triple- 18. Ow SY, Salim M, Noirel J et al (2009) iTRAQ
negative breast cancer. Cell Rep 11:630–644 underestimation in simple and complex mix-
11. Kitata RB, Dimayacyac-Esleta BRT, Choong tures: “the good, the bad and the ugly”.
W-K et al (2015) Mining missing membrane pro- J Proteome Res 8:5347–5355
teins by high-pH reverse-phase StageTip frac- 19. Werner T, Sweetman G, Savitski MF et al (2014)
tionation and multiple reaction monitoring mass Ion coalescence of neutron encoded TMT 10-plex
spectrometry. J Proteome Res 14:3658–3669 reporter ions. Anal Chem 86:3594–3601
Chapter 9
Multi-Lectin Affinity Chromatography for Separation,

Identification, and Quantitation of Intact Protein
Glycoforms in Complex Biological Mixtures
Sarah M. Totten, Majlinda Kullolli, and Sharon J. Pitteri
Abstract
Protein glycosylation is considered to be one of the most abundant post-translational modifications and is rec-
ognized for playing key roles in cellular functions. Aberrant N-linked glycosylation has been associated with
several human diseases and has prompted the development and constant improvement of analytical tools to
separate, characterize, and quantify glycoproteins in complex mixtures extracted from various biological
samples (such as blood and tissue). Lectins, or carbohydrate-binding proteins, have been used as valuable
tools for enriching for glycoproteins and selecting for specific types of glycosylation. Herein a method using
multidimensional intact protein fractionation and LC-MS/MS analysis is described. Immunodepletion is
used to remove highly abundant proteins from human plasma, followed by glycoform separation using
multi-lectin affinity chromatography, in which specific lectins are chosen to capture and elute specific types
of glycosylation. Reversed-phase chromatography prior to digestion is used for further fractionation, allow-
ing for an increased number of protein identifications of moderate- to low-abundant proteins detectable in
plasma . This method also incorporates isotopic labeling during alkylation for relative quantitation between
two samples (such as a case and control). A bottom-up, tandem mass spectrometry-based proteomics
approach is used for protein identification and quantitation, and allows for screening glycoform-specific
changes across hundreds of plasma proteins.
Key words Multi-lectin affinity chromatography, Glycoproteomics, Protein glycosylation, Plasma

proteomics
1 Introduction
Glycosylation is generally thought to be the most widespread

post-translational modification and it is predicted that more than
half of mammalian proteins are glycosylated [1, 2]. In eukaryotes,
most glycans are attached to proteins at asparagine (N-glycans) or
serine/threonine (O-glycans). The method presented herein is
focused on the study of N-linked glycoproteins. N-linked glycosyl-
ation is a prevalent form of glycans covalently attached to an aspar-
agine given an N-X-S/T motif, where X is any amino acid excluding
proline.
99
100 Sarah M. Totten et al.
Protein glycosylation is a biologically relevant post-translational

modification that plays an important role in disease. For example,
glycosylation is aberrant in cancer and glycosylation changes are a
widespread feature of tumor progression and are related to cancer
grade, invasion, and metastasis [3]. In cancer, key cellular pathways
and interactions are disrupted by altered protein glycosylation,
resulting in opportunities for metastasis. In addition to a variety of
cancers, aberrant glycosylation has been associated with several
other human diseases, including Alzheimer’s disease, multiple scle-
rosis, inflammatory bowel disease, and a number of autoimmune
diseases [4–9]. New tools for systematic studies of protein glyco-
sylation would substantially improve our understanding of its
involvement in cellular functions and its relationship to disease.
Mass spectrometry is an emerging tool of choice for studying gly-
cosylation. Among the most well developed approaches include
glycomics, in which glycans are released from the proteins and
characterized [10–16]. The disadvantage to this method is the loss
of protein- and site-specific information. Strategies for the in-depth
characterization of protein-specific glycosylation are being devel-
oped and are more widely available, however are often limited to
studying less complex mixtures, single proteins of interest, or
highly abundant serum proteins, making it difficult to apply such
strategies to systematically studying complex biological mixtures
[17–19]. In recent years, new technologies have been developed
to analyze intact glycopeptides, and mass spectrometry-based
technologies and data analysis tools are constantly improving
[20–22].
Lectins are sugar-binding proteins with specificities for particu-
lar glycan structures. Lectins have been used in a variety of formats
to study glycoproteins. For example, serial lectin chromatography
columns have been used to separate protein glycoforms [23, 24].
Multi-lectin affinity columns have also been used to capture and
elute glycoproteins in separate fractions [25–27]. Lectin-based
glycoproteomic analysis techniques have been applied for bio-
marker identification in plasma [27–31], including, for example,
lectin arrays for the analysis of fucosylated proteins in ovarian can-
cer serum [32] and lectins for the analysis of pancreatic cancer
serum [33].
The method described here is a protocol for using multidi-
mensional intact protein fractionation and liquid chromatography
coupled with tandem mass spectrometry (LC-MS/MS) to detect
changes in the relative quantitation of specific glycoforms of mod-
erate- to low-abundant plasma proteins. This protocol incorpo-
rates immunodepletion, multi-lectin affinity chromatography
(M-LAC), and reversed-phase (RP) chromatography (see Fig. 1) to
address the issue of the complexity and wide dynamic range of
biological mixtures while simultaneously separating lower abun-
dant proteins into separate fractions containing specific glycoforms.
MLAC for Protein Glycoforms 101
200 µL 200 µL
Sample 1 (Case) Sample 2 (Control)
Immunodepletion Immunodepletion
Take Flow-through Take Flow-through
Reduction/Alkylation Reduction/Alkylation
Light Acrylamide Labeling Heavy Acrylamide Labeling
Combine Sample 1 and

Sample 2
Concentrate/Buffer Exchange
Multi-Lectin Affinity
Chromatography
Unbound AAL PHA-L/E
Reversed-Phase
Fractionation (x3)
Tryptic Digestion
LC-MS/MS of RP Fractions
Protein Identification and

Relative Quantitation (H/L)
Fig. 1 Multidimensional intact protein fractionation workflow. Plasma is first

immunodepleted of 14 abundant proteins, followed by reduction and alkylation.
Shown here is an example of how two depleted and isotopically labeled samples
can be mixed together to achieve relative quantitation of glycoproteins. Isotopic
labeling is performed during the alkylation step using heavy (13C) and light (12C)
acrylamide. After samples 1 and 2 are labeled and combined, the protein mixture
is separated by multi-lectin affinity chromatography (M-LAC). Glycoproteins are
eluted in series from the M-LAC column and further separated by reversed-
phase (RP) chromatography. RP fractions are then digested with trypsin and ana-
lyzed by LC-MS/MS
By using an M-LAC approach, in which several lectins are com-

bined onto a single column and packed in-house, we can select for
specific types of glycosylation, as shown in Fig. 2. Furthermore, by
eluting the bound glycoproteins from each individual lectin in
multiple elution steps (versus eluting all captured glycoproteins
together), specific glycoforms of a given protein are kept in distinct
fractions. This provides a way to screen glycosylation changes
across hundreds (possibly thousands) of proteins and retain
protein-specific information. With a wide variety of agarose bound
Protein Mixture
AAL
PHA-L/E
1x PBS Acetic Acid

(UNB) (PHA-L/E)
L-Fucose
N (AAL)
or
or
N
N
Fig. 2 Multi-lectin affinity chromatography schematic. A mixture of depleted and

isotopically labeled proteins is separated on an M-LAC column containing
agarose-bound lectins Aleuria aurantia lectin (AAL) and Phaseolus vulgaris leu-
coagglutinin (PHA-L) and Phaseolus vulgaris erythroagglutinin (PHA-E). Captured
glycoproteins are released in a series of elution steps. The flow-through (mobile
phase 1× PBS) containing the non- or otherwise-glycosylated proteins is col-
lected first. Next, an L-fucose buffer is used to release fucosylated glycoforms
from AAL. AAL preferentially binds core-fucosylated structures, and has a weak
affinity for α1-4-linked fucose on the antennae. In the third elution step, low-pH
mobile phase (100 mM acetic acid, pH 3.8) is used to release the remaining
glycoproteins bound to the PHA-L/E lectins
lectins now commercially available, it is possible to personalize a

multi-lectin column by choosing lectins that will select for specific
types of protein glycoforms (fucosylation, sialylation, oligoman-
nose, etc.), tailoring the analysis for the specific needs of certain
applications. Furthermore, the collected M-LAC fractions are sub-
sequently separated by reversed-phase chromatography on the
intact protein level, allowing for deeper analysis and increased
number of protein identifications, as shown in Fig. 3. Although
this protocol has been optimized for glycoprotein analysis in
plasma, it has also been successfully applied to protein extractions
from tissue and cell culture extracts.
2390 A. Immunodepleon
1990 Flow-through
1590
1190
790
390
-10
-1 5 11 17 23 29 35 41 47 53 59 65
150
B. M-LAC PHA
130
110 Unbound AAL
Absorbance (mAU)
90
70
50
30
10
-10 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34
35
30 C. Reversed-Phase
25
20
15
10
5
0
-5
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40
Retenon Time (minutes)
Fig. 3 A series of chromatographic separations on intact plasma proteins. In our
multidimensional separation approach, immunodepletion (a) is used to deplete
plasma of highly abundant proteins. The flow-through from that separation con-
tains moderate- to low-abundant plasma proteins, which are then separated by
glycoforms using M-LAC (b). Boxes are inset to show suggested fraction collection.
Three separate fractions are collected from M-LAC and are further fractionated at
the intact protein level by reversed-phase chromatography (c). Inset vertical lines
represent the fractions collected in 1 min intervals
2 Materials
2.1 Depletion 1. CaptureSelect™ HumanPlasma14 affinity resin (ThermoFisher)

of Human Plasma (see Note 1 for alternatives).
2.
Omnifit® glass columns, 5.6 mL bed volume,
100 mm × 10 mm L × internal diameter (ID) One column has
a capacity for 125 μL of plasma. In this protocol, two of these
columns are connected and used in tandem for a capacity of
250 μL of plasma.
3. Binding Buffer A: Phosphate-buffered saline (PBS) 1× solution.
4. Elution Buffer B: 100 mM glycine, pH 2.5 with hydrochloric
acid (HCl)
5. 200 μL of human plasma of each sample.
6. 3000 nominal molecular weight limit (NMWL) centrifugal fil-
ters, 15 mL (Amicon® Ultra—15, Millipore).
2.2 Reduction 1. Bradford protein assay kit (including protein standards and
and Alkylation Coomassie dye) and spectrophotometer for measuring protein
of Depleted Plasma concentration.
2. Protein denaturation buffer: 8 M urea, 50 mM Tris–HCl,
0.05 % octyl β-d-glucopyranoside, pH 7.5, prepared in
100 mM ammonium bicarbonate.
3. Dithiothreitol (DTT) for protein reduction.
4. Acrylamide for protein alkylation. For isotopic labeling, both
light (12C) and heavy (13C) acrylamide is required. 1,2,3-13C3
(heavy) acrylamide can be purchased from Cambridge Isotope
Laboratories. If not labeling, alternative alkylating agents can
be used, such as iodoacetamide or iodoacetic acid (not covered
in this protocol, see Note 2).
5. 3000 NMWL centrifugal filters, 4 mL (Amicon® Ultra—4,
Millipore) for buffer exchange.
6. Phosphate-buffered saline, 1× solution
2.3 Multi-Lectin 1. Agarose-bound Lectins Aleuria aurantia lectin (AAL) and

Affinity Phaseolus vulgaris leucoagglutinin (PHA-L) and Phaseolus vul-
Chromatography garis erythroagglutinin (PHA-E) (see Note 3 for alternatives).
2. Omnifit® glass columns, 5.6 mL bed volume, 100 mm × 10 mm
L × ID.
3. M-LAC loading buffer: 1× PBS (see Note 4).
4. 200 mM L-fucose (see Note 5).
5. 100 mM acetic acid.
6. 3K NMWL centrifugal filters, 4 mL (Amicon® Ultra—4,
Millipore).
2.4 Reversed-Phase 1. Reversed-phase HPLC analytical column (C8 or similar,

Chromatography 100 mm × 2.1 mm ID).
of Intact Proteins 2. Buffer A: 0.1 % trifluoroacetic acid (TFA) in water.
from M-LAC Fractions
3. Buffer B: 0.1 % TFA in acetonitrile.
2.5 Tryptic Digestion 1. 50 mM ammonium bicarbonate in 4 % acetonitrile.

of Lyophilized Protein 2. 200 μL of 0.1 μg/μL trypsin (Promega, approximately 1:20
from Reversed-Phase enzyme to protein).
(RP) Fractions
3. Trypsin resuspension buffer (provided by manufacturer).
2.6 LC-MS/MS 1. Conical mass spectrometry vials (250 μL, with snap septum
Analysis of Tryptic lids, or other autosampler vials).
Peptides 2. Buffer A: 0.1 % formic acid (FA) in water.
and Glycopeptides
3. Buffer B: 0.1 % FA in acetonitrile.
4. C18 analytical column (Picofrit 75 μm ID, New Objective,
packed in-house with MagicC18 AQ 5 μm, 100 A resin, packed
to 25 cm, or similar).
5. C18 trap column (Thermo Scientific Acclaim PepMap 100,
5 μm particle size, 5 mm length, 300 μm ID).
2.7 Equipment 1. Depletion, M-LAC separation, and reversed-phase chroma-

and Instrumentation tography require a HPLC. The HPLC separations described in
this protocol were optimized on an Agilent 1260 Infinity Bio-
Inert Quaternary HPLC, equipped with an automatic fraction
collector, multiple wavelength UV detector (MWD), and a
manual bio-inert injector. Agilent’s ChemStation Software
(version C.01.05) was used for data collection and analysis.
2. LC-MS/MS experiments require an online nano-HPLC system
coupled to a high resolution mass spectrometer with MS/MS
capabilities. Protocols described herein were performed on a
Dionex Ultimate 3000 RSLC nanoLC system coupled to a
LTQ-Orbitrap Elite mass spectrometer (Thermo Scientific).
3. Additional equipment include a microplate absorbance reader
or spectrophotometer, a benchtop centrifuge, and a lyophilizer
or freeze drying system.
3 Procedures
3.1 Depletion CaptureSelect® HumanPlasma14 depletion material is comprised

of Human Plasma of affinity ligands derived from Camelidae antibodies used for the
removal of 14 abundant human plasma proteins (albumin, IgG,
IgM, IgA, IgE, IgD, and free light chains, transferrin, fibrinogen,
α-1 antitrypsin, α-2 macroglobulin, α-1 acid glycoprotein, apoli-
poprotein A1, and haptoglobulin). CaptureSelect® HP14
immobilized on agarose support resin is commercially available

(ThermoFisher Scientific). The efficiency, reproducibility, and
specificity of a 5.6 mL CaptureSelect® HP14 column have been
previously assessed in our laboratory [34]. The loading capacity of
5.6 mL glass column was determined to be 125 μL of plasma (or
250 μL on two 5.6 mL columns in tandem) [34]. This protocol
was optimized on an Agilent 1260 Infinity HPLC, however any
HPLC system equipped with two pumps, and a UV detector
(monitoring at 280 nm) that can pump up to 5 mL/min can be
used. See Note 6 for comments regarding chromatographic gradi-
ents and fraction collection times. An example chromatogram of
an immunodepletion is shown in Fig. 3a.
1. Gravity pack agarose bound CaptureSelect® HumanPlasma14
resin onto a 5.6 mL Omnifit glass column. Column should be
stored at 4 °C and operated at 10 °C. Maximum pressure
should not exceed 600 psi. Proteins are detected with UV at
280 nm.
2. For plasma depletion, dilute 200 μL of human plasma 1:5 with
1× PBS. The depletion is performed on two 5.6 mL columns
set up in tandem.
3. Load 1 mL of diluted plasma at 0.5 mL/min with Binding
Buffer A (1× PBS).
4. Once loaded, maintain the flow rate at 0.5 mL/min for 32 min
and then increase to 2.5 mL/min (see Note 6). Collect flow-
through fraction containing the moderate- to low-abundance
proteins of interest of the depleted plasma.
5. Elute bound (high-abundance) proteins with 100 % Elution
Buffer B (100 mM glycine, pH 2.5) at 2.5 mL for 15 min.
Collect eluted fraction and freeze for later use if desired.
6. Re-equilibrate column at 2.5 mL/min for 15 min with Binding
Buffer A (1× PBS), diverting the flow to waste.
7. Using a 4 mL, 3K NMWL centrifugal filter, concentrate the
flow-through fraction (already in 1× PBS) to 100 μL and trans-
fer to a 1.5 mL Eppendorf tube.
8. In isotopic labeling experiments, repeat on a 200 μL aliquot of
a second sample to be compared to Sample 1 (i.e., the case
sample, as shown in Fig. 1). The two depleted and isotopically
labeled samples will be combined, as described in detail in the
following sections (see Note 2 for alternative labeling study
design).
3.2 Reduction The amount of depleted protein from plasma is determined using
and Alkylation a standard Bradford protein assay. Typically between 0.75 and
of Depleted Plasma 1.0 mg of protein is recovered from the depletion of 200 μL of
human plasma. Before separation by multi-lectin affinity chroma-
tography, proteins must be reduced and alkylated.
1. Perform a Bradford protein assay to determine protein con-

centration in depleted and concentrated plasma samples.
2. To the depleted plasma mixture, add the Denaturation Buffer
10:1 (buffer to sample, by volume).
3. Reduce the protein disulfide bonds using DTT by adding
0.616 mg of DTT per mg of protein. Let the reaction run for
2 h at room temperature.
4. Alkylate the reduced protein by adding light acrylamide to
Sample 1 at 7.1 mg per mg of reduced protein, and heavy
acrylamide at 7.4 mg per mg of reduced protein to Sample 2,
as shown in Fig. 1. Leave reaction in the dark at room tem-
perature for one hour (see Note 2).
5. Once reduced and alkylated, mix the heavy-labeled sample and
its paired light-labeled sample together. Concentrate the mix-
ture and buffer exchanged using a 3K Millipore filter into 1×
PBS to 1 mL in preparation for multi-lectin affinity
chromatography
3.3 Multi-Lectin Multi-lectin affinity chromatography is used to separate moderate-

Affinity to low-abundance plasma proteins by specific glycoforms. The
Chromatography method described here combines agarose-bound Aleuria aurantia
lectin (AAL), Phaseolus vulgaris erythroagglutinin and Phaseolus
vulgaris leucoagglutinin (PHA-E and PHA-L) to capture core
fucosylated glycoproteins, and proteins carrying highly branched
complex type glycans, respectively, as shown in Fig. 2 (see Note 3).
By eluting each individual fraction in series (versus releasing all
glycoproteins bound to all lectins in a single elution step), specific
types of glycoforms are separated into distinct fractions depending
on their affinity for each lectin (see Note 7). M-LAC columns are
gravity packed with agarose-bound lectins purchased from Vector
Labs (Burlingame, CA). Chromatography is performed on an
HPLC equipped with a quaternary pump (three pumps are
required for M-LAC experiments that use a three buffer system).
See Fig. 3b for an example M-LAC chromatographic separation
and fraction collection.
1. To pack a 5.5 mL Omnifit glass column, combine 2 mL of a
1:1 slurry of AAL in suspension buffer (2 mg of protein per
mL of gel) and 2 mL of a 1:1 slurry of PHA-E and PHA-L (for
a combined 4 mL of PHA lectin, each 3 mg per mL of gel) in
a 15 mL conical tube with gentle swirling (vortexing the slurry
could destabilize the media). Rinse each bottle with 1× PBS
and add the rinse to the lectin mixture to ensure all of the
slurry is used. Slowly load the lectin mixture onto the column
and gravity pack a few milliliters at a time, as per the manufac-
turers’ instructions, being sure to never let the column go
completely dry.
2. Run a 1× PBS blank across the newly packed column to equili-

brate and to check for leaks and pressure issues.
3. When the column is ready, load 1 mL of sample (approximately
1.0–1.5 mg of protein) into a 2 mL loop and inject at 0.5 mL/
min with M-LAC loading buffer 1× PBS for 15 min. Collect
the flow-through (see Note 6).
4. Elute the unbound material containing non- or otherwise-
glycosylated proteins with 1× PBS at 2.5 mL/min. Collect the
eluent. Maximum pressure should not exceed 100 bar.
5. Elute glycoproteins containing core fucosylated glycans (those
bound to AAL) from the M-LAC column with 200 mM L-fucose
in 1× PBS at 2.5 mL/min. Collect the eluent (see Notes 7
and 8).
6. Elute glycoproteins containing highly branched, complex-type
glycans (those bound to PHA-L/E) by lowering the pH of the
mobile phase with 100 mM acetic acid, pH 3.8 at 2.5 mL/
min. Collect the eluent.
7. Divert the flow to waste and re-equilibrate the column with 1×
PBS for 15 min.
8. Concentrate the three collected fractions (unbound, AAL, and
PHA-L/E) using 3K NMWL Amicon centrifugal filters to
200 μL in preparation for reversed-phase fractionation.
3.4 Reversed-Phase Each of the three serial-eluted lectin fractions collected from M-LAC
Chromatography are then further separated by reversed-phase fractionation at the
of Intact Proteins intact protein level. By separating at the protein level, the mixture of
from M-LAC Fractions proteins is further de-complexed, allowing for deeper analysis and
more protein identifications. Furthermore protein glycoforms are
confined to more distinct reversed-phase fractions prior to digestion,
allowing separation of protein isoforms. Reversed-phase fraction-
ation is performed on a 100 mm × 2.1 mm ID stainless steel column
pre-packed with POROS®R2 (Applied Biosystems), 2000 Å particle
size poly styrene-divinylbenzene immobile phase. Maximum pres-
sure should not exceed 170 bar.
1. Load the concentrated sample onto a 500 μL sample loop at
0.8 mL/min for 5 min with reversed-phase Solvent A (0.1 %
trifluoroacetic acid in water).
2. Separate the proteins using a gradient of increasing organic
content in the mobile phase. Between 5 and 38 min, bring
reversed-phase Buffer B (0.1 % trifluoroacetic acid in aceto-
nitrile) to 90 %. Hold the gradient at 90 % Buffer B for 2 min,
then re-equilibrate the column with 95 % Buffer A for
10 min.
3. Collect reversed-phase fractions in an automatic analytical scale
fraction collector (Agilent 1260) into 1 mL racked microtubes.
In total 24, 0.8 mL fractions are collected (one per minute from
8 to 31 min, as shown in Fig. 3c). Fractions can be collected
manually if automated fraction collection is unavailable.
4. Repeat steps 1–3 for remaining M-LAC fractions. Overall,
three RP separations are performed (one per M-LAC frac-
tion—unbound, AAL, and PHA-L/E). Per sample, there are 3
sets of 24 RP fractions.
5. Freeze RP fractions at −80 °C and lyophilize.
3.5 Tryptic Digestion 1. Reconstitute each of the 24 RP fractions in 50 μL of 50 mM

of Lyophilized Protein ammonium bicarbonate in 4 % acetonitrile. RP fractions
from RP Fractions containing small amounts of protein can be combined as
needed.
2. Digest each RP fraction with 0.5 μg of trypsin at 37 °C for
18 h (or overnight).
3.6 LC-MS/MS RP fractions are then analyzed by nano LC-MS/MS. In total, each
Analysis of Tryptic sample will have 72 RP fractions in total (24 unbound, 24 AAL,
Peptides and 24 PHA-L/E—or less if some were combined, which is rec-
and Glycopeptides ommended to improve signal in the mass spectrometer).
1. Transfer the 50 μL of each RP fraction to a mass spectrometry
vial.
2. Load 15 μL (approximately between 5 and 10 μg of peptides)
onto a C18 trap column at 5 μL/min, briefly de-salt and con-
centrate, then subsequently separate on a 25 cm C18 analytical
column (Picofrit 75 μm ID, New Objective, packed in-house
with MagicC18 AQ resin). Tryptic peptides are separated using
a multistep gradient at a flow rate of 0.6 μL/min in which Buffer
B (0.1 % FA in acetonitrile) is increased from 0 % (100 % Buffer
A, 0.1 % FA in water) to 85 % over 120 min. Re-equilibrate the
analytical column for 20 min at 98 % Buffer A.
3. Ionize nano-LC eluent by electrospray ionization at 2.25 kV
with the capillary temperature set to 200 °C.
4. In each MS/MS experiment, perform an initial MS1 scan over
an m/z range of 400–1800, followed by 10 data-dependent
collision-induced dissociation fragmentation events on the 10
most intense +2 or +3 ions from the MS1 spectrum over an
acquisition time of 140 min.
5. Acquired data can be processed by the Computational Proteomics
Analysis System (CPAS) [35] pipeline using X! Tandem search
algorithm [36, 37] for peptide identifications and the Q3 quan-
titation algorithm to calculate heavy-to-light ratios for cysteine-
containing acrylamide labeled peptides [38]. PeptideProphet
[39, 40] and ProteinProphet can be used to validate peptide and
protein identifications respectively.
4 Notes
1. Alternative materials for immunodepletion of plasma are avail-

able and can be used in this protocol, for example multiple
affinity removal column human 14 (Agilent Technologies) and
Seppro IgY14 (Sigma-Aldrich). Both of these chemistries
deplete plasma of 14 highly abundant proteins, although with
slightly different specificities. Serum samples can also be used
in lieu of plasma.
2. As described above, isotopic labeling for relative quantitation is
incorporated into this protocol during protein alkylation.
Heavy (13C) and light (12C) acrylamide are used to label cyste-
ine residues yielding relative ratios of peptides, from which
protein levels can be inferred, between two samples, as
described in Faca et al. [38]. This method can be used to com-
pare case versus control samples, changes between two time
points, or case and controls can be compared against a refer-
ence pool. With three 13C, the heavy acrylamide will change
the mass of the peptide by 3 Da. The workflow described
herein can also be applied to experiments not incorporating
isotopic labeling, in which case other alkylating agents, such as
iodoacetamide or iodoacetic acid, can be used.
3. One of the great advantages of packing your own multi-lectin
affinity column in-house is the ability to incorporate different
combinations of lectins to suit the needs of a particular experi-
ment. In this protocol, AAL and PHA-L/E were used to spe-
cifically enrich for core-fucosylated glycoforms and glycoforms
containing highly branched complex glycan structures, respec-
tively, as shown in Fig. 2. However, there are many additional
commercially available agarose-bound lectins (Vector
Laboratories) with affinity for a variety of glycan substrates,
ranging from those that broadly capture all N-glycan types to
those with affinity for very specific structures differing only by
glycosidic linkages. Some commonly used lectins include
Concanavalin A (ConA), which binds mannose and has a broad
affinity for oligomannose, hybrid, and biantennary glycan
structures, and wheat germ agglutinin (WGA) that bind
N-acetylglucosamine and sialic acid as substrates, recognizing
chitobiose at the core of many glycans. These lectins are used
to widely capture and enrich for all N-glycans. Alternatively,
glycoproteins containing specific glycan structures can be
enriched for using lectins such as Ulex europaeus agglutinin I
(UEA I) or Lotus tetragonolobus lectin (LTL) that bind Fuc(α-
1,2)Gal (H antigen of ABO blood groups), or Sambucus nigra
lectin that has preferential affinity for N-acetylneuraminic acids
with an α-2,6 linkage to a terminal galactose (see vectorlabs.
com for useful information regarding the characteristics of
their lectin products).
4. If lectins other than those described here are used in a multi-

lectin column, it is important to note they may require different
loading buffers. Some lectins do not perform as well in PBS and
often require metal ions to activate carbohydrate binding. For
example, the widely used ConA lectin performs optimally in
loading buffers containing CaCl2, MnCl2, and NaCl.
Additionally, the eluting or inhibiting sugar is also specific to
the affinity of each lectin. For example, competitive saccharide
binding elution buffer for ConA requires α-methylmannoside
and α-methylglucoside (whereas L-fucose was the sugar used
for release of proteins bound to AAL in this protocol).
5. Fucose buffer should be prepared fresh, ideally the day of use,
to prevent bacterial growth. Bacteria will consume the fucose,
thereby reducing the buffer concentration.
6. General elution times for each step of the chromatography
performed in this protocol is described here, however it is
advisable that exact timing for chromatographic gradients and
collection time points are optimized to individual columns,
sample types, and HPLC systems.
7. Serial elution of glycoproteins from specific lectins during
multi-lectin chromatography has advantages and disadvan-
tages. One advantage is being able to collect multiple fractions
containing specific glycoforms as opposed to having one
“bound” fraction in which all glycoproteins captured by the
multiple lectins present in the column are eluted together.
Separating proteins by specific glycoforms provides insight
into changes in glycosylation between case and control sam-
ples. Additionally, collecting multiple fractions during M-LAC
separations further fractionate and de-complex biological mix-
tures containing thousands of proteins and their glycoforms. It
should be noted, however, that some of the disadvantages of
this method include the nonspecific binding of lectins and
ambiguity surrounding the types of glycoforms present in each
M-LAC fraction. Elution by competitive saccharide binding
may not always be entirely complete, resulting in any remain-
ing glycoproteins bound to that lectin being eluted in the sub-
sequent fraction eluted by low pH. Furthermore, glycoproteins
containing multiple types of glycans (for example both fucosyl-
ated and highly branched structures) may end up in either or
multiple M-LAC fractions depending on how strong the affin-
ity is for either lectin.
8. When bound glycoproteins are serial eluted from a multi-lectin
column, the order in which they are eluted is important.
Glycoproteins being released from lectins via competitive sac-
charide binding should always be eluted before the glycopro-
teins being released from another lectin via low-pH elution
(acetic acid mobile phase). The low-pH mobile phase will release
all glycoproteins bound to all lectins.
References
1. Wang H, Hanash S (2011) Intact-protein 10. Zaia J (2010) Mass spectrometry and glycomics.
analysis system for discovery of serum-based OMICS 14(4):401–418
disease biomarkers. Methods Mol Biol 11. Mechref Y, Hu Y, Desantos-Garcia JL, Hussein
728:69–85 A, Tang H (2013) Quantitative glycomics strat-
2. Apweiler R, Hermjakob H, Sharon N (1999) egies. Mol Cell Proteomics 12(4):874–884
On the frequency of protein glycosylation, as 12. An HJ, Kronewitter SR, de Leoz MLA, Lebrilla
deduced from analysis of the SWISS-PROT CB (2009) Glycomics and disease markers.
database. Biochim Biophys Acta 1473(1):4–8 Curr Opin Chem Biol 13(5-6):601–607
3. Varki A, Cummings RD, Esko JD, et al., edi- 13. Fujitani N, J-i F, Araki K, Fujioka T, Takegawa
tors. Cold Spring Harbor (NY): Cold Spring Y, Piao J, Nishioka T, Tamura T, Nikaido T,
Harbor Laboratory Press; 2009. Bookshelf ID: Ito M, Nakamura Y, Shinohara Y (2013) Total
NBK1963 PMID: 20301279 cellular glycomics allows characterizing cells
4. Mkhikian H, Grigorian A, Li CF, Chen H-L, and streamlining the discovery process for cel-
Newton B, Zhou RW, Beeton C, Torossian S, lular biomarkers. Proc Natl Acad Sci U S A
Tatarian GG, Lee S-U, Lau K, Walker E, 110(6):2105–2110
Siminovitch KA, Chandy KG, Yu Z, Dennis JW, 14. Hu Y, Zhou S, Yu C-Y, Tang H, Mechref Y
Demetriou M (2011) Genetics and the environ- (2015) Automated annotation and quantita-
ment converge to dysregulate N-glycosylation tion of Glycan by LC-ESI-MS analysis using
in multiple sclerosis. Nat Commun 2:334 MultiGlycan-ESI computational tool. Rapid
5. Palmigiano A, Barone R, Sturiale L, Sanfilippo Commun Mass Spectrom 29(1):135–142
C, Bua RO, Romeo DA, Messina A, Capuana 15. Ruhaak LR, Miyamoto S, Lebrilla CB (2013)
ML, Maci T, Le Pira F, Zappia M, Garozzo D Developments in the identification of glycan
(2016) CSF N-glycoproteomics for early diag- biomarkers for the detection of cancer. Mol
nosis in Alzheimer’s disease. J Proteomics Cell Proteomics 12(4):846–855
131:29–37 16. Zhou S, Hu Y, DeSantos-Garcia JL, Mechref Y
6. Theodoratou E, Campbell H, Ventham NT, (2015) Quantitation of permethylated
Kolarich D, Pucic-Bakovic M, Zoldos V, N-glycans through multiple-reaction monitor-
Fernandes D, Pemberton IK, Rudan I, ing (MRM) LC-MS/MS. J Am Soc Mass
Kennedy NA, Wuhrer M, Nimmo E, Annese V, Spectrom 26(4):596–603
McGovern DPB, Satsangi J, Lauc G (2014) 17. Wu S-W, Pu T-H, Viner R, Khoo K-H (2014)
The role of glycosylation in IBD. Nat Rev Novel LC-MS2 product dependent parallel data
Gastroenterol Hepatol 10:588–600, advance acquisition function and data analysis workflow
online publication for sequencing and identification of intact glyco-
7. Liang H-C, Russell C, Mitra V, Chung R, Hye A, peptides. Anal Chem 86(11):5478–5486
Bazenet C, Lovestone S, Pike I, Ward M (2015) 18. Saba J, Dutta S, Hemenway E, Viner R (2012)
Glycosylation of human plasma clusterin yields a Increasing the productivity of glycopeptides
novel candidate biomarker of Alzheimer’s dis- analysis by using higher-energy collision
ease. J Proteome Res 14(12):5063–5076 dissociation-accurate mass-product-dependent
8. Goulabchand R, Vincent T, Batteux F, J-f E, electron transfer dissociation. Int J Proteomics
Guilpain P (2014) Impact of autoantibody gly- 2012:560391
cosylation in autoimmune diseases. Autoimmun 19. Hong Q, Ruhaak LR, Stroble C, Parker E,
Rev 13(7):742–750 Huang J, Maverakis E, Lebrilla CB (2015) A
9. Lauc G, Huffman JE, Pučić M, Zgaga L, method for comprehensive glycosite-mapping
Adamczyk B, Mužinić A, Novokmet M, and direct quantitation of serum glycoproteins.
Polašek O, Gornik O, Krištić J, Keser T, Vitart J Proteome Res 14(12):5179–5192
V, Scheijen B, Uh H-W, Molokhia M, Patrick 20. Mayampurath A, Yu C-Y, Song E, Balan J,
AL, McKeigue P, Kolčić I, Lukić IK, Swann O, Mechref Y, Tang H (2014) Computational
van Leeuwen FN, Ruhaak LR, Houwing- framework for identification of intact glycopep-
Duistermaat JJ, Slagboom PE, Beekman M, de tides in complex samples. Anal Chem
Craen AJM, Deelder AM, Zeng Q, Wang W, 86(1):453–463
Hastie ND, Gyllensten U, Wilson JF, Wuhrer
M, Wright AF, Rudd PM, Hayward C, 21. Hu H, Khatri K, Zaia J (2016) Algorithms and
Aulchenko Y, Campbell H, Rudan I (2013) design strategies towards automated glycopro-
Loci associated with N-glycosylation of human teomics analysis. Mass Spectrom Rev n/a-n/a
immunoglobulin G show pleiotropy with auto- 22. Mayampurath A, Song E, Mathur A, Yu C-y,
immune diseases and haematological cancers. Hammoud Z, Mechref Y, Tang H (2014)
PLoS Genet 9(1):e1003225 Label-free glycopeptide quantification for
biomarker discovery in human sera. J Proteome Press, Totowa, NJ, pp 373–396.

Res 13(11):4821–4832 doi:10.1007/978-1-60327-064-9_29
23. Drake PM, Schilling B, Niles RK, Braten M, 32. Wu J, Xie X, Liu Y, He J, Benitez R,
Johansen E, Liu H, Lerch M, Sorensen DJ, Li Buckanovich RJ, Lubman DM (2012)
B, Allen S, Hall SC, Witkowska HE, Regnier Identification and confirmation of differentially
FE, Gibson BW, Fisher SJ (2011) A lectin expressed fucosylated glycoproteins in the
affinity workflow targeting glycosite-specific, serum of ovarian cancer patients using a lectin
cancer-related carbohydrate structures in array and LC-MS/MS. J Proteome Res
trypsin-digested human plasma(). Anal 11(9):4541–4552
Biochem 408(1):71–85 33. Zhao J, Qiu W, Simeone DM, Lubman DM
24. Jung K, Cho W, Regnier FE (2009) (2007) N-linked glycosylation profiling of
Glycoproteomics of plasma based on narrow pancreatic cancer serum using capillary liquid
selectivity lectin affinity chromatography. phase separation coupled with mass spectro-
J Proteome Res 8(2):643–650 metric analysis. J Proteome Res
25. Gbormittah FO, Hincapie M, Hancock WS 6(3):1126–1138
(2014) Development of an improved fraction- 34. Kullolli M, Warren J, Arampatzidou M, Pitteri
ation of the human plasma proteome by a com- SJ (2013) Performance evaluation of affinity
bination of abundant proteins depletion and ligands for depletion of abundant plasma pro-
multi-lectin affinity chromatography. teins. J Chromatogr B Analyt Technol Biomed
Bioanalysis 6(19):2537–2548 Life Sci 939:10–16
26. Lee LY, Hincapie M, Packer N, Baker MS, 35. Rauch A, Bellew M, Eng J, Fitzgibbon M,
Hancock WS, Fanayan S (2012) An optimized Holzman T, Hussey P, Igra M, Maclean B,
approach for enrichment of glycoproteins Lin CW, Detter A, Fang R, Faca V, Gafken P,
from cell culture lysates using native multi- Zhang H, Whiteaker J, States D, Hanash S,
lectin affinity chromatography. J Sep Sci Paulovich A, McIntosh MW (2006)
35(18):2445–2452 Computational proteomics analysis system
27. Kullolli M, Hancock WS, Hincapie M (2008) (CPAS): an extensible, open-source analytic
Preparation of a high-performance multi-lectin system for evaluating and publishing pro-
affinity chromatography (HP-M-LAC) adsor- teomic data and high throughput biological
bent for the analysis of human plasma glyco- experiments. J Proteome Res 5(1):112–121
proteins. J Sep Sci 31(14):2733–2739 36. Craig R, Beavis RC (2004) TANDEM: match-
28. Fanayan S, Hincapie M, Hancock WS (2012) ing proteins with tandem mass spectra.
Using lectins to harvest the plasma/serum Bioinformatics 20(9):1466–1467
glycoproteome. Electrophoresis 33(12): 37. Craig R, Cortens JP, Beavis RC (2004) Open
1746–1754 source system for analyzing, validating, and
29. Madera M, Mechref Y, Klouckova I, Novotny storing protein identification data. J Proteome
MV (2007) High-sensitivity profiling of gly- Res 3(6):1234–1242
coproteins from human blood serum 38. Faca V, Coram M, Phanstiel D, Glukhova V,
through multiple-lectin affinity chromatog- Zhang Q, Fitzgibbon M, McIntosh M, Hanash
raphy and liquid chromatography/tandem S (2006) Quantitative analysis of acrylamide
mass spectrometry. J Chromatogr B labeled serum proteins by LC-MS/MS. J
845(1):121–137 Proteome Res 5(8):2009–2018
30. Song E, Zhu R, Hammoud ZT, Mechref Y 39. Keller A, Nesvizhskii AI, Kolker E, Aebersold
(2014) LC–MS/MS quantitation of esopha- R (2002) Empirical statistical model to esti-
gus disease blood serum glycoproteins by mate the accuracy of peptide identifications
enrichment with hydrazide chemistry and lec- made by MS/MS and database search. Anal
tin affinity chromatography. J Proteome Res Chem 74(20):5383–5392
13(11):4808–4820 40. Nesvizhskii AI, Keller A, Kolker E, Aebersold
31. Mechref Y, Madera M, Novotny MV (2008) R (2003) A statistical model for identifying
Glycoprotein enrichment through lectin affin- proteins by tandem mass spectrometry. Anal
ity techniques. In: Posch A (ed) 2D PAGE: Chem 75(17):4646–4658
sample preparation and fractionation. Humana
Chapter 10
Parallel Exploration of Interaction Space by BioID

and Affinity Purification Coupled to Mass Spectrometry
Geoffrey G. Hesketh, Ji-Young Youn, Payman Samavarchi-Tehrani,
Brian Raught, and Anne-Claude Gingras
Abstract
Complete understanding of cellular function requires knowledge of the composition and dynamics of
protein interaction networks, the importance of which spans all molecular cell biology fields. Mass
spectrometry-based proteomics approaches are instrumental in this process, with affinity purification coupled
to mass spectrometry (AP-MS) now widely used for defining interaction landscapes. Traditional AP-MS
methods are well suited to providing information regarding the temporal aspects of soluble protein–
protein interactions, but the requirement to maintain protein–protein interactions during cell lysis and AP
means that both weak-affinity interactions and spatial information is lost. A more recently developed
method called BioID employs the expression of bait proteins fused to a nonspecific biotin ligase, BirA*,
that induces in vivo biotinylation of proximal proteins. Coupling this method to biotin affinity enrichment
and mass spectrometry negates many of the solubility and interaction strength issues inherent in traditional
AP-MS methods, and provides unparalleled spatial context for protein interactions. Here we describe the
parallel implementation of both BioID and FLAG AP-MS allowing simultaneous exploration of both spatial
and temporal aspects of protein interaction networks.
Key words Mass spectrometry, BioID, Biotin, Streptavidin, FLAG tag, Proximity labeling, Affinity
purification, Proteomics, Protein interactions, Protein network, Protein identification
1 Introduction
Understanding how proteins associate with one another is paramount

to discovering their molecular function. While multiple experi-
mental strategies have been devised for this purpose [1], the cou-
pling of immunoprecipitation or other affinity purification strategy
with mass spectrometry (referred to as AP-MS) is particularly pow-
erful at detecting interactions for proteins of interest under near-
endogenous conditions [2]. Building on initial successes in the
model organism S. cerevisiae [3, 4], the use of AP-MS in human
cells has grown from initially small-scale experiments consisting of
only one or a few “bait” proteins, to scales that are approaching
the genomics realm [5, 6]. Multiple computational tools have been
115
116 Geoffrey G. Hesketh et al.
developed that can help identify “true” interactions from con-

taminants, by comparing the quantitative mass spectrometry results
to other purifications performed under the same conditions [7, 8],
by comparing against well-defined negative control purifications
[2], or a combination of both [9]. As a result, the quality as well as
the quantity of protein interactions identified by AP-MS has grown
considerably over the past several years.
A key strength of AP-MS is its versatility: it is compatible with
a number of tagging strategies and cell types or, provided that
proper controls are designed, with immunoprecipitation of endog-
enous proteins [10]. Providing that the interactions withstand the
purification conditions, AP-MS can capture relative changes in
interactions imparted by changes in growth conditions, for exam-
ple stimulation by growth factors, or treatment with pharmaco-
logical inhibitors. This enables the analysis of protein interaction
kinetics, especially when coupled to robust quantitative approaches
such as Multiple/Selected Reaction Monitoring [11, 12], Data
Independent Acquisition (DIA, e.g., SWATH, [13, 14]), MS1
intensity measurements [15, 16], or isobaric and isotopic labeling
strategies [17–19].
While the temporal aspects of protein–protein interactions can
be captured by AP-MS, using this approach to understand the spa-
tial component of interactions has been much more challenging.
The central problem lies in the fact that cell lysis must be per-
formed prior to affinity purification, which inevitably results in
mixing of the cellular subcompartments and loss of spatial identity.
In some small scale studies, this has been to some extent mitigated
by combining subcellular fractionation with AP-MS [20], but this
approach is neither very robust, nor scalable, and can only provide
information on the few subcellular fractions that are generated. We
note that emerging approaches such as those employing cell-
permeable cross-linkers can in principle generate spatial informa-
tion for protein interactions (e.g., [21]), but to date, the coverage
of crosslinking information for any protein (or protein complex) is
sparse. Alternative approaches that do not involve a mass spec-
trometer at all (e.g., Proximity Ligation Assay, PLA, Fluorescence
Resonance Energy Transfer, FRET, or Protein-fragment
Complementation Assay, PCA) have been classically used for char-
acterizing the spatial components of protein–protein interactions
[1], but these methods require prior knowledge of the two interac-
tion partners or the capability to execute large scale screens.
Besides the loss of spatial identity in AP-MS (and all biochemi-
cal strategies employing post-cell lysis purification), another chal-
lenge has been the analysis of interactions for proteins that are
refractory to solubilization in standard lysis buffers used for
AP-MS. These notably include membrane proteins, which require
higher concentrations of detergents for extraction from the lipid
bilayer than is typically used to maintain protein–protein interac-
tions in AP-MS. While it is indeed possible to systematically alter
Parallel BioID and FLAG Affinity-Purification Coupled to Mass Spectrometry 117
detergent types and concentrations prior to performing AP-MS

[22], or to screen for multiple conditions of detergents, salts and
buffers [23], it is not clear what interactions are maintained under
these conditions. Furthermore, this systematic condition screen-
ing, while sometimes itself informative [23], drastically reduces the
overall throughput of the process.
Complementary approaches to AP-MS with the potential to
improve spatial resolution have emerged in the past few years. The
first of these, which will be reviewed here, is termed BioID and
consists of the fusion of a bait protein to an enzyme, namely a
mutated bacterial biotin ligase (BirA*) capable of inducing the
activation of biotin to biotin-AMP, but does not directly couple
this activated biotin to a specific sequence [24]. As such, BirA*
creates a cloud of activated biotin-AMP in the vicinity of the bait
that can covalently react with lysine ε-amines, tagging the proxim-
ity partners for the bait of interest. A related approach, APEX, uses
a peroxidase that can mediate the transfer of biotin (from biotin-
phenol) to tyrosine residues upon stimulation with peroxide [25].
Since the bait’s neighbors become covalently labeled with biotin,
there is no need to maintain protein–protein interactions during
lysis and purification, and indeed, harsher lysis conditions can be
employed. In the original BioID manuscript, proximity partners
for the nuclear lamina component LMNA were identified [24],
and this work was followed by a number of papers describing the
applicability of the approach for the identification of components
of the nuclear pore [26], chromatin [27, 28], plasma membrane or
junctions [29, 30], centrosome and cilium [30, 31].
Parallel AP-MS and BioID across several studies revealed their
complementary nature [27, 29, 32]. Notably, the long labeling
times of BioID (6–24 h is typical) make it less than ideal for study-
ing interaction dynamics (e.g., growth factors elicit multiple inter-
actomic changes in less than a minute), but it is superior to AP-MS
for probing the interactomes of insoluble structures. Interestingly,
BioID can also be used to capture cycling/transient interactions,
such as those involving a posttranslational recognition event such
as phosphorylation [29], and to identify substrates for certain
classes of enzymes such as E3 ligases [32] and phosphatase regula-
tory subunits [33].
The protocols presented here build upon the complementarity
of BioID and AP-MS [27, 29, 32] (see Fig. 1 for overall workflow).
We describe the cloning and generation of a single pool of cells
that harbors both the BirA* tag for BioID and a FLAG tag for
detection of bait expression and for AP-MS. Here, we provide a
version of the BioID protocol, which we have optimized for
membrane-associated proteins. We also briefly describe the FLAG
protocol (which only bears minor modification from our previous
method article [34]; note that this tag is significantly larger than
the single or 3× FLAG used in our previous Methods article, which
may affect folding and/or expression of some proteins).
A.
Gateway
entry clone
Tet repressor Flp-In T-REx
attL1 attL2
ORF
Gateway inhibits expression expression cell line
expression clone
attB1 attB2
co-transfect into HygR PCMV
BirA* ORF
/2X TetO 2
KANR/SpecR BirA* ORF Flp-In T-REx host
FLAG
LR clonase cell line
2
FLAG
tO
Hy
/2X TeV
M
g
R
PC
+ tetracycline
AmpR pOG44
Gateway (encodes Flp
pDEST-pcDNA5/FRT/TO-BirA*-FLAG-ORF recombinase)
destination vector HygR PCMV
/2X TetO 2
BirA* ORF
attR1 attR2 FLAG

BirA* ccdB
2
FLAG
Hy
tO
/2X TeV
g
M
R
PC
AmpR
pDEST-pcDNA5/FRT/TO-BirA*-FLAG
B. Expand Hygromycin
resistant cell population
Freeze 15 cm plate
BioID FLAG-AP
Induce with Induce with

BirA* Tetracycline + Tetracycline BirA*
Biotin
FLAG FLAG
PREY BAIT PREY BAIT

Harvest cell
PREY PREY
pellets
rep 1 rep 2 rep 1 rep 2

Strong solubilizing conditions Weak solubilizing conditions
Protein-protein interactions lost Protein-protein interactions maintained
BirA*
sepharose FLAG magnetic
PREY BAIT
PREY
anti-FLAG antibody
streptavidin biotinylated prey
biotin
On-bead tryptic digestion and

mass spectrometric analysis
Fig. 1 (a) Simplified diagram shows steps to generate pDEST-pcDNA5-FRT/TO-BirA*-FLAG-ORF expression

plasmid and to integrate this construct at a specific genomic locus in Flp-In™ T-REx™ cell line. In the first
step, the Open Reading Frame (ORF) of your gene of interest in a Gateway entry vector is transferred to the
pDEST-pcDNA5/FRT/TO-BirA*-FLAG vector through LR reaction. Confirmed expression plasmid (pDEST-
pcDNA5/FRT/TO-BirA*-FLAG-ORF) can then be co-transfected into Flp-In™ T-Rex™ host cell line with pOG44
plasmid, which facilitates integration of the expression plasmid at a predestined FRT locus. Successful inte-
gration at the FRT locus provides resistance to hygromycin. Once integrated, BirA*-FLAG-ORF expression is
2 Materials
2.1 Expression 1. Gateway entry vector for the protein(s) of interest (see
Constructs Subheading 3). Either “open” (no stop codon) or “closed”
(with stop codon) ORFs should be used depending on whether
C-terminal or N-terminal fusions, respectively, are desired.
2. Gateway destination vector (pDEST-pcDNA5-BirA*-FLAG_
Nterm for N-terminal fusions or pDEST-pcDNA5- BirA*FLAG_
Cterm for C-terminal fusions). These can be requested through
the Gingras lab. Non-Gateway versions of these vectors can be
requested through the Raught lab.
3. Gateway LR Clonase II (Invitrogen, Cat #11791-100).
4. Competent bacteria (e.g., DH5α).
5. SOC and LB media.
6. LB-ampicillin (100 μg/mL) agar plates.
7. Ampicillin stock (100 mg/mL).
8. DNA miniprep kit.
9. BsrGI restriction enzyme and standard DNA agarose gel system
(for clone validation).
2.2 Generation 1. Flp-In T-REx HEK293 cells (Invitrogen, for cell culture).
of Pooled Stable Cell 2. DMEM (high glucose, with pyruvate and l-glutamine) supple-
Lines mented with 10 % FBS, 100 U/mL penicillin/streptomycin
(Flp-In and T-REx cassettes can be selected for by the addition
of Zeocin (100 μg/mL) and Blasticidin (3 μg/mL), respec-
tively, to the media).
3. pOG44 vector (encodes Flp recombinase) (1 μg/mL), puri-
fied expression clone (100 ng/μL), transfection reagent.
4. Standard growth media supplemented with 200 μg/mL
Hygromycin B.
Fig. 1 (continued) constitutively repressed by the Tet repressor until tetracycline is supplemented. For complete
details, please see the Gateway and Flp-In™ T-Rex™ manual available on the Invitrogen website. (b)
Schematic representation of the parallel BioID and FLAG AP-MS workflow. Hygromycin resistant clones are
pooled and expanded to five 15 cm plates; one for generating frozen stock, and two biological replicates for
BioID, the two biological replicates for FLAG-AP. For BioID, pooled clones expressing BirA*-FLAG-ORF construct
and parallel control cell lines are grown to approximately 80 % confluence, and treated with tetracycline and
biotin (4–24 h) to induce bait expression and biotinylation of proximal proteins. Cells are lysed under strong
solubilizing conditions (protein–protein interactions are lost) and AP is performed with streptavidin-sepharose
beads (buffer conditions are outlined in Fig. 2). For FLAG-AP, cells are grown to approximately 80 % confluence,
and treated with tetracycline (24 h) to induce bait expression. Cells are lysed under gentle solubilizing condi-
tions (protein–protein interactions are maintained) and AP is performed with anti-FLAG magnetic beads. In
both cases, proteins are digested on-bead for subsequent mass-spectrometric analysis
2.3 Biotin– Use “clean” (non-autoclaved) pipette tips and tubes throughout
Streptavidin Affinity procedure (see Note 1). Use ultrapure/HPLC grade water for all
Purification (BioID) buffers and reagents.
1. BioID Lysis Buffer Stock, for 50 mL: 50 mM Tris pH 7.5
(2.5 mL of 1 M stock), 150 mM NaCl (1.5 mL of 5 M stock),
0.4 % SDS (2 mL of 10 % stock), 1 % IGEPAL CA-630 (or other
Nonidet P-40 substitute) (5 mL of 10 % stock), 1.5 mM MgCl2
(150 μL of 0.5 M stock), 1 mM EGTA (200 μL of 0.25 M
stock), fill up to 50 mL with water.
2. BioID Lysis Buffer Complete (see Note 2): Per 1 mL Lysis
Buffer add (freshly) 2 μL of 500× stock of Protease Inhibitors
(PI) (Cat #P8340, Sigma-Aldrich), and 1 μL Benzonase
(250 U/μL) (Cat #71205, EMD-Millipore) (alternatively
Turbo Nuclease (Cat #9207, BioVision) may be used).
3. BioID Wash Buffer (for 10 mL): 2 % SDS (2 mL of 10 % stock),
50 mM Tris pH 7.5 (500 μL of 1 M stock), fill up to 10 mL
with water.
4. Biotin stock solution (for 20 mL of 20 mM which is a 400×
stock): 100 mg Biotin, add 2 mL of 30 % NH4OH and place
on ice. Slowly add 5 mL of 1 N HCl, wait 5 min, and repeat
for a total of 18 mL added. Store at 4 °C protected from light
(see Note 3).
5. Tetracycline stock solution (10 mg/mL which is a 10,000×
stock): 100 mg in 10 mL 50 % EtOH (anhydrous) in water.
Store at −20 °C protected from light (see Note 4).
6. Streptavidin-sepharose beads “high performance” (Cat #17-
5113-01, GE Healthcare).
7. Ammonium bicarbonate (ABC) (50 mM, pH 8.0): 200 mg in
water (see Note 5).
8. Trypsin from porcine pancreas (proteomics grade) (see Note 6):
for the solution to digest on the Streptavidin-sepharose beads,
resuspend one 20 μg tube (Cat #T6567, Sigma-Aldrich) in a
final volume of 2 mL ABC (concentration 10 ng/μL). For
spike-in trypsin addition, resuspend a 20 μg tube of trypsin in
100 μL ABC (concentration 0.2 μg/μL). When using trypsin
“singles” (Cat #T7575, Sigma-Aldrich) instead, resuspend
each in 10 μL ABC and dilute appropriately to achieve desired
final concentration.
9. Formic acid (mass spectrometry grade): 5 % and 10 % stocks
freshly prepared in HPLC grade water and stored in clean glass
vials (Cat #14-955-319, Fisher Scientific).
10. Autosampler vials (Cat #160134, Dionex) and caps (Cat #03-
391-43, Fisher Scientific).
2.4 FLAG Affinity 1. FLAG Lysis Buffer: 50 mM Hepes–KOH pH 8.0, 100 mM
Purification (FLAG-AP) KCl, 2 mM EDTA, 0.1 % NP40, and 10 % glycerol, supple-
mented with 1 mM PMSF, 1 mM DTT, and 1× protease
inhibitor cocktail (Cat #P8340, Sigma-Aldrich) immediately

prior to using.
2. FLAG Rinsing Buffer: 20 mM Tris–HCl (pH 8.0) and 2 mM
CaCl2.
3. Trypsin Digest Buffer: 20 mM Tris–HCl (pH 8.0).
4.
Anti-FLAG M2 magnetic beads (Cat #M8823,
Sigma-Aldrich).
5. Trypsin from porcine pancreas (proteomics grade): resuspend
a 20 μg tube (Cat #T6567, Sigma-Aldrich) in 200 μL Trypsin
Digest Buffer (concentration 0.1 μg/μL). If using trypsin
“singles” (Cat #T7575, Sigma-Aldrich), resuspend each tube
in 10 μL Trypsin Digest Buffer (concentration 0.1 μg/μL).
6. Formic acid: (mass spectrometry grade) 50 % stock prepared in
HPLC grade water and stored in clean glass vial (Cat #14-955-
319, Fisher Scientific).
3 Methods
3.1 Preparation A Gateway “entry” clone is defined as the open reading frame
of Expression (ORF) for the protein of interest cloned into a Gateway donor vec-
Constructs Using tor such as pDONR233 (Invitrogen). For N-terminal fusions, we
Gateway Cloning recommend using an entry clone harboring a stop codon to pre-
vent cloning scars [35]; C-terminal fusions should be generated
with “open” entry vectors. The protocol assumes that an entry
clone for Gateway cloning is available for the protein of interest
(and that it has been sequence-verified); a miniprep provides a
sufficient amount. New “Entry” clones can be generated through
PCR amplification of inserts with flanking attB sites and recombi-
nation into a donor vector with attP sites by Gateway “BP” reac-
tion. Entry clones may also be generated via a BP reaction with an
expression clone and a donor vector. Alternatively, BirA*-FLAG
vectors compatible with standard restriction enzyme/ligation cloning
can be used.
1. For the Gateway LR reaction, combine the following: (a)
Entry clone (for LR reaction) (ideally 150 ng, although less
can be used); (b) Destination vector (1 μL of a 150 ng/μL
stock); (c) Water or TE pH 8.0 up to 4 μL.
2. Thaw LR Clonase on ice and briefly mix/vortex, then add
1 μL of LR Clonase to reaction (total volume of 5 μL), mix
and spin down. Incubate reaction at RT for 1 hour up to over-
night (see Note 7).
3. Add 1 μL of Proteinase K solution to reaction and incubate for
10 min at 37 ° C.
4. To transform, add 1–2 μL of LR reaction to 20–50 μL chemi-
cally competent E. coli (e.g., DH5α) and incubate on ice for
approximately 15 min. Heat shock at 42 °C for 30 s and return

to ice for 2 min. Add a 10× volume of SOC medium at RT and
place in a 37 °C shaking incubator for 60 min. Plate appropriate
volumes (typically 40–100 μL will yield good colony density)
onto LB-agar plates containing 100 μg/mL ampicillin.
Incubate plates at 37 °C overnight.
5. Expand single colonies in 2 mL LB + ampicillin media over-
night in 37 °C shaking incubator. Prepare DNA mini-preps of
the clones, following the manufacturer’s instructions. A diag-
nostic digest can be performed with BsrGI restriction enzyme
to check insert size by analysis with DNA-gel electrophoresis
(Gateway recombination sites flanking the ORF contain BsrGI
cut sites. Be sure to check if insert contains an internal BsrGI
site in order to appropriately interpret the digestion results).
If the original entry clone was sequenced, we do not sequence
the destination vector. Label and store the constructs
(Laboratory Information Management Systems such as
OpenFreezer [36] are useful for this process).
3.2 Generation Our groups have been primarily using the Flp-In T-REx HEK293
of Pools of Stable Cell cell system from Invitrogen for interaction proteomics (other Flp-In
Lines T-Rex cells can alternatively be used with the same constructs). The
system enables tetracycline-inducible expression of the transgenes
expressed at a single copy (from the same Flp Recombination Target
(FRT) containing locus) through a Flp-mediated recombination
event. This system enables robust establishment of cell lines and
pools of cell lines, even for proteins whose expression is toxic for
cell growth (see Note 8).
1. On DAY 0, plate approximately 6 × 105 cells in wells of 6-well
plates such that they are approximately 60–80 % confluent at
the time of transfection (use media that does not contain anti-
biotics for optimal cell viability). In order to transfect with
pcDNA5-FRT-TO based vectors, ensure that media does not
contain Zeocin, to which resistance is lost upon successful
recombination.
2. On DAY 1, transfect cells with 1 μg pOG44 plasmid (expresses
the Flp recombinase) and 100 ng pDEST-pcDNA5-BirA*-
FLAG-ORF (aka, the expression clone) construct using mam-
malian cell transfection reagent of choice according to
manufacturer’s directions (see Note 9). Remember to transfect
your selected controls in parallel (see Note 10).
3. On DAY 2, split each well into a 10 cm dish in complete media
(see Note 11).
4. On DAY 3, aspirate media and replace with complete media
supplemented with 200 μg/mL Hygromycin B to begin selection
of integrated cells. Be careful to avoid cross-contamination
during the selection process.
5. Allow selection to occur for approximately 1–2 weeks, changing

selection media as required to remove dead cells and to allow
stable colonies to grow.
6. Once colonies are clearly visible by eye (a few mm in diameter)
pool them into either 6 cm (<20 colonies), 10 cm (approxi-
mately 50–100 colonies), or 15 cm (high colony density) plates
and continue to expand in selection media (see Note 12).
7. Scale cells up until a near confluent 15 cm dish is achieved, and
then split this plate into additional 15 cm plates. If parallel
BioID and FLAG AP-MS experiments are to be performed in
biological duplicates, re-plate into 5 × 15 cm plates (one plate
to be frozen, and two to be processed by BioID and two to be
processed by FLAG AP-MS; see Fig. 1. Distribute the volume
into the destination plates such that the two replicates for
FLAG-AP or BioID are neither induced at the same time nor
harvested at the same time. For example: resuspend trypsin-
ized plate into 12 mL and distribute 3 mL to freeze back,
3 mL for each of the biological replicates 1 (BioID and FLAG),
and 1.5 mL for each of the biological replicates 2. If only
FLAG or BioID is to be performed, or if a different number of
replicates is needed, adjust this last splitting accordingly.
3.3 Preparing Cell The protocol for the preparation of the cell pellets and the BioID
Pellets and FLAG-AP purification assumes that only identification and
quantification of the proteins is desired. Note that cell pellets are
typically harvested at room temperature for BioID, whereas pel-
lets should be harvested on ice with chilled PBS for FLAG-AP.
For projects that involve the characterization of specific posttransla-
tional modifications, extra care should be taken (e.g., to inactivate
phosphatases in the case of phosphorylation). This can be done
through the addition of inhibitors, handling all steps with ice-cold
reagents, and shortening some of the incubation times. Also note
that if the analysis of specific peptides is needed, it is recommended
to include alkylation and reduction steps prior to mass spectrometric
identification.
1. Grow 1 × 15 cm plate of cells stably expressing Flp-In BirA*-
FLAG tagged construct (or suitable control) to approximately
80 % confluence in complete media (see Note 13).
2. Add 1 μg/μL of tetracycline (2 μL of a 10 mg/mL stock per
20 mL media) for both BioID and FLAG-AP analysis and
50 μM biotin (50 μL of a 20 mM stock per 20 mL media) for
BioID only. Incubate the cells for 24 h (see Note 14). For
BioID, note that you can pre-build protein expression by incu-
bating with tetracycline for 12–24 h prior to adding biotin for
4–12 h. Keep in mind, however, that you should always time
the incubation of your controls and samples carefully to enable
comparative studies.
3. Pre-weigh pre-labeled 2 mL tubes and write weight on side

(see Note 15). This is important, as the quantity of lysis buffer
used is based on the weight of the “dry” cell pellet. Prepare
other equipment to harvest cells: beaker of PBS, beaker of 70 %
EtOH, beaker for waste, bottle of PBS (at room temperature),
clean rubber spatula, paper towels and Kimwipes, 1 mL pipette
and tips.
4. Remove media into waste (either by pouring media into beaker,
or using vacuum aspirator), then rinse cells gently and briefly
with approximately 20 mL PBS and remove to waste. Tilt plate
on its side for approximately 30 s and pipette/aspirate off
remaining PBS.
5. Add 1 mL fresh PBS and scrape cells with spatula (pre-wetted
in PBS beaker) to one edge of the tilted plate. Suspend cells in
the PBS by repeated aspirations with a 1 mL pipette and add
suspension to labeled 2 mL tube
6. Clean spatula well with EtOH and wipe with Kimwipes. Return
the spatula to the beaker of PBS.
7. Repeat with remaining plates, washing spatula as above
between each plate.
8. Centrifuge cells at a maximum of 500 × g for 5 min at room
temperature (as low as 200–300 × g for 2–3 min can also be
used). 500 × g corresponds to 2300 rpm on an Eppendorf
5415D benchtop centrifuge.
9. Carefully remove all visible supernatant with 1 mL pipette (see
Note 16) and freeze pellet on dry ice and store at −80 °C until
ready to process.
3.4 Parallel BioID The BioID protocol has been optimized for the capture of proxim-
and FLAG-AP ity partners for membrane-associated proteins and other non-
Purification soluble cellular structures (such as RNA granules and bodies), and
differs slightly from those we previously reported for signaling
molecules, identification of E3 ligase substrates, chromatin associ-
ated proteins or the centrosome/cilia, notably in the concentra-
tion of detergents used ([27, 29, 31, 32]; see Fig. 2, and Note 17).
The protocol starts from 1 × 15 cm plate per replicate (normally
corresponds to a dry cell pellet weighing approximately
75–150 mg), which is sufficient for up to six injections on AB Sciex
TripleTOF mass spectrometers (e.g., enabling acquisition of tech-
nical replicates, or parallel Data Dependent and Data Independent
Acquisition runs [13]; see Note 18).
3.4.1 Biotin–Streptavidin Steps 1–8 should be performed at 4 °C and/or on ice (ensure that
Affinity Purification (BioID) all solutions are chilled prior to use), with the remaining steps
and on-Bead Trypsin being carried out at RT.
Digest
A. B.
1. RIPA 2. BioID Lysis Buffer 1. mild wash 2. harsh wash
Tris-HCl (pH 7.5) 50 mM 50 mM 1x RIPA buffer BioID Wash Buffer
NaCl 150 mM 150 mM 2x FLAG Lysis Buffer BioID Lysis Buffer
SDS 0.1% 0.4%
3x ABC Buffer ABC Buffer
IGEPAL CA-630 - 1%
Triton X-100 1% -
sodium deoxycholate 0.5% - 1 2
EDTA 1 mM - C.
EGTA 1 mM 1 mM
MgCl2 - 1.5 mM 10 20 31
Fig. 2 (a) Detailed comparison of the composition of buffers previously reported: (1) RIPA lysis buffer used for
BioID in our previous study [29], and (2) the modified BioID Lysis Buffer described in this protocol. (b).
Comparison of wash steps after streptavidin affinity-purification (AP): (1) mild wash steps as previously
reported [9], and (2) the harsh wash steps described in this protocol. (c) Number of high confidence preys
identified using BirA-FLAG-DCP1A as bait using the two different methods outlined in A. and B.; (1) RIPA
lysis + mild wash post-AP and (2) BioID Lysis Buffer + harsh wash post-AP
1. Weigh tubes and subtract tube weight to determine cell pellet

weight.
2. Thaw cell pellets on ice and add BioID Lysis Buffer (complete
with protease inhibitors and benzonase added (see Note 19) at
a 4:1 ratio vol:wt, e.g., 400 μL buffer for a 100 mg pellet).
3. Break up pellet with pipette (by repeated aspirations) and then
freeze–thaw once by freezing samples on dry ice, then imme-
diately moving tube to a 37 °C water bath with agitation, just
until the visible ice is about to thaw.
4. Immediately place the samples on an end-over-end rotator at
4 °C and incubate for 30 min to allow benzonase to fully
degrade nucleic acids and for complete solubilisation. Save a
20 μL aliquot to monitor bait expression by Western blot. To
determine the percentage of solubilisation, centrifuge this
20 μL as in step 5 below and analyze the supernatant and the
pellet separately (there should be essentially no visible pellet or
insoluble proteins with this protocol).
5. Centrifuge for 20 min at 16,100 × g at 4 °C (see Note 20).
6. Collect supernatant in new 1.5 mL tube.
7.
Prepare the Streptavidin-sepharose beads. Streptavidin-
sepharose beads “high performance” are supplied as a 60 %
slurry (in a 20 % ethanol solution). 20 μL bed volume of beads
are used per sample (this corresponds to approximately 35 μL
slurry). Wash the desired volume of beads three times with 1 mL
BioID Lysis Buffer (in a 1.5 mL tube). Resuspend in BioID
Lysis Buffer to the original 60 % slurry volume (e.g., by adding
15 μL to the 20 μL packed beads). Beads are best pipetted using
a tip cut at an angle with a clean razor blade; when preparing
beads for multiple samples, prepare a bit extra to account for
losses due to pipetting.
8. Add 20 μL dry bead equivalent as bead slurry (35 μL) to each
sample. Ensure that bead slurry is homogeneously mixed prior
to pipetting beads for each sample. This is effectively achieved
by gently vortexing the tube (at lowest vortex intensity required
to effectively suspend beads) for a few seconds immediately
prior to pipetting. Rotate samples “end-over-end” overnight
at 4 °C (see Note 21).
9. Centrifuge beads at 500 × g for 2 min (see Note 22).
10. Remove supernatant with a vacuum line equipped with a fine
pipette tip (e.g., P20 or “gel loading” tip), being careful to not
disturb or lose beads.
11. Transfer beads in 500 μL of BioID Lysis Buffer into a new
1.5 mL tube (see Note 23). Centrifuge at 500 × g for 30 s and
remove the supernatant.
12. For all washes, add 500 μL of the indicated wash solution, and
suspend beads with brief, gentle vortexing/mixing. Centrifuge
beads at 500 × g for 30 s and remove supernatants as above.
Perform the following washes: (a) Wash once with BioID Wash
Buffer; (b) Wash twice with BioID Lysis Buffer; (c) Wash three
times with 50 mM ammonium bicarbonate (ABC) solution
(see Note 24).
13. Add 100 μL trypsin solution (i.e., 1 μg trypsin in 100 μL ABC)
to beads and incubate 4 h with rotation at 37 °C (see Notes 25
and 26).
14. Add a fresh 1 μg of trypsin (5 μL if stock is prepared as described
in Materials) to each sample and continue incubation over-
night with rotation at 37 °C.
15. Centrifuge beads at 500 × g for 2 min and retrieve 100 μL of
the supernatant (containing the digested peptides) into a new
1.5 mL tube.
16. Add 100 μL of HPLC grade water to the beads and gently
vortex/mix. Centrifuge and collect 100 μL of supernatant as
above, and pool with peptide solution collected in previous
step.
17. Add formic acid to digested peptides to a final concentration of
2 % (use 50 μL of 10 % stock yielding 250 μL total solution)
and vortex briefly.
18. Centrifuge at 16,100 × g for 5 min to eliminate debris and any
residual beads that may have carried over, then transfer exactly
230 μL (being careful not to disrupt the debris “pellet” in the
remaining 20 μL) to new 1.5 mL tubes.
19. Dry by centrifugal evaporation without heat (in a Speed-Vac

vacuum centrifuge or similar).
20. Store dried peptides at −80 °C until ready to analyze by mass
spectrometry.
21. When ready to analyze, warm tube to RT and centrifuge dry
tube at 16,100 × g for 5 min to ensure that the peptides are
concentrated at the bottom of the tube.
22. Resuspend peptides in 10 μL 5 % formic acid (FA) and vortex
aggressively for at least 20–30 s, then centrifuge at 16,100 × g
for 5 min.
23. Mass spectrometric analysis (also see below): For analysis on
AB Sciex Triple-TOF instruments, pipette 4.5 μL of 5 % FA
into a new autosampler tube and then add 1.5 μL of resus-
pended/centrifuged peptide solution for a single injection. Set
MS method to pick up 5 μL of this solution (see Note 27 for
analysis on other instruments).
24. Any unused sample should be stored (do not re-dry) at
−80 °C. To rerun stored samples, thaw tubes to RT, vortex,
centrifuge, and take sample for injection (as in step 23).
3.4.2 FLAG Affinity This protocol was developed to perform affinity purification from
Purification (FLAG-AP) 1 × 15 cm plate prepared in Subheading 3.3. In this procedure cells are
and on-Bead Trypsin lysed by passive lysis assisted by freeze–thaw and affinity purification is
Digest performed on a magnetic bead support (see ref. [34] for alternative
protocol using an agarose bead support and for protocol optimiza-
tion). Ensure that all solutions are chilled on ice prior to use.
1. Weigh tubes and subtract tube weight to determine cell pellet
weight.
2. Thaw cell pellets on ice and add FLAG Lysis Buffer at a 4:1
ratio vol (μL): wt (mg) (e.g., 400 μL buffer to a 100 mg
pellet).
3. Break up pellet with pipette (by repeated aspirations) and then
freeze–thaw once by freezing samples on dry ice, then imme-
diately moving tube to a 37 °C water bath with agitation, just
until the visible ice is about to thaw.
4. Centrifuge at 16,100 × g for 20 min at 4 °C and transfer the
supernatants to fresh tubes (remove or avoid the lipid layer on
top of the lysate if present). Save a 20 μL aliquot to monitor
bait expression by Western blot.
5. Prepare the anti-FLAG M2 magnetic beads. Use 25 μL per
sample (supplied by vendor as a 50 % slurry). Wash the required
volume of beads 3–4× with 1 mL of FLAG Lysis Buffer (using
a magnetic tube rack). Resuspend the beads in FLAG Lysis
Buffer to make a 50 % slurry and distribute 25 μL of this into
each sample.
6. Incubate this mixture for 2–3 h at 4 °C with “end-over-end”

rotation.
7. Centrifuge samples at 500 × g for 1 min to recover beads that
may have stuck to the lid. Remove the supernatant with a
pipette and transfer the beads to a clean 1.5 mL tube using
1 mL of FLAG Lysis Buffer.
8. Using a magnetic tube rack, wash beads once with 1 mL FLAG
Lysis Buffer, followed by one wash with FLAG Rinsing Buffer
(resuspending the beads in buffers by pipetting up and down
four times). The washing steps should be done as quickly as
possible while allowing approximately 30 s for beads to mag-
netize before aspirating supernatant—a complete wash cycle
should take between 1 and 2 min (see Note 28).
9. After the last wash is aspirated, centrifuge the sample at 500 × g
for 1 min to pellet the beads and remove any remaining liquid
with a fine pipette.
10. Add 7.5 μL of trypsin (100 ng/μL solution = 750 ng trypsin)
to the beads and incubate at 37 °C overnight on a rotator.
11. The following day, centrifuge the samples at 500 × g for 1 min,
magnetize the beads for 30 s and transfer the supernatant to a
fresh tube, then add an additional 2.5 μL of trypsin (250 ng)
to the collected supernatant and incubate for 4 h at 37 °C (no
agitation required).
12. Add 1 μL of 50 % formic acid solution to the samples (approxi-
mately 5 % final concentration). The sample can be directly
analyzed by mass spectrometry (centrifuge at 16,100 × g for
5 min prior to retrieving sample).
13. Mass spectrometric analysis: For analysis on AB Sciex Triple-
TOF instruments load approximately 5 μL of sample.
14. Any unused sample may be stored at −80 °C.
3.4.3 Mass Spectrometry Mass spectrometric analysis uses a standard LC-MS/MS set-up,
and Data Analysis which we have also described elsewhere [13, 27, 29]. Briefly, using
an autosampler (Eksigent), samples are loaded onto fused silica
capillary columns (0.75 μm ID, 350 μm OD) pre-loaded with
10 cm of C18 reversed phase material (3.5 μm diameter). Ionized
peptides are emitted by nanoelectrospray ion source in-line with a
nano-HPLC system and analyzed using either Data Dependent
(DDA) or Data Independent (DIA) Acquisition methods.
Important considerations include identifying and mitigating
potential carryover issues, and being able to identify which of the
identified proteins are likely to represent bona fide interactors
(or proximity partners in the case of BioID) and which are likely to
be contaminants.
We attempt to address these considerations by: (1) the inclusion
of negative controls in our experimental design (see Note 10); (2) the
processing of at least two biological replicates per bait analyzed,
reporting only those interactions confidently detected across both

replicates; (3) the inclusion of long wash cycles of the LC column
between samples (see Note 29); (4) the randomization of acquisition
order on the LC-MS/MS across both replicates such that if carryover
is present (column retention of the most abundant peptides contami-
nating subsequent samples), it should not carry through to our
final interaction list; 5) the acquisition of quantitative data which,
when analyzed through the pipeline briefly described below, enables
consistent scoring and reporting of the interactions.
All data processing and analysis is done using software devel-
oped and maintained in-house (ProHits [37]), which is open-
source and can be run using a “virtual machine” [38], or on a local
server. ProHits enables storing, tracking and annotating experi-
ments; simple visualization interfaces permit rapid analysis of each
sample side-by-side, and the identification of possible issues (car-
ryover, low recovery of the bait or its known interactors, etc.).
High confidence interactions are determined by comparing bait
samples against appropriate control samples (see Note 10) using
the statistical tool SAINT [9, 39]. A user-friendly way to run
SAINT analysis and other interaction scoring tools without having
to install any software is through the Contaminant Repository for
Affinity Purification (CRAPome) website [40]. Our laboratories
have deposited to the CRAPome multiple negative controls from
HEK293 FLAG and BioID experiments that can be used to sup-
plement user-generated controls in order to aid in the removal of
nonspecific / background preys. Results from SAINT analysis
(through ProHits, the CRAPome, or as a stand-alone version) can
all be prepared for publication through a suite of visualization tools
[41] developed by our group.
4 Notes
1. Autoclaving may result in a residue being deposited on to plas-

ticware, which may result in contamination during MS analy-
sis. As sterility is not essential for the protocols (other than
during cell culture), use tips and tubes directly as received from
the manufacturer. Keep all plasticware and reagents protected
from dust and other environmental contaminants as much as
possible and use gloves for all preparation steps.
2. Protease inhibitors and nucleases are only added to the work-
ing volume of lysis buffer that will be directly added to the cell
pellets. Sum the weights of all pellets and prepare only the
volume of lysis buffer that will be required (plus extra volume
for pipetting error).
3. Extreme care must be taken in preparing biotin solutions, as it
can easily precipitate from the solution if the pH is raised too
rapidly or is not kept adequately chilled during preparation
(this may be reversed by the readdition of a small amount of

ammonium hydroxide). In our experience the biotin stock
solution is very stable and can be used for approximately 6
months to a year without any change in potency when pre-
pared and stored appropriately.
4. Tetracycline stocks can be stored long term at −20 °C, but a
small amount will gradually precipitate out of solution. Warm
the solution to RT prior to use, avoiding the settled particu-
late. This gradual loss of tetracycline from solution has no
apparent impact on the induction of expression, since the con-
centration used is well in excess of what is required to inhibit
the Tet repressor. To minimize this, consider aliquoting to
avoid multiple freeze–thaw cycles.
5. The pH of ammonium bicarbonate solution rises upon storage
(with pH ~7.8 when freshly made and ≥ 8.5 after storage for a
few weeks). Make solution freshly and readjust pH as necessary
by mixing older solutions to have a working pH around 8–8.2
for reproducible trypsin digestion.
6. We have not noticed major differences in digestions when
using the 20 μg trypsin vial or the trypsin Singles from Sigma-
Aldrich. The selection of the enzyme is primarily based on cost
efficiency for the number of samples to be analyzed.
7. The LR reaction is very efficient and we routinely scale the
enzyme quantity in half.
8. Some degree of leaky expression can be expected in the absence
of tetracycline induction, and this may be heterogeneous across
the cell population (i.e., some sporadic cells will exhibit much
higher expression than the general population as assessed by
immunofluorescence). If leakiness of a toxic protein is detected,
consider using a tetracycline-free source of Fetal Bovine Serum
(such as the Clontech “Tet Approved FBS” product). Because
of the integration at a single locus, expression across individu-
ally isolated clones is relatively uniform. As such, we use pools
of clones for all proteomics experiments described here. If iso-
lated clones are desired (e.g., for functional studies), they can
easily be obtained by letting colonies on selection grow
to > 1 mm diameter and picking them using sterile filter papers
dipped in trypsin (we use the Scienceware 3.2 mm disks
(Cat #Z374431, Sigma-Aldrich) for this purpose and transfer
each clone into a prepared 96- or 48-well plate).
9. We use JetPrime (Cat #11407, Polyplus Transfection) which is
a cost effective reagent to use with the HEK293 Flp-In T-REx
system. For a 6 well format we use 200 μL JetPrime buffer and
2 μL transfection reagent. Other transfection reagents can be
used instead.
10. The inclusion of negative controls in the experimental design is

essential to distinguish the true interactors of the protein of inter-
est from the nonspecific (contaminant, or background) proteins
in both FLAG-AP and BioID experiments. There are a number
of sources of background, including nonspecific protein binding
to the resin, the affinity tag, or to any overexpressed protein. For
BioID, additional sources of background contaminants include
ubiquitous endogenously biotinylated proteins and nonspecific
proximal biotinylation by BirA*. A FLAG pull-down experiment
would typically include at least one type of negative control (at
the minimum, the “empty” BirA*-FLAG construct, expressed at
levels at least as great as the FLAG-BirA*-tagged bait). However,
to properly model the background of the BioID experiment—
and especially the endogenous biotinylation—it is necessary to
also include cells not transfected with BirA* (e.g., transfected
with another construct such as FLAG-GFP, 3×FLAG or left
untransfected). These non-BirA* cells should be treated with
biotin to maximize endogenous biotinylation. Depending on
the biological question, it may be important to also include
other controls for the BioID experiments; for example, to
increase the stringency of filtering, subcellular compartment
specific controls can be generated through fusion of BirA* to
sequence tags for the compartment of interest (e.g., a nuclear
localization signal [27]). These “compartment-specific” con-
trols may be treated as true negative controls when scoring
interactions, but can also be used as a secondary filter (e.g., by
analyzing the quantitative fold enrichment of each potential
interactor against the “compartment-specific” controls).
11. HEK293 Flp-In T-REx cells are loosely adherent and therefore
care must be taken not to disrupt cells during media changes.
12. Cells in the center of dense colonies will have slower prolifera-
tion rates and therefore it is desirable to split plates before
allowing colonies to grow too large in order to allow cells to
proliferate more efficiently. The selection of the destination
vessels should be determined empirically (the estimates pro-
vided are only guidelines).
13. Hygromycin may be omitted from the plates used for induc-
tion of protein expression.
14. We have reduced biotin labeling times to as low as 4 h.
However, if bait expression is low, more material (i.e., 2 or
more 15 cm plates) may need to be combined to achieve an
adequate signal if short labeling times are used.
15. In our experience, most brands of tubes exhibit variation in
their weights (by as much as 50 mg or more) and therefore
accurate cell weight can only be achieved by obtaining the tare
weight of each tube individually.
16. It is important to remove all residual liquid in order to obtain

an accurate cell pellet weight.
17. The power of the BioID protocol rests on the strength of the
biotin–streptavidin interaction, which is one of the strongest
known biological interactions (Kd ~10−14 M) and is highly
resistant to disruption by SDS. Due to the strength of this
interaction, the lysis buffer employed here contains a higher
concentration of SDS (0.4 %) and employs a high concentra-
tion SDS wash step (2 %). This is in contrast to typical buffers
used in affinity purification protocols, which depend on native
protein–protein interactions to be maintained. The higher
concentration of SDS in the lysis buffer, which is highly dena-
turing, ensures near complete solubilization of membrane pro-
teins (including those with multiple transmembrane domains)
and other insoluble cellular structures (including RNA gran-
ules and bodies). The high concentration of SDS in the wash
helps remove contaminating (i.e., non-biotinylated) proteins,
thus improving the signal-to-noise ratio. Note that we have
also omitted EDTA (which chelates Mg2+) and included MgCl2
which improves the activity of Benzonase in our revised buffer
(see Fig. 2).
18. We provided here the amounts loaded on the AB Sciex
TripleTOF instruments equipped with Eksigent Ultra systems
and 10 cm C18 fused silica capillary columns (0.75 μm ID,
350 μm OD). Other instrument configurations will require
optimization of this amount. Notably, we are loading more
peptides for analysis on ThermoScientific Velos Orbitrap or
Orbitrap Elite mass spectrometers (see Note 27).
19. Benzonase and protease inhibitors are only added to the volume
of BioID Lysis Buffer that will specifically be used to lyse the cells
(aka “complete”). For washing steps with BioID Lysis Buffer,
Benzonase and protease inhibitors are not added (aka “stock”).
20. When establishing the protocol in the laboratory for the first
time, or if initial results are unsatisfactory and troubleshooting
must be performed, take an aliquot of the unbound fraction
and of the washes to analyze by immunoblotting. A white,
fibrous looking pellet should result (only 1–2 mm in diameter),
and likely corresponds to highly insoluble aggregated protein.
21. Note that incubations as short as 3 h are routinely performed
in our group (see notes in the cell pellet preparation regarding
the analysis of posttranslational modifications). Ensure how-
ever, that you are consistent across any experiments that need
to be compared.
22. Changing the tube at this step helps eliminate residues that
may have adhered onto the tube walls.
23. The general idea of the washes in ABC is to eliminate the

detergents that would interfere with the mass spectrometric
identification and to equilibrate beads in a buffer suitable for
trypsin digestion that is volatile and will be evaporated upon
sample drying. If residual detergents are detected in the mass
spectrometer, consider using fresh tubes for the last two washes
in ABC.
24. For on-bead trypsin digestion we use a “rotating drum” style
rotator (Cel-Gro Tissue Culture Rotator, Thermo Scientific)
placed inside a 37 °C incubator.
25. For protein-level analysis, no alkylation/reduction steps are
performed. This will result in a loss of cysteine-containing pep-
tides from the analysis. If the identification of cysteine-
containing peptides is desired, introduce a reduction and
alkylation step prior to mass spectrometric analysis.
26. For the analysis of BioID samples on Thermo OrbiTrap Velos
or Orbitrap Elite instruments, pipette 5 μL of undiluted pep-
tide solution into new autosampler tube and set MS method to
pick up 1 μL less than volume in tube (e.g., 5 μL of 6 μL total
solution).
27. Alternatively, the stringency of the washes can be increased by
washing the beads 3x with 1 mL FLAG Lysis Buffer, followed
by 2× with 1 mL FLAG Rinsing Buffer.
27. Performing database searches of bovine serum albumin (BSA)
or other control samples after washes and between runs enables
further identification of carryover issues and optimization of
your washing conditions.
Acknowledgment
We thank Wade H Dunham and Zhen-Yuan Lin for optimization

of the FLAG protocol, all members of the Gingras and Raught
laboratories for help in optimizing the BioID protocol and for
helpful discussions, and Boris Dyakov and Cassandra Wong for
comments on the manuscript. This work is funded by the Canadian
Institutes of Health Research (Foundation grant FDN143301 to
A.-C.G.; salary support to PST), the Natural Sciences and
Engineering Research Council of Canada (Discovery grant to
A.-C.G; salary support to JYY), and a Genome Canada Genome
Innovation (GIN) network (through the Ontario Genomics
Institute OGI-069 to A.-C.G.). Salary awards are from the Canada
Research Chairs Program (ACG and BR) and a Basic Research
Fellowship from Parkinson Canada (GGH).
References
1. Snider J, Kotlyar M, Saraon P, Yao Z, Jurisica I, 7. Breitkreutz A, Choi H, Sharom JR, Boucher L,
Stagljar I (2015) Fundamentals of protein Neduva V, Larsen B, Lin ZY, Breitkreutz BJ,
interaction network mapping. Mol Syst Biol Stark C, Liu G, Ahn J, Dewar-Darch D, Reguly
11(12):848. doi:10.15252/msb.20156351 T, Tang X, Almeida R, Qin ZS, Pawson T,
2. Gingras AC, Gstaiger M, Raught B, Aebersold Gingras AC, Nesvizhskii AI, Tyers M (2010) A
R (2007) Analysis of protein complexes using global protein kinase and phosphatase interac-
mass spectrometry. Nat Rev Mol Cell Biol tion network in yeast. Science 328(5981):1043–
8(8):645–654. doi:10.1038/nrm2208 1046. doi:10.1126/science.1176495
3. Gavin AC, Bosche M, Krause R, Grandi P, 8. Sowa ME, Bennett EJ, Gygi SP, Harper JW
Marzioch M, Bauer A, Schultz J, Rick JM, (2009) Defining the human deubiquitinating
Michon AM, Cruciat CM, Remor M, Hofert enzyme interaction landscape. Cell 138(2):389–
C, Schelder M, Brajenovic M, Ruffner H, 403. doi:10.1016/j.cell.2009.04.042
Merino A, Klein K, Hudak M, Dickson D, 9. Choi H, Larsen B, Lin ZY, Breitkreutz A,
Rudi T, Gnau V, Bauch A, Bastuck S, Huhse B, Mellacheruvu D, Fermin D, Qin ZS, Tyers M,
Leutwein C, Heurtier MA, Copley RR, Gingras AC, Nesvizhskii AI (2011) SAINT:
Edelmann A, Querfurth E, Rybin V, Drewes probabilistic scoring of affinity purification-mass
G, Raida M, Bouwmeester T, Bork P, Seraphin spectrometry data. Nat Methods 8(1):70–73.
B, Kuster B, Neubauer G, Superti-Furga G doi:10.1038/nmeth.1541
(2002) Functional organization of the yeast 10. Dunham WH, Mullin M, Gingras AC (2012)
proteome by systematic analysis of protein Affinity-purification coupled to mass spectrom-
complexes. Nature 415(6868):141–147. etry: basic principles and strategies. Proteomics
doi:10.1038/415141a 12(10):1576–1590. doi:10.1002/pmic.
4. Ho Y, Gruhler A, Heilbut A, Bader GD, Moore 201100523
L, Adams SL, Millar A, Taylor P, Bennett K, 11. Bisson N, James DA, Ivosev G, Tate SA,
Boutilier K, Yang L, Wolting C, Donaldson I, Bonner R, Taylor L, Pawson T (2011) Selected
Schandorff S, Shewnarane J, Vo M, Taggart J, reaction monitoring mass spectrometry reveals
Goudreault M, Muskat B, Alfarano C, Dewar the dynamics of signaling through the GRB2
D, Lin Z, Michalickova K, Willems AR, Sassi adaptor. Nat Biotechnol 29(7):653–658.
H, Nielsen PA, Rasmussen KJ, Andersen JR, doi:10.1038/nbt.1905
Johansen LE, Hansen LH, Jespersen H, 12. Zheng Y, Zhang C, Croucher DR, Soliman
Podtelejnikov A, Nielsen E, Crawford J, MA, St-Denis N, Pasculescu A, Taylor L, Tate
Poulsen V, Sorensen BD, Matthiesen J, SA, Hardy WR, Colwill K, Dai AY, Bagshaw
Hendrickson RC, Gleeson F, Pawson T, Moran R, Dennis JW, Gingras AC, Daly RJ, Pawson
MF, Durocher D, Mann M, Hogue CW, Figeys T (2013) Temporal regulation of EGF signal-
D, Tyers M (2002) Systematic identification of ling networks by the scaffold protein Shc1.
protein complexes in Saccharomyces cerevisiae Nature 499(7457):166–171. doi:10.1038/
by mass spectrometry. Nature 415(6868):180– nature12308
183. doi:10.1038/415180a
13. Lambert JP, Ivosev G, Couzens AL, Larsen B,
5. Hein MY, Hubner NC, Poser I, Cox J, Nagaraj Taipale M, Lin ZY, Zhong Q, Lindquist S,
N, Toyoda Y, Gak IA, Weisswange I, Mansfeld Vidal M, Aebersold R, Pawson T, Bonner R,
J, Buchholz F, Hyman AA, Mann M (2015) A Tate S, Gingras AC (2013) Mapping differen-
human interactome in three quantitative tial interactomes by affinity purification cou-
dimensions organized by stoichiometries and pled with data-independent mass spectrometry
abundances. Cell 163(3):712–723. acquisition. Nat Methods 10(12):1239–1245.
doi:10.1016/j.cell.2015.09.053 doi:10.1038/nmeth.2702
6. Huttlin EL, Ting L, Bruckner RJ, Gebreab F, 14. Collins BC, Gillet LC, Rosenberger G, Rost
Gygi MP, Szpyt J, Tam S, Zarraga G, Colby HL, Vichalkovski A, Gstaiger M, Aebersold R
G, Baltier K, Dong R, Guarani V, Vaites LP, (2013) Quantifying protein interaction dynam-
Ordureau A, Rad R, Erickson BK, Wuhr M, ics by SWATH mass spectrometry: application
Chick J, Zhai B, Kolippakkam D, Mintseris J, to the 14-3-3 system. Nat Methods 10(12):
Obar RA, Harris T, Artavanis-Tsakonas S, 1246–1253. doi:10.1038/nmeth.2703
Sowa ME, De Camilli P, Paulo JA, Harper JW,
Gygi SP (2015) The bioplex network: a sys- 15. Hubner NC, Bird AW, Cox J, Splettstoesser
tematic exploration of the human interac- B, Bandilla P, Poser I, Hyman A, Mann M
tome. Cell 162(2):425–440. d oi:10.1016/j. (2010) Quantitative proteomics combined
cell.2015.06.043 with BAC TransgeneOmics reveals in vivo
protein interactions. J Cell Biol 189(4):739–754. screening. Nat Methods 12(6):553–560.

doi:10.1083/jcb.200911091 doi:10.1038/nmeth.3395
16. Roncagalli R, Hauri S, Fiore F, Liang Y, Chen 24. Roux KJ, Kim DI, Raida M, Burke B (2012) A
Z, Sansoni A, Kanduri K, Joly R, Malzac A, promiscuous biotin ligase fusion protein iden-
Lahdesmaki H, Lahesmaa R, Yamasaki S, Saito tifies proximal and interacting proteins in
T, Malissen M, Aebersold R, Gstaiger M, mammalian cells. J Cell Biol 196(6):801–810.
Malissen B (2014) Quantitative proteomics doi:10.1083/jcb.201112098
analysis of signalosome dynamics in primary T 25. Rhee HW, Zou P, Udeshi ND, Martell JD,
cells identifies the surface receptor CD6 as a lat Mootha VK, Carr SA, Ting AY (2013)
adaptor-independent TCR signaling hub. Nat Proteomic mapping of mitochondria in living
Immunol 15(4):384–392. doi:10.1038/ cells via spatially restricted enzymatic tagging.
ni.2843 Science 339(6125):1328–1331. doi:10.1126/
17. Blagoev B, Kratchmarova I, Ong SE, Nielsen science.1230593
M, Foster LJ, Mann M (2003) A proteomics 26. Kim DI, Birendra KC, Zhu W, Motamedchaboki
strategy to elucidate functional protein-protein K, Doye V, Roux KJ (2014) Probing nuclear
interactions applied to EGF signaling. Nat pore complex architecture with proximity-
Biotechnol 21(3):315–318. doi:10.1038/ dependent biotinylation. Proc Natl Acad Sci U
nbt790 S A 111(24):E2453–E2461. doi:10.1073/
18. Hilger M, Mann M (2012) Triple SILAC to pnas.1406459111
determine stimulus specific interactions in the 27. Lambert JP, Tucholska M, Go C, Knight JD,
Wnt pathway. J Proteome Res 11(2):982–994. Gingras AC (2015) Proximity biotinylation
doi:10.1021/pr200740a and affinity purification are complementary
19. Pagliuca FW, Collins MO, Lichawska A, approaches for the interactome mapping of
Zegerman P, Choudhary JS, Pines J (2011) chromatin-associated protein complexes.
Quantitative proteomics reveals the basis for J Proteomics 118:81–94. doi:10.1016/j.
the biochemical specificity of the cell-cycle jprot.2014.09.011
machinery. Mol Cell 43(3):406–417. 28. Dingar D, Kalkat M, Chan PK, Srikumar T,
doi:10.1016/j.molcel.2011.05.031 Bailey SD, Tu WB, Coyaud E, Ponzielli R,
20. Lavallee-Adam M, Rousseau J, Domecq C, Kolyar M, Jurisica I, Huang A, Lupien M, Penn
Bouchard A, Forget D, Faubert D, Blanchette LZ, Raught B (2015) BioID identifies novel
M, Coulombe B (2013) Discovery of cell com- c-MYC interacting partners in cultured cells and
partment specific protein-protein interactions xenograft tumors. J Proteomics 118:95–111.
using affinity purification combined with tan- doi:10.1016/j.jprot.2014.09.029
dem mass spectrometry. J Proteome Res 29. Couzens AL, Knight JD, Kean MJ, Teo G,
12(1):272–281. doi:10.1021/pr300778b Weiss A, Dunham WH, Lin ZY, Bagshaw RD,
21. Kaake RM, Wang X, Burke A, Yu C, Kandur W, Sicheri F, Pawson T, Wrana JL, Choi H,
Yang Y, Novtisky EJ, Second T, Duan J, Kao A, Gingras AC (2013) Protein interaction net-
Guan S, Vellucci D, Rychnovsky SD, Huang L work of the mammalian Hippo pathway reveals
(2014) A new in vivo cross-linking mass spec- mechanisms of kinase-phosphatase interac-
trometry platform to define protein-protein tions. Sci Signal 6(302):rs15. doi:10.1126/
interactions in living cells. Mol Cell Proteomics scisignal.2004712
13(12):3533–3543. doi:10.1074/mcp. 30. Firat-Karalar EN, Rauniyar N, Yates JR 3rd,
M114.042630 Stearns T (2014) Proximity interactions among
22. Babu M, Vlasblom J, Pu S, Guo X, Graham C, centrosome components identify regulators of
Bean BD, Burston HE, Vizeacoumar FJ, centriole duplication. Curr Biol 24(6):664–670.
Snider J, Phanse S, Fong V, Tam YY, Davey M, doi:10.1016/j.cub.2014.01.067
Hnatshak O, Bajaj N, Chandran S, Punna T, 31. Gupta GD, Coyaud E, Goncalves J, Mojarad
Christopolous C, Wong V, Yu A, Zhong G, Li BA, Liu Y, Wu Q, Gheiratmand L, Comartin
J, Stagljar I, Conibear E, Wodak SJ, Emili A, D, Tkach JM, Cheung SW, Bashkurov M,
Greenblatt JF (2012) Interaction landscape of Hasegan M, Knight JD, Lin ZY, Schueler M,
membrane-protein complexes in Saccharomyces Hildebrandt F, Moffat J, Gingras AC, Raught
cerevisiae. Nature 489(7417):585–589. B, Pelletier L (2015) A dynamic protein inter-
doi:10.1038/nature11354 action landscape of the human centrosome-
23. Hakhverdyan Z, Domanski M, Hough LE, cilium interface. Cell 163(6):1484–1499.
Oroskar AA, Oroskar AR, Keegan S, Dilworth doi:10.1016/j.cell.2015.10.065
DJ, Molloy KR, Sherman V, Aitchison JD,
32. Coyaud E, Mis M, Laurent EM, Dunham
Fenyo D, Chait BT, Jensen TH, Rout MP, WH, Couzens AL, Robitaille M, Gingras AC,
LaCava J (2015) Rapid, optimized interactomic Angers S, Raught B (2015) BioID-based
identification of Skp cullin F-box (SCF)beta- Nat Biotechnol 28(10):1015–1017.

TrCP1/2 E3 ligase substrates. Mol Cell doi:10.1038/nbt1010-1015
Proteomics 14(7):1781–1795. doi:10.1074/ 38. Liu G, Zhang J, Choi H, Lambert JP, Srikumar
mcp.M114.045658 T, Larsen B, Nesvizhskii AI, Raught B, Tyers
33. Cheng YS, Seibert O, Kloting N, Dietrich A, M, Gingras AC (2012) Using ProHits to store,
Strassburger K, Fernandez-Veledo S, Vendrell annotate, and analyze affinity purification-mass
JJ, Zorzano A, Bluher M, Herzig S, Berriel spectrometry (AP-MS) data. Curr Protoc
Diaz M, Teleman AA (2015) PPP2R5C cou- Bioinformatics Chapter 8:Unit8.16.
ples hepatic glucose and lipid homeostasis. doi:10.1002/0471250953.bi0816s39
PLoS Genet 11(10):e1005561. doi:10.1371/ 39. Teo G, Liu G, Zhang J, Nesvizhskii AI, Gingras
journal.pgen.1005561 AC, Choi H (2014) SAINTexpress: improve-
34. Kean MJ, Couzens AL, Gingras AC (2012) ments and additional features in significance
Mass spectrometry approaches to study mam- analysis of INTeractome software. J Proteomics
malian kinase and phosphatase associated pro- 100:37–43. doi:10.1016/j.jprot.2013.10.023
teins. Methods 57(4):400–408. doi:10.1016/j. 40. Mellacheruvu D, Wright Z, Couzens AL,
ymeth.2012.06.002 Lambert JP, St-Denis NA, Li T, Miteva YV,
35. Banks CA, Boanca G, Lee ZT, Florens L, Hauri S, Sardiu ME, Low TY, Halim VA,
Washburn MP (2015) Proteins interacting Bagshaw RD, Hubner NC, Al-Hakim A,
with cloning scars: a source of false positive Bouchard A, Faubert D, Fermin D, Dunham
protein-protein interactions. Sci Rep 5:8530. WH, Goudreault M, Lin ZY, Badillo BG,
doi:10.1038/srep08530 Pawson T, Durocher D, Coulombe B,
36. Olhovsky M, Williton K, Dai AY, Pasculescu A, Aebersold R, Superti-Furga G, Colinge J, Heck
Lee JP, Goudreault M, Wells CD, Park JG, AJ, Choi H, Gstaiger M, Mohammed S, Cristea
Gingras AC, Linding R, Pawson T, Colwill K IM, Bennett KL, Washburn MP, Raught B,
(2011) OpenFreezer: a reagent information Ewing RM, Gingras AC, Nesvizhskii AI (2013)
management software system. Nat Methods The CRAPome: a contaminant repository for
8(8):612–613. doi:10.1038/nmeth.1658 affinity purification-mass spectrometry data.
37. Liu G, Zhang J, Larsen B, Stark C, Breitkreutz Nat Methods 10(8):730–736. doi:10.1038/
A, Lin ZY, Breitkreutz BJ, Ding Y, Colwill K, nmeth.2557
Pasculescu A, Pawson T, Wrana JL, Nesvizhskii 41. Knight JD, Liu G, Zhang JP, Pasculescu A,
AI, Raught B, Tyers M, Gingras AC (2010) Choi H, Gingras AC (2015) A web-tool for
ProHits: integrated software for mass visualizing quantitative protein-protein inter-
spectrometry- based interaction proteomics. action data. Proteomics 15(8):1432–1436.
doi:10.1002/pmic.201400429
Chapter 11
LUMIER: A Discovery Tool for Mammalian Protein

Interaction Networks
Miriam Barrios-Rodiles, Jonathan D. Ellis, Benjamin J. Blencowe,
and Jeffrey L. Wrana
Abstract
Protein–protein interactions (PPIs) play an essential role in all biological processes. In vivo, PPIs occur
dynamically and depend on extracellular cues. To discover novel protein–protein interactions in mamma-
lian cells, we developed a high-throughput automated technology called LUMIER (LUminescence-based
Mammalian IntERactome). In this approach, we co-express a Luciferase (LUC)-tagged fusion protein
along with a Flag-tagged protein in an efficiently transfectable cell line such as HEK-293T cells. The inter-
action between the two proteins is determined by co-immunoprecipitation using an anti-Flag antibody,
and the presence of the LUC-tagged interactor in the complex is subsequently detected via its luciferase
activity. LUMIER can easily detect transmembrane protein partners, interactions that are signaling- or
splice isoform-dependent, as well as those that may occur only in the presence of posttranslational modifi-
cations. Using various collections of Flag-tagged proteins, we have generated protein interaction networks
for several TGF-β family receptors, Wnt pathway members, and have systematically analyzed the effect of
neural-specific alternative splicing on protein interaction networks. The results have provided important
insights into the physiological and functional relevance of some of the novel interactions found. LUMIER
is highly scalable and can be used for both low- and high-throughput strategies. LUMIER is thus a valu-
able tool for the identification and characterization of dynamically regulated PPIs in mammalian systems.
Here, we describe a manual version of LUMIER in a 96-well format that can be easily implemented in any
laboratory.
Key words Protein–protein interaction, LUMIER, Mammalian cells, Signaling pathways,

Transmembrane proteins, Binary complex, Ternary complex
1 Introduction
All cells respond to extracellular cues from their environment. The

cellular reponse to such triggers is orchestrated via diverse signal-
ing pathways that rely on dynamic interactions between intracel-
lular proteins. These intracellular effectors may possess enzymatic
activities (i.e., kinases, phosphatases, ubiquitin ligases) or contain
specific structural domains that allow their association with other
proteins. Over the last few decades, several methods have been
137
138 Miriam Barrios-Rodiles et al.
developed to screen for novel binary protein interactions. An

important and low cost assay has been the yeast-two-hybrid method
(Y2H) that relies on a transcriptional reporter readout that is stim-
ulated when two proteins that are independently fused to a DNA
binding domain or a transcription activation domain interact in a
yeast cell [1]. Y2H was first applied to yeast protein–protein inter-
actions (PPIs) and later applied to mammalian proteins [2, 3].
Other examples of methods exploiting the Y2H principle of PPIs
indirectly detected through the transcriptional activation of a
reporter gene include MAPPIT [4], MYTH [5] and MaMTH [6].
Another class of PPI detection methods are based on the recon-
struction of an enzyme or a fluorescent protein that has been “split”
in two parts with each part fused to the potential interacting part-
ners. When the putative partners come into close proximity, the
enzymatic activity or fluorescence is restored. In this category, Split
dihydrofolate reductase, split-fluorescent protein (such as GFP) or
split-luciferase methods have been reported [7–9]. Finally, examples
of a different class of methods to detect PPIs which rely on the phys-
ical interaction of partners that have been fused to fluorescent tags
or an enzyme for detection include FRET/BRET [10], NAPPA
[11], and LUMIER [12]. The evidence for the interaction is pro-
vided by detection of the attached fluorescent signal or enzymatic
activity fused to one of the partners.
We first reported LUMIER, which stands for LUminescence-
based Mammalian IntERactome as a high-throughput assay with
a luminescence readout [12]. For this method, we co-express a
Luciferase (LUC)-tagged fusion protein (Bait) along with a Flag-
tagged protein (Prey) in HEK-293T cells, or any other efficiently
transfectable cell line. The interaction between the two proteins is
determined by co-immunoprecipitation using an anti-Flag anti-
body, and subsequently detecting the presence of the LUC-tagged
interactor in the complex by measuring luciferase activity (Fig. 1a).
We have used different luciferases to perform LUMIER assays in
manual (low throughput) or automated screens (high-through-
put). During the development of LUMIER, we used Renilla
luciferase (RLUC or RL, 36 kDa) fused to TGF-β core pathway
components to uncover novel interactions in this pathway, includ-
ing the associations between the TGF-β receptor with Occludin or
Par6 [12, 13]. Furthermore, we observed the dynamic gain and
loss of interactions for pathway components Smad2 and Smad4
upon signal activation [12]. We also used Firefly luciferase
(64 kDa) on Baits to unveil novel interactions for members of the
Wnt signaling pathway such as the interaction between β-catenin
and Ube2m [14]. Moreover, the introduction of genetically mod-
ified and smaller luciferases such as NanoLuc (19 kDa) [15], with
an optimized substrate plus enhanced sensitivity, allows the use of
luminometers without ultra-sensitive photomultipliers and facili-
tates an increased throughput, a key feature for network biology
studies and drug screens.
LUMIER and Mammalian Protein Interaction Networks 139
Fig. 1 (a) LUMIER strategy for the detection of protein–protein interactions. A protein of interest (Bait) tagged
with Renilla luciferase (RL or RLUC) is co-expressed in mammalian cells along with a 3Flag-tagged putative
partner (Prey). If the two proteins interact (in the presence or absence of a stimulus), the complex can be
immunoprecipitated using the anti-Flag antibody. The presence of the Bait is detected by light emission upon
Renilla substrate addition. (b) LUMIER strategy for domain mapping. TBRI-HA-RL wild type was expressed
alone or together with wild type (WT) or mutant Occludin (OCLN) harboring the indicated domain deletions (see
schematic), and interactions were measured by LUMIER (top panel) or by immunoblotting with antibody against
HA. Expression of proteins was confirmed by immunoblotting total cell lysates, as indicated. From Barrios-
Rodiles M. et al., 2005 Science (307):1621–1625. Reprinted with Permission from AAAS. (c) LUMIER strategy
for the detection of a ternary complex. The RLUC-tagged Bait is co-expressed with a 3Flag-tagged Prey1 and
a second 3HA-tagged Prey2. A first immunoprecipitation is performed using the anti-Flag antibody and an
aliquot is taken to perform a luciferase assay and western blot to confirm the presence of the Bait and the
Flag-tagged Prey1. Following elution using Flag-peptide to release the complex from the beads, a second
immunoprecipitation is conducted using the anti-HA antibody. A second luciferase assay and western blot is
perfomed to determine the presence of the Bait and the HA-tagged Prey2 (see ref. 19)
In addition to its simple set up, LUMIER has several advantages.

It can easily detect interactions for transmembrane proteins [16]
including seven-transmembrane domain receptors (GPCRs) [17].
LUMIER detects PPIs dependent on posttranscriptional modifica-
tions such as phosphorylation that change dynamically during cell
signaling [12]. Moreover, mapping the domain responsible for a

PPI can be quickly achieved with this assay [12, 18] (Fig. 1b),
while a ternary protein complex can be detected using two-step
immunoprecipitations [19] (Fig. 1c). We have taken advantage of
the semiquantitative nature of LUMIER to detect differential PPIs
due to tissue-dependent alternative splicing events [20]. This work
revealed that approximately one-third of neural- specific exons
either disrupt or promote specific PPIs, many of which are impor-
tant for neural cell functions. Furthermore, recent protocol modi-
fications combine LUMIER with ELISA measurements to quantify
the Flag-tagged protein in the immunoprecipitation to generate a
more quantitative evaluation of interaction strength. Quantification
of the Flag-tagged partner also allows the immunoprecipitation of
the Flag-tagged protein to be verified in each well and thus false
negatives can be identified [21]. In summary, since we first devel-
oped the assay in 2005 [12], LUMIER has proven to be a useful
tool for the systematic mapping and functional characterization of
mammalian PPI networks. This method has been employed in
several laboratories for manual or high-throughput applications
[22–25], and we have previously described the protocol to con-
duct LUMIER automated screens [3, 12]. In this chapter we focus
on the implementation of LUMIER for manual experiments that
can be easily performed in any research laboratory.
2 Materials
2.1 Cell Culture 1. Human Embryonic Kidney 293T (HEK-293T) cells.

Reagents 2. Plasmids for subcloning Baits and Preys (Fig. 2; see Notes 1
and 2).
(a) pCMV5-hRLUC AmpR, N or C terminal humanized
Renilla luciferase tag (AF362545).
(b) pCMV5-3Flag AmpR, N or C terminal 3Flag tag.
(c) 3Flag-hRLUC AmpR, this plasmid encodes Renilla lucifer-
ase tagged with 3Flag at the N-terminus.
(d) pCMV5 AmpR, plasmid without any insert.
3.
Dulbeccos’s Modified Eagles’s Medium—high glucose
(DMEM): DMEM (500 ml, SIGMA), 10 % fetal calf serum,
5 ml penicillin/streptomycin (100×, 10,000 U/ml, Life
Technologies). Ready-to-use media are stored at 4 °C.
4. Trypsin–EDTA (0.5 %, 10×, Life Technologies), diluted to 1×
in sterile Dulbecco’s Phosphate-Buffered Saline (DPBS).
5. Sterile DPBS, no calcium, no magnesium.
6. Serum- and antibiotic-free DMEM.
7. QIAGEN PolyFect transfection reagent.
Fig. 2 Plasmids used in LUMIER. The vectors are pCMV5-based for high protein expression when transiently
transfected into HEK-239T cells. Only unique restriction enzyme sites at the multiple cloning site (MCS) are
shown. (a and b) NhRLUC and N3Flag are used to tag the Bait and Prey at the N-terminus. There is no stop codon
at the 3′ end sequence of Renilla and N3Flag. (c and d) ChRLUC and C3Flag are used to tag the Bait and Prey at
the C-terminus. The MCS of ChRLUC is a composite of pCMV5 and pCDNA3 polylinker regions. There is a stop
codon at the 3′ end of Renilla and 3Flag. The GenBank accession number for the humanized Renilla luciferase
(hRLUC) sequence is AF362545. The amino acid sequence for the 3Flag is: MDYKDHDGDYKDHDIDYKDDDDK.
The cytomegalovirus (CMV) promoter drives high levels of gene expression for the Bait or Prey. The AmpR gene
encodes beta-lactamase for ampicillin resistance selection in bacteria
2.2 LUMIER 1. Monoclonal M2 anti-Flag-antibody (Sigma).

and ELISA Reagents 2. Renilla-Glo Assay System (Promega).
3. Monoclonal Anti-Flag M2-Peroxidase (HRP) clone M2
(Sigma).
4. ELISA pico substrate (Pierce).
2.3 Equipment 1. HandEvac aspirator 8-channel with ejector for disposable tips
(Argos EV514 and EV520).
2. Multichannel pipets (for 10 μl, 20–300 μl, and 1 ml).
3. Transparent seals (QIAGEN).

4. Orbital shaker for plates.
5. Luminometer.
6. Corning, poly-L-lysine coated 96-well plates for cell culture.
7. Greiner Lumitrac 600 flat-bottom 96-well plates.
8. COSTAR White round-bottom 96-well plates.
2.4 Buffers 1. PBS: 137 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4, and
1.8 mM KH2PO4. Adjust pH to 7.4 with HCl.
2. Blocking Buffer: 3 % bovine serum albumin, 5 % sucrose, 0.5 %
Tween 20 in PBS.
3. Lysis Buffer: 50 mM Tris–HCl pH 7.4, 150 mM NaCl, 1 mM
tetrasodium EDTA, 0.5 % Triton X-100. A 10× stock can be
stored at 4 °C and diluted to 1× just before lysis when adding
phosphatase and protease inhibitors.
4. Phosphatase Inhibitors (final concentrations): 10 mM sodium
pyrophosphate tetrabasic decahydrate, 1 mM sodium orthovan-
adate, 25 mM NaF. Stock solutions at 10×, 100× and 20×
respectively, can be prepared and stored at 4 °C to be added
just before lysis.
5. Protease Inhibitors: 1 mM phenylmethanesulfonylfluoride
(PMSF; stock 100 mM in ethanol), 10 μg/ml pepstatin A
(stock 1 mg/ml in DMSO), 100 μg/ml soy trypsin inhibitor
(stock 10 mg/ml in Tris–EDTA buffer). 10 μg/ml leupeptin,
10 μg/ml antipain, 50 μg/ml aprotinin, 100 μg/ml benzami-
dine hydrochloride (stocks 1 mg/ml, 1 mg/ml, 5 mg/ml, and
10 mg/ml, respectively in Tris–EDTA buffer). Stocks for pro-
tease inhibitors are stored at −20 °C.
6. Tris–EDTA Buffer for protease inhibitors: 10 mM Tris–HCl
pH 7.4, 1 mM tetrasodium EDTA.
7. Wash buffer: 50 mM Tris–HCl pH 7.4, 150 mM NaCl, 1 mM
tetrasodium EDTA, and 0.1 % Triton X-100.
8. Working Renilla-Glo substrate: Dilute the 100× substrate to
1× using the provided buffer in the kit.
9. PBS-T: 0.05 % Tween 20 in PBS.
10. Buffer to dilute Flag-HRP antibody (Flag-HRP buffer): 5 %
Tween 20 and 1 % fetal bovine serum in PBS.
11. Working Super signal ELISA pico substrate: Reagent A and B
are mixed 1:1 ratio and then diluted 1:5 with distilled water.
3 Methods
See Fig. 3a for assay workflow.

Fig. 3 (a) A LUMIER workflow diagram. Forty-eight hours post-transfection in 96-well plates, the cells are
lysed. Lysate is split into Immunoprecipitation (IP) plate and Totals plate. After 1 h incubation the IP plate is
washed and Renilla substrate added to IP and Totals plates to assess the interaction and to measure the
expression levels of the Bait. Subsequently, an anti-Flag-HRP-conjugated antibody is added to the IP plate to
measure the expression levels of the 3Flag-tagged Prey. After lysis, LUMIER results can be obtained within 4 h.
(b) LUMIER data shows that exon 7 of Bridging Integrator 1 (Bin1) promotes an interaction with Dynamin 2
(Dnm2). HEK-293T cells were transfected as indicated. Dnm2-3Flag was used as prey while Bin1-RL (without
exon 7) or Bin1 + ex7-RL (with exon 7) were used a baits. Averages for Renilla luciferase signals from IP plate,
Totals plate and Flag-HRP measurements are shown on the left-side graphs (error bars show ± standard
deviations from four technical replicates). The LIR IP, NLIR, and NtFLIR values calculated as described in Data
Analysis are shown on the right-side graphs indicating that exon 7 promotes the interaction between Bin1 and
Dnm2
3.1 Day 1: Plate 1. Manually plate 15,000 HEK293T cells per well in 70 μl/well,
HEK-293T Cells in poly-l-lysine coated 96-well plates (see Note 3) in DMEM
and Coat Plates with with 10 % fetal bovine serum and antibiotics (penicillin/strep-
Anti-Flag Antibody tomycin). Incubate overnight at 37 °C.
2. Coat white-flat-bottom Lumitrac plates with 100 μl/well of

anti-Flag antibody in 1× PBS at 10 ng/μl (IP plate). Seal plates
and store at 4 °C overnight. No shaking is necessary.
3.2 Day 2: Transfect 1. Transfect every Bait-Prey combination at least in duplicate

Cells with LUMIER (quadruplicates are recommended). Dilute 100 ng of 3Flag-
Constructs and Block tagged cDNA (Prey) and 100 ng of the Renilla luciferase (RL)
Plates tagged cDNA (Bait) in DMEM without serum or antibiotics.
Transfection is performed using PolyFect (see Note 4) accord-
ing to 96-well plate manufacturer’s protocol, using a 1:4 ratio
of DNA (μg): PolyFect (μl). Maintain cells at 37 °C for 48 h.
Include as negative controls the following conditions: empty
vector only (pCMV5) and a second condition where the
RLUC-tagged Bait is present, while the 3Flag-tagged Prey is
substituted by empty vector (pCMV5). As a positive control to
ensure the assay is working transfect 3Flag-hRLUC (see Notes 5
and 6 for this and other recommended controls).
2. Aspirate the liquid from Lumitrac plates using the 8-channel
aspirator and pat-dry plate on a paper towel to remove all liquid.
Dispense 250 μl/well of Blocking Buffer. Shake at low speed in
orbital shaker for 1 h at room temperature. Aspirate all liquid
and pat-dry on paper towel. Store sealed plate at 4 °C.
3.3 Day 4: Lyse Cells 1. Forty-eight hours post-transfection aspirate media from cells
and Perform and dispense 150 μl/well of Lysis Buffer that includes phos-
Immunoprecipitation phatase and protease inhibitors (see Note 7). Shake plate at low
(IP) and Flag-Tagged speed for 40 min at 4 °C.
Prey Measurement 2. Transfer 90 μl of lysate to anti-Flag-coated plate (Lumitrac; IP
plate) and 20 μl to a round bottom-white plate (white round-
bottom; Totals plate). Incubate both plates at 4 °C for 1 h,
without shaking.
3. Aspirate liquid from IP plate using the 8-channel aspirator and
dispense 250 μl of Wash buffer for each wash. Repeat wash step
5–6 times. Aspirate liquid from last wash and flick off remain-
ing liquid. Dispense 100 μl/well of Working Renilla-Glo sub-
strate to IP plates, incubate for 10 min at room temperature,
and read in ENVISION luminometer 100 ms/well to detect
interactions (see Note 8).
4. Measure expression levels of the Renilla-tagged proteins.
Dispense 20 μl/well of Working Renilla-Glo substrate to
Totals plate containing 20 μl of lysate, incubate for 10 min at
room temperature, and read in ENVISION luminometer
100 ms/well.
5. Aspirate the liquid from IP plate and flick to remove all liquid.
Wash IP plate eight times with 250 μl/well of PBS-T each
time. Dispense 100 μl/well of anti-Flag-HRP antibody at
1:20,000 dilution (see Note 9) in Flag-HRP Buffer. Incubate
for 30 min at 4 °C without shaking.
6. Aspirate all liquid and wash eight times with 250 μl/well PBS-
T. Aspirate liquid from last wash and flick off remaining liquid.
Dispense 100 μl/well of Working Super signal ELISA pico
substrate. Incubate at room temperature for 7 min and read in
ENVISION for 20 ms/well.
3.4 Data Analysis The Normalized LUMIER Intensity Ratio (NLIR) value for each
interaction is calculated as follows: The LUMIER Intensity Ratio
for each Bait/Prey combination is first determined from the lucif-
erase activity in the immunoprecipitates (LIR IP). This LIR IP is
calculated as the luciferase intensity in IP plate when each Bait
(RLUC-fusions) and Prey (3Flag-tagged cDNAs) are co-
transfected, over the luciferase intensity from immunoprecipitates
when the Bait (RLUC-fusion) is co-transfected with empty vector
(pCMV5).
LIR IP = signal IP (Bait + Prey)/signal IP (Bait + pCMV5).
The LIR TOT is caculated in a similar manner from the lucif-
erase activity measured in the total cell lysates (Totals plate) of each
Bait/Prey combination over the luciferase activity from total cell
lysates of each Bait/empty vector.
LIR TOT = signal Totals (Bait + Prey)/signal Totals (Bait + pCMV5).
The Normalized LIR (NLIR) is the ratio of LIR IP/LIR TOT
for each Bait/Prey combination. In general, an NLIR value equal
or higher than 3 is considered as a positive interaction.
To give a more quantitative interaction score a Flag Intensity
Ratio (FIR) is calculated to account for changes in the immuno-
precipitation of Flag-tagged Prey protein with different RLUC-
tagged Baits. This FIR is calculated with the Flag-HRP signal value
from immunoprecipitates when Bait and Prey are co-transfected,
over the Flag-HRP signal value from immunoprecipitates when
the Prey is co-transfected with empty vector.
FIR = signal Flag-HRP (Bait + Prey)/signal Flag-HRP
(Bait + pCMV5).
A final Normalized to Flag LIR value (NtFLIR) is the ratio of
NLIR/FIR for each interaction tested. See Fig. 3b for an example
on data analysis.
4 Notes
1. The plasmids to tag the Bait with Renilla luciferase and the
Prey with 3Flag are pCMV5-based. There is also a Gateway®
version for each.
2. The Baits and Preys can be tagged at either N- or C-terminus.
This is an important consideration when designing a LUMIER
experiment. If the Bait of interest has an important domain at
one end (i.e., PDZ-binding domain or a kinase domain), place

the luciferase tag at the opposite end. Alternatively, place the
tag at both ends and test the performance of the two Baits (N-
and C-terminally tagged) in preliminary experiments using a
known Flag-tagged partner as positive control.
3. Regular 96-well TC plates can be coated with poly-l-lysine
(Sigma cat. P-1399) at 50 μg/ml dissolved in PBS and filter-
sterilized. Dispense 100 μl/well of diluted poly-l-lysine and
incubate at room temperature for 1 h. Rinse twice with sterile
PBS and plate cells immediately after PBS has been removed.
Stock of poly-l-Lysine can be made at 2 mg/ml in PBS and
stored at −20 °C. Coating the plates is strongly recommended
because during transfection the HEK-293T cells can easily
dettach and increase variability.
4. Other transfection reagents alternative to PolyFect can be used
with HEK-293T cells, including the Calcium Phosphate method
which is low cost, or Lipofectamine 2000 (Life Technologies).
PolyFect is recommended because of its high transfection effi-
ciency, cells can be transfected in the presence of antibiotics and
there is no need to change media the next day.
5. To ensure that the assay is working, a very good positive control
to use is a plasmid encoding 3Flag-hRLUC. When this plasmid
is transfected in HEK-293T cells, a high signal for Renilla
luciferase activity is detected in the IP plate and Totals plate
upon addition of Renilla-Glo substrate, but also from Flag-
HRP after the addition of the ELISA pico substrate.
6. To distinguish whether a 3Flag-tagged partner is interacting
with Renilla luciferase within the Bait fusion, include as control
a condition where the 3Flag-tagged cDNA is co-transfected
with the ChRLUC plasmid (Fig. 2c) which has luciferase
activity in total lysates but is not fused to any Bait. In high-
throughput screens out of ~560 Preys only few proteins were
found to interact with Renilla luciferase (UBA52, UBC, UBB,
RPS27A, SQSTM1, and ECSIT).
7. A 10× lysis buffer stock can be prepared and diluted to 1×
while adding phosphatase and protease inhibitos just before
use. Instead of PMSF, AEBSF (Sigma cat. A8456) can be used
at a final concentration of 0.5 mM, because it is a more stable
serine protease inhibitor.
8. A Luminometer with injectors can be used if a luminometer
with a high sensitivity detector like ENVISION (Perkin Elmer)
is not available. For this, the Renilla Luciferase Assay system
kit (Promega), which contains a ‘flash-kinetics’ substrate with
high sensitivity must be used. This substrate has to be dis-
pensed by a luminometer’s injectors because the signal obtained
with this kit has a very short half-life (~2 min), in contrast to
the Renilla-Glo substrate, whose signal half-life is greater than

60 min at room temperature.
9. Comparison between different lot numbers of a rabbit poly-
clonal Flag antibody conjugated to HRP from Abcam to a
mouse monoclonal Flag antibody conjugated to HRP from
Sigma showed that the Sigma antibody performs significantly
better in ELISA experiments in terms of signal to noise ratio.
Acknowledgments
This work was supported by funding from the Canadian Institutes

of Health Research (CIHR) (J.L.W. and B.J.B.), the Alzheimer’s
Society of Canada (B.J.B.), and the Krembil Foundation (J.L.W.).
We would also like to thank Dr. Saranya Kittanakom for the plas-
mid schematics framework.
References
1. Fields S, Song O (1989) A novel genetic sys- 7. Pelletier JN, Campbell-Valois FX, Michnick
tem to detect protein-protein interactions. SW (1998) Oligomerization domain-directed
Nature 340(6230):245–246 reassembly of active dihydrofolate reductase
2. Lievens S, Lemmens I, Tavernier J (2009) from rationally designed fragments. Proc Natl
Mammalian two-hybrids come of age. Trends Acad Sci U S A 95(21):12141–12146
Biochem Sci 34(11):579–588 8. Wilson CG, Magliery TJ, Regan L (2004)
3. Braun P, Tasan M, Dreze M, Barrios-Rodiles Detecting protein-protein interactions with
M, Lemmens I, Yu H, Sahalie JM, Murray RR, GFP-fragment reassembly. Nat Methods
Roncari L, de Smet AS, Venkatesan K, Rual JF, 1(3):255–262
Vandenhaute J, Cusick ME, Pawson T, Hill 9. Remy I, Michnick SW (2006) A highly sensi-
DE, Tavernier J, Wrana JL, Roth FP, Vidal M tive protein-protein interaction assay based on
(2009) An experimentally derived confidence Gaussia luciferase. Nat Methods 3(12):
score for binary protein-protein interactions. 977–979
Nat Methods 6(1):91–97 10. Boute N, Jockers R, Issad T (2002) The use of
4. Eyckerman S, Verhee A, der Heyden JV, resonance energy transfer in high-throughput
Lemmens I, Ostade XV, Vandekerckhove J, screening: BRET versus FRET. Trends
Tavernier J (2001) Design and application of a Pharmacol Sci 23(8):351–354
cytokine-receptor-based interaction trap. Nat 11. Ramachandran N, Hainsworth E, Bhullar B,
Cell Biol 3(12):1114–1119 Eisenstein S, Rosen B, Lau AY, Walter JC,
5. Stagljar I, Korostensky C, Johnsson N, te LaBaer J (2004) Self-assembling protein
Heesen S (1998) A genetic system based on microarrays. Science 305(5680):86–90
split-ubiquitin for the analysis of interactions 12. Barrios-Rodiles M, Brown KR, Ozdamar B,
between membrane proteins in vivo. Proc Natl Bose R, Liu Z, Donovan RS, Shinjo F, Liu Y,
Acad Sci U S A 95(9):5187–5192 Dembowy J, Taylor IW, Luga V, Przulj N,
6. Petschnigg J, Groisman B, Kotlyar M, Taipale Robinson M, Suzuki H, Hayashizaki Y, Jurisica
M, Zheng Y, Kurat CF, Sayad A, Sierra JR, I, Wrana JL (2005) High-throughput mapping
Mattiazzi Usaj M, Snider J, Nachman A, of a dynamic signaling network in mammalian
Krykbaeva I, Tsao MS, Moffat J, Pawson T, cells. Science 307(5715):1621–1625
Lindquist S, Jurisica I, Stagljar I (2014) The 13. Ozdamar B, Bose R, Barrios-Rodiles M, Wang
mammalian-membrane two-hybrid assay HR, Zhang Y, Wrana JL (2005) Regulation of
(MaMTH) for probing membrane-protein the polarity protein Par6 by TGFbeta receptors
interactions in human cells. Nat Methods controls epithelial cell plasticity. Science
11(5):585–592 307(5715):1603–1609
14. Miller BW, Lau G, Grouios C, Mollica E, embryonic stem cells. Cell Rep 5(6):
Barrios-Rodiles M, Liu Y, Datti A, Morris Q, 1611–1624
Wrana JL, Attisano L (2009) Application of an 20. Ellis JD, Barrios-Rodiles M, Colak R, Irimia
integrated physical and functional screening M, Kim T, Calarco JA, Wang X, Pan Q,
approach to identify inhibitors of the Wnt O’Hanlon D, Kim PM, Wrana JL, Blencowe
pathway. Mol Syst Biol 5:315 BJ (2012) Tissue-specific alternative splicing
15. Hall MP, Unch J, Binkowski BF, Valley MP, remodels protein-protein interaction networks.
Butler BL, Wood MG, Otto P, Zimmerman K, Mol Cell 46(6):884–892
Vidugiris G, Machleidt T, Robers MB, Benink 21. Taipale M, Krykbaeva I, Koeva M, Kayatekin
HA, Eggers CT, Slater MR, Meisenheimer PL, C, Westover KD, Karras GI, Lindquist S
Klaubert DH, Fan F, Encell LP, Wood KV (2012) Quantitative analysis of HSP90-client
(2012) Engineered luciferase reporter from a interactions reveals principles of substrate rec-
deep sea shrimp utilizing a novel imidazopyr- ognition. Cell 150(5):987–1001
azinone substrate. ACS Chem Biol 7(11): 22. Ryzhakov G, Teixeira A, Saliba D, Blazek K,
1848–1857 Muta T, Ragoussis J, Udalova IA (2013)
16. Xu G, Barrios-Rodiles M, Jerkic M, Turinsky Cross-species analysis reveals evolving and con-
AL, Nadon R, Vera S, Voulgaraki D, Wrana JL, served features of the nuclear factor kappaB
Toporsian M, Letarte M (2014) Novel protein (NF-kappaB) proteins. J Biol Chem 288(16):
interactions with endoglin and activin receptor- 11546–11554
like kinase 1: potential role in vascular net- 23. Blasche S, Mortl M, Steuber H, Siszler G, Nisa
works. Mol Cell Proteomics 13(2):489–502 S, Schwarz F, Lavrik I, Gronewold TM, Maskos
17. Kittanakom S, Barrios-Rodiles M, Petschnigg K, Donnenberg MS, Ullmann D, Uetz P, Kogl
J, Arnoldo A, Wong V, Kotlyar M, Heisler LE, M (2013) The E. coli effector protein NleF is a
Jurisica I, Wrana JL, Nislow C, Stagljar I caspase inhibitor. PLoS One 8(3):e58937
(2014) CHIP-MYTH: a novel interactive pro- 24. Deng Q, Wang D, Li F (2014) Detection of
teomics method for the assessment of agonist- viral protein-protein interaction by microplate-
dependent interactions of the human format luminescence-based mammalian inter-
beta(2)-adrenergic receptor. Biochem Biophys actome mapping (LUMIER). Virol Sin 29(3):
Res Commun 445(4):746–756 189–192
18. Varelas X, Miller BW, Sopko R, Song S, Gregorieff 25. Tahoun A, Siszler G, Spears K, McAteer S, Tree
A, Fellouse FA, Sakuma R, Pawson T, Hunziker J, Paxton E, Gillespie TL, Martinez-Argudo I,
W, McNeill H, Wrana JL, Attisano L (2010) The Jepson MA, Shaw DJ, Koegl M, Haas J, Gally
Hippo pathway regulates Wnt/beta-catenin sig- DL, Mahajan A (2011) Comparative analysis of
naling. Dev Cell 18(4):579–591 EspF variants in inhibition of Escherichia coli
19. Beyer TA, Weiss A, Khomchuk Y, Huang K, phagocytosis by macrophages and inhibition of
Ogunjimi AA, Varelas X, Wrana JL (2013) E. coli translocation through human- and
Switch enhancers interpret TGF-beta and bovine-derived M cells. Infect Immun 79(11):
Hippo signaling to control cell fate in human 4716–4729
Chapter 12
Dual-Color, Multiplex Analysis of Protein Microarrays

for Precision Medicine
Solomon Yeon, Florian Bell, Michael Shultz, Grace Lawrence,
Michael Harpole, and Virginia Espina
Abstract
Generating molecular information in a clinically relevant time frame is the first hurdle to truly integrating precision
medicine in health care. Reverse phase protein microarrays are being utilized in clinical trials for quantifying
posttranslationally modified signal transduction proteins and cellular signaling pathways, allowing direct com-
parison of the activation state of proteins from multiple specimens, or individual patient specimens, within the
same array. This technology provides diagnostic and therapeutic information critical to precision medicine. To
enhance accessibility of this technology, two hurdles must be overcome: data normalization and data acquisi-
tion. Herein we describe an unamplified, dual-color signal detection methodology for reverse phase protein
microarrays that allows multiplex, within spot data normalization, reduces data acquisition time, simplifies
automated spot detection, and provides a stable signal output. This method utilizes Quantum Nanocrystal fluo-
rophore labels (Qdot) substituted for organic fluorophores coupled with an imager (ArrayCAM) that captures
images of the microarray rather than sequentially scanning the array. Streamlining and standardizing the data
analysis steps with ArrayCAM high-resolution, dual mode chromogenic/fluorescent array imaging overcomes
the data acquisition hurdle. The spot location and analysis algorithm provides certain parameter settings
that can be tailored to the particular microarray type (fluorescent vs. colorimetric), resulting in greater than 99 %
spot location sensitivity. The described method demonstrates equivalent sensitivity for a non-amplified Qdot
immunoassay when using automated vs. manual immunostaining procedures.
Key words Fluorescence, Multicolor detection, Precision medicine, Posttranslational modification,

Protein microarray, Protein phosphorylation, Qdot nanocrystal, Receptor tyrosine kinase, Reverse
phase protein microarray, Signal transduction
1 Introduction
How do we provide functional molecular information about an

individual person’s tumor or disease state? What is the state of a
patient’s cellular signal transduction cascades before, during, and
after therapy? What are the on and off target effects of molecular
targeted inhibitors? These questions represent a small fraction of
the complex proteomic information that can be gleaned from
reverse phase protein microarrays (RPPA) [1–6].
149
150 Solomon Yeon et al.
Precision medicine requires rapid, precise, and reproducible

technology for integrating genetic, proteomic, and phenotypic
information into actionable treatment strategies [7–10]. Our labo-
ratory has developed and standardized Reverse Phase Protein
Microarrays (RPPAs) as a technology for studying posttranslation-
ally modified proteins and their unmodified protein forms in the
context of complex signal transduction networks [6, 11–17].
RPPAs provide functional epigenetic protein information that is
not discernable from gene expression arrays. Several laboratories in
the USA offer RPPA technology for precision medicine (Table 1).
Table 1
Examples of US laboratories performing reverse phase protein microarray technology
Services and Sample requirements/method of

Laboratory tests lysate prep Turnaround time
George Mason Translational Sample: Investigator/project
Universitya Research and 1. Cell/tissue lysates dependent
Center for Applied clinical trials 2. Macro-dissected tissue samples
Proteomics and 3. Laser capture microdissected
Molecular Medicine samples
4. Stem cells/cell lines
5. Body fluids (serum, urine,
vitreous)
Lysate Prep: Laser Capture
Microdissection
Baylor College of Research for BMC Sample: Information not
Medicineb Faculty only; 1. Cell/tissue lysates provided
The Cancer Prevention Two-tiered 2. Macro-dissected tumor
and Research Institute services samples
of Texas (CPRIT) 3. Laser capture microdissected
Cancer proteomics samples
and metabolomic 4. Stem cells
Core Facility 5. Body fluids (serum and urine)
MD Anderson RPPA Clinical Sample: 8–10 weeks
Core Facilityc diagnostics and 1. Cultured cells, mouse
research xenografts and frozen human
Functional tissue.
Proteomics 2. Not currently accepting FFPE
1. Human samples.
samples—panel 3. Client must prepare RPPA
of ≥200 samples for submission
antibodies according to prescribed
2. Mouse protocol
samples—panel Lysate Prep: Lysate sample
of ≥150 preparation from tissue by
antibodies electric homogenizer
Lysate sample preparation from
frozen tissue by Precellys
Homogenizer
(continued)
Dual-Color Protein Microarray Analysis 151
Table 1
(continued)
Services and Sample requirements/method of

Laboratory tests lysate prep Turnaround time
Theranostics Healthd Clinical Sample: Performed weekly
Diagnostics; 1. Tumor from core needle biopsy on Tuesday—final
Research and or open biopsy which have been report available
development fixed in 10 % neutral buffered electronically in
formalin within 20 min of 10 business days
biopsy
2. Paraffin blocks—cut 4–5 μm,
mounted on uncharged slides
Lysate Prep: Laser Capture
Microdissection
a
http://capmm.gmu.edu/
b
http://www.mdanderson.org/education-and-research/resources-for-professionals/scientific-resources/core-
facilities-and-services/functional-proteomics-rppa-core/index.html
c
https://www.bcm.edu/centers/cancer-center/research/shared-resources/cprit-cancer-proteomics-and-metabolo-
mics/reverse-phase-proteinarray
d
http://www.theranosticshealth.com/protemic-services/
RPPA technology can be applied to any cellular lysate, or

protein-containing body fluid, including bacterial lysates. The
lysate/protein specimen is immobilized (printed) on a nitrocellu-
lose substratum, typically using a robotic printer [12, 18]. Controls,
calibrators, and standards can be printed on the same array, provid-
ing a complete analytical test on one array. RPPAs enable compari-
son of the functional state of multiple or single specimens within
the same array. Each spot printed on an array is representative of
the proteome in that specimen. Robotic arraying devices and auto-
mated slide stainers facilitate reliable technical precision of the
RPPA [12, 18–20]. The RPPA detection methods depend on the
analyte concentration, the type of microarray imaging system avail-
able in the laboratory, and the type of sample being investigated.
Typically, identification of low abundance cell signaling proteins
and their posttranslationally modifications, such as phosphoryla-
tion, have required signal amplification and optimized detection
methods (fluorescent or colorimetric).
To enhance accessibility of RPPA technology, two hurdles must
be overcome: data normalization and data acquisition [2, 21–25].
The most labor intensive process in RPPA methodology is the spot
finding and quantification. Generating health-related information
in a clinically relevant time frame is the first hurdle to providing
precision medicine. Molecular information regarding the specific
functional defects in a tumor is worthless unless that information
can be delivered and utilized promptly. Herein we describe an
unamplified, dual-color signal detection methodology for reverse
phase protein microarrays that allows multiplex, within spot data

normalization, reduces data acquisition time, simplifies automated
spot detection, and provides a stable signal output. This method
utilizes Quantum Nanocrystal fluorophore labels (Qdot) substi-
tuted for organic fluorophores coupled with an imager (ArrayCAM)
that captures images of the microarray rather than sequentially
scanning the array.
Quantum nanocrystals (Qdots) are spherical compound semi-
conductor particles that exhibit luminescence from electronic
states related to the particle’s material composition and size [26,
27]. Due to the quantum confinement of electronic states in the
crystals, Qdots absorb light within a broad band from the ultravio-
let to visible range and emit fluorescent light within narrow bands
from the visible to near infrared. Each Qdot species is tuned to
emit light in a specific band for the purpose of fluorescence detec-
tion [28]. Qdot emission bands are narrow compared to emission
bands from organic fluorophores. When used as labels for immu-
noassay detection, Qdots have a number of advantages relative to
organic fluorophores, including greater brightness, resistance to
photobleaching and quenching, and long-term stability.
Additionally, all Qdots can be excited with a common laser wave-
length and this can contribute to simplified and lower-cost instru-
mentation for fluorescence detection.
Because of their narrow fluorescence emission bands, Qdots may
be applied to the specific problem of multiplexed labeling within the
same microarray [29] or the same well of a microtiter plate [30].
Individual interrogation molecules may be conjugated to particular
Qdot emitters and all emitters can be activated with a common exci-
tation source. Through the use of separate emission filters, each label
can be individually detected. This method is particularly effective
when a digital camera is used to image the microarray. In the current
work, we have established a methodology for multiplexing the detec-
tion of several analytes simultaneously and within the same spot in
the microarray, using Qdot labels and ArrayCAM, a high-resolution,
high-sensitivity imaging system (Fig. 1).
This chapter describes two new methods for processes relevant
to protein microarrays: (1) manual and automated non-amplified
immunostaining procedures using Qdot nanocrystal conjugated
antibodies, and (2) spot finding/quantification with an ArrayCAM
chromogenic/fluorescent array imager. Manual immunostaining
methods describe RPPA detection strategies without the use of an
automated slide stainer, whereas the automated immunostaining
method describes adaptation of immunostaining to an automated
slide stainer (Autostainer, Dako). This chapter also describes use of
the ArrayCAM microarray imager to sequentially identify multiple
fluorescent endpoints within the same microarray spot. The
method provides equivalent results to separate assays with a single
fluorescent/colorimetric endpoint (Fig. 2).
Fig. 1 Dual-color quantum dot (Qdot) detection scheme. Qdot nanocrystals with different emission wave-
lengths are conjugated to either anti-mouse immunoglobulin (IgG) (Qdot 655 nm) or to anti-rabbit IgG (Qdot
800 nm). Both Qdot conjugated antibodies are mixed in a single detection cocktail, permitting within spot
detection of two different analytes
Fig. 2 Equivalent resolution in the detection of relevant biologically pathways previously observed using colorimet-
ric detection can be achieved on the ArrayCAM platform. Strong concordance in spot signal intensities between
a β-Actin stained array imaged on a flatbed scanner (x-axis, Image Quant ver 5.1 software spot analysis) and
the ArrayCAM imager (y-axis, ArrayCAM software spot analysis). β-Actin was detected with a catalyzed signal
amplification method using colorimeteric detection [18]
2 Materials
Reverse phase protein microarray technology consists of four main

steps: sample preparation/cell lysis, array printing, array immunos-
taining, and spot finding and quantification. The focus of this
chapter is on immunostaining with dual-color, intra-spot signal
detection by manual and automated methods, and spot quantifica-
tion using an ArrayCAM multimodal imager. The staining reagents
are designed for simultaneous detection of two IgG antibody
probes of different species at two spectrally distinct wavelengths.
RPPA slides containing immobilized protein specimens are
simultaneously probed with a cocktail of mouse and rabbit primary
antibodies against the targets of interest. One protein target (anti-
body) is used for data normalization, while the second target (anti-
body) binds to an analyte of interest. The primary antibodies are
detected with species specific anti-IgG conjugated to either 800 nm
Qdot (rabbit) or 655 nm Qdot (mouse). After staining, the RPPA
slides are read using an ArrayCAM instrument with excitation by a
violet laser and imaging at the Qdot’s respective wavelength.
2.1 Manual Dual- 1. Reverse phase protein microarrays printed with whole cell
Color Qdot lysates, serum, vitreous, laser capture microdissected cell
Immunostaining lysates, peripheral blood mononuclear cells, or other protein
containing body fluids.
2. ProPlate slide chambers and spring clips (Grace Bio-Labs):
1-well format.
3. Orbital shaker.
4. Pipettes (1 μL–50 mL delivery volume).
5. Plastic trays/petri dishes with lids.
6. Type 1 reagent grade water (dH2O).
7. Antigen retrieval solution, 10 % sodium hydroxide in water,
10×. Store at 4 °C. Dilute to 1× prior to use: Prepare 100 mL
of 1× Antigen Retrieval by adding 10 mL of 10× Antigen
Retrieval to 90 mL dH2O. Mix thoroughly. Store at 4 °C.
8. Q-Block blocking buffer, proprietary composition (Grace Bio-
Labs). Store at 4 °C.
9. Primary antibodies against analytes of interest (anti-rabbit
and/or anti-mouse, unconjugated) (see Note 1).
10. Antibody diluent proprietary composition (Grace Bio-Labs).
Store at 4 °C.
11. 10× Wash Buffer Q, proprietary composition (Grace Bio-
Labs). Store at 4 °C. Prior to use prepare 400 mL of 1× Wash
Buffer Q by adding 40 mL of 10× Wash Buffer Q to 360 mL
of type 1 reagent grade water (dH2O). Mix thoroughly. Store

at 4 °C.
12. 10× Rinse buffer, proprietary composition (Grace Bio-Labs).
Store at 4 °C. Prior to use, prepare 400 mL of 1× Rinse Buffer
by adding 40 mL of 10× Rinse Buffer to 360 mL of dH2O. Mix
thoroughly. Store at 4 °C.

13. Qdot nanocrystals conjugated to secondary antibodies:
800 nm goat anti-rabbit and 655 nm goat anti-mouse. Store at
4 °C. Do not freeze.
14. Detection reagent diluent, proprietary composition (Grace
Bio-Labs). Store at 4 °C.
2.2 Automated 1. Reverse phase protein microarrays printed with whole cell
Dual-Color Qdot lysates, serum, vitreous, laser capture microdissected cell
Immunostaining lysates, peripheral blood mononuclear cells, or other protein
containing body fluids.
2. Automated slide stainer with open-source reagents (e.g., Dako
Autostainer).
3. Orbital shaker.
4. Plastic trays/petri dishes with lids.
5. Pipettes (1 μL–50 mL delivery volume).
6. Type 1 reagent grade water (dH2O).
7. Antigen retrieval solution 10×, 10 % sodium hydroxide in
water. Store at 4 °C. Dilute to 1× prior to use as described in
item 7 of Subheading 2.1.
8. Q-Block blocking buffer, proprietary composition (Grace Bio-
Labs). Store at 4 °C.
9. Primary antibodies (anti-rabbit and/or anti-mouse, unconju-
gated) (see Note 1).
10. Antibody diluent, proprietary composition (Grace Bio-Labs).
Store at 4 °C.
11. 10× Wash Buffer Q, proprietary composition (Grace Bio-
Labs). Store at 4 °C. Dilute to 1× prior to use as described in
item 11 of Subheading 2.1.
12. 10× Rinse buffer, proprietary composition (Grace Bio-Labs).
Store at 4 °C. Dilute to 1× prior to use as described in item 12
of Subheading 2.1.
13. Qdot nanocrystals conjugated to secondary antibodies: 800 nm
goat anti-rabbit and 655 nm goat anti-mouse. Store at 4 °C.
Do not freeze.
14. Detection reagent diluent, proprietary composition (Grace
Bio-Labs). Store at 4 °C.
2.3 ArrayCAM 1. ArrayCAM imager with ArrayCAM software (Grace Bio-Labs),

Scanning of Reverse computer (Windows 7 or 8).
Phase Protein 2. Printed and stained reverse phase protein microarray. RPPAs
Microarrays can be stained with Sypro Ruby Protein Blot stain, Fast Green,
Qdot nanocrystals (800 nm, 655 nm, and/or 585 nm) conju-
gated to rabbit or mouse immunoglobulin, or colorimetric
(e.g., diaminobenzidine (DAB)) detection molecules [18].
3 Methods
Personalized medicine requires analysis of single patient samples in

a timely manner. RPPAs can accommodate many samples, stan-
dards, and controls on a single array (single nitrocellulose pad) or
can be constructed with one or more samples on a series of smaller
nitrocellulose pads (sectors) on a single slide. Regardless of the
array format, analyte detection strategies must be specific, sensi-
tive, and unaltered by contaminating proteins [21]. To compare
analyte levels across samples on an array, a normalizer molecule,
usually total protein, single stranded DNA or a housekeeping pro-
tein, is measured on a separate array. However, a more precise nor-
malization procedure would be to measure the normalizer molecule
in the same spot, on the same array, in which the analyte is mea-
sured. Multiplex detection of more than one target per slide is not
possible using colorimetric catalyzed signal amplification methods
[31–34]. Non-amplified signal detection capitalizes on the ability
to combine two or more quantum nanocrystal (Qdot) fluorescent
molecules in a single detection solution. This dual-color detection
system permits two different analytes to be measured within the
same array spot (Fig. 1). Qdots have narrow emission spectra, per-
mitting Qdots with different wavelengths to be combined in a
single detection cocktail. Qdots provide high signal:noise ratios,
resist photobleaching and quenching, and provide a stable signal.
Herein we describe manual and automated slide stainer detection
methods for RPPAs using non-amplified signal detection with
intra-spot, multicolor analysis.
3.1 Manual Dual- 1. Retrieve RPPA slides from storage (−20 °C). Allow the slides
Color QDot RPPA to reach room temperature. The number of slides to stain
Immunostaining equals the number of primary antibodies plus 1 for a secondary
antibody only control [12, 18].
2. Wash RPPA slides 4 × 15 min in fresh dH2O. Place slides with
the nitrocellulose film facing up, in a plastic dish. Add adequate
amounts of water to completely immerse the slides. Place the
dish on an orbital rocker with gentle rocking for 15 min.
Remove and discard the water. Repeat the wash steps three
times. Do not allow the slides to dry out at any point during
the staining.
3. Attach a ProPlate chamber to the slide. Apply one ProPlate

chamber at a time. Fill the chamber with fresh dH2O.
4. Remove and discard the dH2O. Add 2 mL of 1× Antigen
Retrieval buffer. Incubate for 15 min at room temperature.
5. Wash 3 × 5 min each with 2 mL of fresh dH2O for each wash
with gentle rocking on an orbital rocker.
6. Discard the water after each wash.
7. Add 2 mL 1× Wash Buffer Q. Wash for 5 min with gentle rock-
ing on an orbital rocker.
8. Remove and discard the Wash Buffer Q. Add 2 mL Q-Block
Blocking Buffer. Incubate for 10 min at room temp with gen-
tle rocking on an orbital rocker.
9. While the slides are blocking, prepare the primary antibody
dilutions in the Antibody Diluent. Each array will be probed
with a cocktail of two antibodies: the anti-mouse antibody and
the anti-rabbit antibody. Prepare an antibody cocktail for each
primary antibody of interest (one cocktail per array). One slide
will not have any primary antibody added—this is the second-
ary antibody only control. This slide will have antibody diluent
only added in place of the primary antibody cocktail.
10. Remove and discard the Q-Block Blocking Buffer. Wash twice
for 2 min with 2 mL 1× Wash Buffer Q, with gentle rocking on
an orbital rocker.
11. Remove and discard the blocking buffer. Add the appropriate
primary antibody cocktail to each slide. Add antibody diluent
only to the secondary antibody only control slide. Incubate at
room temperature for 90 min with gentle rocking on an orbital
rocker.
12. Remove and discard the antibody cocktail/antibody diluent.
13. Wash 3 × 5 min with 2 mL 1× Wash Buffer Q with gentle rock-
14. While the slides are washing, prepare the Detection cocktail
(Qdot conjugated anti-IgG, mouse and rabbit). For each slide,
prepare 2100 μL of a 1:500 dilution of each Qdot conjugated
anti-IgG (800 and 655 nm) in 1× Detection Reagent Diluent.
Example: for one slide, add 4.2 μL of 500× Qdot 800 nm anti-
rabbit IgG and 4.2 μL of 500× Qdot 655 nm anti-mouse IgG to
2091.6 μL of 1× Detection Reagent Diluent. Mix thoroughly.
15. Remove and discard the wash buffer. Wash 2 × 2 min with
2 mL 1× Rinse Buffer.
16. Remove and discard the Rinse Buffer. Add 2100 μL of
Detection cocktail. Incubate at room temperature for 90 min
17. Wash 3 × 5 min with 2 mL 1× Rinse Buffer with gentle rocking
on an orbital rocker.
18. Remove and discard the Rinse Buffer. Remove the ProPlate
chamber from each array slide.
19. Rinse the slides in copious volumes of fresh dH2O.
20. Allow the slides to dry at room temperature, protected from
light.
3.2 Automated 1. Program the Dako Autostainer according to Table 2. Determine

Dual-Color Qdot RPPA the volume of Wash Buffer Q and water required for the stain-
Immunostaining ing run. Typically five slides require approximately 2000 mL of
(Fig. 3) Wash Buffer Q in the buffer carboy.
2. Add the diluted Wash Buffer Q to the Autostainer buffer
carboy.
3. Retrieve RPPA slides from storage (−20 °C). Allow the slides
to reach room temperature. The number of slides to stain
equals the number of primary antibodies plus 1 for a secondary
antibody only control [12, 18].
4. Wash RPPA slides 2 × 5 min in fresh dH2O. Place slides with
the nitrocellulose film facing up, in a plastic dish. Add adequate
amounts of water to completely immerse the slides. Place the
dish on an orbital rocker with gentle rocking for 15 min.
Remove and discard the water. Repeat the wash steps three
times. Do not allow the slides to dry out at any point during
the staining.
Fig. 3 Workflow for automated dual-color, non-amplified staining of reverse

phase protein microarrays. The slides are rinsed, washed, and blocked prior to
loading them on the Autostainer. On the Autostainer, the slides are rinsed, incu-
bated with primary antibody cocktails, rinsed, incubated with secondary anti-
body detection reagents, with final rinses in buffer and water. The total staining
time on the Autostainer is 285 min
Table 2
Autostainer program for multiplex, two-color Qdot RPPA staining
Reagent category Reagent Time (min)

Rinse Wash buffer Q
Auxiliary Wash buffer Q 2
Rinse Wash buffer Q
Rinse Wash buffer Q
Primary Antibody Primary Antibodies (species: rabbit and 90
mouse)
Rinse Wash buffer Q
Rinse Wash buffer Q
Rinse Wash buffer Q
Rinse Wash buffer Q
Tertiary reagent Rinse buffer 2
Rinse Wash buffer Q
Rinse Wash buffer Q
Substrate Qdot 655 nm + Qdot 800 nm 90
Rinse Wash buffer Q
Rinse Wash buffer Q
Rinse dH2O
Rinse dH2O
Rinse dH2O
Rinse dH2O
Auxiliary dH2O 3
Rinse dH2O
5. Remove and discard the dH2O. Add 2 mL of 1× Antigen

Retrieval buffer. Incubate for 15 min at room temperature.
6. Wash 3 × 5 min each with 2 mL of fresh dH2O for each wash
7. Discard the water after each wash.
8. Add 2 mL 1× Wash Buffer Q. Wash for 5 min with gentle rock-
9. Remove and discard the Wash Buffer Q. Add 2 mL QBlock
Blocking Buffer. Incubate for 30 min at room temp with gen-
tle rocking on an orbital rocker.
10. While the slides are blocking, prepare the primary antibody
dilutions in the Antibody Diluent. Each array will be probed
with a cocktail of two antibodies: the anti-mouse antibody and
the anti-rabbit antibody. Prepare an antibody cocktail for each
primary antibody of interest (one cocktail per array). One slide
will not have any primary antibody added—this is the second-
ary antibody only control. This slide will have antibody diluent
added in place of the primary antibody cocktail.
11. Prepare the Detection cocktail (Qdot conjugated anti-IgG,
mouse and rabbit). For each slide, prepare 800 μL of a 1:500
dilution of each Qdot conjugated anti-IgG (800 and 655 nm)
in 1× Detection Reagent Diluent. Example: for one slide, add
1.6 μL of 500× Qdot 800 nm anti-rabbit IgG and 1.6 μL of
500× Qdot 655 nm anti-mouse IgG to 796.8 μL of 1×
Detection Reagent Diluent. Mix thoroughly.
12. Remove the slides from the QBlock Blocking Buffer. Place the
slides in the Autostainer slide rack and start the Autostainer
program (Table 2). Do not allow the slides to dry out. 1×
Wash Buffer Q can be poured on to the slides to keep them
wet while they are on the Autostainer.
13. Program the Autostainer to dispense 800 μL of reagent per
slide. Prime the water and buffer on the Autostainer. Start the
Autostainer.
14. Remove the Autostainer rack with the stained arrays.
15. Allow the slides to dry at room temperature, protected from
light.
3.3 ArrayCAM Image acquisition for fluorescent or colorimetric detection systems

Imaging of Reverse can be conducted on the ArrayCAM imager. All Qdots can be
Phase Protein excited with a common laser wavelength which contributes to sim-
Microarrays (Fig. 4) plified and lower-cost instrumentation for fluorescence detection.
The ArrayCAM uses a violet semiconductor diode laser as the excita-
tion source. Streamlining and standardizing the data analysis steps
can be achieved using an ArrayCAM high-resolution, dual mode
chromogenic/fluorescent array imager. The ArrayCAM design facil-
itates adoption of RPPA technology in any size laboratory.
Fig. 4 ArrayCAM spot finding features. The ArrayCAM spot finding software depicts spots that meet the spot
search parameters as green circles and those spots that fail the search parameters are depicted as red circles
with lines through them. The 3-dimensional features of the spot can be readily visualized with a 3-D spot
inspection tool
3.3.1 ArrayCAM Setup 1. Make sure ArrayCAM is connected to the computer.

2. Turn on ArrayCAM, blue power light will indicate that the
machine is operational.
3. Click on ArrayCAM shortcut icon to launch program.
4. After program has opened a green “Imager Ready…” status
bar should be visible at the top right corner of the program to
confirm that the computer is connected to the ArrayCAM.
3.3.2 Image Acquisition 1. Load a slide by pushing the ArrayCAM lid backwards to expose
the slide stage.
2. Orient the slide face down with the label on the left (see Note 2).
3. Secure the slide on the stage by pushing the two clasps backwards
and close the lid.
4. Click on “File Info” tab (Fig. 1).
5. Under “Select Image Type”, select TIFF.

6. Enter a file name prefix if desired.
7. Under “Image Storage Directory Path”, select the file destination
where the scanned images will be sent, by navigating to the pre-
ferred directory and pressing “Select folder.”
8. Under “Misc File Items”, check the “Save Images?” box and
the “Use universal time?” box.
9. Click on the “Configure Imager” tab. Under image profiles
select the “full slide” tab and Select All. The slide icon will be
highlighted in green (Fig. 1) (see Note 3).
10. Click on the “Image Control” tab.
11. Under “Label” select the correct setting for the slide being
analyzed (specific wavelength, Sypro Ruby, or colorimeteric).
12. Set exposure to 200 ms, acquisition time to 1 s, and gain at 50
for the first array scan (see Note 4).
13. Set palette to gray scale, Reticule off and invert color off
(Fig. 1) (see Note 5)
14. Click on the Capture Image button to acquire an image of the
slide (see Note 6).
15. After the image is obtained, the Brightness, Contrast, and
Gamma controls can be used to enhance or reduce the appear-
ance of spots within the image (see Note 7).
3.3.3 Image Analysis 1. Click the “File Info” tab, and load a previously created naviga-
tional .gal file by clicking on the “Load .gal File” button.
2. Select the “1 × 1 Sub-array” and click ok, then choose the
appropriate .gal file.
3. Click on the “Image Control” tab. Click on the “Invert Color”
button if the image represents a colorimetric assay.
4. Under the “Auto tab”, set values as (see Note 8):
• Intensity Measure: Volume • Annular Ring Thk %: 75

• Bkg Method: Annular Ring • Intensity Correction: Sig-Bkg
• Bkg Diameter: 2.0 • Avg Bkg Reference: Fiducial (optional)
5. Click the “Params” button and set the values as (see Note 9):
• Spot Search Diameter: 1.5 • Spot Diameter Flag (Upper

• Spot Evaluation Diameter: 1.25 Limit): 2.0
• Spot Detection Sensitivity: 25 % • Spot Circularity (%): 75
• Spot Diameter Flag (Lower • All “Spot Types to Analyze”
Limit): 0.5 options should be highlighted
• Max Spot Offset: 1.5 green
• Normalize values: Yes • Printer Vertical Drift: 1.0 (see
• Search mode: One Time Note 10)
6. Click the “Identify” button, then the “Manual” button. Select

“Continue” to create a gal file. A navigational gal file may be
created in the ArrayCAM software, or can be copied and pasted
from another source (see Note 11). The ArrayCAM software
creates the navigational gal file in the correct format when the
array parameters are entered. Enter information for the array:
# of columns, # of rows, column pitch, row pitch, spot diam-
eter, left origin, and top origin. The arrays are analyzed in a
vertical position, with the barcode label at the bottom of the
image (Fig. 2).
7. The key stroke sequences for pasting columns of text into the
sample ID and spot type columns in the gal file template are:
Type in your value, click enter, highlight the text in the first cell
that you wish to copy, hold the shift key, scroll down, click in the
last box (while still holding the shift key), then click Fill Down.
8. Click Set, click Done, then save the gal file.
9. On the File Info tab, load the gal file.
10. Click on the Image Control tab. Ensure that the correct image
is loaded by verifying the image file name in the image box.
Click Verify.
11. Use the rectangle box tool to draw a rectangle that fits closely
around the first row of spots (e.g. draw the rectangle so it is
just touching them at their perimeters) and click “ok” to ana-
lyze the spots. Use the magnifying glass tool to zoom in and
out when drawing the box, left mouse click will zoom in, left
mouse click + shift will zoom out.
12. A new window will appear with the spot boundaries marked as
circles (see Note 12).
13. Use the ellipse tool button in the newly created window to
readjust any spot boundaries whose locations have been identi-
fied incorrectly. Draw the new circle where the spot location is
intended and click the “Set” button. A new, adjusted circle will
appear and the old circle will disappear. If many circles require
adjustment, the spot evaluation parameters should be adjusted
in “Params” window (see Note 13).
14. After readjustments have been completed, click the “recalcula-
tion” button to reanalyze the slide (see Note 14).
15. Data from analysis will be transferred to previously selected
destination folder (see Note 15).
4 Notes
1. The antibodies directed against the analytes of interest should

be from a different species than the antibody used to detect the
protein to be used as the normalizer. For example, if β-actin is

selected as the endpoint to be used for normalization between
spots, select an anti-mouse β-actin antibody, which will be
detected with Qdot 655 nm conjugated to anti-mouse IgG. All
other antibodies for detecting individual analytes should be
anti-rabbit. Primary antibodies must be validated by western
blotting to verify specificity to the protein of interest. A domi-
nant band on the western blot, at the specified molecular
weight, provides a means of determining antibody specificity.
The primary antibody concentration (dilution) with the
non-amplified staining method is approximately fourfold
greater than with catalyzed signal amplification (CSA) meth-
ods. For example, if the recommended primary antibody dilu-
tion for the CSA method is 1:100, use the antibody at 1:400
with the non-amplified method.
2. Viewing the microarray slide as a roadmap, the narrow edge of
the slide is west-to-east (left-to-right) and the long edge is
north-to-south (top-to-bottom); the label or barcode is on the
“south” side. When placing the slide in the ArrayCAM car-
riage, the film slide is always oriented so that the nitrocellulose
surface is downward and the label is on the left.
3. Slide configuration settings are compatible with single pad or
multi-pad nitrocellulose film slides. Single-pad film slides are
imaged as five separate images and the resulting images are
stitched to form a montage.
4. The optimal exposure, acquisition, and gain settings must be
determined empirically for each slide. Typical settings for DAB
(colorimetric) stained arrays are 20 ms exposure, 1 s acquisition,
20 gain.
5. Images can be viewed with various palette settings. Monochrome
and grayscale will provide lesser contrast. Rainbow and
Spectrum will provide greater variation in color contrast for
highlighting detail. Adjustments to these controls do not affect
the actual image data, only the displayed image.
6. The acquisition progress will be indicated with a progress bar.
Acquisition time is less than 1 min for a full slide and approxi-
mately 10 s for a single region of a slide. When the acquisition is
finished, an image will be displayed. If several sections of the slide
are being imaged, the last image acquired will be displayed.
7. Contrast between spots and background can be enhanced by
adjusting the Image Enhance and Image Diminish controls.
Adjustments to these controls do not affect the actual image
data, only the displayed image.
8. Intensity Measure is normally set to Volume or Median.
Volume provides slightly better sensitivity while Median is
more tolerant to outlier pixels. The ArrayCAM software allows
the user to change the type of background measurement,

methods for determining the local area background, and signal
intensity corrections for the local area background. Annular
Ring background is highly tolerant to spot bleed into the back-
ground area.
9. Parameter settings allow the user to define the stringency of
the spot finding algorithms. Of these, the Detection Sensitivity
has the most effect, as it determines the discrimination of sig-
nal vs. background in defining the spot boundary. Spot Search
Diameter, Spot Evaluation Diameter, and Maximum Spot
Offset all limit the search algorithm to the immediate region of
the expected spot location. These settings reduce the probabil-
ity that spot centers will need to be adjusted manually.
Definitions of spot search parameters:
Spot Search Diameter: This sets the maximum distance
about the expected spot center to search for the actual spot
“centroid”. The centroid location enables the spot location
algorithm to narrow the search to an area close to the centroid.
The units are in spot diameters. For example, a setting of 1.5
will permit searching within 1.5 diameters of the expected spot
location.
Spot Evaluation Diameter: This sets the maximum dis-
tance about the located spot centroid to locate the actual spot
using the spot location algorithm. The units are in spot diam-
eters. For example, a setting of 1.5 will permit searching within
1.5 diameters of the spot centroid.
Spot Detection Sensitivity: This tells the algorithm how
sensitive to be in finding the spot. A high-sensitivity will locate
spots of very low intensity but will sometimes react to “noise”.
Spot Diameter Flag (Lower Limit): Flags spots in the
results text file when the evaluated diameter is less than the flag
limit. Units are in diameters.
Spot Diameter Flag (Upper Limit): Flags spots in the
results text file when the evaluated diameter is greater than the
flag limit. Units are in diameters.
Spot Circularity: Flags spots in the results text file when
the circularity (aspect ratio) exceeds the percentage set.
Max Spot Offset: Limits the location algorithm to finding
spot within the limits set. For example, a limit of 1.5 will limit
the offset relative to the expected location to 1.5 diameters.
If a spot is not found within the setting limit, a spot circle is
drawn at the expected location and is flagged with a slash.
Search Mode: Performs the spot location once or twice,
depending on whether the switch is set to “One Time” or
“Iterative.”
10. During printing, any variation in the position of the slide,
printer pins, or sample plates can cause the deposited sample
spots to vary slightly in their horizontal/vertical position.
Printer Vertical Drift compensates for this drift. If the

ArrayCAM circles are lower than the actual printed spots, set
the Printer Vertical Drift to <1 (i.e., 0.993). If the ArrayCAM
circles are higher than the actual printed spots, set the Printer
Vertical Drift to >1 (i.e., 1.005).
11. A microarray block configuration is normally printed accord-
ing to a .gal file that controls the printing process. ArrayCAM™
uses a “navigational gal file” for auto analysis that is similar to
a normal gal file and obeys the rules for creating gal files. The
gal file created for printing the arrays can simply be modified
and saved as the navigational gal for the purpose of the auto-
mated image analysis. However, the format of the navigational
gal file must follow a specific format as described below.
The following example demonstrates the configuration of
the navigational .gal file. In this example, only the header and
the first several rows are included in the list. The navigational
gal file must conform to the general rules for gal file construc-
tion. In addition to the basic rules, a few modifications are
required as indicated below.
●● Add a new line in the header that indicates the slide type
and include the corresponding numeric code.
Slide type Numeric

Four-pad 0
16-pad 1
Full slide pad 2
●● Add column #6 to indicate the type of spot.
The number of header rows can be unlimited. However,
the numerical value in the first column of the second row must
be equal to the total number of header rows minus three. In
the example below, there are eleven (11) header rows and the
numerical value is eight (8):
11 header rows − 3 required rows = 8 ← enter this number
Additionally, the numerical value in the second column of
the second row must be “6” as there are exactly six columns.
The block designator row must be enclosed in double quotes.
For example:
“Block1 = 4400,1400,200,13,433,13,433”
More than one block designator row can be provided as
long as the entire list is consistent with the individual block
designators.
The “Type” designator in column 6 can have any of the
following values:
fiducial
analyte
positive
negative
buffer
blank
control
content
blocker
Each type designator can be of all lower case letters and the
first letter can be capitalized. No other case options are permit-
ted. Each must be spelled exactly as shown.
The user may construct any navigational .gal file as long as
it obeys the general rules for .gal file construction and includes
the additional information as described above.
Save the navigational .gal file to the directory C:\GBL\Gal\
Controls
ATF 1
8 6
Type = GenePix ArrayList V1.0
Supplier = Grace Bio-Labs
ArrayerSoftwareName = None
ArrayerSoftwareVersion = None
BlockCount = 1
BlockType = 0
SlideType = 1
“Block1 = 4400,1400,200,13,433,13,433”
Block Column Row ID Name Type
1 1 1 ID FIDUCIAL fiducial
1 2 1 ID BLANK blank
1 1 2 ID Content analyte
(continued)
12. Green circles indicate satisfactorily located spot boundaries.

Orange circles with horizontal slash indicate spots are either
not found or have been found with peculiar shapes. Regardless
of the spot fidelity, all spots will be listed in the results text file.
Those with peculiar shapes or diameters out of limit will be
flagged accordingly in the text file.
13. Spot boundaries that have been manually adjusted will be dis-
played in red.
14. If multiple circles require adjustment, the adjusted circle layout
may be saved as a “map” and loaded onto subsequent images,
obviating the need to manually adjust the circles for each
image. Adjust the circles as desired, then click “save map”.
Name the map file for future use. To use the saved map, open
the image and click “verify” to start the analysis. After the anal-
ysis is complete, click “load map”, choose the saved map file.
The image reanalysis proceeds automatically. Click OK to save
the data analysis.
15. The data file is linked to the image name. The existing data file
will be over-written with the most current data when the image
is reanalyzed/recalculated. If you wish to analyze an image with
different spot analysis parameters and to save both the original
and modified data analysis, make a copy of the image and save
it with a different file name prior to repeating the analysis.
Acknowledgments
This work was funded in part by George Mason University and

Grace Bio-Labs, Inc.
References
1. Holmes FA, Espina V, Liotta LA, Nagarwala study utilizing multi-omic molecular profiling
YM, Danso M, McIntyre KJ (2013) Pathologic to find potential targets and select individual-
complete response after preoperative anti- ized treatments for patients with previously
HER2 therapy correlates with alterations in treated metastatic breast cancer. Breast Cancer
PTEN, FOXO, phosphorylated Stat5, and Res Treat 147:579–588
autophagy protein signaling. BMC Res Notes 14. Pierobon M, Silvestri A, Spira A, Reeder A, Pin
6:507 E, Banks S (2014) Pilot phase I/II personal-
2. Korf U, Derdak S, Tresch A, Henjes F, ized therapy trial for metastatic colorectal can-
Schumacher S, Schmidt C (2008) Quantitative cer: evaluating the feasibility of protein pathway
protein microarrays for time-resolved measure- activation mapping for stratifying patients to
ments of protein phosphorylation. Proteomics therapy with imatinib and panitumumab.
8:4603–4612 J Proteome Res 13:2846–2855
3. Liotta LA, Espina V, Mehta AI, Calvert V, 15. Robertson FM, Petricoin EF III, Van Laere SJ,
Rosenblatt K, Geho D (2003) Protein micro- Bertucci F, Chu K, Fernandez SV (2013)
arrays: meeting analytical challenges for clinical Presence of anaplastic lymphoma kinase in
applications. Cancer Cell 3:317–325 inflammatory breast cancer. Springerplus 2:497
4. MacBeath G, Schreiber SL (2000) Printing pro- 16. Wulfkuhle JD, Speer R, Pierobon M, Laird J,
teins as microarrays for high-throughput function Espina V, Deng J (2008) Multiplexed cell sig-
determination. Science 289:1760–1763 naling analysis of human breast cancer applica-
5. Paweletz CP, Charboneau L, Bichsel VE, tions for personalized therapy. J Proteome Res
Simone NL, Chen T, Gillespie JW (2001) 7:1508–1517
Reverse phase protein microarrays which cap- 17. Xia W, Petricoin EF III, Zhao S, Liu L, Osada
ture disease progression show activation of T, Cheng Q (2013) An heregulin-EGFR-
pro-survival pathways at the cancer invasion HER3 autocrine signaling axis can mediate
front. Oncogene 20:1981–1989 acquired lapatinib resistance in HER2+ breast
6. Petricoin EF III, Espina V, Araujo RP, Midura cancer models. Breast Cancer Res 15:R85
B, Yeung C, Wan X (2007) Phosphoprotein 18. Gallagher RI, Silvestri A, Petricoin EF III,
pathway mapping: Akt/mammalian target of Liotta LA, Espina V (2011) Reverse phase
rapamycin activation is negatively associated protein microarrays: fluorometric and colori-
with childhood rhabdomyosarcoma survival. metric detection. Methods Mol Biol
Cancer Res 67:3431–3440 723:275–301
7. Gallagher RI, Espina V (2014) Reverse phase 19. Tibes R, Qiu Y, Lu Y, Hennessy B, Andreeff
protein arrays: mapping the path towards per- M, Mills GB (2006) Reverse phase protein
sonalized medicine. Mol Diagn Ther array: validation of a novel proteomic technol-
18:619–630 ogy and utility for analysis of primary leukemia
8. Huels C, Muellner S, Meyer HE, Cahill DJ specimens and hematopoietic stem cells. Mol
(2002) The impact of protein biochips and Cancer Ther 5:2512–2521
microarrays on the drug development process. 20. VanMeter AJ, Rodriguez AS, Bowman ED, Jen
Drug Discov Today 7:S119–S124 J, Harris CC, Deng J (2008) Laser capture
9. Jameson JL, Longo DL (2015) Precision med- microdissection and protein microarray analysis
icine—personalized, problematic, and promis- of human non-small cell lung cancer: differen-
ing. N Engl J Med 372:2229–2234 tial epidermal growth factor receptor (EGPR)
10. Mueller C, Liotta LA, Espina V (2010) Reverse phosphorylation events associated with
phase protein microarrays advance to use in mutated EGFR compared with wild type. Mol
clinical trials. Mol Oncol 4:461–481 Cell Proteomics 7:1902–1924
11. Cremona M, Espina V, Caccia D, Veneroni S, 21. Chiechi A, Mueller C, Boehm KM, Romano A,
Colecchia M, Pierobon M (2014) Stratification Benassi MS, Picci P (2012) Improved data nor-
of clear cell renal cell carcinoma by signaling malization methods for reverse phase protein
pathway analysis. Expert Rev Proteomics microarray analysis of complex biological
11:237–249 samples. Biotechniques 1–7
12. Espina V, Liotta LA, Petricoin EF III (2009) 22. Mannsperger HA, Gade S, Henjes F, Beissbarth
Reverse-phase protein microarrays for ther- T, Korf U (2010) RPPanalyzer: analysis of
anostics and patient tailored therapy. Methods reverse-phase protein array data. Bioinformatics
Mol Biol 520:89–105 26:2202–2203
13. Jameson GS, Petricoin EF, Sachdev J, Liotta 23. Stanislaus R, Carey M, Deus HF, Coombes K,
LA, Loesch DM, Anthony SP (2014) A pilot Hennessy BT, Mills GB (2008) RPPAML/
RIMS: a metadata format and an information dots as reporters in multiplexed immunoassays

management system for reverse phase protein for biomarkers of exposure to agrochemicals.
arrays. BMC Bioinformatics 9:555 Anal Lett 40:1423–1433
24. Troncale S, Barbet A, Coulibaly L, Henry E, 30. Goldman ER, Clapp AR, Anderson GP, Uyeda
He B, Barillot E (2012) NormaCurve: a HT, Mauro JM, Medintz IL (2004) Multiplexed
SuperCurve-based method that simultaneously toxin analysis using four colors of quantum dot
quantifies and normalizes reverse phase protein fluororeagents. Anal Chem 76:684–688
array data. PLoS One 7:e38686 31. Bobrow MN, Harris TD, Shaughnessy KJ, Litt
25. von der Heyde S, Sonntag J, Kaschek D, Bender GJ (1989) Catalyzed reporter deposition, a
C, Bues J, Wachter A (2014) RPPanalyzer tool- novel method of signal amplification.
box: an improved R package for analysis of Application to immunoassays. J Immunol
reverse phase protein array data. Biotechniques Methods 125:279–285
57:125–135 32. Bobrow MN, Litt GJ, Shaughnessy KJ, Mayer
26. Michalet X, Pinaud FF, Bentolila LA, Tsay JM, PC, Conlon J (1992) The use of catalyzed
Doose S, Li JJ (2005) Quantum dots for live reporter deposition as a means of signal ampli-
cells, in vivo imaging, and diagnostics. Science fication in a variety of formats. J Immunol
307:538–544 Methods 150:145–149
27. Resch-Genger U, Grabolle M, Cavaliere- 33. Bobrow MN, Shaughnessy KJ, Litt GJ (1991)
Jaricot S, Nitschke R, Nann T (2008) Quantum Catalyzed reporter deposition, a novel method
dots versus organic dyes as fluorescent labels. of signal amplification. II. Application to mem-
Nat Methods 5:763–775 brane immunoassays. J Immunol Methods
28. Shao L, Gao Y, Yan F (2011) Semiconductor 137:103–112
quantum dots for biomedicial applications. 34. King G, Payne S, Walker F, Murray GI (1997)
Sensors (Basel) 11:11736–11751 A highly sensitive detection method for immu-
29. Nichkova M, Dosev D, Davies AE, Gee SJ, nohistochemistry using biotinylated tyramine.
Kennedy IM, Hammock BD (2007) Quantum J Pathol 183:237–241
Chapter 13
Quantitative Proteomics Using SILAC

Kian Kani
Abstract
The ability to enumerate all of the proteins in a cell is quickly becoming a reality. Quantitative proteomics
adds an extra dimension to proteome-wide discovery experiments by enabling differential measurements
of protein concentrations, characterization of protein turnover, increased stringency of co-
immunoprecipitation reactions, as well as many other intriguing applications. One of the most widely used
techniques that enable relative protein quantitation is stable isotope labeling by amino acids in cell culture
(SILAC) (Ong et al., Mol Cell Proteomics 1(5):376–386, 2002). Over the past decade, SILAC has
become the preferred approach for proteome-wide quantitation by mass spectrometry. This approach
relies on the metabolic incorporation of isotopically enriched amino acids into the proteome of cells—the
proteome of “light” (1H, 12C, 14N) cells can then be compared to “heavy” (2H, 13C, 15N) cells as the iso-
topically labeled proteins and peptides are easily distinguished in a mass spectrometer. Since cellular uptake
and response to isotopically different amino acid(s) is naïve, it is without impact on cell physiology. We
provide a detailed step-by-step procedure for performing SILAC-based experiment for proteome-wide
quantitation in this chapter.
Key words SILAC, AMT, Quantitative, Proteomics, Multiplex
1 Introduction
Recent advances in mass spectrometers and pre-fractionation tech-

niques have enabled near complete coverage of various proteomes
[2]. Researchers are confronted with the challenge of determining
the biological significance of long lists of proteins. Quantitative
proteomics enables researchers to extract more impactful data from
their data-sets. A number of different techniques are available to
researchers to implement quantitative proteomic experiments. The
determining factor for choosing the right technique is based on the
type of sample, the number of controls and cases, and the type of
mass spectrometer (Fig. 1). The most widely used technique for
quantitative proteomics with cell line models is the SILAC
approach. The reliability and reproducibility of this technique is
superior to other methods of quantitation [3, 4].
171
172 Kian Kani
Quantitation
Immortalized Cell lines Biological tissue or

(rapidly dividing) plasma
>5
Number of samples Isobaric mass tags
Metabolic labeling >3

(2H, 15N, 13C) Number of samples iTRAQ/TMT
<2
ICAT/Acrylamide/di-
methyl
Fig. 1 Quantitative proteomics flowchart for either metabolic labeling or chemical addition of isobaric mass tags
SILAC is ideal for experiments that compare the differential

expression of proteins in up to five samples [5], however, the major-
ity of SILAC based experiments utilize no more than three samples
[6]. The main principle for metabolic labeling involves the incorpo-
ration of amino acids with unusual elemental isotopes (e.g., one or
more 13C replacing the more natural 12C) into proteins in cells that
cannot endogenous produce that specific amino acid. The reliance
of exogenous amino acids into biochemical pathways is termed aux-
otrophy. For this reason, modern SILAC experiments for human
cell lines rely on the incorporation of arginine and lysine. Most two-
plex SILAC experiments utilize 13C versus 12C labeled lysine for a
mass shift of 6 Da. In general, we recommend using combinations
of “light” and “heavy” peptides with an expected mass shift greater
than 4 Da. Metabolically labeling with arginine offers greater flexi-
bility in the mass due to the number of “heavy” carbons, deuterium
and/or nitrogen atoms. For experiments that require the compari-
son of three conditions, the most widely used combination of amino
acids is 12C lysine, 13C lysine, and 13C/15N arginine which allows for
a mass shift of 6 and 10 Da from the control.
One of the complications of this technique is the unintended
interconversion of isotopic amino acids during the labeling process
into other byproducts. This conversion alters MS/MS signals from
Quantitative Proteomics Using SILAC 173
one amino acid and introduces various artifacts that affect quanti-
fication by increasing the number of unwanted peptide ion peaks.
Of particular concern is the biosynthesis of proline from excess
arginine [7]. This impacts all SILAC experiments that utilize argi-
nine. The best workaround for this complication is to supplement
SILAC media with standard l-proline, thereby suppressing the
endogenous conversion enzymes [8]. We suggest that all SILAC
media contain 100 mg/L of l-proline in order to prevent the con-
version of isotope-coded arginine to proline.
The application of the SILAC approach is based on the research
aims of the investigators. Some of the early works focused on dif-
ferential expression of proteins in various model systems (yeast, E.
coli, C. elegans, Arabidopsis, etc.) [9–11]. The combination of
SILAC based quantitation with enrichment strategies that enable
measurement of protein posttranslational modifications has been a
powerful application of this technique. For example, Bose et al.
discovered novel downstream phosphorylation targets of the onco-
gene HER2 [12]. A number of other reports have applied SILAC
based approaches to measure changes in glycosylation [13], ubiq-
uitinylation [14], methylation [15], and histone modification [16].
Global changes in various host systems may also be monitored by
the SILAC approach. For example, Kruger et al. metabolically
labeled a mouse with heavy lysine and demonstrated the applica-
tion of quantitative proteomics with knockout mice [17]. A num-
ber of other experimental strategies involving SILAC have been
developed. For instance, Schwanhausser et al. utilized pulsed
incorporation of amino acids to measure kinetics of protein turn-
over (pulsed-SILAC) [18]. SILAC may also be used to enhance
the reliability of absolute protein quantitation [19] and facilitate
improvement in reducing the false positive identification of con-
taminant proteins in CO-IP LC-MS/MS experiments [20, 21]. In
the following section, we describe a step-by-step procedure for a
typical two-plex based SILAC experiment (Fig. 2).
2 Materials
All cell culture material should be sterile and cell lines should be
tested for mycoplasma contamination prior to amino acid incorpo-
ration with a suitable kit (Mycoprobe Mycoplasma detection kit,
R&D Systems, Minneapolis, MN). All buffers and solvents should
be prepared with Milli-Q water (Millipore, Billerica, MA).
1. Appropriate media for your cell line deficient in arginine,
lysine, and glutamine (see Note 1). For example, RPMI 1640
Medium deficient in arginine, lysine, and glutamine (Life
Technologies, Carlsbad, CA).
A. Ctrl Treatment
13 12
Heavy ( C Lysine) Light( C Lysine)
6 6
Cell lysis Cell lysis
Mix lysates
(equal mass)
Reduction/Alkylation
Digest (trypsin/lys-c)
Analysis
MS
MS/MS
6 Da
Intensity
M/Z
Fig. 2 Schematic of a typical two-plex SILAC experiment. Metabolic labeling of cells occurs in rapidly dividing
cultures. Cell are passaged six times in SILAC media. Cells are lysed in an appropriate buffer and subject to
protein concentration determination by Bradford or BCA. Equal masses of proteins are mixed and subject to
reduction and alkylation. Lysates are treated with trypsin or lys-c and analyzed by LC-MS/MS. The area of the
“light” and “heavy” peptides are used for quantitation
2. Dialyzed Fetal Bovine Serum (FBS) (Life Technologies,

Carlsbad, CA) (see Note 2).
3. Select “Heavy” and “Light” amino acids based on the level of
multiplexing (Table 1). For example, for a two-plex experiment
obtain “light” lysine and arginine as well as the isotopically
“heavy” versions.
Table 1
Recommended amino acids for SILAC experiments
Sample 1 Sample 2 Sample 3 Sample 4 Sample 5

2 plex 12
C/14N light C6 Lysine and
13
amino acids 13
C6, 15N4
arginine
3 plex 1
H/12C/14N light D4 lysine and C6 15N2 Lysine and
13
amino acids 13
C6 arginine 13
C6, 15N4 arginine
4 plex 1
H/12C/14N light D4 lysine and 13
C6 15N2 Lysine and 13
C6 15N2 D9
amino acids 13
C6 arginine 13
C6, 15N4 arginine Lysine and D7
arginine
5 plex 12
C6, 14N4 12
C6, 15N4 13
C6, 14N4 arginine 13
C6, 15N4 arginine D7 13C6, 15N4
arginine arginine arginine
(a) “Heavy” and “Light” lysine (Cambridge Isotope

Laboratories, Tewksbury, MA), if needed, should be dis-
solved to a final concentration of 146 mg/mL in media to
create a 1000× stock (see Note 3).
(b) “Heavy” and “Light” arginine (Cambridge Isotope
Laboratories, Tewksbury, MA), if needed, should be dis-
create a 1000× stock (see Note 3).
4. Unlabeled (i.e., “normal” or “light”) l-proline should be dis-
make a 1000× stock.
5. Unlabeled l-glutamine should be dissolved in media to a final
concentration of 200 mM to yield a 100× stock solution.
6.
Penicillin/streptomycin (10,000 U/10,000 μg stock
solution).
7. 1 M HEPES buffer.
8.
Trypsin–EDTA solution (trypsin, 500 mg/L) (Life
Technologies, Carlsbad, CA).
9. Phosphate buffered saline (PBS).
10. 100 mm (10 cm) tissue culture plates.
11. Sterile Vacuum Filter Units, 1 L (EMD Millipore Stericup,
0.22 mm, Fisher Scientific, St. Louis, MO).
12. Microcentrifuge tubes, 1.5 mL.
13. Spin filter, 0.22 μm (e.g., 0.22 μm Corning Costar Spin-X cen-
trifuge tube filters , Sigma-Aldrich, Pittsburg, PA).
14. BCA protein analysis kit.
176 Kian Kani
15. Dithiothreitol (DTT) 1 M stock made fresh in water (100×).

16. Iodoacetamide (IAA) 1 M stock made fresh in water (18.18×)
(Note: stock solutions of iodoacetamide at 1 M may require soni-
cation in order to obtain complete solubility).
17. Sequencing grade Trypsin and Lys-C (optional).
18. Glacial acetic acid.
19. Trifluoroacetic acid (TFA).
20. Formic acid.
21. Urea lysis buffer: 6 M urea, 2 M thiourea in 100 mM ammo-
nium bicarbonate (pH 8) (Note: must be made fresh).
3 Methods
One of the major concerns for accurate proteome quantitation is

to ensure that metabolic incorporation of amino acids is complete.
We outline a sample protocol that may be adapted to both smaller
and larger scale experiments. It is highly recommended to use low
passage cells for any SILAC experiments. This is especially impor-
tant because the cells need to be cultured in the SILAC media for
at least six passages.
3.1 Cell Culture 1. Seed one 10 cm plate per condition in normal media at 25–50 %
and Labeling confluence the day before starting the SILAC experiment (Day
0, see Note 4 for comments on experiments requiring a larger
quantity of cells for final analysis).
2. Obtain one (1 L) Sterile Vacuum Filter for each growth condi-
tion (e.g., “Heavy” and “Light”) inside a sterile laminar bio-
logic safety bench.
3. Prepare SILAC media by adding the following items to the top
chamber of the vacuum filter:
(a) 860 mL of media.
(b) 1 mL of 1000× proline (to suppress arginine conversion).
(c) 1 mL of heavy or light 1000× amino acids as appropriate
(see Note 5).
(d) 10 mL 100× glutamine.
(e) 18 mL 1 M HEPES.
(f) 10 mL 100× of Penicillin/streptomycin.
(g) 100 mL dialyzed FBS.
4. Allow components to filter through the bottom chamber using
vacuum, discard the upper section. Label the media appropri-
ately (e.g., “Heavy” or “Light”).
5. On Day 1, aspirate the media and wash cells with PBS thrice.
6. Add 10 mL of SILAC media to each 10 cm plate.

7. Monitor the growth kinetics of the cells during the course of
metabolic labeling to ensure that the SILAC media does not
interfere with cell physiology (see Note 6).
8. Passage cells at appropriate times.
(a) Aspirate media off plates.
(b) Wash off residual media with 10 mL of PBS.
(c) Remove residual PBS by vacuum aspirator.
(d) Add 1 mL of trypsin and incubate cells for 5 min in a tem-
perature controlled incubator at 37 °C (5 % CO2).
(e) Add 9 mL of appropriate SILAC media to quench the
trypsin and transfer 1 mL to a new 10 cm plate.
(f) Add 9 mL of fresh SILAC media.
9. Repeat until the cells have gone through five successive pas-
sages (six total passages).
10. Confirm full incorporation of amino acids by performing a
small LC-MS/MS experiment.
3.2 Cell Lysis Each lysis condition is application and protocol dependent. We will
describe a simple urea lysis procedure that is applicable to many
subsequent analysis pipelines.
1. Seed the final passage of cells such that they reach 85 % conflu-
ence 48 h after passaging.
2. For Adherent cells, place 10 cm plates on ice and remove by
aspiration.
3. Wash cells thrice with 10 mL of ice cold PBS. Aspirating PBS
between washes (Note: Do not wash cells off the plate).
4. After the last wash, tilt plate at 45° angle for 1 min thereby
allowing the remaining PBS to drain to bottom of plate.
5. Completely aspirate remaining PBS.
6. Add 1 mL of urea lysis buffer.
7. Scrape cells of plate with a Nunc Cell scraper (Fisher Scientific,
Pittsburg, PA).
8.
Remove cells and pipette into a prechilled 1.5 mL
microcentrifuge.
9. Freeze cells at −20 °C for at least 2 h (Note: cells can be kept
frozen for later analysis).
10. Thaw cells on ice.
11. Prepare cell lysate by passing cell material through a 27 gauge
needle at least ten times (BD Sciences, Thermo Scientific,
Pittsburg, PA).
178 Kian Kani
12. Centrifuge cell lysate at 4 °C for 10 min at 14,000 × g.

13. Apply supernatant to 0.22 μm Corning Costar Spin-X centri-
fuge tube filters (Sigma-Aldrich, Pittsburg, PA) and centrifuge
at 5,000 × g for 10 min.
14. Place filtrate in a prechilled Eppendorf microcentrifuge tube.
15. Determine protein concentration using the BCA (Note: most
lysates must be diluted 1:4 to 1:10 in order to fall into the linear
range of the assay).
16. Mix equal mass of “heavy” and “light” labeled cells in a new
Eppendorf microcentrifuge tube.
(a) Example: 1 mg of “light” labeled lysates should be mixed
with 1 mg of “heavy” labeled lysates.
17. Samples must be reduced with 10 mM final concentration of
DTT (Sigma-Aldrich, St. Louis, MO). Incubate samples at
65 °C for 45 min.
18. Allow samples to cool to room temperature and add iodoacet-
amide (Sigma-Aldrich, St. Louis, MO) at a final concentration
of 55 mM for 30 min in the dark.
19. Protein digestion is accomplished by addition of sequencing
grade trypsin (Promega, Madison, WI) and/or Lys-c (Sigma-
Aldrich, St. Louis, MO). Efficient protein digestion occurs at
1 M urea, diluted with 55 mM ammonium bicarbonate sup-
plemented with 1:100 (mass ratio) trypsin. Incubate at 37 °C
for 16 h with gentle agitation or rotation.
20. Add additional 1:100 (mass ratio) trypsin for 4 h at 37 °C.
21. Trypsin digest should be quenched by addition of 1 μL of gla-
cial acetic acid (Sigma-Aldrich, St. Louis, MO).
22. Lyophilize samples in a Speed-Vac.
23. Resuspend lyophilized peptides in 20 μL of 0.5 % trifluoroace-
tic acid (TFA) in 5 % acetonitrile (ACN) (Sigma-Aldrich, St.
Louis, MO).
24. Desalt samples by applying peptides through a C18 spin col-
umn (Life Technologies, Carlsbad, CA). Wash columns with
400 μL of 0.5 % TFA, 5 % acetonitrile thrice. Elute peptides in
50 % methanol, 50 % acetonitrile, 0.1 % formic acid.
25. Proceed to LC-MS/MS.
3.3 Practical The following section will outline some of the practical consider-
Example ations that are involved in a typical two-plex SILAC based experi-
ment. This example can be modified to fit various experimental
conditions, cell lines, and controls. One of the key challenges in
clinical oncology is to determine which patient is going to benefit
from a particular therapeutic regime. Biomarkers that indicate
therapy response are useful clinical tools for patient stratification.
We will utilize the cellular dependence of the A431 cell line (ATCC,
Manassas, VA) to EGFR signaling as a surrogate for lung cancer
patients that harbor EGFR activating mutations in their tumors.
These patients have dramatic response rates to tyrosine kinase
inhibitors (TKIs) that target EGFR. The goal of the following
experiment is to identify proteins that are either upregulated or
downregulated in the EGFR dependent A431 cell line upon treat-
ment with an EGFR TKI (Iressa).
1. Two-plex SILAC experiments can be done with various com-
binations of “light” and “heavy” amino acids. In this example,
utilize SILAC with 13C6 lysine which results in a mass shift of
6 Da.
2. Maintain two cultures of A431 cells in either heavy or light
SILAC media and passage them for at least six generations (see
Note 7).
3. Perform technical replicates by dosing “heavy” labeled cells
with drug (in this case Iressa) in order to comparing the pro-
teome with “light” labeled cells treated with vehicle. The
reciprocal experiment is “heavy” labeled cells treated with
vehicle and compared to “light” labeled cells treated with
drug.
4. Two 15 cm plates of “heavy” labeled A431 cells and two 15 cm
plates of “light” labeled A431 cells are seeded at 50 % conflu-
ence (Note: the confluence and general health of cells is an impor-
tant component to SILAC based in vitro experiments. Proper
care should be placed on cell health).
5. The following day, dose cells “heavy” labeled cells with 100 nM
Iressa and “light” labeled cells with vehicle control.
6. Repeat the dosing on the reciprocal labeled cells.
7. After 16 h, proceed to cell lysis step as detailed in Subheading 3.2.
8. Perform LC-MS/MS analysis; in general, peptides identifica-
tions are based on the MS/MS spectra and the m/z of the
monoisotopic peak.
9. Peptide quantitation is obtained from the extracted ion chro-
matogram (XIC) of two differently labeled versions of the pep-
tide. The XIC is the contribution of a given m/z to the total
ion chromatogram (TIC). The main components of the XIC
are the m/z value, the elution time profile, and the intensity of
the peak. A number of different software packages are available
including Proteowizard [22], MSQuant [23], MaxQuant [24,
25], and Census [26, 27].
10. In order to assess the robustness of each quantitative peptide,
manual inspection of the XIC may be performed (Fig. 3).
Peptides with poor XIC overap and or low intensity should be
excluded.
180 Kian Kani
A. B.
Good quantitation Poor quantitation
500,000 500,000
450,000 450,000
Light 400,000 Light 400,000
350,000
Intensity
350,000
Intensity
Scans: 1172 - 1215 Scans: 1172 - 1215
300,000 300,000
Mass: 1122.63 Mass: 1122.63
250,000 250,000
Area: 1,300,000 Area: 75,000
200,000 200,000
150,000 150,000
100,000 100,000
50,000 50,000
1110 1130 1150 1170 1190 1210 1230 1250 1290 1310 1110 1130 1150 1170 1190 1210 1230 1250 1290 1310
Scan Scan
500,000 500,000
450,000 450,000
Heavy 400,000 Heavy 400,000
Intensity
Scans: 1172 - 1215 350,000
Intensity
Scans: 1172 - 1215 350,000
Mass: 1128.65 300,000 300,000
Mass: 1128.65
Area: 1,360,000 250,000 250,000
Area: 890,000
200,000 200,000
150,000 150,000
100,000 100,000
50,000 50,000
1110 1130 1150 1170 1190 1210 1230 1250 1290 1310 1110 1130 1150 1170 1190 1210 1230 1250 1290 1310
Scan Scan
500,000 500,000
450,000 450,000
Combined 400,000 400,000
Combined
Intensity
350,000
Intensity
H to L ratio: 1.05:1 350,000
300,000 H to L ratio: 11.9:1 300,000
L to H ratio: 0.95:1 250,000 L to H ratio: 0.08:1 250,000
200,000 200,000
150,000 150,000
100,000 100,000
50,000 50,000
1110 1130 1150 1170 1190 1210 1230 1250 1290 1310 1110 1130 1150 1170 1190 1210 1230 1250 1290 1310
Scan Scan
Fig. 3 (a) Accurate quantitation is obtained when the elution profile of the light and heavy peptide overlap in
time but differ in m/z. (b) In this example, the elution profile of the heavy and light peptide show traces of a
secondary peak which results in poor overlap of the XIC and unreliable quantitation
11. The final ratio for the protein expression reflects the expression
change as a function of Iressa dosing. The SILAC ratio of any
given protein can be determined by averaging the area of the
“light” and “heavy” peptides. The statistical significance of the
SILAC ratio must include the following parameters:
(a) The number of quantitative peptides (at least two).
(b) The intensity of the quantitative peptides should be greater
than 1000.
(c) The mean ratio for all peptides from this protein.
(d) The standard deviation (lower the better).
(e) Error determined by the student’s t-test (less than 0.05).
12. Lists of proteins may be exported to a suitable spreadsheet for
further analysis.
13. The initial step is to convert the “heavy” and “light” fold
changes into a log2 scale (Fig. 4a).
14. In order to compare results from different proteomic experi-
ments (reciprocal mixing) and eliminate to systematic errors in
the mixing step the data must be normalized using the median
of the log2 ratio. This may be achieved by dividing the ratio of
each protein by the median value.
A. 8.0
6.0 Increased protein level upon Iressa treatment
4.0
Fold change (log )

2
2.0
0.0
2.0
4.0
6.0 Decreased protein level upon Iressa treatment
8.0
Proteins
B.
# Pep. Avg H/L p value # Pep. Avg H/L p value
CBFB 18 2.29 0.003 5 0.49 0.02
Heavy treated Light treated
C.
1000 nM
100 nM
500 nM
CTRL
p-EGFR-Y1068
CBFB
Actin
16 hours (Iressa)
Fig. 4 (a) Distribution of protein fold change for the heavy treated sample. (b)
Example of several experimental parameters used to determine robustness of
data. (c) Western blot validation of protein fold change with loading control
15. We assess biological significance by performing a mock pro-

teomic experiment with untreated “heavy” mixed with
untreated “light” cells. Based on the distribution of proteins in
this experiment, we suggest a cutoff of at least 1.3. Most pro-
teomic publications utilize a threshold of 1.5 in SILAC ratios.
16. The results for the “heavy” Iressa treated sample should be
compared to the “light” Iressa treated sample. In an ideal
experiment, the ratios of the two experiments will be inversed
(Fig. 4b).
182 Kian Kani
17. Validation of proteomic experiments is a key step in assuring

reproducibility of proteomic experiments. Techniques that rely
on protein fold change are suitable for verification of pro-
teomic results (ELISA, and Western blots being the most com-
mon) (Fig. 4c).
4 Notes
1. Cell lines often require a unique formulation of media for opti-

mum survival. We recommend checking with an online deposi-
tory (http://www.atcc.org/) for the selection of correct
media. Most immortalized cell lines may be used with this
application, but, try to avoid cells that divide slower than once
in a 48 h period.
2. Some cell lines require a unique combination of growth factors
that are either lost during the dialysis step or not present in
sufficient concentrations to promote cell growth. In these con-
ditions, we suggest exogenous addition of growth factors or
dialysis of any preferred FBS with a lower molecular weight
cutoff. In this case, we suggest dialysis with a 3500 Da cutoff
Snakeskin tubing (Life Technologies, Carlsbad, CA) with
Phosphate Based Saline (PBS).
3. Small differences in the molecular mass of the “heavy” isotopes
may be ignored when preparing the stock solutions. As these are
1000× solutions, 1 mL of stock is sufficient to make 1 L of final
media which is the usual amount of media to make at one time.
4. Since isotopically “heavy” reagents are expensive, we suggest
starting the labeling procedure in 10 cm dishes. Large-scale
experiments may increase the number of plates after the third
passage.
5. For a two condition differential proteomics experiment, one
condition will be grown in all light media and the other condi-
tion will be grown in media supplemented with heavy lysine and
arginine. Sequential protein digestion with Lys-c and trypsin
provide increased peptide and protein coverage compared to
Lys-c alone; while single protease digestion with Lys-c increases
quantification of lysine-labeled proteins [28]. Media will have to
be supplemented with the heavy or light versions of all amino
acids absent in the media. That is, you will add 1 mL of 1000×
lysine (heavy or light) and 1 mL of 1000× arginine (heavy or
light) to every bottle of media used in the experiment.
6. Changes in glucose, pyruvate, growth factors, and steroids may
be optimized if growth kinetics become altered during labeling.
7. Prior to the start of any large-scale proteomics experiment, it is
highly advisable to plan out positive and negative controls. In this
example, the use of the A431 cell line was motivated by their dra-
matic response to Iressa treatment. In order to optimize drug dos-

age and incubation time we performed a number of mock
experiments to ensure our experimental strategy was robust. Based
on these results, we dosed A431 cells for 16 h with 100 nM of the
drug (Note: we utilized the phosphorylation of EGFR as a readout
to optimize dosing regimens. A similar approach should be used to
ensure validity of any large scale proteomics experiment).
References
1. Ong SE, Blagoev B, Kratchmarova I, ments with embryonic stem cells. Mol Cell
Kristensen DB, Steen H, Pandey A, Mann M Proteomics 7(9):1587–1597. doi:10.1074/
(2002) Stable isotope labeling by amino acids mcp.M800113-MCP200, M800113-
in cell culture, SILAC, as a simple and accurate MCP200 [pii]
approach to expression proteomics. Mol Cell 9. de Godoy LM, Olsen JV, de Souza GA, Li G,
Proteomics 1(5):376–386 Mortensen P, Mann M (2006) Status of com-
2. Beck S, Michalski A, Raether O, Lubeck M, plete proteome analysis by mass spectrom-
Kaspar S, Goedecke N, Baessmann C, Hornburg etry: SILAC labeled yeast as a model system.
D, Meier F, Paron I, Kulak NA, Cox J, Mann Genome Biol 7(6):R50. doi:10.1186/gb-
M (2015) The impact II, a very high-resolution 2006-7-6-r50, gb-2006-7-6-r50 [pii]
quadrupole time-of-flight instrument (QTOF) 10. Gruhler A, Schulze WX, Matthiesen R, Mann
for deep shotgun proteomics. Mol Cell M, Jensen ON (2005) Stable isotope label-
Proteomics 14(7):2014–2029. doi:10.1074/ ing of Arabidopsis thaliana cells and quantita-
mcp.M114.047407, M114.047407 [pii] tive proteomics by mass spectrometry. Mol Cell
3. Zhang G, Fenyo D, Neubert TA (2009) Proteomics 4(11):1697–1709. d oi:10.1074/mcp.
Evaluation of the variation in sample prepa- M500190-MCP200, M500190-MCP200 [pii]
ration for comparative proteomics using 11. Gruhler A, Olsen JV, Mohammed S, Mortensen
stable isotope labeling by amino acids in cell P, Faergeman NJ, Mann M, Jensen ON (2005)
culture. J Proteome Res 8(3):1285–1292. Quantitative phosphoproteomics applied to the
doi:10.1021/pr8006107 yeast pheromone signaling pathway. Mol Cell
4. Bantscheff M, Schirle M, Sweetman G, Rick J, Proteomics 4(3):310–327. doi:10.1074/mcp.
Kuster B (2007) Quantitative mass spectrome- M400219-MCP200, M400219-MCP200 [pii]
try in proteomics: a critical review. Anal Bioanal 12. Bose R, Molina H, Patterson AS, Bitok JK,
Chem 389(4):1017–1031. doi:10.1007/ Periaswamy B, Bader JS, Pandey A, Cole PA
s00216-007-1486-6 (2006) Phosphoproteomic analysis of Her2/
5. Tzouros M, Golling S, Avila D, Lamerz J, neu signaling and inhibition. Proc Natl Acad
Berrera M, Ebeling M, Langen H, Augustin Sci U S A 103(26):9773–9778. doi:10.1073/
A (2013) Development of a 5-plex SILAC pnas.0603948103, 0603948103 [pii]
method tuned for the quantitation of tyro- 13. Boersema PJ, Geiger T, Wisniewski JR, Mann
sine phosphorylation dynamics. Mol Cell M (2013) Quantification of the N-glycosylated
Proteomics 12(11):3339–3349. doi:10.1074/ secretome by super-SILAC during breast cancer
mcp.O113.027342, O113.027342 [pii] progression and in human blood samples. Mol
6. Hilger M, Mann M (2012) Triple SILAC to Cell Proteomics 12(1):158–171. doi:10.1074/
determine stimulus specific interactions in the mcp.M112.023614, M112.023614 [pii]
Wnt pathway. J Proteome Res 11(2):982–994. 14. Dhungana S, Merrick BA, Tomer KB, Fessler
doi:10.1021/pr200740a MB (2009) Quantitative proteomics analy-
7. Van Hoof D, Pinkse MW, Oostwaard DW, sis of macrophage rafts reveals compartmen-
Mummery CL, Heck AJ, Krijgsveld J (2007) talized activation of the proteasome and of
An experimental correction for arginine- proteasome- mediated ERK activation in
to-
proline conversion artifacts in SILAC- response to lipopolysaccharide. Mol Cell
based quantitative proteomics. Nat Methods Proteomics 8(1):201–213. doi:10.1074/mcp.
4(9):677–678. doi:10.1038/nmeth0907-677, M800286-MCP200, M800286-MCP200 [pii]
nmeth0907-677 [pii] 15. Ong SE, Mittler G, Mann M (2004) Identifying
8. Bendall SC, Hughes C, Stewart MH, Doble and quantifying in vivo methylation sites by
B, Bhatia M, Lajoie GA (2008) Prevention heavy methyl SILAC. Nat Methods 1(2):119–
of amino acid conversion in SILAC experi- 126. doi:10.1038/nmeth715, nmeth715 [pii]
184 Kian Kani
16. Zhang K, Li L, Zhu M, Wang G, Xie J, Zhao Y, development. Bioinformatics 24(21):2534–

Fan E, Xu L, Li E (2015) Comparative analysis 2536. doi:10.1093/bioinformatics/btn323,
of histone H3 and H4 post-translational modi- btn323 [pii]
fications of esophageal squamous cell carcinoma 23. Mortensen P, Gouw JW, Olsen JV, Ong SE,
with different invasive capabilities. J Proteomics Rigbolt KT, Bunkenborg J, Cox J, Foster LJ,
112:180–189. doi:10.1016/j.jprot.2014.09.004, Heck AJ, Blagoev B, Andersen JS, Mann M
S1874-3919(14)00419-9 [pii] (2010) MSQuant, an open source platform
1 7. Kruger M, Moser M, Ussar S, Thievessen for mass spectrometry-based quantitative
I, Luber CA, Forner F, Schmidt S, Zanivan proteomics. J Proteome Res 9(1):393–403.
S, Fassler R, Mann M (2008) SILAC doi:10.1021/pr900721e
mouse for quantitative proteomics uncov- 24. Cox J, Mann M (2008) MaxQuant enables
ers kindlin-3 as an essential factor for red high peptide identification rates, individualized
blood cell function. Cell 134(2):353–364. p.p.b.-range mass accuracies and proteome-
doi:10.1016/j.cell.2008.05.033, S0092- wide protein quantification. Nat Biotechnol
8674(08)00695-8 [pii] 26(12):1367–1372. doi:10.1038/nbt.1511,

18. Schwanhausser B, Gossen M, Dittmar G, nbt.1511 [pii]
Selbach M (2009) Global analysis of cellular pro- 25. Cox J, Matic I, Hilger M, Nagaraj N, Selbach
tein translation by pulsed SILAC. Proteomics M, Olsen JV, Mann M (2009) A practical guide
9(1):205–209. doi:10.1002/pmic.200800275 to the MaxQuant computational platform

19. Hanke S, Besir H, Oesterhelt D, Mann M for SILAC-based quantitative proteomics.
(2008) Absolute SILAC for accurate quantita- Nat Protoc 4(5):698–705. doi:10.1038/
tion of proteins in complex mixtures down to nprot.2009.36, nprot.2009.36 [pii]
the attomole level. J Proteome Res 7(3):1118– 26. Moore RE, Young MK, Lee TD (2002)
1130. doi:10.1021/pr7007175 Qscore: an algorithm for evaluating SEQUEST
20. Rees JS, Lilley KS, Jackson AP (2015) SILAC- database search results. J Am Soc Mass
iPAC: a quantitative method for distinguishing Spectrom 13(4):378–386. doi:10.1016/
genuine from non-specific components of protein S1044-0305(02)00352-5
complexes by parallel affinity capture. J Proteomics 27. Qiao Y, Zhang H, Bu D, Sun S (2011) PI: an
115:143–156. doi:10.1016/j.jprot.2014.12.006, open-source software package for validation
S1874-3919(14)00559-4 [pii] of the SEQUEST result and visualization of

21. Tackett AJ, DeGrasse JA, Sekedat MD, mass spectrum. BMC Bioinformatics 12:234.
Oeffinger M, Rout MP, Chait BT (2005) doi:10.1186/1471-2105-12-234, 1471-2105-
I-DIRT, a general method for distinguish- 12-234 [pii]
ing between specific and nonspecific protein 28. Ma J, Li W, Lv Y, Chang C, Wu S, Song
interactions. J Proteome Res 4(5):1752–1756. L, Ding C, Wei H, He F, Jiang Y, Zhu Y
doi:10.1021/pr050225e (2013) A new insight into the impact of
22. Kessner D, Chambers M, Burke R, Agus
different proteases on SILAC quantitative
D, Mallick P (2008) ProteoWizard: open proteome of the mouse liver. Proteomics.
source software for rapid proteomics tools 13(15):2238–2242
Chapter 14
Relative Protein Quantification Using Tandem

Mass Tag Mass Spectrometry
Lichao Zhang and Joshua E. Elias
Abstract
Measuring protein changes over time or following stimuli is one of the important tasks of proteomics. In
the past decade, several strategies have been developed for the relative quantification of proteins using mass
spectrometry (MS). Isobaric labeling strategies for relative quantitative proteomics allow for parallel mul-
tiplexing of quantitative experiments. With this technique, multiple peptide samples are chemically labeled
with isobaric chemical tag variants and each variant has the same molecular structure and mass. Each vari-
ant, however, is designed to produce a unique “reporter ion” when fragmented inside a mass spectrometer.
Once peptide samples are labeled, combined, and analyzed using MS, differentially labeled peptides are
indistinguishable in a first, MS spectrum of intact peptides. However, since each tag variant contains a
labile component with different mass, “reporter ions” can be generated and recorded in a subsequent MS2
spectrum. Intensities from each variant are recorded to represent the relative abundances of the peptide in
each sample. Isobaric tags for relative and absolute quantitation (iTRAQ) and tandem mass tags (TMT)
are commercially available reagents for performing this technique. Here, we describe the general workflow
of relative quantification of proteins using TMT by MS2, or an additional MS3 spectrum.
Key words Mass spectrometry, Quantification, Tandem mass tag
1 Introduction
Several labeling techniques measure relative abundances of pro-

teins in different biological samples by mass spectrometry. Two
general labeling strategies—in vivo and in vitro labeling—have
been widely applied to quantitative protein analysis. In vivo label-
ing requires heavy isotopes to be incorporated into the proteins of
living organisms by feeding them isotopically labeled nutrients in
the growth medium or food [1]. This approach has limitations,
such as the requirement for large amounts of costly growth
medium, and difficulties in controlling of labeling efficiency [2]. In
vitro labeling overcomes these limitations by chemically labeling
any peptide sample with high efficiency. Isotope-coded affinity tag
(ICAT) introduced the concept of in vitro, quantitative peptide
185
186 Lichao Zhang and Joshua E. Elias
labeling for MS-based detection [3]. With this method, chemically

identical probes with distinct masses are utilized to label two pep-
tide samples. Their relative abundances are then determined from
the each form’s MS1 peak intensity. The identity of each peptide is
inferred from a MS2 spectrum derived from at least one of the
labeled forms.
The introduction of Isobaric tags such as TMT and iTRAQ
enables more accurate and multiplexed quantification by using MS2
spectra for quantitative analysis [2, 4]. These reagents are typically
composed of a mass reporter, a mass normalizer and an amine reac-
tive group (Fig. 1a). The mass reporter and mass normalizer moi-
eties incorporate stable isotopes in multiple configurations such
that each mass reporter’s mass can be resolved in a MS2 spectrum.
The intact mass of each isobaric tag variant, however, is the same.
With this labeling strategy, digested peptides from multiple samples
are first labeled in parallel with as many as ten tag variants (TMT).
They are then mixed and analyzed using reversed phase high per-
formance liquid chromatograph (HPLC) coupled with a mass spec-
trometer capable of tandem MS analysis (Fig. 2). The mass reporter
can be cleaved off the labeled peptide through collision-induced
dissociation (CID) prior to detection with a time-of-flight (TOF)
mass analyzer, or through higher energy collision dissociation
(HCD) prior to detection with an orbitrap mass analyzer. Intensities
measured in the resulting MS2 spectrum report relative peptide
quantification (Fig. 3). The isobaric and chemically identical nature
of isobaric tags ensure that identical peptides labeled with different
tag variants will have the same chromatographic elution profile, and
experience identical ionization processes in the mass spectrometer.
This leads to a more accurate quantification in comparison to the
original ICAT strategy. In addition, this strategy often improves the
signal-to-noise ratio of quantitative measurements by enabling
quantification at the MS2 level, which is inherently less noisy than
MS1 scans of intact peptides.
However, in highly complex peptide mixtures, co-isolation
and co-fragmentation of multiple ions is increasingly likely to occur
with MS2, resulting in distorted ratios of isobaric tag reporter ions
[5]. In this case, distorted reporter ion ratios no longer reflect the
true proportions of a selected peptide precursor’s components.
Combining ion trap and orbitrap technologies allows reporter sig-
nals to be measured following a third stage of ion isolation and
fragmentation (MS3). This technique was shown to greatly reduce
ratio distortion resulting from interfering ions [5]. Moreover, a
“MultiNotch MS3” method (also referred to as “synchronous pre-
cursor selection”) was demonstrated to increase detection
sensitivity by simultaneously selecting multiple fragment ions from
MS2 to enhance reporter ion intensities at the MS3 level [6].
This chapter describes a protocol for protein quantification
using commercially available TMT reagents, which include 2plex,
TMT Quantitation Mass Spectrometry 187
Fig. 1 Structure of TMT reagents. (a) TMT reagent contains mass reporter, mass
normalizer and amine reactive group. A HCD cleavable linker is between the
mass reporter and mass normalizer. (b) 10plex TMT. Heavy labeled C12 and N13
are indicated by asterisks. Figures are adapted from Thermo Scientific instruc-
tions for TMT10plex mass tag labeling kits and reagents
6plex, and 10plex (Fig. 1b). The TMT 10plex technique is used to

describe the analysis procedure, including protein digestion, TMT
reaction, sample purification, fractionation, HPLC-MS, and data
analysis.
Fig. 2 Workflow for relative quantification using 10plex TMT. Ten different protein samples are denatured,
reduced, alkylated and digested into peptide samples. These peptide samples are then labeled with 10plex
TMT reagent kit and then combined into one sample. The combined sample is fractionated using HPRP liquid
chromatography and the fractions are concatenated into 12 samples. After cleaning up, each sample is sub-
jected to HPLC-MS analysis and the data are then analyzed to obtain the identity and relative quantification of
proteins
2 Materials
2.1 Reduction, 1. Ammonium bicarbonate.

Alkylation and Protein 2. HPLC grade water.
Digestion
3. Dithiothreitol.
4. Iodoacetamide.
5. Sequencing grade modified trypsin.
6. Formic acid, 99 %.
2.2 TMT Labeling 1. Acetonitrile (ACN), anhydrous, 99.8 %.

2. HEPES sodium salt (Na-HEPES), minimum 99.5 % titration.
3. TMT 10plex Mass Tag Labeling Kit (Thermo Scientific,
Rockford, IL).
4. Hydroxylamine solution 50 wt%.
2.3 Sample 1. Sep-Pak C18 1 cc vac cartridge, 50 mg sorbent per cartridge
Purification (Waters, Milford, MA).
and Fractionation 2. Vacuum manifold (Varian, Palo Alto, CA).
3. Vortex mixer.
Fig. 3 MS analysis of 10plex TMT samples for peptide sequencing and quantification. Top figure shows a typi-
cal MS2 HCD spectrum. Bottom left figure shows the zoomed spectrum of the reporter ions region. Theoretical
HCD monoisotopic reporter ion masses for 10plex TMT reagent are listed at bottom right
4. Centrifuge.
5. Methanol.
6. Acetonitrile.
7. Acetic acid, glacial.
8. Ammonium formate for HPLC ≥ 99.0.
9. Ammonium hydroxide (Fisher Scientific, Fair Lawn, NJ).
10. Centrivap concentrator (Labconco, Kansas City, MO).
11. Empore™ SPE Disks, C18, 47 mm (Sigma-Aldrich, St. Louis, MO).
12. ZORBAX Extend-C18 column (Agilent technologies, Santa
Clara, CA).
13. HPLC binary pump.
2.4 HPLC-MS 1. Nano-flow C18 HPLC columns are commercially available

Analysis such as Acclaim PepMap100 C18 column of 75 μm in i.d. and
250 mm in length, packed with 3 μm 100 Å beads (Fisher
Scientific, Fair Lawn, NJ).
2. The following equipment are needed for in-house packed
nano-flow C18 column:
Laser-based micropipette puller P-2000 (Sutter Instrument,
Novato, CA).
Fused Silica Capillary Tubing, 100 μm (Polymicro
Technologies, Phoenix, AZ).
Reprosil-Pur 120 C18-AQ, 5 μm and 3 μm (Dr. Maisch
GmbH, Ammerbuch, Germany).
3. Ekspert nanoLC 425 (AB Sciex, Redwood City, CA).
4. LTQ-Obitrap Velos/Elite Mass spectrometer, or Obitrap
Fusion Tribrid Mass Spectrometer (Thermo Scientific, San
Jose, CA).
3 Methods
3.1 Reduction, 1. Suspend 100 μg of protein in 100 mM ammonium bicarbon-

Alkylation and Protein ate (pH ~ 8) (see Notes 1–4).
Digestion 2. Prepare 0.5 M DTT in HPLC grade water. Add 0.5 M DTT to
the protein sample to obtain a final concentration of 5 mM and
incubate for 60 min at 37 °C to reduce disulfide bonds.
3. Allow sample to cool to room temperature. Prepare 0.5 M
iodoacetamide in HPLC grade water and add to the sample to
obtain a final concentration of 14 mM. Incubate for 45 min at
room temperature in the dark to alkylate cysteines.
4. Add trypsin or LysC (see Note 5) to the sample at an enzyme
to protein ratio of 1:20 to 1:50 and digest for 6 h to overnight
at 37 °C (see Note 4).
5. Acidify the sample to pH 3 with formic acid to quench diges-
tion (see Note 3).
6. Desalt the digest using C18 solid-phase extraction (Sep-Pak).
The C18 columns are conditioned and equilibrated with 1 mL
of methanol (three times), 1 mL of 70 % ACN in 0.5 % Acetic
acid (one time), 1 mL of 40 % ACN in 0.5 % acetic acid, and
1 mL of 0.5 % Acetic acid (three times). Add the sample, wash
with 1 mL of 0.5 % acetic acid (three times) and elute the pep-
tides with 1 mL of 40 % ACN in 0.5 % acetic acid. The eluted
samples are dried in a centrivap concentrator and stored in
−20 °C (see Note 6).
3.2 TMT Labeling 1. Resuspend ten samples in 100 μL of 50 mM Na-HEPES

(Using TMT 10plex (pH 8).
as Example) 2. Add 30 μL of anhydrous ACN to each sample.
3. Equilibrate TMT reagent at room temperature for 10–15 min
before opening lids of reagent vials (see Note 7). Quickly spin
down the reagent vial to collect any residue to the bottom of
the tube and then resuspend the TMT reagent in 40 μL of
anhydrous ACN. Allow reagent to dissolve for 5 min with
occasional vortexing, and spin down briefly at the end.
4. Add 10 μL of TMT reagent to each corresponding peptide
sample and mix with a very gentle and quick vortex (see Notes
8 and 9).
5. Incubate the samples for 1 h at room temperature.
6. Prepare 5 % hydroxylamine solution by diluting 50 % hydroxyl-
amine with 50 mM Na-HEPES. Add 10 μL of 5 % hydroxyl-
amine to each sample and incubate for 15 min at room
temperature to quench the reaction.
7. Acidify the samples to pH 2 with 60 μL of 25 % formic acid.
3.3 Sample 1. Combine 1 % of each labeled sample and dry down in a cen-
Preparation for Ratio trivap concentrator.
and Labeling 2. Dry down the rest of the labeled samples and store in −20 °C.
Efficiency Check
3. Purify the combined sample using STAGE-Tip [7]. Cut out
(See Note 10) four cookies (2 mm o.d.) from the Empore C18 Disks and
pack into a P200 pipette tip. The STAGE-Tip column is con-
ditioned and equilibrated with 80 μL of methanol, 40 μL of
40 % ACN in 5 % formic acid, 40 μL of 5 % formic acid (two
times). Resuspend the dried sample in 40 μL of 0.1 % formic
acid. Add the sample, wash with 40 μL of 5 % formic acid (two
times) and elute the peptides with 40 μL of 40 % ACN in 5 %
formic acid. The eluted samples are dried in a centrivap con-
centrator and stored in −20 °C.
4. Perform HPLC-MS analysis described in Subheading 3.5 and
data analysis in Subheading 3.6.
5. Verify if the reporter ion distributions follow desired overall
ratio (e.g., 1:1:1:1:1:1:1:1:1:1). Calculate sample ratio adjust-
ments, if necessary (see Note 10).
3.4 Sample 1. Resuspend the samples in 50 μL of 0.1 % formic acid and combine
Purification the samples with the adjusted ratio (Subheading 3.3, step 5).
and Fractionation 2. Purify the combined sample using Sep-Pak.
See Note 11)
3. Fractionate the sample using offline high pH reversed phase

(HPRP) liquid chromatography (see Note 11). Buffer A is
10 mM ammonium formate and buffer B is 90 % acetoni-
trile/10 % water (pH for both buffers is adjusted to 10 with
ammonium hydroxide). Solubilize the labeled and mixed pep-
tides in buffer A and separate on an Agilent ZORBAX 80 Å
Extend-C18 column (5 μm particles, 4.6 × 250 mm) with an
Agilent 1200 binary pump equipped with a photodiode array
detector. Use an 84 min linear gradient from 0 to 40 % buffer
B in 66 min to separate the peptide mixture into 96 fractions.
Concatenate the 96 fractions into 12 samples by combining
eight fractions with equal time interval [8]. Acidify each sam-
ple with formic acid to pH 3, and dry the samples in a cen-
trivap concentrator.
4. Redissolve each sample in 0.1 % formic acid and purify with
Sep-Pak. Vacuum-dry and reconstitute the samples in 0.1 %
formic acid for HPLC-MS analysis.
3.5 HPLC-MS 1. Prepare the following buffers: buffer A is 0.2 % formic acid in
Analysis water; buffer B is 0.2 % formic acid in ACN.
2. Peptides are separated on a 100 μm inner diameter microcapil-
lary column. The tip of the column is pulled in-house with
laser puller and the column is packed with Reprosil C18 resin
(0.5 cm of 5 μm resin close to the tip followed by 18 cm of
3 μm resin). There are also commercially available nano-flow
C18 columns that can be used.
3. Run the HPLC gradient as follows: 98 % A to 60 % A in
130 min, to 40 % A in 20 min, and then to 2 % A in 18 min.
The total gradient time is 180 min.
4. The TMT reporter ions can be analyzed at either MS2 or MS3
level on a LTQ-Orbitrap instrument (Velos or Elite), and the
“synchronous precursor selection” MS3 method is available with
Obitrap Fusion Tribrid Mass Spectrometer (see Note 12) [5, 6].
(a) MS2 method. The MS1 scans are performed in the Orbitrap
in the mass range of 300–1500 m/z and the resolution is
set to 60,000. Ten most intense ions (intensity above 500
counts) are selected for HCD fragmentation. The precur-
sor isolation width is set to 2 m/z. The AGC settings are
1E6 and 5E4 for FTMS1 and FT MSn scans, respectively.
Maximum ion times for FTMS1 and FTMSn are both
250 ms. The normalized collision energy for HCD is 45 %
at 30 ms activation time. Monoisotopic precursor selection
and charge state rejection are enabled. Singly charged ion
species and ions with no unassigned charge states are
excluded from MS2 analysis. Ions within ±10 ppm m/z
window around ions selected for MS2 are excluded from
further selection for fragmentation for 90 s.
(b) MS3 method. The MS1 scans are performed in a similar

way. Ten most intense ions (intensity above 500 counts)
are selected for CID fragmentation in the ion trap with
precursor ion isolation width of 2 m/z, AGC setting of
5E3, maximum ion time of 150 ms and normalized colli-
sion energy of 35 % at activation time of 20 ms. Following
each MS2 analysis, the most intense fragment ion (intensity
above 500 counts) is selected for HCD MS3 analysis with
isolation width of 2 m/z, normalized collision energy of
60 % at activation time of 50 ms and resolution of 60,000.
Charge state screening is disabled to allow fragment ions
to be selected for MS3 (see Note 13).
(c) MS3 method with synchronous precursor selection. This
method can be selected from the available predefined work-
flows. Similarly, the full MS scans are performed in the
Orbitrap in the mass range of 300–1500 m/z and the reso-
lution is set to 120,000. The AGC settings is 4E5 and max-
imum ion time for FTMS1 is 50 ms. Data dependent mode
is set to top speed and precursor priority is set to most
intense. Precursors with charge states 2–7 for MS2 CID are
isolated in quadrupole with isolation width of 0.7 m/z,
AGC setting of 1E4 and maximum ion time of 50 ms. CID
collision energy is set to 35 % and the mass analysis is per-
formed in ion trap. Ions within ±10 ppm m/z window
around ions selected for MS2 are excluded from further
selection for fragmentation for 90 s. Following each MS2
CID, a MS3 HCD fragmentation is performed with syn-
chronous precursor selection enabled (the number of pre-
cursors set to 5), AGC setting of 1E5, and maximum ion
time of 105 ms. HCD collision energy is set to 65 % and the
fragment ions are detected in Orbitrap in the scan range of
120–500 m/z with resolution of 60,000 (see Note 14).
3.6 Data Analysis 1. Several software packages such as Proteome Discoverer,

Mascot, and MaxQuant, are available to process TMT data for
protein identification and quantification. Here we describe a
protocol using Proteome Discoverer 2.1.
2. Create a workflow including “Spectrum Files”, “Spectrum
Selector”, “Sequest HT” or “Mascot”, “Percolator”, and
“Reporter Ions Quantifier”. Trypsin or LysC is selected as the
enzyme with at most two missed cleavage sites. Precursor mass
tolerance is set to ±10 ppm and fragment mass tolerance is set
to ±0.6 Da. At most four dynamic modifications are allowed
per peptide. Carbamidomethylation of cysteine (+57.021 Da)
is set as static modification, oxidation of methionine
(+15.995 Da) is set as differential modification, TMT-labeled
N-terminus and lysine (+229.163) are set as differential
modifications or static modifications (see Note 15). Percolator

is applied to filter out the false MS2 assignments at a strict false
discovery rate of 1 % and relaxed false discovery rate of 5 %. The
maximum delta Cn is set to 0.05. For quantification, a mass
tolerance of ±20 ppm window is applied to the integration of
report ions using the most confident centroid method.
3. Create a consensus workflow (see Note 16). In the “peptide
and protein quantification” node, set “apply quan value cor-
rections” to “true” to apply correction for the isotopic impu-
rity of each reporter ions (values provided by the manufacturer
in the product insert). Reporter abundances can be present
using S/N values in Proteome Discoverer 2.1. Peptides are
normally filtered based on average reporter S/N and isolation
efficiency. The threshold for average reporter S/N can be set
to 10 and the threshold for co-isolation can be set to 30 %.
4. Data normalization is often necessary for more accurate quanti-
fication. Proteome discoverer provides the function of normal-
ization and scaling in the “peptide and protein quantifier” node
in the consensus workflow. The normalization mode can be set
to “total peptide amount” or “specific protein amount” (see
Note 17). The scaling mode is normally set to “on channels
average” (see Note 18). For ratio check analyses (Subheading 3.3)
normalization is not needed but the scaling can still be per-
formed, then the average values of all the scaled reporter ion
abundances for each channel are calculated and compared. The
rest of the samples should be combined with adjusted ratio.
5. Protein abundances can be calculated as the sum, mean, or
median of peptide abundances. Proteome Discoverer 2.1 uses
the sum of peptide abundances as protein abundances. Proteins
are then evaluated by the extent to which they increased or
decreased in one sample relative to another. A scatter plot
comparing the magnitude of a log-transformed ratio (x-axis)
versus the significance of the ratio (y-axis) can be a useful way
to intuitively select meaningful changes between any two
experimental variables. Significance can be calculated based on
the variance across biological replicates if TMT channels
encode such replicates. Alternatively, significance can be esti-
mated from the variance in ratios recorded from each peptide
that contributed to a protein’s quantification.
4 Notes
1. It is recommended to measure protein amount using a Bradford

or BCA protein assay, or SDS-PAGE with standard amount of
protein before protein digestion.
2. This protocol is based on labeling 100 μg of peptide sample per

condition. If more than 100 μg of protein is digested, an aliquot
of 100 μg of sample can be prepared for TMT labeling.
3. Digestion protocols often use primary amine-containing buf-
fers such as ammonium bicarbonate or 50 mM Tris–HCl
(pH ~ 8.8). Such buffers will partially or completely quench the
TMT reagent, leading to incomplete peptide labeling. Thus,
this protocol desalts peptides by solid phase extraction, prior to
TMT labeling. If no such cleanup step is possible or desired,
non-primary amine-based buffers (such as TEAB, HEPES)
should be used as the digestion buffer. With this modification,
there is no need to quench digestion before labeling.
4. If the protein sample is difficult to solubilize, prepare 8 M urea
in 100 mM ammonium bicarbonate as the denaturing buffer
instead. After reduction and alkylation, dilute urea to 1 M with
100 mM ammonium bicarbonate for trypsin digestion.
However, exposure of protein to urea may lead to carba-
mylation on peptide N-terminus and lysine residues [9]. To
avoid carbamylation, the sample in urea buffer should not be
heated to above 37 °C. Normally room temperature or 37 °C
is efficient for reduction and digestion.
5. Digestion with LysC instead of trypsin is recommended for
TMT quantification at the MS3 level [5]. Since TMT is reactive
to peptide N-terminus and lysine residues, digestion with LysC
ensures all b- and y-type ions contain at least one TMT reporter.
Therefore when performing conventional MS3 analysis, the most
abundant ion from an MS2 spectrum always contains TMT and
its MS3 fragment ions can always provide quantitative informa-
tion. However, since LysC generates longer peptides with higher
charge states, the number of peptides identified using collision-
induced dissociation (CID) will decrease. If “synchronous pre-
cursor selection” MS3 analysis is applied, tryptic digestion
becomes an option again since multiple b- and/or y-type ions will
be selected for MS3 analysis and there is large chance that one or
more of the selected ions contain TMT.
6. It is recommended to perform quality control on the digested
sample using HPLC-MS to make sure the digestion is com-
plete and there are no chemical modifications on N-terminus
and lysine (such as carbamylation) that could affect TMT label-
ing. Also, the concentration of protein digest can be measured
using quantitative peptide assay such as Thermo Scientific
Pierce Quantitative Florescent Peptide Assay or Thermo
Scientific Pierce Quantitative Colorimetric Peptide Assay.
7. TMT reagents are sensitive to moisture. Equilibrate the
reagents to room temperature before opening lid to avoid
forming condensation inside the reagent vial.
8. Thermo scientific instructions for TMT10plex mass tag label-

ing kits and reagents suggests that 25–100 μg of protein can be
labeled with one vial of reagent. However one vial of reagent is
actually enough to label up to 400 μg of peptides. Therefore in
this protocol when labeling 100 μg of peptides, one quarter of
the reagent is used. If more TMT reagent is desired, the vol-
ume of anhydrous ACN added to the sample should be
decreased in order to maintain the same the final concentration
of ACN in the labeling reaction.
9. If any TMT reagent remains once the vials have been opened,
it can be stored in anhydrous ACN at −20 °C for up to 1 week.
After 1 week, the reagent will degrade and the labeling effi-
ciency will decrease.
10. After TMT labeling, it is highly recommended to combine the
same small amount of sample from each condition and perform
a HPLC-MS run to check the ratio of different conditions as
well as TMT labeling efficiency. Since the measurement of pro-
tein amount may be inaccurate, the digestion may have differ-
ent efficiencies and the post-digestion purification may have
different peptide recoveries, the actual ratio of labeled peptide
samples may be not at 1:1. Normally ratio variation of 1.5-fold
or less is acceptable and can be normalized later. However if the
ratio variation is too large, the channels with low reporter ion
intensities will be skewed towards zero. For final TMT sample,
samples of different conditions can be combined with the ratio
adjusted based on ratio check result. Deviant test ratios could
also reflect material loss affecting one or more samples prior to
labeling. It could also reflect inaccurate protein quantification
(see Note 1), or nonuniform protein abundance distributions
across the experimental samples (e.g., one dominant protein in
one sample). To resolve these possibilities, SDS-PAGE analysis
of each sample should be considered. The ratio check run is also
helpful to evaluate the TMT labeling efficiency. Labeling effi-
ciency is defined as the percentage of fully labeled unique pep-
tides in all the peptides identified. Using this protocol 99 % of
labeling efficiency can be achieved routinely. If the labeling effi-
ciency is lower than 90 %, contaminants in the sample or in the
buffer used for TMT labeling could have quenched the TMT
reagent. Samples can be relabeled with fresh TMT reagent in
newly made buffers after being purified. Also if the amount of
peptides is suspected to be higher than expected, larger amount
of TMT reagents should be used.
11. Two-dimensional liquid chromatography is necessary for com-
plex samples to further reduce sample complexity before MS
analysis so that the dynamic range and proteome coverage can
be improved, and the co-eluting/co-fragmented signal interfer-
ences for quantification can be reduced. Offline, concatenated
HPRP liquid chromatography is recommended for its ability to

maximize the quantity of identifications a mass spectrometer can
make per minute when combined with standard HPLC-MS [8].
Alternative fractionation techniques include strong cation
exchange, isoelectric focusing, and hydrophilic interaction liq-
uid chromatography.
12. The “MultiNotch MS3” method from ref. 6 was implemented
on an Orbitrap-Elite mass spectrometer with experimental
software. Essential aspects of this method were implemented
commercially on the Obitrap Fusion platform and called “syn-
chronous precursor selection”.
13. With charge state screening disabled, ions with charge state +1
and no charge state assigned can still be excluded for MS2 anal-
ysis. Charge state exclusion is only applied to precursor selec-
tion for MS2 but not for MS3.
14. For TMT 10plex reporter ions detection, resolution of 60,000
is recommended. However for TMT 6plex or less, lower reso-
lution can be applied for orbitrap mass analysis or low resolu-
tion analysis in ion trap can be performed.
15. TMT-labeled N-terminus and lysine (+229.163) can be set as
differential modifications for the purpose of checking labeling
efficiency.
16. An established workflow template for TMT quantification is
named “CWF_Comprehensive_Enhanced Annotation_Quan”
in Proteome Discoverer.
17. If data is normalized with total peptide amount, it is based on
the assumption that the majority of peptides are essentially
invariant across samples while only a small portion of peptides
increase or decrease among different samples. This assumption
is not necessarily valid for all experiments. After normalization
with total peptide amount, the total abundance of peptides is
the same for all channels. This step is to correct different
amounts of sample input in each channel resulting from indi-
vidual sample preparation and final mixing.
18. In this mode the abundances of each channel are scaled so that
for every peptide and protein the average of all channels is 100.
This step is to correct the differences of reporter ion abun-
dance resulting from different amount of peptides detected.
Acknowledgments
The authors would like to thank Dr. Lihua Jiang from Stanford
University and Dr. Xiaoyue Jiang from Thermo Scientific for help-
ful discussions about TMT methods on Obitrap Fusion Tribrid
Mass Spectrometer and data processing using Proteome Discoverer.
References
1. Oda Y, Huang K, Cross F, Cowburn D, Chait plexed quantitative proteomics. Nat Methods
B (1999) Accurate quantitation of protein 8:937–940
expression and site-specific phosphorylation. 6. McAlister GC, Nusinow DP, Jedrychowski MP,
Proc Natl Acad Sci U S A 96:6591–6596 Wuehr M, Huttlin EL, Erickson BK, Rad R,
2. Thompson A, Schafer J, Kuhn K, Kienle S, Haas W, Gygi SP (2014) MultiNotch MS3
Schwarz J, Schmidt G, Neumann T, Hamon C enables accurate, sensitive, and multiplexed
(2003) Tandem mass tags: a novel quantifica- detection of differential expression across cancer
tion strategy for comparative analysis of com- cell line proteomes. Anal Chem 86:7150–7158
plex protein mixtures by MS/MS. Anal Chem 7. Rappsilber J, Mann M, Ishihama Y (2007)
75:1895–1904 Protocol for micro-purification, enrichment,
3. Gygi S, Rist B, Gerber S, Turecek F, Gelb M, pre-fractionation and storage of peptides for
Aebersold R (1999) Quantitative analysis of proteomics using StageTips. Nat Protoc
complex protein mixtures using isotope-coded 2:1896–1906
affinity tags. Nat Biotechnol 17:994–999 8. Wang Y, Yang F, Gritsenko MA, Wang Y, Clauss
4. Ross P, Huang Y, Marchese J, Williamson B, T, Liu T, Shen Y, Monroe ME, Lopez-Ferrer
Parker K, Hattan S, Khainovski N, Pillai S, Dey D, Reno T, Moore RJ, Klemke RL, Camp DG
S, Daniels S, Purkayastha S, Juhasz P, Martin S, II, Smith RD (2011) Reversed-phase chroma-
Bartlet-Jones M, He F, Jacobson A, Pappin D tography with multiple fraction concatenation
(2004) Multiplexed protein quantitation in strategy for proteome profiling of human
Saccharomyces cerevisiae using amine-reactive MCF10A cells. Proteomics 11:2019–2026
isobaric tagging reagents. Mol Cell Proteomics 9. Stark G, Stein W, Moore S (1960) Reactions of
3:1154–1169 cyanate present in aqueous urea with amino
5. Ting L, Rad R, Gygi SP, Haas W (2011) MS3 acids and proteins. J Biol Chem
eliminates ratio distortion in isobaric multi- 235:3177–3181
Chapter 15
Pathway-Informed Discovery and Targeted Proteomic

Workflows Using Mass Spectrometry
Caroline S. Chu, Christine A. Miller, Andy Gieschen,
and Steve M. Fischer
Abstract
Recent advancements in mass spectrometry (MS) and data analysis software have enabled new strategies
for biological discovery using proteomics. Proteomics has evolved from routine discovery and identifica-
tion of proteins to integrated multi-omics projects relating specific proteins to their genes and metabolites.
Using additional information, such as that contained in biological pathways, has enabled the use of tar-
geted protein quantitation for monitoring fold changes in expression as well as biomarker discovery. Here
we discuss a full proteomic workflow from discovery proteomics on a quadrupole Time-of-Flight (Q-TOF)
MS to targeted proteomics using a triple quadrupole (QQQ) MS. A discovery proteomics workflow
encompassing acquisition of data-dependent proteomics data on a Q-TOF and protein database searching
will be described which uses the protein abundances from identified proteins for subsequent statistical
analysis and pathway visualization. From the active pathways, a protein target list is created for use in a
peptide-based QQQ assay. These peptides are used as surrogates for target protein quantitation. Peptide-
based QQQ assays provide sensitivity and selectivity allowing rapid and robust analysis of large batches of
samples. These quantitative results are then statistically compared and visualized on the original biological
pathways with a more complete coverage of proteins in the studied pathways.
Key words Mass spectrometry, Proteomics, QTOF, QQQ, Time-of-flight, Quadrupole, Proteins,
Informatics, Data-dependent acquisition, Protein quantitation
1 Introduction
Here we discuss a full proteomic workflow using MS from discov-

ery proteomics LC/MS on a quadrupole Time-of-Flight (Q-TOF)
MS to targeted proteomics using a triple quadrupole (QQQ) LC/
MS. For the discovery proteomics on the Q-TOF both ultra-high
performance liquid chromatography (UHPLC) and nano-LC con-
ditions will be included in the discussion depending on the user’s
sample matrix. The multistep discovery proteomics workflows
(Fig. 1) encompass acquisition of the data-dependent proteomics
data on the Q-TOF, protein database searching using Spectrum
199
200 Caroline S. Chu et al.
Fig. 1 Workflow for discovery proteomics
Fig. 2 Targeted proteomics workflow
Mill to the statistical comparison using Mass Profiler Professional

and visualization of the experimental results in biological pathways
in Pathway Architect used for the development of the targeted
proteomic experiments. For the targeted proteomics workflows
Pathway-Informed Discovery and Targeted Proteomic Workflows… 201
(Fig. 2) the targeted proteins obtained from Pathway Architect are

transformed to a targeted peptide method using Spectrum Mill or
Skyline for use on the QQQ, and the data is then statistically com-
pared using Mass Profiler Professional.
2 Materials
Prepare all solutions in water of 18 MΩ-cm resistivity or better and

free of organics. For formic acid, we use LC/MS grade and prefer
to buy small volume bottles to maintain contamination-free acid.
2.1 Protein Digest 1. Protein Lysate: As a quality assessment sample, we use either a
Reagents Pyrococcus furiosus (Pfu, Agilent p/n 400510) [1, 2] or E. coli
protein lysate (BioRad p/n 163-2110).
2. Ammonium bicarbonate (100 mM): add 100 mL water to
0.7906 g ammonium bicarbonate.
3. Dithiothreitol (DTT, 200 mM) >99 + %: add 1 mL water to
0.031 g DTT in a 1.5 mL Eppendorf tube.
4. Iodoacetamide (IAM, 200 mM) 97 %: add 1 mL water to
0.037 g IAM in a 1.5 mL Eppendorf tube.
5. Trypsin Stock Solution: Sequencing-grade trypsin and resus-
pension buffer (Agilent p/n 204310): Prepare a 1 μg/μL solu-
tion in a) 50 mM acetic acid if not all the solution will be used
(store remainder in freezer) or b) 50 mM ammonium bicar-
bonate if it will all be used immediately.
2.2 Discovery 1. AdvanceBio Peptide Mapping Column, 2.1 × 250 mm, 2.7 μm

Proteomics Standard- (Agilent p/n 651750-902) for 60, 90, 120, and 150 min LC
Flow HPLC Mass gradients (Tables 1, 2, 3, and 4)
Spectrometry 2. Mobile Phase Buffer A: 0.1 % formic acid in water, for a 1 L
solution add 1 mL formic acid to a clean bottle and 999 mL of
water and mix thoroughly.
3. Mobile Phase Buffer B: 0.1 % formic acid in 90 % acetonitrile in
water, for a 1 L solution add 1 mL of formic acid to a clean
bottle, 900 mL acetonitrile, 99 mL of water, mix thoroughly.
4. Needle Wash: 0.1 % formic acid in 50 % (v/v) methanol in
water, for a 1 L solution, add 1 mL formic acid to a clean bot-
tle, add 500 mL of methanol, 499 mL water, and mix
thoroughly.
2.3 Discovery 1. Agilent 1260 Infinity HPLC-Chip/MS for use with all Agilent
and Targeted 6000 series Mass Spectrometers.
Proteomics Nanoflow 2. Q-TOF Internal Reference Mass (IRM) Low Mass Solution
HPLC-Chip Mass (1000 μg/mL methyl stearate in acetonitrile): Put 10 mg
Spectrometry methyl stearate into a 15 mL Falcon conical tube, add 10 mL
of acetonitrile, vortex to get methyl stearate into solution.
Table 1
60 min Gradient with the Agilent 1290 UPLC
LC conditions
Column AdvanceBio Peptide Mapping, 2.1 × 250 mm, 2.7 μm (Agilent p/n
651750-902)
Column temperature 50 °C
Injection volume 20 μL
Autosampler temp 4 °C
Needle wash 10 s in wash port (50:50 water:methanol with 0.1 % formic acid)
Mobile phase A = 0.1 % formic acid in water
B = 0.1 % formic acid in 90 % acetonitrile in water
Flow rate 0.40 mL/min
Gradient program Time, mins %B
0.0 3
52.0 35
55.0 70
57.0 70
58.0 3
Stop time 60.0 min
Post time 5.0 min
Table 2
LC conditions
651750-902)
0.0 3
82.0 35
85.0 70
87.0 70
88.0 3
Stop time 90.0 min
Post time 5.0 min
Table 3
LC conditions
651750-902)
Gradient program Time, min %B
0.0 3
110.0 40
115.0 70
117.5 70
118.0 3
Stop time 120.0 min
Post time 5.0 min
Table 4
150 min GRADIENT with the Agilent 1290 UPLC
LC conditions
651750-902)
0.0 3
140.0 40
145.0 70
147.5 70
148.0 3
Stop time 150.0 min
Post time 5.0 min
3. Q-TOF IRM, High Mass Solution: Add 1 mL acetonitrile to a

1.5 mL Eppendorf tube. Add 20 μL HP-1221 (Agilent p/n
G1982-85001) high mass compound, mix on a vortex mixer.
Both solutions should be refrigerated when not in use.
4. For nanoflow, typically a short column HPLC-Chip (ProtID-
Chip-43 (II), Agilent p/n G4240-62005, 43 mm 300 Å C18
chip with 40 nL trap column) is used for simple samples such
as 2D-gel spots or single protein digests. The Polaris-HR-Chip
3C18 (Agilent p/n G4240-62030, High Resolution Chip
150 mm 180 Å 3 μm C18 Chip with 360 nL trap column),
offers improved resolutions and peak capacities for peptides in
complex protein digests.
5. As an alternative the HPLC-Chip interface and columns, the
Agilent G1992A Nanospray Ion Source with gas distributor
(Agilent p/n G1964-20303) and spray shield kit (Agilent p/n
G1988-60007) can be used to interface any nanocolumn to
Agilent mass spectrometers.
2.4 Targeted 1. AdvanceBio Peptide Mapping Column, 2.1 × 100 mm, 2.7 μm

Proteomics Standard- (Agilent p/n 655750-902) for 60, 90, 120 min LC gradients.
Flow HPLC Mass (Tables 1–4)
Spectrometry 2. Mobile Phase Buffer A: 0.1 % formic acid in water, for a 1 L
solution add 1 mL formic acid to a clean bottle and 999 mL of
water and mix thoroughly.
3. Mobile Phase Buffer B: 0.1 % formic acid in 90 % acetonitrile in
water, for a 1 L solution add 1 mL of formic acid to a clean
bottle, 900 mL acetonitrile, 99 mL of water, mix thoroughly.
4. Needle Wash: 0.1 % formic acid in 50 % (v/v) methanol in water,
for a 1 L solution, add 1 mL formic acid to a clean bottle, add
500 mL of methanol, 499 mL water, and mix thoroughly.
2.5 Mass 1.
ESI-L Low Concentration Tuning Mix (Agilent p/n
Spectrometry: 6500 G1969-8500) (for Dual ESI or ESI with Jet Stream Technology).
Series Q-TOF and 6400 2. Glass Calibrant Delivery System (CDS) bottle (Agilent p/n
Series QQQ 9300-2576) and cap (Agilent p/n 9300-2575).
3. Tuning and calibrating with the ESI with Jet Stream Technology
on the 6500 series iFunnel Q-TOFs in positive ion mode, add
10 mL ESI-L to a clean CDS bottle, add 88.5 mL acetonitrile,
1.5 mL water, and 5 μL of 0.1 mM HP-0.321 (included in the
Biopolymer Reference Mass Kit (Agilent p/n G1969-85003).
Mix thoroughly and place in position Bottle B in the CDS.
4. Tuning and calibrating with the dual ESI on the 6500 series
iFunnel Q-TOFs in positive ion mode, add 25 mL ESI-L to a
clean CDS bottle, add 71.25 mL acetonitrile, and 3.75 mL
water. Mix thoroughly and place in position Bottle B in the
CDS.
5. Internal Reference Mass Solution for the ESI with Jet Stream
Technology on the 6500 Series Q-TOFs; ES-TOF Reference
Mass Solution Kit (Agilent p/n G1969-85001) containing
two ampoules (2.2 mL/ampoule) of the following reference
ions: 100 mM ammonium trifluoracetate (TFA-NH4) in 90:10
acetonitrile:water, 5 mM purine in 90:10 acetonitrile:water,
and 2.5 mM hexakis(1H, 1H, 3H-tetrafluoropropoxy)phosp-
hazine (HP-0921) in 90:10 acetonitrile:water (see Note 1). To
a 1 L Nalgene bottle, add 950 mL acetonitrile, 50 mL water,
0.4 mL purine, and 1.0 mL HP-0921. Cap and invert the bot-
tle several times to mix the reference solution. Pour 100 mL
into a CDS bottle and place onto Bottle A in the
CDS. Alternatively an isocratic pump can be used with the 1 L
stock solution with a 1:100 splitter (Agilent p/n G1607-
60000) connected to the reference nebulizer.
6. Internal Reference Mass Solution for the ESI on the 6500
Series Q-TOFs; ES-TOF Reference Mass Solution Kit (Agilent
p/n G1969-85001) containing two ampoules (2.2 mL/
ampoule) of the following reference ions: 100 mM TFA-NH4
in 90:10 acetonitrile:water, 5 mM purine in 90:10
acetonitrile:water, and 2.5 mM HP-0921 in 90:10
acetonitrile:water. To a 1 L Nalgene bottle, add 950 mL ace-
tonitrile, 50 mL water, 0.5 mL TFA-NH4, 1.0 mL purine, and
0.45 mL HP-0921 (see Note 2) Tuning and calibrating with
the ESI with Jet Stream Technology and ESI on the 6400
series QQQs in positive ion mode, add 100 mL ESI-L to a
clean CDS bottle and place in position Bottle B in the CDS.
2.6 Peptide 1. Bovine Serum Albumin (BSA) Stock: Prepare a 1 pmol/μL

Quantitation Checkout stock of trypsinized bovine serum albumin, BSA digest (Agilent
on the 6400 Series p/n G1900-85000) by adding 500 μL of 15 % acetoni-
QQQ trile/85 % water with 0.1 % formic acid to the standard (500
pmol per vial). This can be aliquoted into 0.5 mL Eppendorf
tubes and frozen.
2. BSA Dilution Solution: Prepare a 10 fmol/ μL solution by
diluting the BSA Stock 1:100 with 15 % acetonitrile/85 %
water with 0.1 % formic acid. For the levels shown below, you
should make 1 mL (10 μL of BSA Stock plus 990 μL of
solvent).
3. Human Serum Albumin (HSA) Stock: Prepare the stock solu-
tion of human serum albumin, HSA, peptides standard (Agilent
p/n G2455-85001) by adding 500 μL of the BSA Dilution
Solution. Vortex well to completely dissolve the standard. The
resulting stock solution is 1 pmol/μL and contains seven pep-
tides. Only one peptide will be used for the quantitation check-
out; however, optimization information for the other six is
shown in the appendix.
4. Prepare dilutions as shown in the table below by adding speci-

fied volumes of HSA standard and BSA Dilution Solution to
vials. (Best practice is to put the BSA Dilution Solution in the
vial first, then add the standard volume.) For convenience,
dilutions can be prepared directly in the conical bottom poly-
propylene autosampler vials (Agilent p/n 5190-3155) and
sealed with the appropriate caps (Agilent p/n 5182-0541).
5. For HPLC-Chip systems, use a range appropriate to the 6400
Series QQQ model. Typically your samples should cover six
orders of linear dynamic range (see Note 3).
6. Prepare a vial containing 60 % acetonitrile/40 % water with 1 %
trifluoroacetatic acid, TFA (or 50 % 2,2,2-trifluorethanol,TFE
in water) to clean the injector before doing low-level samples.
This solution works well for solubilizing hydrophobic pep-
tides. If a clean blank cannot be achieved after running several
injections of this solution, the best action is to remove the
needle seat (with seat capillary) and place it seat-side down in
a beaker with this solution, then sonicate for 5–10 min.
7. For the HPLC Chip Cube interface, the ProtID-Chip-43 (II)
(Agilent p/n G4240-62005) 43 mm × 0.075 mm chip with
40 nL enrichment column is used. For standard flow, use the
Eclipse Plus EC-C18 RRHD 2.1 × 50 mm, 1.8 μm column
(Agilent p/n 857750-902).
8. The mobile phases used for Channel A: 0.1 % formic acid in 3 %
acetonitrile in water (v/v) and Channel B: 0.1 % formic acid in
90 % acetonitrile in water (v/v) (see Note 4)
9. Commercial Targeted Proteomics Kits:
●● PeptiQuant™ MRM-MS Workflow Performance Kit: LC/
MRM/MS PEPTIQUANT WORKFLOW PERFORM
KIT FOR AGILENT 6495 1 RUN (Cambridge Isotope
Laboratories, Inc., Item #: WFPK-A6495-1).
●● PeptiQuant™ LC/MS Platform Performance Kit: LC/
MRM/MS PEPTIQUANT PLATFORM PERFORM. KIT
FOR AGILENT 6495/UPLC 1290 1WK SUPPLY
(Cambridge Isotope Laboratories, Inc., Item #:
LCMSP-D-A6495-1).
2.7 Data Analysis 1. Agilent G2721AA/G2733AA Spectrum Mill MS Proteomics

Workbench.
2. Agilent G3835AA MassHunter Mass Profiler Professional
Software.
3. Agilent G6825AA Pathway Architect.
4. Skyline(https://brendanx-uw1.gs.washington.edu/labkey/
project/home/software/Skyline/begin.view).
2.8 Instrumentation 1. 1290 UPLC, consisting of

●● G4220A 1290 Infinity Binary Pump.
●● G1330B 1290 Thermostat.
●● G4226A 1290 Infinity Autosampler.
●● G1316C 1290 Infinity Thermal Column Compartment.
2. 1260 Infinity HPLC-Chip/MS System, consisting of
●● G4225A 1260 HiP Degasser (quantity 2).
●● G1376A 1260 Infinity Capillary Pump.
●● G2226A 1260 Infinity Nanopump.
●● G1377A Infinity Autosampler.
●● G1330B 1290 Thermostat.
●● G4240A 1260 HPLC Chip Cube.
3. G6550AA 6550 iFunnel Q-TOF LC/MS.
4. G6495AA 6495 QQQ LC/MS.
3 Methods
All procedures are performed under room temperature. Biological

samples and hazardous chemicals should be handled with caution.
3.1 Protein Digestion 1. Dissolve sample in 50 % TFE in 50 mM ammonium bicarbon-
ate buffer to yield a 1.35 mg/mL solution (see Note 5) For
larger amounts of sample (such as the E. coli lysate), aliquot
100 μL per tube into 1.5 mL Eppendorf tubes (E. coli sample
yields 18 tubes; 135 μg per tube).
2. Add 2.5 μL DTT stock solution (200 mM) to each tube and
vortex to mix. Heat at 60 °C for 45 min.
3. Add 10 μL IAM stock solution (200 mM). Vortex briefly.
Allow to stand at room temperature for 45 min in the dark
(foil covered rack).
4. Add 2.5 μL DTT stock solution (200 mM) to remove excess
IAM. Allow to stand at room temperature for 30 min in the
dark.
5. Add 600 μL water and 200 μL ammonium bicarbonate to each
vial (Note: pre-mix this to reduce pipetting).
6. Add 6 μL trypsin stock solution at 1:20 or 1:50 enzyme:substrate.
Vortex briefly. Incubate overnight at 37 °C.
7. Add 4 μL neat formic acid or TFA to stop trypsin activity.
Vortex briefly. Digest is ready to analyze by LC/MS. Final
concentration should be ~150 ng/μL. For larger amounts of
sample, such as the quality assessment samples, it is convenient
to mix all the digest vials together (15 mL Falcon tube), then
aliquot to Eppendorf tubes (100 μL per vial) and store at
−80 °C (or −20 °C if you don’t have a −80 °C freezer).
3.2 Discovery 1. With the ESI with Jet Stream Technology source [3] on the
Proteomics: Data- Q-TOF, in MassHunter Acquisition, go to the Tune page.
Dependent Acquisition Tune and calibrate the Q-TOF in the Extended Mass Range,
(DDA) on an Agilent 2 GHz at mass range m/z 3200.
6500 Series Q-TOF 2. Select Quadrupole Tune and perform a Quad Tune.
3. Once the tune is completed, change the mass range to m/z
1700 and select “Apply”.
4. Once the 20 min equilibration is complete, select m/z 118 to
m/z 1622, and select calibrate the TOF.
5. Save the Tune File before switching back to acquisition
window.
6. In the acquisition window, generate the following acquisition
method for standard flow with the ESI with Jet Stream
Technology source:
Liquid Chromatography Gradient: Table 2.
Source: Table 5.
MS Acquisition: Table 6.
7. Save method as “DDA_90min_AJS.m”.
8. In the acquisition window, generate the following acquisition
method for nanoflow with the HPLC-Chip source [4, 5]:
Liquid Chromatography Gradient: Table 7.
Table 5
Source conditions for the 6500 series QTOF with the ESI with Agilent Jet
Stream source
Ion source/mode Agilent Jet Stream, positive

Gas temperature 250 °C
Drying gas flow 14 L/min
Nebulizer gas 35 psi
Sheath gas temperature 250 °C
Sheath gas flow 11 L/min
Capillary voltage 3500 V
Nozzle voltage 0 V
Fragmentor 360 V
Reference mass m/z 322.0481 and 1221.9906
Table 6
Discovery proteomics acquisition method for the 6500 series QTOF
Parameter Setting
Acquisition mode Extended dynamic range (2 GHz), high sensitivity, low mass range
m/z 1700
Mass range m/z 300–1700
Acquisition rate/time 8 spectra/s
Auto MSMS range m/z 50-1700
MSMS acquisition rate/time 3 spectra/s (max)
Isolation width Narrow (~1.3 Hz)
Precursors/cycle Top 20
Collision energy 3.6*(m/z)/100–4.8
Threshold for MSMS 1000 counts and 0.001
Dynamic exclusion On; 1 repeat then exclude for 0.2 min
Precursor abundance-based scan Yes
speed
Target 25,000
Use MS/MS accumulation time Yes
limit
Purity 100 % stringency, 30 % cutoff
Isotope model Peptides
Sort precursors By abundance only; +2, +3, > + 3
Source: Table 8.
MS Acquisition: Table 9.
9. Save method as “DDA_130min_Chip.m.
3.3 Discovery 1. Move your DDA data from the QTOF to a new folder under
Proteomics: Data the “smdata” folder on your Spectrum Mill server. Typically
Analysis Using data is organized in subfolders under smdata to organize proj-
Spectrum Mill (Fig. 3) ects (see Note 6).
2. Open Internet Explorer and load Spectrum Mill, navigate to
the Data Extractor page.
3. In the Data Directories section, click the Select… button to
select the folder or folders that contain your files.
4. For Agilent Q-TOF data and most extractions, adjust the
parameters as outlined below:
●● Click the Choose… button to select the
“Carbamidomethylation (C) “modification”.
Table 7
Source and LC gradient conditions for the 6500 series QTOF with the HPLC Chip
LC conditions HPLC-Chip, positive

Column Agilent Polaris-HR-Chip-3C18 with a 360 nL enrichment
and 150 mm × 0.075 mm analytical column
Injection volume Adjusted to load 1 μg of total protein per injection
Needle wash 10 s in wash port (50:50 water:methanol
with 0.1 % formic acid)
Flow rate Loading: 2 μL/min with 3 % B
Analytical: 300 nL/min
Gradient program Time, min %B
0.0 3
90.0 25
120.0 40
125.0 90
130.0 90
130.1 3
Stop time 133.0 min
Post time 5.0 min
Table 8
Source conditions for the 6500 series QTOF with HPLC Chipcube
Ion source/mode HPLC ChipCube, positive

Gas temperature 250 °C
Drying gas flow 11 L/min
Capillary voltage 1800–1950 V
Fragmentor 360 V
Reference mass 299.2944 and 1221.9906
●● MS/MS Spectral Feature Finding:

MH+ 600.0-6000.0 Da.
Scan Time Range: 0–300 min.
Sequence tag length >0.
Ignore spectra with dissociation mode: Disable both CID
and ETD.
●● Merge nearby MSn scans with same precursor m/z:
Table 9
Discovery proteomics acquisition method for the 6500 series QTOF with HPLC-Chip
Parameter Setting
Acquisition mode Extended Dynamic Range (2GHz), High Sensitivity, Low Mass Range m/z
1700
Mass range m/z 275–1700 (MS) and 50–1700 (MS/MS)
Acquisition rate/time 8 spectra/s
Auto MSMS range m/z 50–1700
MSMS acquisition rate/ 3 spectra/s (max)
time
Isolation width Narrow (~1.3 Hz)
Precursors/cycle Top 20 precursors per cycle using precursor abundance-based acquisition
rate with accumulation time limit enabled; active exclusion after one
spectrum for 0.5 min
Collision energy 3.6*(m/z)/100–4.8
Threshold for MSMS 1000 counts and 0.001
Dynamic Exclusion On; 1 repeat then exclude for 0.2 min
Precursor abundance- Yes
based scan speed
Target 25,000
Use MS/MS Yes
accumulation time
limit
Purity 100 % stringency, 30 % cutoff
Isotope model Peptides
Sort precursors By abundance only; +2, +3, > + 3
Retention time & m/z tolerance: ±45 s, ±1.4 m/z

General MS/MS Merging Constraints: Select “Spectral
Similarity & RT & m/z” from the pull down menu.
●● Precursor m/z and Charge Assignment:
Precursor Charge: Select “Find”.
Maximum (z): 7.
Minimum MS1 S/N: 25.
Find 12C precursor m/z: Enabled.
MS Noise threshold: 400.
5. Navigate to MS/MS Search. Adjust the parameters as outlined
below:
Fig. 3 Spectrum Mill workflow
●● Select the data directory where the Data Extraction was

performed.
●● Search Parameters:
Validation filter: Select “spectrum-not marked-sequence-
not-validated”.
Batch size: 150.
Search previous hits: Disabled.
Max reported hits: 5.
Database: Select the appropriate database for your
sample.
Digest: Select “Trypsin”.
Species: Select “All”.
Maximum # missed cleavages: 2.
Modifications: Select “Carbamidomethylation (C)”.
Search Criteria: Matching Tolerances.

Minimum matched peak intensity: 50 %.
Masses are: Select “Monoisotopic”.
Precursor mass tolerance: ±20 ppm.
Product mass tolerance: ±40 ppm.
Maximum ambiguous precursor charge: 3.
Search Criteria: Spectral Quality.
Sequence tag length: Disabled.
Minimum detected peaks: 4.
Search Criteria: Search Mode.
Calculate revered database scores: Enabled.
Dynamic peak thresholding: Enabled.
Discriminant scoring: Select “Disable (same as Score).
Search mode: Select “Variable modifications” and add the
appropriate modifications for your sample such as
phosphorylated S, T, or Y. Typical sample handling
modifications can be included such as oxidized M and
deamidated N.
Search Criteria: Data Files.
Fragmentation mode: Select “All”.
Spectrum files (./cpick_in/): Enter “*.pkl”.
6. It is recommended that you require your peptide spectral
matches meet a specified false discovery rate (FDR) by navigat-
ing to MS/MS Autovalidation, then selecting the following
parameters:
●● Select the data directory where the Data Base Search was
performed.
●● Strategy: Auto thresholds.
●● Mode: Peptide.
●● Optimize score and R1-R2 score thresholds with max
FDR: 1.2 % across each LC Run.
●● Precursor charge range: 2–4 (see Note 7).
●● Min sequence length: 6.
●● Required AAs: any.
●● Disallowed AAs: none.
●● Filtering: None (ppm) and None (SC/pI).
7. To view your search results and prepare for cross-sample com-
parison, navigate to Protein/Peptide Summary.
8. Adjust the parameters as outlined below:
●● Select the data directory or directories where the Data

Base Search was performed.
●● Summarize Results for Review:
Select MPP Generic Export from pulldown menu.
Mode: Select “Protein-Protein Comparison”.
Validation and Sorting:
Filter results by: Select “valid”.
Validation preset: Select “none”.
Protein grouping method: Select “1 shared, expand
subgroups”.
Sort proteins by: Select “Score”.
Filter by protein score: Select “0”.
Sort proteins by: Select “Score”.
Filter peptides by:
Score: Select “>0”.
% SPI: Select “>0”.
Required AAs: Select “any”.
Disallowed AAs: Select “none”.
●● Review Fields:
Enable the following fields for review:
Filename.
Score.
Subgroup specific (see Note 8).
Accession #.
Protein name.
Intensity: Total.
Protein Quantitation Options: Enable Exclude isotope
quality Precursor XICs (<0.85 Chi2 vs. Averagine).
●● Select Summarize. The exported file can be found under
the data subdirectory (or the first of the subdirectories
listed if more than one directory was used). Copy this file
to a known location for subsequent analysis.
3.4 Discovery Data Analysis Using Mass Profiler Professional (MPP) to perform
and Targeted statistical and correlation analysis between samples groups. Using
Proteomics the optional Pathway Architect module allows the results to be
mapped to publically available biological pathways databases along
with metabolomics and gene data, if available.
1. Open MPP and create a new project.
2. Create a New Experiment within MPP.
3. Use the following parameters for discovery data:

●● Analysis type: Mass Profiler Professional.
●● Experiment type: Identified.
●● Workflow type: Data Import Wizard.
●● Click OK, then select data source: Generic and specify the
organism used.
●● Click Next, then select the generic file exported from
Spectrum Mill.
●● Click Next to confirm that the appropriate samples are
present, then click Next again.
●● For Experimental Groupings, you need to select “Add
Parameter”, then specify the Parameter Name. For each
sample, specify the group. To perform statistics, a mini-
mum of n = 3 is required. Click OK after all information is
added. Another parameter may be entered at this time if
you have multiple parameters you wish to evaluate.
●● Click Next to align which is done by the protein name
across all samples and groups.
●● Click Next to specify Normalization Criteria if desired.
Typically this is not used in proteomics.
●● Click Next to select Baselining Options. Z-transform can
be used when samples from different sources are used.
Baseline to median of all samples works well when the sam-
ples are very similar, which is often the case with pro-
teomics samples.
●● Under the Workflow: Experiment Setup, select Create
Interpretation. Select the experiment parameter of interest
and click Next, then Next to create the Interpretations
(Averaged and non-averaged) for that parameter.
●● Under Workflow: Navigate to the Quality Control menu,
Select Filter by Abundance. Select All Entities for the
Entity List and your non-averaged interpretation. Select
Next to input parameters. Change the lower cutoff to 2 as
MPP changes all zero values (where a protein is not found)
into a 1 when importing the data. Specify the required
reproducibility. Depending on the number of samples, you
can adjust the % of samples in any group that must have
values greater than 1. Select Next to view the results, then
Next to save the new filtered entity list.
●● Correlation analysis or statistical analysis can now be per-
formed on this filtered entity list. Data can be viewed using
a variety of tools including hierarchical clustering, principal
component analysis etc.
●● After analysis is completed, the processed entity list can be

mapped to available pathways using the optional Pathway
Architect module (see Note 9) Under Workflow: Navigate
to Pathway Analysis, select Single Experiment Analysis.
Select the pathway source of interest and click Next. After
selecting the interpretation and entity list, click Next. After
reviewing the results, select Next to save the pathway
results. Selecting any pathway of interest from the list will
display the pathway. The relative abundances of each
experimental group will be displayed as a heatstrip next to
the protein entry on the pathway.
●● From the pathway view, all proteins, metabolites and/or
genes in the selected pathway can be exported. To export
the SwissProt list, right click in the pathway view and
choose Select All Entities. Then right click again and select
Copy to Clipboard as SwissProt/UniProt IDs. This can be
pasted into Spectrum Mill or Skyline software for pathway
directed targeted analysis (see Note 10).
3.5 Targeted 1. Create a method named “HSA_BSA_Checkout_AJS.m” with

Proteomics: HSA the following:
in BSA Checkout
Autosampler:
on the 6400 Series
QQQ using Standard Injection volume: 1 μL
Flow with the AJS
Needle wash: Enabled for 10 s in flushport
Needle flush solvent: 50 % methanol / 50 % H2O with
0.1 % formic acid
Bottom sensing: On
Vial offset: 0
ALS Therm: On, 4 °C
Binary pump:
Solvents: A1: water with 0.1 % formic acid

B1: 90 % acetonitrile / 10 % water with
0.1 % formic acid
Flow: 0.6 L/min
Max 600 bar
pressure:
Stop time: 3.5 min
Post time: 0 min
Gradient: Time %B
0 5
1 25
1.5 70
1.6 5
Column temperature: 35 °C

QQQ MRM Method:
Ionization mode: Positive Agilent

Jet Stream,
positive
Time filter: ON
Time Filter Width: 0.03
Dry gas flow: 15 L/min.
Dry gas temp: 150 0 C
Sheath gas flow: 11 L/min
Sheath gas temp: 200 °C
Nebulizer: 30 psig
Nozzle voltage: 0
Capillary Voltage: 3500 V is typical
Time segments: 1
Delta EMV: 0–200 V
MS1 Res/MS2 Res: Wide/unit
Dwell: 60 ms
Fragmentor: 380 V
Cell Accelerator Voltage: 4 V
Polarity: Positive
MRM scan segments: (Precursor Ion, Product Ion, Collision

Energy)
m/z 575.3111, m/z 937.4625, 16 V.
m/z 575.3111, m/zm/z 823.4196, 20 V.
m/z 575.3111, m/z 694.3770, 20 V.
2. Create a worklist with a BSA Blank as the first entry followed
by Sample H to A, where Sample A is the last sample injected.
Use the above method for all entries with a 1 μL injection vol-
ume. For the sample type, use “Blank” for the BSA blank and
“Calibration” for the HSA spiked into BSA. For the Level for
Table 10
HSA in BSA dilution table for 6400 Series QQQ checkout
Sample Final concentration Volume of HSA standard Volume of BSA dilution solution (μL)
A 100 fmol/μL 10 μL of HSA Stock 90
B 10 fmol/μL 10 μL of A 90
C 1 fmol/μL 10 μL of B 90
D 100 amol/μL 10 μL of C 90
E 10 amol/μL 10 μL of D 90
F 5 amol/μL 20 μL of E 20
G 2 amol/μL 10 μL of E 40
H 1 amol/μL 10 μL of E 90
each sample enter 1 for Sample H and go up in increments of

1 till you get to Sample A with at Level 8 (Table 10).
3.6 Targeted 1. Create a method named “HSA_BSA_Checkout_Chip.m” with

Proteomics: HSA the following parameters:
in BSA Checkout
Nanoflow pump:
on the 6400 Series
QQQ using Standard
Flow
with the HPLC-Chip Solvents: A1: water with 0.1 % formic acid
B1: 90 % acetonitrile / 10 % water with
0.1 % formic acid
Flow: 600 nL/min (primary flow 500 μL/min)
Max pressure: 200 bar
Stop time: 9 min
Post time: 0 min
Gradient: Time %B
0 3.0
7 70.0
7.1 3.0
Capillary pump:
Solvents: A1: water with 0.1 % formic acid

B1: 90 % acetonitrile / 10 % water with 0.1 % formic acid
Flow: 3.00 μL/min (primary flow 200 μL/min)
Max pressure: 200 bar
Micro wellplate sampler:
Injection volume: 1 μL

Needle wash: Enabled for 10 s in flushport
Needle flush solvent: 20 % methanol / 80 % water with
0.1 % formic acid
Bottom sensing: On
Vial offset: 0
ALS Therm: On, 4 °C
HPLC-Chip MS Interface:
Injection flush volume: 4 μL.

Pumps configured so Intelligent Sample Loading is activated.
Chip Cube Timetable: Valve to enrichment at 7.5 min.
QQQ MRM:
Ionization mode: HPLC-Chip, positive

Time filter: On
Time Filter Width: 0.07
Dry gas flow: 11 L/min.
Dry gas temp: 150 °C
Capillary Voltage: 1750–1900 V is typical
Time segments: 1
Delta EMV: 200 V
MS1 Res/MS2 Res: Wide/unit
Dwell: 60 ms
Fragmentor: 380 V
Cell accelerator voltage: 4 V
Polarity: Positive
MRM scan segments: (Precursor Ion, Product Ion, Collision

Energy).
m/z 575.3111, m/z 937.4625, 16 V.
m/z 575.3111, m/z 823.4196, 20 V.
m/z 575.3111, m/z 694.3770, 20 V.
2. Create a worklist with a BSA Blank as the first entry followed
by Sample H to A, where Sample A is the last sample injected.
Use the above method for all entries with a 1 μL injection vol-
ume. For the sample type, use “Blank” for the BSA blank and
“Calibration” for the HSA spiked into BSA. For the Level for
each sample enter 1 for Sample H and go up in increments of
1 till you get to Sample A with at Level 8 (Table 10).
3.7 Targeted 1. For initial performance checkout of the QQQ for daily or
Proteomics: monthly assessment of the LC/MS platform, use the
Commercial Kits PeptiQuant™ LC/MS Platform Performance kit. Each
for Quality Control Platform Performance kit includes seven tryptic-digested
Using Standard Flow plasma samples spiked with stable-isotope labeled standards
on the 6400 Series (SIS). These samples are rehydrated and ready for use with the
QQQ [6, 7, 8] LC/MRM/MS. Protocols are included within each kit.
2. For quality control of the bottom-up workflow (from denatur-
ation through to detection), use the PeptiQuant™ MRM-MS
Workflow Performance Kit. Each Workflow Performance kit
contains: human plasma, trypsin, and a SIS peptide mixture.
The protocol is included with the kit providing the detailed
procedure for the sample preparation. The protocol describes
the sequential reduction, alkylation, and quenching of the 10×
diluted plasma prior to digestion overnight with trypsin.
Digestion is stopped by the addition of the chilled SIS peptide
mixture from 250 to 0.025 fmol/μL for standard samples G to
A (A being the lowest concentration) and chilled formic acid
solution. The samples are allocated into separate Eppendorf
tubes, centrifuged, and the peptide supernatant is removed for
desalting. The desalted supernatant is lyophilized and rehy-
drated in 0.1 % formic acid for a final concentration of 1 μg/μL
for LC-MRM/MS analysis.
4 Notes
1. Before you break open each ampoule, invert the ampoule sev-
eral times to mix. Inspect the ampoule’s contents to ensure
that all the solution is contained in the lower cylindrical base.
Shake the ampoule, if needed, to dislodge any air pocket that
may prevent solution from settling in the lower portion of the
ampoule.
2. Cap and invert the bottle several times to mix the reference
solution. Pour 100 mL into a CDS bottle and place onto
Bottle A in the CDS. Alternatively an isocratic pump can be
used with the 1 L stock solution with a 1:100 splitter (Agilent
p/n G1607-60000) connected to the reference nebulizer.
3. System cleanliness is the biggest challenge as this peptide will
exhibit some carryover.
4. Best practice is to use fresh mobile phases for studies. Purge

both channels when mobile phases are changed.
5. The solution may appear opaque prior to digestion. Mix well
to ensure homogeneity before proceeding.
6. Organizing data in subfolders can be important for subsequent
database searching.
7. You should have at least 200 peptide spectra matches for all
charge states in the specified range for the FDR calculation.
8. This summarizes the protein abundance only from peptides
unique to that protein subgroup thus ensuring that peptides
which occur in more than one protein are not used.
9. Pathways must be downloaded from publically available
sources using the Annotations: Import Pathways menu prior
to doing pathway analysis.
10. Alternatively, protein database search results from Spectrum
can also be exported as a peptide spectral library for import
into Skyline. The user can then use Skyline to select target
proteins with corresponding transitions for direct import into
the 6400 series QQQ for validation and optimization on the
QQQ. The user should review the manuals and training videos
provided by Skyline for familiarization.
References
1. Wong CC, Cociorva D, Miller CA et al (2013) proteomics results. Agilent Technical Overview
Proteomics of Pyrococcus Furiosus (Pfu): identifi- 5991-0735EN http://www.agilent.com/cs/
cation of extracted proteins by three indepen- librar y/technicalover views/Public/5991-
dent methods. J Proteome Res 12(2):763–770 0735EN.pdf Accessed 31 August 2015
2. Vaudel M, Burkhart JM, Breiter D et al (2012) 6. Percy AJ, Chambers AG, Borchers CH, (2014)
A complex standard for protein identification, Application kits for standardizing MRM-based
designed by evolution. J Proteome Res quantitative plasma proteomic workflows on
11(10):5065–5071 the Agilent 6490 LC/MS system. Agilent
3. Yang Y, Bhat V, Miller CA. (2015) Jet Stream Application Note 5991-3601EN https://
proteomics for sensitive and robust standard w w w. a g i l e n t . c o m / c s / l i b r a r y /
flow LC/MS. Agilent Technical Overview applications/5991-3601EN.pdf Accessed 31
5991-5687EN http://www.agilent.com/cs/ August 2015
librar y/technicalover views/public/5991- 7. Percy AJ, Chambers AG, Yang J et al (2012)
5687EN.pdf Accessed 31 August 2015 Comparison of standard-and nano-flow liquid
4. Miller CA, Jenkins S, Sana TR, et al. (2013) chromatography platforms for MRM-based
Proteomics in multi-omics workflows using quantitation of putative plasma biomarker
yeast as a model system. Agilent Application proteins. Anal Bioanal Chem
Note 5991-2484EN https://www.agilent. 404(4):1089–1101
com/cs/library/applications/5991-2484EN. 8. Percy AJ, Mohammed Y, Yang J, Borchers CH
pdf 31 August 2015 (2015) A standardized kit for automated
5. Buckenmaier S, Mora J, van de GoorT, et al. (2012) quantitative assessment of candidate protein
Enhanced chromatography with the Agilent biomarkers in human plasma. Bioanalysis,
Polaris-HR-Chip-3C18 improved LC/MS/MS 7(23):2991–3004
Chapter 16
Generation of High-Quality SWATH® Acquisition Data

for Label-free Quantitative Proteomics Studies Using
TripleTOF® Mass Spectrometers
Birgit Schilling, Bradford W. Gibson, and Christie L. Hunter
Abstract
Data-independent acquisition is a powerful mass spectrometry technique that enables comprehensive MS
and MS/MS analysis of all detectable species, providing an information rich data file that can be mined
deeply. Here, we describe how to acquire high-quality SWATH® Acquisition data to be used for large
quantitative proteomic studies. We specifically focus on using variable sized Q1 windows for acquisition of
MS/MS data for generating higher specificity quantitative data.
Key words Mass spectrometry, SWATH acquisition, Data-independent acquisitions, Variable win-
dows, Quantitation, Proteomics
1 Introduction
The goal of quantitative proteomics is to both identify and quan-

tify a broad range of peptides and proteins. The extreme complex-
ity and dynamic range of proteins in cells, tissues, and fluids
challenges traditional data-dependent workflows to reproducibly
and deeply interrogate complex samples. There has been resur-
gence in the use of data-independent acquisition (DIA) mass spec-
trometry strategies to increase the reproducibility and
comprehensiveness of data collection, enabled by recent techno-
logical innovations in both MS hardware and software. This acqui-
sition strategy can now be routinely applied to proteomic samples
to collect high quality quantitative data for high numbers of pep-
tides/proteins. These data-independent acquisitions are widely
performed on orthogonal time-of-flight (QqTOF or TripleTOF®)
platforms, and the workflow is referred to as SWATH Acquisition
[1–4]. However, DIA acquisitions can also be executed on other
platforms, such as on Q-Exactive mass spectrometers [5]. In
SWATH Acquisition on TripleTOF platforms, Q1 isolation
223
224 Birgit Schilling et al.
windows are stepped across the mass range in an LC timescale,

transmitting populations of peptides for fragmentation, and high-
resolution composite MS/MS spectra are acquired at each step.
Gillet et al. [2] described their initial SWATH experiments setting
the Q1 quadrupole to transmit Δ25 m/z windows, with a 1 m/z
overlap between windows. The MS/MS spectra generated during
a SWATH acquisition are of much greater complexity than the
MS/MS spectra in typical data-dependent acquisition (DDA)
experiments when using unit resolution (1 m/z) Q1 windows. In
addition, an MS1 scan is typically collected in each cycle, and this
cycle of MS1 and SWATH MS/MS scans is repeated throughout
the entire LC-MS acquisition. Fragment ion information from the
obtained SWATH MS/MS spectra can be used to uniquely con-
firm detection of specific peptides, typically by comparisons to ref-
erence spectra or spectral libraries as described by Gillet et al. [2]
and Röst et al. (Chapter 20).
The complexity of the chimeric SWATH MS/MS spectra and
the resulting specificity depends on the number of peptides eluting
off the column at the same retention time within the same m/z
isolation window. Initial approaches used fixed sized Q1 windows
of 20–25 m/z; however, more recent work has shown that decreas-
ing the Q1 window size, as well as varying the window width as a
function of precursor ion density [6] can provide higher specificity,
less interferences and improvements in the amount of quantitative
data extracted to ultimately increase the depth of coverage.
Data-independent SWATH acquisitions are compatible with
several downstream data processing algorithms, and recently a vari-
ety of new software tools have emerged. One widely used approach
is to match SWATH data against pre-assembled MS/MS spectral
libraries, using tools such as OpenSWATH [7], Skyline [5, 8],
SWATH 2.0 [9], and Spectronaut [10]. These tools analyze the
untargeted DIA acquisition data with targeted data processing tak-
ing advantage of existing spectral libraries. Another strategy,
MSPLIT-DIA (or mixture-spectrum partitioning using libraries of
identified tandem mass spectra) is a spectral matching tool for
untargeted peptide identification in DIA data [11]. Lastly, two
recently published algorithms DIA-Umpire [12], and Group-DIA
[13] apply an untargeted data analysis approach in which SWATH
or DIA data is first processed by detecting precursor and fragment
ion features and subsequently assembling them into pseudo-
tandem MS/MS spectra, which then can be searched by any typical
database search engine.
SWATH Acquisitions Using TripleTOF Mass Spectrometers 225
2 Materials
2.1 Samples 1. Retention Time Calibration Peptides: The iRT peptides (kit
for Analysis P/N Ki-3003, Biognosys, Switzerland) are spiked into com-
plex samples, typically at dilutions of 1/10 (see Note 1).
2. LC-MS quality control (QC) sample: Beta-Galactosidase (BGal)
Stock Solution. Add 625 μL of buffer (10 % acetonitrile/0.1 %
formic acid) to the BGal vial (kit P/N 4465867, SCIEX) to cre-
ate a 1 pmol/μL stock solution. Aliquot the stock solution into
10–50 μL volumes and freeze at −20 °C for future use.
3. SWATH system suitability sample: pre-digested cell lysate simi-
lar to the study sample described under step 4 to assess SWATH
performance.
4. Pre-digested Cell Lysate samples: a typical study sample used
here consists of a tryptic digestion of a yeast cell lysate.
Proteomic samples were previously reduced, alkylated,
digested, and desalted, providing a solution of ~1 μg/μL
digested protein lysate for analysis.
2.2 Chromatography 1. LC System—nanoLC™ 425 System (SCIEX) combined with a

cHiPLC® System (see Note 2).
2. Two analytical cHiPLC columns set up in serial mode—each
75 μm × 15 cm ChromXP™ C18-CL chip, 3 μm, 300 Å (P/N
804-00001, SCIEX).
3. LC Buffer A: 98 % water with 0.1 % formic acid and 2 %
acetonitrile;
4. LC Buffer B: 98 % acetonitrile with 0.1 % formic acid and 2 %
water.
2.3 Mass 1. MS analysis is performed using the NanoSpray® Source on the

Spectrometry TripleTOF® 6600 System (see Note 3) operating with Analyst®
Software TF 1.7 (SCIEX).
2.4 Data Processing 1. SWATH® Acquisition data is processed with the SWATH
Acquisition MicroApp 2.0 in PeakView® Software [9], using a
spectral ion library [4] generated from prior data-dependent
acquisitions (see Note 4).
3 Methods
3.1 Set 1. Using the nanoLC 425 LC system with a cHiPLC System
Up LC-MS System (SCIEX), set up in standard ‘trap elute’ mode, to desalt first
then separate the peptides from the complex tryptic digestions.
Set up the autosampler (AS-3) valve with a 10 μL sample loop,
and connect to the loading pump, and the cHiPLC system.
Connect the analytical gradient to the cHiPLC system, which

is set up with two analytical C18 chip columns in serial mode
(see below and see Note 5 and 6).
2. Place a trap elute jumper chip (PN 800-00389, SCIEX) in slot
1 in the cHiPLC system. Place two analytical chips in slots 2
and 3 (SCIEX), (see Note 7). Set the temperature on the
cHiPLC system to 35 °C for both to maintain a constant tem-
perature (see Note 8).
3. Connect the eluant from the column to the nanoflow spray tip
(New Objectives FS 360-20-10-N-20) mounted on the
NanoSpray Ion Source (SCIEX). The source is connected to a
TripleTOF 6600 system operated using Analyst Software TF
1.7 (SCIEX).
3.2 Create a Basic 1. Click Build Acquisition Method to start a new method. Set
LC-MS experiment 1 to TOF MS and set the m/z range to 400 – 1500
Acquisition Method and the Accumulation Time to 250 msec. Next, open the Edit
Parameters and set the source conditions that were optimized
for the system, these should typically be in the following ranges.
On the Compound tab, ensure the Declustering Potential (DP) is
set to between 80 and 100 V. On the Source/Gas tab, typical
conditions will be 2300–2600 V for the IonSpray Voltage (IVSF),
3-6 for the Ion Source Gas 1 (GS1), 0 for the Ion Source Gas 2
(GS2), 20–25 for the Curtain Gas (CUR), and 100–150 for the
Interface Heater Temperature (IHT).
2. Create a trap loading method for desalting and add this to the
LC-MS acquisition method; for example using a loading pump
flow rate of 0.5 μL/min of 100 % Buffer A for 30 min (see Note 6).
3. Create a μL-pickup method for the autosampler, injecting 2 μL
of sample using a 10 μL loop and add this to the LC-MS acqui-
sition method.
4. Create the analytical gradient method and add this to the
LC-MS method. An example gradient for a complex sample is
to use a flow rate of 300 nL/min and the following gradient:
5–30 % (v/v) solvent B with solvent A making up the differ-
ence (from 0 to 120 min), 30–80 % solvent B (from 120 to
130 min), and at 80 % solvent B (from 130 to 136 min) then
back to 5 % solvent B from 136 to 138 min for mobile phase
equilibration, with a total run-time of 155 min (see Note 9).
5. Set the MS acquisition Duration to 150 min. Save the LC-MS
acquisition method.
3.3 Convert 1. Click on the Create SWATH Exp button in the top right hand
to a SWATH side of the LC-MS acquisition method window. The Create
Acquisition Method SWATH Experiment UI appears (Fig. 1). Click on the ‘Manual’
Tab of the UI on the right to set up a custom SWATH
experiment.
Fig. 1 Build a Variable Window SWATH Method. Any window strategy can be constructed in text file format and
loaded into the SWATH Acquisition method editor for automatic method building. (a) A text file describing the
Q1 isolation window strategy can be constructed. (b) Load into SWATH Acquisition method editor and adjust
the MS and MS/MS settings as shown. (c) Click OK to automatically build a SWATH Acquisition method
Prepare the SWATH Variable Window text file that defines the
Q1 isolation window strategy as shown in Fig. 1a (see Note 10).
Generate the Variable Window Table manually in Excel or
compute from previously acquired MS data on the specific
sample of interest (see Note 11). The Variable Window excel

file should be set up to define the varying windows width across
the entire m/z range with 3 Excel columns ‘Q1 Start m/z’,
‘Q1 Stop m/z’, and ‘collision energy spread, CES’ (see Fig. 1a).
Save the document as a *.txt file. The variable window text file
is imported into the LC-MS acquisition method (described
below in step 2 of Subheading 3.3), to build the Variable
Window SWATH Acquisition Method in the Analyst software.
2. From the Manual tab of the Create SWATH Experiment UI
(see Fig. 1b) check the Read SWATH Windows from Text File
option at the very bottom of the UI and load the SWATH
Variable Window text file using the Browse button. Note that
the SWATH Analysis Parameters Section (top of window) is
grayed out and not used once the Read SWATH Windows from
Text File option is selected.
3. Next, under the Fragmentation Conditions Section click on
the Rolling Collision Energy option and set the charge state to
2+ (Fig. 1b). The collision energy (CE) equations are defined
in an Analyst TF 1.7 Script called ‘IDA CE Parameters’.
Software defaults can be used or SWATH optimized CE equa-
tions can be downloaded (http://sciex.com/community/
entity/11856).
4. Finally, set the SWATH Detection Parameters section to the
following values, MS/MS scans acquired from m/z 100–1500
with 25 msec accumulation times in high MS/MS sensitivity
mode, as shown in Fig. 1b (see Note 10). Note the overall
cycle time will not be computed until after you click OK.
5. Click OK at the bottom to build the LC-MS acquisition
method (Fig. 1c). The method shows the TOF MS scan
(experiment 1), followed by all 100 Product Ion scans (vari-
able Q1 window MS/MS scans—experiments 2–101).
6. After method building, click on the TOF MS experiment (Exp
1) defining the TOF MS1 scan, and confirm that the mass
range is set to m/z 400–1500 and that the accumulation time
is 250 msec. This ensures the MS1 scan is also of high quality
for future use [14].
7. The total cycle time in this case is 2.8 s which allows enough
measured points across an eluting peak with the given chro-
matographic setup (see Note 12). This completes the building
of the Variable Window SWATH Acquisition method. For
information on creating a fixed window acquisition method
(see Note 13).
3.4 Creating 1. In order to obtain consistent and reliable data from study sam-
a Practical SWATH ples, regular LC-MS tests need to be performed before and
Study during the entire SWATH study. Initially, use the pre-digested
Acquisition Batch BGal standard (see Note 5) and perform QC acquisitions, or
mass calibration acquisitions, that are typically used in your

laboratory.
2. In addition to the simple LC-MS system QC checks discussed
above, also use complex standard samples similar to the study
samples and test the generated SWATH acquisition method
and performance of the system in SWATH mode. Perform
before starting the study and also intersperse such SWATH
specific tests throughout the entire SWATH study. In addition,
retention time calibration (iRT) peptides spiked into the
SWATH test samples and the study samples by themselves can
be used to monitor retention time stability (see Note 14).
3. Randomize study samples to avoid systematic errors in the
study, block-randomization of biologically different samples
often is applied in proteomics studies.
4. Acquire data for the study samples with the acquisition meth-
ods generated above, injecting 2 μL of samples at concentra-
tions of ~0.5–1.5 μg/μL yielding an amount of ~1–3 μg of
sample on column.
3.5 Data Processing Once the SWATH data is acquired, several data processing pipe-
lines can be used to process the data (see Note 4). As an example,
a SWATH data set was acquired using the above methods and a
number of different Q1 window strategies, then processed using a
spectral library approach as previously described for the SWATH
2.0 algorithm by Lambert et al. [9], also referred to as targeted
extraction. Figure 2 highlights that higher numbers of peptides are
robustly identified and quantified when more, smaller (variable)
Q1 windows are used for SWATH acquisitions [6].
4 Notes
1. iRT peptides can also be spiked at 1:20 dilution instead,

depending on study sample and preference.
2. Different nano-flow LC systems can be used, such as NanoLC
Ultra® 2Dplus (or 1D) HPLC system (Eksigent), or nano-flow
LC systems from other vendors.
3. Method setup and acquisitions can similarly be performed on
a TripleTOF 5600 or 5600+ system. All steps described will
equally apply, as all of the TripleTOF systems use the same
acquisition software, Analyst TF Software 1.7.
4. Other SWATH data processing software can be used as dis-
cussed in the introduction, i.e., OpenSWATH, Skyline,
Spectronaut, DIA Umpire, and others. Please refer to the

Chapter by Röst et al. in this book for a detailed description of
data processing.
25000
Pepdes with Equal or Beer CV

20000
15000
100 VW
10000
80 VW
60 VW
5000
0
0 10 20 30 40 50
% CV
Fig. 2 Increasing Window Numbers/Decreasing Window Size Provides More

Robust Peptide Detection. Using a yeast cell lysate (3 μg on column), SWATH
data was collected (five replicates) in serial column mode (2 h gradient) using
variable Q1 windows. The number of windows used to cover the mass range was
increased from 60 to 100 windows and the reproducibility curves were plotted.
The number of confident peptide detections (<1 % FDR) with ≤20 % CV further
increased as the window size decreases [6]
5. All instructions for operating the LC System and building pump

and autosampler methods can be found in the Operators Guide
(http://sciex.com/Documents/user%20guides/ekspert-
nanolc-400-systems-operator-guide-eng.pdf). All protocols for
performing the BetaGalactosidase tests can be found in the
System Integration Test (http://sciex.com/Documents/
user%20guides/nanolc-system-integration-test-en.pdf).
6. Here, the cHiPLC system setup consists of two analytical col-
umn chips in serial mode; in principle this is a “trap–elute”
experiment, which uses the first analytical column chip for
trapping and desalting of the sample with the Loading solvent,
while then upon valve switching the analytical gradient will
flow through both analytical column chips (in serial mode) to
separate and elute the sample, providing a 30 cm column
length. Due to two analytical column chips in serial mode,
loading solvent flow rates should be relatively low (~0.5 μL/
min) to keep column pressures reasonable (see step 2 of
Subheading 3.2). Alternatively, operators can use a more tradi-
tional trap elute set up using a trap chip (with 200 μm × 6 mm
ChromXP™ C18-CL chip, 3 μm, 300 Å, P/N 5015841,
SCIEX) instead of the first analytical column chip (see Note 5).
7. The chromatographic configuration described here is one
example of a possible configuration. It is possible to use differ-
ent “trap and elute” configurations and fused silica packed col-
umns for separations. There are different column length
options that can be used, to provide increased peak capacity

and potentially increased depth of coverage. Several chromato-
graphic setups are possible and feasible; however, it is essential
to perform high quality chromatography to obtain the best
separation and best peak shape for SWATH quantitation.
8. Retention time reproducibility is a key consideration when col-
lecting SWATH Acquisition data. It is recommended that col-
umn heaters are used to thermostat the analytical column and
ideally both the analytical column and trap for highest
reproducibility.
9. Sufficient column re-equilibration is important for high reten-
tion time reproducibility. Ensure that the column is fully re-
equilibrated in the starting conditions of the gradient before
injecting the next sample.
10. There are a number of key parameters to consider when decid-
ing on the best variable window strategy to use. First, it is
important to determine the average chromatographic peak
width, as this will define the cycle time. A typical cycle time
estimate would be to take the chromatographic peak width at
half height and divide by 6–8 to get an appropriate cycle time
and to provide sufficient chromatographic peak sampling for
good quantitation (i.e., for a peak width of 15 s at half height
therefore a good cycle time would be between 1.8 and 2.5 s).
Once the cycle time is determined, the number of windows can
be maximized. Accumulation times as fast as 25–30 ms per
MS/MS have been shown to produce good quality quantita-
tive data on the TripleTOF systems, so divide the cycle time by
25 ms to get the number of windows (i.e., a cycle time of 2.5 s
would allow for 100 SWATH Q1 windows).
11. To easily set up variable windows, the ‘variable window calcu-
lator’ at http://sciex.com/software-downloads-x2110
(SCIEX) or the swathTUNER tool [15] (http://swathtuner.
sourceforge.net) can be used to create the variable window
text file for loading into Analyst Software. In general, the
instrument operator has great flexibility in designing the vari-
able window text file, and can tailor SWATH window sizes
either specifically for the type and complexity of the sample
measured in the study (Fig. 3), or just using a once established
generally applicable window pattern. Within a quantitative
study, use the same variable window pattern for all samples. An
example of a 100 Q1 variable window text file can be found
here (http://sciex.com/community/entity/1217).
12. The number of variable windows, accumulation times, and
total cycle times can be further optimized by the individual
operator to allow for the optimal number of points across the
eluting sample peak to allow >10 points across each peak for
Variable Window Calculation

1 120
0.9
100
0.8
Window Width (m/z)

Normalized Density
0.7
80
0.6
0.5 60
0.4
Input Histogram 40
0.3
Var Windows
0.2
20
0.1
0 0
400 600 800 1000 1200 1400
m/z
Fig. 3 Variable Q1 Window Widths for SWATH Acquisition. To achieve better spec-
ificity in complex matrices, smaller Q1 windows are desirable especially in the
m/z dense regions where many peptide precursors are measured. The m/z den-
sity histograms constructed from the TOF MS data for the proteome of interest
(blue line) can be used to construct variable sized windows (red line), where the
density of precursors in each of the isolation windows is equalized across the
m/z range
optimal quantitation. This is very dependent on the chroma-

tography and length of gradient, which can also be optimized
for the specific sample type.
13. To build a fixed SWATH acquisition method instead of the
variable SWATH window method described above, go to the
SWATH Experiment UI and select the ‘Manual’ tab (see
Fig. 1b). First set the SWATH Analysis Parameters on the top:
set the Start Mass to 400, the Stop Mass to 1250, and the
SWATH Width to desired width (typically use 10 m/z), the #
of SWATH MS/MS scans per cycle is then automatically com-
puted to 85 windows. As no text file is imported (as for vari-
able windows), and the box Read SWATH Windows from text
file is unchecked, the UI can directly calculate SWATH win-
dows under the SWATH Analysis Parameters. Set the
Fragmentation Conditions as defined in step 3 of
Subheading 3.3 and set the SWATH Detection Parameters as
defined in step 4 of Subheading 3.3. The cycle time for the
MS/MS portion is computed to be 2.2 s. Click OK. Confirm
the TOF MS settings as defined in step 6 of Subheading 3.3.
14. Spiked synthetic peptides can also be used to adjust for slight
retention time drifts during data processing, and to perform
retention time alignments using libraries from other research-
ers and other LC-MS systems. Endogenous peptides already
present in the sample can also be used to perform retention
time alignment between library and sample.
Acknowledgments
We acknowledge support from the NIH shared instrumentation

grant for the TripleTOF 6600 system at the Buck Institute (1S10
OD016281, B.W.G.).
References
1. Collins BC, Gillet LC, Rosenberger G et al tor for creating and analyzing targeted pro-
(2013) Quantifying protein interaction dynam- teomics experiments. Bioinformatics
ics by SWATH mass spectrometry: application 26:966–968
to the 14-3-3 system. Nat Methods 9. Lambert JP, Ivosev G, Couzens AL et al (2013)
10:1246–1253 Mapping differential interactomes by affinity
2. Gillet LC, Navarro P, Tate S et al (2012) purification coupled with data-independent
Targeted data extraction of the MS/MS spec- mass spectrometry acquisition. Nat Methods
tra generated by data-independent acquisition: 10:1239–1245
a new concept for consistent and accurate pro- 10. Bruderer R, Bernhardt OM, Gandhi T et al
teome analysis. Mol Cell Proteomics (2015) Extending the limits of quantitative
11(O111):016717 proteome profiling with data-independent
3. Liu Y, Buil A, Collins BC et al (2015) acquisition and application to acetaminophen-
Quantitative variability of 342 plasma proteins treated three-dimensional liver microtissues.
in a human twin population. Mol Syst Biol Mol Cell Proteomics 14:1400–1410
11:786 11. Wang J, Tucholska M, Knight JD et al (2015)
4. Selevsek N, Chang CY, Gillet LC et al (2015) MSPLIT-DIA: sensitive peptide identification
Reproducible and consistent quantification of for data-independent acquisition. Nat Methods
the Saccharomyces cerevisiae proteome by 12(12):1106–1108, Online
SWATH-mass spectrometry. Mol Cell 12. Tsou CC, Avtonomov D, Larsen B et al (2015)
Proteomics 14:739–749 DIA-Umpire: comprehensive computational
5. Egertson JD, MacLean B, Johnson R et al framework for data-independent acquisition
(2015) Multiplexed peptide analysis using proteomics. Nat Methods 12:258–264, 7 p
data-independent acquisition and Skyline. Nat following 264
Protoc 10:887–903 13. Li Y, Zhong CQ, Xu X et al (2015) Group-
6. Hunter CL, Collins B, Gillet L, et al. (2014) DIA: analyzing multiple data-independent
Increasing depth of coverage in data indepen- acquisition mass spectrometry data files. Nat
dent acquisition with acquisition improve- Methods 12(12):1105–1106, Online
ments and higher sample loads. Proccedings of 14. Rardin MJ, Schilling B, Cheng LY et al (2015)
the 61st Annual ASMS Conference on Mass MS1 peptide ion intensity chromatograms in
Spectrometry & Allied Topics, Baltimore, MD, MS2 (SWATH) data independent acquisitions.
June 15–19, 2014 improving post acquisition analysis of pro-
7. Rost HL, Rosenberger G, Navarro P et al teomic experiments. Mol Cell Proteomics
(2014) OpenSWATH enables automated, tar- 14:2405–2419
geted analysis of data-independent acquisition 15. Zhang Y, Bilbao A, Bruderer T et al (2015)
MS data. Nat Biotechnol 32:219–223 The use of variable Q1 isolation windows
8. MacLean B, Tomazela DM, Shulman N et al improves selectivity in LC-SWATH-MS acqui-
(2010) Skyline: an open source document edi- sition. J Proteome Res 14:4359–4371
Chapter 17
Annotating Mutational Effects on Proteins and Protein

Interactions: Designing Novel and Revisiting Existing
Protocols
Minghui Li*, Alexander Goncearenco*, and Anna R. Panchenko
Abstract
In this review we describe a protocol to annotate the effects of missense mutations on proteins, their func-
tions, stability, and binding. For this purpose we present a collection of the most comprehensive databases
which store different types of sequencing data on missense mutations, we discuss their relationships, pos-
sible intersections, and unique features. Next, we suggest an annotation workflow using the state-of-the
art methods and highlight their usability, advantages, and limitations for different cases. Finally, we address
a particularly difficult problem of deciphering the molecular mechanisms of mutations on proteins and
protein complexes to understand the origins and mechanisms of diseases.
Key words Protein–protein interactions, Databases, Mutations
1 Introduction
The era of genome sequencing has unraveled a large number of

human genetic variations, as illustrated by the milestone 1000
Genomes project [1]. Mutations and genetic recombinations may
occur naturally during the cell division and at the same time may
be caused by extrinsic factors. A single nucleotide substitution that
results in a codon change encoding a different amino acid is called
“missense point mutation” (called “mutation” thereafter).
Germline missense mutations are passed to progeny, whereas
somatic mutations are not inherited. Due to predominantly neutral
effects of many genetic variations, they have accumulated in human
population and can be responsible for many individual phenotypic
traits in humans and may be used for genetic fingerprinting.
Generally, a variant frequently occurring in a population is termed
a polymorphisms and single nucleotide polymorphisms (SNP) are
one of the most common types of genetic variations.
*
Author contributed equally with all other contributors.
235
236 Minghui Li et al.
Missense mutations can render proteins nonfunctional and

may be responsible for many diseases. From the clinical perspec-
tive, these non-neutral mutations affecting human health represent
the main interest. For some diseases and genes, particularly follow-
ing the Mendelian inheritance patterns, the causal genotype–phe-
notype relationship has been already established, while for complex
polygenic diseases involving multiple factors it is still unknown.
Moreover, genetic variants with low penetrance, weakly associated
with disease phenotypes, can only be annotated for large samples
and for many diseases their genetic determinants have to be
discovered.
2 Materials
2.1 Software 1. Molecular dynamics packages: NAMD (http://www.ks.uiuc.

edu/Research/namd/) [2] and CHARMM (http://www.
charmm.org/) [3].
2. Structural visualization packages: VMD (http://www.ks.uiuc.
edu/Research/vmd/) [4], Chimera (http://www.cgl.ucsf.
edu/chimera/) [5] and CN3D (http://www.ncbi.nlm.nih.
gov/Structure/CN3D/cn3d.shtml) [6].
2.2 Online Resources 1.

FTP site at NCBI (ftp://ftp.ncbi.nih.gov/pub/panch/
Mutation_binding/) includes examples of configuration and
runfiles for VMD, NAMD, and CHARMM used in protocols
of Subheadings 3.5 and 3.6.
2. Databases collecting human genetic variations, mutations, and
data on their clinical relevance (Table 1).
3. Web servers for characterization of structural features of muta-
tions (Table 2).
4. Web servers for predicting the phenotypic effects of mutations
(Table 3)
5. Web servers and programs for predicting the effects of muta-
tions on protein stability (Table 4).
6. Web servers and programs for predicting the effects of muta-
tions on protein–protein binding affinity (Table 5).
3 Methods
Recent advances in experimental methods reduced the cost of the

DNA sequencing and equipped many labs and hospitals with geno-
typing and sequencing assays, so that these data along with the
clinical profiles of patients can be deposited into the central archi-
val facilities. Some of them include the databases of Genotypes and
Effects of Mutations on Protein-Protein Interactions 237
Table 1
A summary of databases integrating the data on human genetic variations, mutations, and their
clinical relevance
Database Description URL Reference

COSMIC Somatic mutations in cancer http://cancer.sanger.ac.uk/cosmic [16]
HGMD Published gene lesions http://www.hgmd.cf.ac.uk/ac/ [52]
responsible for human
inherited disease
TCGA Cancer Genome Atlas http://cancergenome.nih.gov/ [17]
OMIM Human genes, inherited genetic http://www.omim.org/ [14]
disorders, and germline
mutations
dbGaP Clinical studies linking genotypes http://www.ncbi.nlm.nih.gov/gap [7]
to disease phenotypes
GTR Genetic Testing Registry http://www.ncbi.nlm.nih.gov/gtr/ [8]
PheGenI Phenotype–genotype integrator http://www.ncbi.nlm.nih.gov/gap/ [12]
phegeni
ClinVar Genomic variations and their http://www.ncbi.nlm.nih.gov/clinvar/ [15]
relationship to human health
dbSNP Single nucleotide polymorphisms http://www.ncbi.nlm.nih.gov/SNP/ [9]
dbVar Large genetic variations http://www.ncbi.nlm.nih.gov/dbvar [10]
SwissVar Disease-related variants in http://swissvar.expasy.org [19]
Uniprot, provides structural
mapping
PharmGKB Associates genes with drugs. https://www.pharmgkb.org/ [53]
Catalogs genetic variations
known to impact drug
response
MutDB Integrates human variations and http://www.mutdb.org/ [18]
COSMIC mutations, maps to
structural data, KEGG
pathways, and includes
predictions of effects of
variations on phenotype
CBioPortal Visualization and analysis of large http://cbioportal.org/ [54]
cancer studies. It is based on
TCGA and incorporates the
overlapping data from
COSMIC
Table 2
Online resources for exploring the structural, cellular, and genomic context of mutated proteins
Database Description URL Reference

HPRD Integrates information http://www.hprd.org [55]
pertaining to domain
architecture,
posttranslational
modifications, interaction
networks, and disease
association
EBI IntAct Molecular interaction data http://www.ebi.ac.uk/intact/ [56]
BioSystems Provides integrated access http://www.ncbi.nlm.nih.gov/biosystems/ [21]
to genes, proteins, small
molecules, and pathways
Reactome Curated and peer-reviewed http://www.reactome.org [57]
pathways
KEGG Manually curated pathways http://www.genome.jp/kegg/pathway.html [58]
CDD Annotates functional and http://www.ncbi.nlm.nih.gov/Structure/ [23]
binding sites in protein cdd/cdd.shtml
domain families
IBIS Annotates protein–protein http://www.ncbi.nlm.nih.gov/Structure/ [27]
and protein–DNA/ ibis/ibis.cgi
RNA/ion/small molecule
interactions and binding
sites. Identifies conserved
binding sites in
homologous protein
complexes
muPIT Interactive exploration of http://mupit.icm.jhu.edu [59]
mutations and their
structural context
DMDM Domain mapping of http://bioinf.umbc.edu/dmdm/ [60]
mutations
PolyDoms Mapping of mutations to https://polydoms.cchmc.org/polydoms/ [61]
protein domains and
prediction of structural
and functional impact of
mutations
Phenotypes (dbGaP) [7] and Genetic Testing Registry (GTR) [8]

developed at NCBI, NIH. The genotypes are mapped to a refer-
ence genome assembly and typically include two major categories
of variations: (1) single nucleotide polymorphisms (SNP) depos-
ited in the dbSNP database [9] and (2) larger structural genetic
variations in genomes, including long indels, inversions, and copy
Table 3
A summary of online resources for predicting the phenotypic effects of mutations
Name Features and methods URL Reference

SIFT Sequence homology and http://sift.bii.a-star.edu.sg [62]
physical properties of
amino acids
PolyPhen-2 Bayesian models based on http://genetics.bwh.harvard.edu/ [31]
substitution scores, pph2/
phylogenetic
conservation, and
structural features
SNPs3D Sequence conservation and http://snps3d.org [63]
changes in physical
properties of amino acids
affecting protein stability
PMut Neural network classifier http://mmb2.pcb.ub.es/PMut/ [64]
SNAP Neural network classifier https://www.rostlab.org/services/ [65]
based on protein structural snap/
properties predicted from
protein sequence
Fathmm Cancer-specific and other http://fathmm.biocompute.org.uk [66]
disease-specific Hidden
Markov Models for
predicting the functional
consequences of coding
and noncoding variants
MutationAssessor Predicts the effects of http://mutationassessor.org/ [67]
mutations on subfamily
specific sites
CHASM Random Forest classifier for http://karchinlab.org/apps/appChasm. [68]
cancer somatic mutations html/
MutPred Uses sequence conservation http://mutpred.mutdb.org/ [18]
and structural features,
posttranslational
modifications
SNPs&GO SVM classifier of disease- http://snps.biofold.org/snps-and-go/ [69]
related variations based
on protein functional
annotation (GO)
PROVEAN Predicts the effects of amino http://provean.jcvi.org/ [30]
acid substitutions,
insertions, and deletions on
protein function, allows
scanning multiple mutations
(continued)
Table 3
(continued)
Name Features and methods URL Reference

FunSAV Random Forest-based http://sunflower.kuicr.kyoto-u.ac.jp/ [70]
classifier, uses structural sjn/FunSA/
features and network
properties of mutated
proteins
nsSNPAnalyzer Random Forest-based http://snpanalyzer.utmem.edu/ [71]
classifier, uses structural
and evolutionary
information
PANTHER Uses phylogenetic and http://www.pantherdb.org/tools/ [72]
evolutionary information csnpScoreForm.jsp/
PhD-SNP SVM classifier, uses http://gpcr.biocomp.unibo.it/cgi/ [73]
sequence profiles predictors/PhD-SNP/PhD-SNP.cgi/
SAAP Precalculated database of http://www.bioinf.org.uk/saap/ [74]
predicted effects of
known variants, considers
structural properties and
clashes resulting from
amino acid substitutions
SusPect Incorporates sequence www.sbg.bio.ic.ac.uk/suspect/ [75]
conservation and
network-based features
KinDriver Annotates driver mutations http://kin-driver.leloir.org.ar/ [76]
in protein kinases with
experimental evidence
demonstrating their
functional role
ProKinO Unified resource for mining http://vulcan.cs.uga.edu/prokino [77]
the cancer kinome
number variants (CNVs), deposited in the dbVar database [10].

The phenotypes in these databases are mainly organized by disease
names or clinical conditions, and diseases are classified according
to the Disease Ontology (DO) [11] and Medical Subject Headings
(MeSH) terms. Additionally, the NCBI phenotype–genotype inte-
grator (PheGenI) [12] merges genetic variations identified by
genome-wide association studies (GWAS) with dbVar, OMIM,
GTR, and dbSNP databases.
Table 4
A summary of online software resources for predicting the effects of mutations on protein stability. The second column indicates the kind of data a
method requires as an input: protein sequence, structure, or any of the two. Here “ΔΔG” refers to ΔΔGfold.
Type of output,
method, and energy
Resource Input function URL Reference
FoldX Structure ΔΔG using empirical http://foldxsuite.crg.eu/ [78]
force fields
PoPMuSiC-2.0 Structure ΔΔG using a http://dezyme.com/ [79]
combination of
statistical potentials
and neural
networks
ERIS Structure ΔΔG using physical http://dokhlab.unc.edu/tools/eris/ [80]
force fields with
atomic modeling
CUPSAT Structure ΔΔG using mean http://cupsat.tu-bs.de/ [81]
force atom pair and
torsion angle
potentials
Hunter Structure ΔΔG using http://bioinfo41.weizmann.ac.il/hunter/ [82]
knowledge-based
potentials
MultiMutate Structure ΔΔG using four-body http://www.math.wsu.edu/math/faculty/bkrishna/DT/Mutate/ [83]
scoring functions
AUTO-MUTE Structure ΔΔG using http://proteins.gmu.edu/automute [84]
knowledge-based
Effects of Mutations on Protein-Protein Interactions
potentials
(continued)
241
Table 4
242
(continued)
Type of output,
method, and energy
Resource Input function URL Reference
NeEMO Structure ΔΔG using residue http://protein.bio.unipd.it/neemo/ [85]
Minghui Li et al.
interaction
networks
DUET Structure ΔΔG using an http://bleoberis.bioc.cam.ac.uk/duet/stability [86]
integrated
computational
approach of mCSM
and SDM
MAESTRO Structure ΔΔG using multi http://biwww.che.sbg.ac.at/MAESTRO [87]
agent stability
prediction
I-Mutant3.0 Structure/Sequence ΔΔG using SVMs http://gpcr2.biocomp.unibo.it/cgi/ [88]
predictors/I-Mutant3.0/I-Mutant3.0.cgi
MUPro Structure/ Predicts qualitative http://mupro.proteomics.ics.uci.edu/ [89]
Sequence decrease/increase
of stability using
SVM
iStable Structure/Sequence ΔΔG using SVM http://predictor.nchu.edu.tw/iStable [90]
MuStab Sequence Predicts qualitative http://bioinfo.ggc.org/mustab/ [91]
decrease/increase
of stability using
SVM
iPTREE-STAB Sequence ΔΔG using decision http://bioinformatics.myweb.hinet.net/iptree.htm [92]
tree methods
Table 5
A summary of online and software resources for predicting the effects of mutations on protein–
protein binding affinity. Here “ΔΔG” refers to ΔΔGbind. All resources need structure as an input
Type of output, method, and

Resource energy function URL Reference
MutaBind ΔΔG using molecular https://www.ncbi.nlm.nih.gov/ [46]
mechanics force fields, projects/mutabind/
statistical potentials, and
fast side-chain optimization
algorithms. Produces a
model of mutant.
BeAtMuSiC ΔΔG using a set of statistical http://babylone.ulb.ac.be/beatmusic [93]
potentials, does not
produce a model of mutant
ELASPIC ΔΔG for mutations located on http://www.kimlab.org/software/ [94]
interface using a elaspic
combination of semi-
empirical energy terms and
molecular features; does not
produce a model of mutant
DrugScorePPI ΔΔG for alanine-scanning http://cpclab.uni-duesseldorf.de/ [95]
mutations located on dsppi/
interface using knowledge-
based scoring functions;
does not produce a model
of mutant
SNP-IN classifies effects of mutations http://andromeda.rnet.missouri.edu/ [96]
on protein–protein snpintool/
interactions using supervised
and semi-supervised
learning; does not produce a
model of mutant
FoldX ΔΔG using empirical force http://foldxsuite.crg.eu/ [78]
field, Produces a model of
mutant.
The clinical implications of genetic variations are recorded in

several other databases. The human gene mutation (HGMD) [13]
and Online Mendelian inheritance in Man (OMIM) [14] are the
main databases that integrate the information on genetic Mendelian
disorders with genes and mutations reported to be causative.
ClinVar [15] is another archive, which collects the data on genetic
variations from dbSNP and dbVar and integrates them with the
available clinical evidence of these variants obtained from multiple
studies. In ClinVar each variant is assigned a score which shows the
consistency of the reported clinical effect in different studies.
ClinVar annotates variants as: benign, likely benign, likely patho-

genic, pathogenic, of uncertain significance, or variants with con-
flicting interpretations. Importantly, it also annotates the variants
associated with individual drug effects.
Germline mutations constitute the bulk of mutations in
OMIM, ClinVar, and HGMD, while somatic missense mutations
are found predominantly in cancer cells and are not inherited. The
information on location, tissue type, and frequency of somatic
mutations in tumor samples together with the cancer type can be
obtained from COSMIC[16] and TCGA databases [17], while
other resources, for instance, MutDB [18] and SwissVar [19],
integrate the data on germline and somatic mutations for specific
genes and diseases.
Figure 1 describes a computational pipeline for exploring
mutations and assessing their effects on protein structure, func-
tion, and interactions. Different sections of this chapter follow this
pipeline and suggest the protocols to solve each specific problem.
Fig. 1 Mutation assessment workflow

3.1 Collecting First, we will describe a protocol for extracting clinically relevant
and Integrating mutations from ClinVar, COSMIC, and TCGA databases to fur-
the Data on Human ther analyze them with respect to their clinical and functional
Polymorphisms impacts. Here we use human monomeric Casitas B-lineage lym-
and Mutations phoma c-CBL protein (CBL) as an example [20]. The links and
references to web resources used in this protocol are listed in
Table 1.
1. NCBI variation viewer (http://www.ncbi.nlm.nih.gov/varia-
tion/view/) [21] can be searched by gene name (CBL), Refseq
accession (NP_005179), or Uniprot ID (P22681.2). A snap-
shot of variation viewer shows the genetic variation from
ClinVar in the locus of CBL gene (Fig. 2a). Different variants
in this viewer are depicted as separate tracks below the CBL
gene locus. The ClinVar track displays multiple variants as
boxes with the number of variants listed within each box.
Variants with benign effect are shown using green color,
whereas the purple boxes show pathogenic variants.
2. Additionally, the user can download all mutations from the
COSMIC ftp server in VCF format (CosmicCodingMuts.vcf.gz)
and display these mutations in a separate track in the variation
viewer (choose menu “your data” and select “add track” option).
Note that it is necessary to select the same genome assembly in
both variation viewer and in COSMIC (e.g., GRCh37).
3. Each variant with a valid ClinVar annotation is linked to a cor-
responding dbVar or dbSNP record. Here we will focus on a
single nucleotide variant of CBL gene with the dbSNP acces-
sion rs267606706. As shown in Fig. 2b, it is a missense muta-
tion where nucleotide substitution T->C results in p.Y371H
amino acid substitution. There is clinical evidence that this
variant can cause a Noonan syndrome-like disorder and/or
juvenile myelomonocytic leukemia. The GTR studies cited in
ClinVar show [22] that p.Y371H is a heterozygous germline
substitution.
4. The dbSNP page for accession rs267606706 contains a link to
“3D structure mapping” (found under the “NCBI resources”
section) which points to the SNP3D page where several syn-
onymous and missense variants are mapped onto the CBL pro-
tein structure. By default, this variant is selected, but it is
possible to select other variants for display by clicking on the
“Cn3D selected” button. Additionally, the SNP3D page con-
tains a link “CD” that shows conserved domains from the
CDD database mapped onto this protein [23]. Figure 2c
depicts the structure of CBL using Cn3D with mutated Tyr
residue side chain colored in yellow.
5. As was shown previously using ClinVar, the germline p.Y371H
mutation may be linked to leukemia, however, many cancer
Fig. 2 Identification of clinically relevant mutations in ClinVar, COSMIC, and TCGA. (a) NCBI Variation Viewer
showing the CBL gene locus on chromosome 11; ClinVar and dbSNP data shown as tracks below the gene,
pathogenic mutations are presented as purple squares and closely-located mutations are grouped together in
ClinVar track; (b) one of the pathogenic mutations p.Y371H shown in dbSNP with the corresponding disease
annotation in GTR; (c) a representative structure of CBL protein visualized in SNP3D with Cn3D software,
mutated p.Y371H is shown with the yellow side chain; (d) cBioPortal view of CBL protein with missense muta-
tions mapped onto the corresponding domains; (e) a representative structure visualized with JMol directly in
cBioPortal with all mapped missense mutations shown in green
mutations are somatic and therefore are not present in ClinVar.

In order to explore somatic mutations we switch to cBioPortal,
which allows exploring mutations from the TCGA and
COSMIC databases. Open cBioPortal web page and submit a
query “CBL” as the user-defined gene set. A summary for dif-
ferent types of cancers will be shown for the CBL gene. Open
the second tab “Mutations”, which display mutations on CBL
protein sequence. Figure 2d depicts missense CBL mutations
mapped by the cBioPortal onto the corresponding CDD
domain context. Additionally, the cBioPortal provides loca-
tions of mutations on protein structures. The blue footprints
in Fig. 2d show that several structures cover CBL protein
sequence and could be explored interactively. All missense
mutations from COSMIC and TCGA databases are mapped
on the representative structure (in green) and are displayed in
the web browser (Fig. 2e).
6. Each mutation in cBioPortal “Mutations” tab is shown as a pin

indicating its position in CBL protein sequence. The height of
the pin corresponds to the number of known mutations. Place
the mouse cursor over the Znf domain (zinc finger, shown in
yellow) and over the first pin from the left in Znf domain. As a
result, a window with a list of mutations and cancer types will
pop up. For zinc finger domain the first two missense muta-
tions, C384R and C384W, are associated with glioblastoma
and melanoma, respectively. By searching for Cys384 residue
in the table below and pressing the “3D” button, the struc-
tural location of this mutation is displayed. We will explain
how this mutation could be interpreted in the context of
molecular interactions in the next section.
3.2 Finding Protein Now we consider the domain context of CBL mutations described
Domains in the previous section (p.Y371H and p.C384R) and annotate
and Functional Sites their impacts on protein interactions and signaling pathways. The
Affected by Mutations web resources used in this protocol are listed in Table 2.
1. Evolutionarily conserved sites in a multiple sequence align-
ment usually correspond to functionally important sites and
mutations in these sites can be harmful to protein function. If
a protein of interest has a known PDB structure, conservation
profiles can be downloaded from the PDBsum resource other-
wise ConSurf server [24] can be used. In addition, the CDD
server can offer functional annotations of sites in conserved
protein domains, whereas IBIS server provides locations of
binding sites for different types of binding partners (protein,
small molecule, nucleic acid, ion, and peptides), and facilitates
the mapping of a comprehensive biomolecular interaction net-
work for a given protein query (with or without structure)
[25–27]. Similar binding sites in IBIS are clustered together
based on their sequence and structure conservation.
2. Open IBIS web page and search for 1FBV structure, chain
A. Go to “protein–protein” tab and click on the balloon with
the annotation “RING” domain to display interactions of the
CBL RING domain with other domains/proteins (Fig. 3a).
Binding sites are shown on CBL sequence as triangles and
highly conserved binding sites are shown in red color. In the
list of interaction partners below, the first conserved binding
site cluster is formed between RING domain of CBL and
ubiquitin conjugating enzyme from UBCc family. By clicking
on the plus sign next to “UBCc”, one can see the correspond-
ing binding site, the alignment of similar binding sites found in
different CBL-UBCc complexes. By opening the link to Cn3D
viewer, one can explore the interfaces and binding sites in these
protein complexes (Fig. 3c).
Fig. 3 Analysis of conserved functional and binding sites in mutated proteins using IBIS method. (a) conserved
protein binding site in the RING domain of CBL; (b) interaction graph of CBL protein (represented by 1FBV PDB
structure); (c) visualization of protein interface between CBL and UBCc in Cn3D software. Position 384 in 1FBV
corresponds to position 338 in the full-length PDB sequence
3. The interaction graph in Fig. 3b shows the observed (black

lines) and predicted interaction partners of CBL. Next, we will
focus on interactions of CBL with zinc ions and UBCc ubiqui-
tin ligase. Note that self-links indicate interactions between
domains in CBL protein within or between CBL chains.
4. The structure in Fig. 3c shows that conserved cysteine residues
in the binding site coordinate two zinc ions, apparently playing
an important structural role. A substitution of cysteine by argi-
nine disrupts the coordination of Zn, which affects the struc-
ture and stability of the zinc finger RING domain and may also
affect CBL function.
5. Structural and biochemical analyses [28] show that CBL inac-

tive state adopts an autoinhibited interaction. Substrate bind-
ing and Tyr371 phosphorylation activates CBL by producing a
large conformational change in order to place the RING
domain and UBCc in close proximity to the substrate neces-
sary for effective catalysis. Importantly, mutation p.Y371H
may prohibit activation of CBL by phosphorylation and may
also affect the interaction with UBCc.
6. The impact of mutations on signaling pathways can be explored
using recently developed PathiVar server [29]. Alternatively, it
is possible to explore all pathways in which CBL interacts with
UBCc using KEGG, Reactome, or NCBI Biosystems
databases.
7. Search for ‘CBL AND Y731 and "Homo sapiens"[Organism]’
in the NCBI Biosystems portal. The first record will point to
the “Regulation of signaling by CBL” pathway and the role of
Y731 phosphorylation will be explained in the pathway
description.
In addition to the NCBI IBIS server that allows analyzing the
domain context and structural determinants of interactions, several
other servers (DMDM, PolyDoms, and muPIT) provide the map-
ping of mutations on protein domains and protein structures.
3.3 Assessing If Many methods have been proposed to predict the effects of mis-
Mutations Have sense mutations on proteins, classifying them as damaging or
Damaging or Benign benign. These methods differ in terms of the properties of muta-
Effects on Proteins tions or proteins used during the training procedure as well as in
terms of the algorithms applied for decision-making. For example,
machine learning algorithms train their models to distinguish
known disease-associated from neutral mutations. Other methods
do not explicitly train their models but almost all methods described
in this section exploit the evolutionary conservation assuming that
changes at conserved positions tend to be more deleterious. Besides
sequence conservation various other sequence and structural fea-
tures are used, which may include: changes in physicochemical
properties between wild-type and substituted amino acid, struc-
tural features (mostly solvent accessibility), site mutability in DNA,
and sequence context of the site.
An unbiased testing and comparison of machine learning meth-
ods is obviously an issue since they are usually trained on all avail-
able datasets of mutations and it is difficult to obtain a test set which
would not overlap with training set. There are several experimental
studies on variants in P53, LacI, and ABCA1 proteins which can be
regarded as unbiased test cases. Comparisons of different methods
on these experimental sets reported up to 70 % TPR (True positive
rate) at 10 % FPR (False positive rate) [30]. Models trained to dis-
tinguish Mendelian variants with pronounced deleterious effects
are more appropriate and accurate for predicting the effects of

Mendelian mutations. The accuracy of these models is much higher
than of those models that aim to assess the effects of mutations
from complex diseases including cancer. This is evident from the
evaluation of PolyPhen-2 performance which yields 0.70–0.77
TPR at 10 % FPR when trained on the HumDiv dataset (Mendelian
disease mutations) and drops to 0.50–0.52 TPR for HumVar (all
human disease causing mutations) trained models [31]. It should
be mentioned that there are several methods that are trained to
distinguish cancer mutations from neutral polymorphisms (Table 3);
however, no existing method can accurately identify driver and pas-
senger mutations within the pool of cancer mutations.
One of the most comprehensive comparisons of different
methods to predict phenotypic effects of mutations was performed
recently [32]. To avoid any bias in evaluation of these methods,
most of which were trained on all available sets of disease muta-
tions and neutral polymorphisms, the authors of this study tested
different methods on an independent set of experimental studies.
They concluded that there was a variation between various meth-
ods in terms of their accuracy and applicability, with SNP&GO and
MuPred being the most reliable predictors. Interestingly, despite
the fact that different methods use similar sets of features, only half
of their correctly predicted cases overlap [32]. Since this study was
published, several new methods have been introduced (see
Table 3). For example, in contrast to many methods, that assess
amino acid frequency distribution in a given site of interest, a
recently developed method PROVEAN accounts for the sequence
context around the site of interest and poorly aligned regions/sites
are assigned very low scores. Overall, the effect of alignment qual-
ity on the performance of all methods is largely undetermined but
suspected to be very large. Therefore, a user-based construction of
accurate alignments of homologous proteins would be very advan-
tageous for accurate annotations of the effects of mutations.
3.4 Predicting While methods, which provide a classification of damaging effects

the Impact of mutations, are widely used by the genomics community, a new
of Mutations level of annotation is needed to offer an explanation of why and
on Protein Stability how these mutations damage proteins. Algorithms and servers
described in the next several sections address these tasks. Proteins
may evolve through the acquisition of new mutations, most of
which are destabilizing but phenotypically neutral. Stability of a
protein may be directly related to its functional activity and incor-
rect folding and decreased stability can be the major consequences
of pathogenic missense mutations [33, 34]. However, protein sta-
bility is necessary but not sufficient for protein function, and pro-
teins do not evolve to maximize their stability. Typically, the
magnitude of effects of mutations on stability can be quantified by
changes in unfolding free energy (ΔΔGfold) (Fig. 4).
Fig. 4 Annotation of the effects of mutations on proteins with available structures
mut WT
DDG fold = DG fold - DG fold (1)
Table 4 lists several state-of-the art methods for predicting the

quantitative changes in unfolding free energy upon mutations and
provides short descriptions and links to corresponding programs/
servers. Methods described in this section differ in terms of energy
functions, procedures used for optimization and sampling, and
algorithms used for training, if applicable. Energy functions may
vary from physics-based force fields, which describe fundamental
physical forces between atoms, to knowledge-based potentials,
which are based on statistical analyses of protein structures and
residue properties. The majority of these methods require the
coordinates of protein structures while methods like MuStab or
iPTREE-STAB do not use structural data but their performance is
also limited. The performance of different methods was evaluated
in several studies [35–37]. In the first study [35] the following
performance ranking was reported: EGAD > CC/PBSA > I-Mutan
t2.0 > FoldX > Hunter > Rosetta with correlation coefficients
between experimental and predicted ΔΔG values in the range of
0.59 and 0.26 and standard deviation in the range of 0.95 and
2.32 kcal mol−1. However, the servers of the top performing meth-

ods EGAD and CC/PBSA are no longer available. In the second
study I-Mutant3.0, Dmutant, and FoldX were found to be the
most reliable predictors [36].
There are several servers to assess the effects of mutations on
stability that are straightforward and easy to use. Here we present
a protocol on how to use FoldX software.
1. Run RepairPDB module of FoldX to correct errors in the
structure produced during the refinement (nonstandard angles
or distances) (Runfile can be obtained from http://foldxsuite.
crg.eu/command/RepairPDB).
2. Run BuildModel module of FoldX to introduce a mutation on
the optimized wild-type structure http://foldxsuite.crg.eu/
command/BuildModel. BuildModel module optimizes the
configurations of the side chains of amino acids in the vicinity
of the mutated site and calculates the difference in unfolding
free energy (ΔΔGfold) between mutant and repaired wild-type
structure. The total unfolding free energy and each energy
term can be obtained from the “Dif_BuildModed_*.fxout”
output file.
3.5 Predicting Crucial prerequisite for proper biological function is a protein’s

the Effects ability to establish highly selective interactions with macromolecu-
of Mutations lar partners. A missense mutation that affects protein interactions
on Protein–Protein [38–40] may cause significant perturbations or complete abolish-
Binding Affinity ment of protein function, potentially leading to diseases. Typically,
the change in binding free energy (DDGbind ) is used to quantify the
magnitude of mutational effects on protein–protein interactions
(Fig. 4).
mut WT
DDGbind = DGbind - DGbind (2)
The binding energy is calculated as a difference between the free

energies of a complex AB and unbound proteins A and B:
DGbind = G AB - G A - G B (3)
There are very few methods that estimate actual ΔΔGbind values
and these methods require all-atom or at least protein backbone
atom coordinates of a wild-type and/or mutated protein. Some of
the methods use coarse-grained predictors based on statistical or
empirical potentials, others apply molecular mechanics force fields
with different solvation models. For example, the molecular
mechanics Poisson–Boltzmann surface area (MM-PBSA) method
and its derivatives have been shown to yield very good agreement
between predicted and experimental values with correlation coef-
ficients up to 0.69 [41]. For all methods, the right choices of mini-
mization protocols, energy functions, and solvation models are
crucial for achieving reasonable prediction accuracy. In addition,

prediction accuracy strongly depends on the type of mutation and
its location in a protein complex. For example, if a residue is located
on the protein–protein interface, its mutation might have a larger
effect on protein–protein interaction and binding affinity com-
pared to a non-interfacial mutation [41]. A location of mutated
sites can be mapped by SPPIDER (http://sppider.cchmc.org/)
[42] or Meta-PPISP (http://pipe.scs.fsu.edu/meta-ppisp.html)
[43] servers. These servers are recommended by two assessments
of computational methods for predicting protein–protein interac-
tion sites [44, 45]. Users can also analyze structures and locations
of mutations using software Chimera or VMD.
Below is a step-by-step protocol reported in our previous paper
[41] to predict the impact of mutations on binding affinity. This
protocol combines molecular mechanics force fields with statistical
(BeAtMuSiC) and empirical (FoldX) energy functions. All files are
provided via the ftp site ftp://ftp.ncbi.nih.gov/pub/panch/
Mutation_binding. The improved version can be available from
our MutaBind server https://www.ncbi.nlm.nih.gov/projects/
mutabind/.
1. Install software VMD, NAMD, and CHARMM.
2. Download a structure for your protein of interest from the
Protein Data Bank (PDB).
3. Add hydrogen atoms, a rectangular box (10 Å) of water mol-
ecules, and Na+ and Cl− ions (ionic concentration of 150 mM)
to the structure using VMD (“vmd.pgn”).
4. Carry out 5000-step energy minimization with harmonic
restraints (with the force constant of 5 kcal mol−1 Å−2) applied
on the backbone atoms of all residues (“minimization1.conf”),
followed by a 35,000-step energy minimization on the whole
system (“minimization2.conf”) with NAMD program using
the CHARMM27 force field.
5. Introduce a mutation using “mutator” plugin of VMD soft-
ware on the final minimized model from step 4.
6. Run an additional 300-step minimization for the whole mutant
structure (“minimization2.conf”).
7. Run CHARMM program using the last frame from step 4 (for
wild-type structure) and step 6 (for mutant structure) to
obtain van der Waals interaction energy (ΔEvdw), polar solva-
tion energy of solute in water (ΔGsolv) for the wild-type and
mutant, and interface area (ΔSAmut) for mutant (Runfile is
“binding_energy.str”).
8. Submit your structure from step 2 and a mutation to
BeAtMuSiC webserver (Table 5) to obtain the binding affinity
change ΔΔGBM.
9. Run AnalyseComplex module of FoldX to obtain the binding

affinity change ΔΔGFD. (http://foldxsuite.crg.eu/command/
AnalyseComplex).
10. Obtain the binding affinity changes using the following com-
bination of energy terms from [41]:
DDGbind = aDDE vdw + bDDG solv + gDSAmut

+ eDDGBM + lDDG FD + d
æ a = 0.122 , b = 0.101 , g = 0.043 , ö

ç ÷
è e = 0.446 , l = 0.168 and d = 1.326 ø.
Recently a new computational method MutaBind [46] was
developed to evaluate the effects of mutations on protein–protein
interactions (http://www.ncbi.nlm.nih.gov/projects/mutabind/).
The MutaBind method uses molecular mechanics force fields, sta-
tistical potentials, and fast side-chain optimization algorithms. It
maps mutations on a protein complex structure, calculates the asso-
ciated changes in binding affinity, determines the deleterious effect
of a mutation, estimates the confidence of this prediction, and pro-
duces a mutant structural model for download.
3.6 Assessing Proteins may adopt different conformations along the pathway of
the Changes in Protein a biochemical reaction and their intrinsic flexibility and ability to
Conformations sample alternative conformations are crucial for protein function.
and Hydrogen Bond Mutations might shift the equilibrium between different confor-
Networks Induced mations (Fig. 4) and as a result, the most populated conformation
by Mutations of a mutated protein can be different in structure, stability, and
functional activity from the wild-type conformation. It is extremely
difficult to model structural changes in a protein backbone pro-
duced by mutations and large conformational shifts can be pre-
dicted correctly only in a few cases. In fact, most algorithms
discussed in the previous sections do not account for the backbone
flexibility. If several conformations of the same protein are available
in the structural databank, all of them ideally should be used to
provide a complete picture of dynamical and energetic effects of
mutations [20].
Mutations can either change the global conformation of an
entire molecule or have a localized effect in a small region. In a
recent study of the NFAT5 transcription factor [47], different
mutations from the same DNA-binding loop were analyzed and it
was shown that effects of these mutations on protein dynamics and
DNA binding were drastically different although they were located
very close to each other in sequence and structure. Protein dynam-
ics can be studied through performing molecular dynamics (MD)
simulations using NAMD [2], CHARMM [3] and Amber [48]
MD packages. NAMD, for example, is fast and easy to use; it can
be applied with CHARMM or Amber force fields, whereas VMD

or CHARMM packages can be used to analyze the MD trajectories
produced by NAMD.
Changes in structure may also be assessed through the analyses
of hydrogen bond networks and their differences between mutant
and wild-type proteins since hydrogen bonds are important in
determining protein stability. A mutation disrupting hydrogen
bonds might have a significant impact on protein conformation,
stability, and dynamics (reviewed in [49], Fig. 4). Hydrogen bonds
can be calculated using HBOND (http://caps.ncbs.res.in/iws/
hbond.html) [50] or PIC servers (http://pic.mbu.iisc.ernet.in/)
[51] and visualized by Chimera.
Below is a step-by-step protocol to assess the conformational
changes induced by mutations:
1. Download the structure and introduce a mutation using VMD
(refer to steps 2 and 5 from Subheading 3.5).
2. Build the model systems with VMD (Refer to step 3 from
Subheading 3.5).
3. Carry out the energy minimizations (Refer to step 4 from
Subheading 3.5). The number of minimization steps can be
chosen based on the size of system.
4. Heat the system to 300 K over 300 ps with harmonic con-
straints applied to protein backbone atoms using NAMD
(“heat.conf”).
5. Perform unconstrained MD simulation on the system with
NAMD (“md.conf”).
6. Load MD trajectories into the VMD software to monitor the
conformational changes and calculate the root-mean-square devi-
ation (RMSD) between the wild-type and mutant structures.
Assessing the effects of mutations on hydrogen bond net-
works using Chimera.
7. Load your protein structure of interest into Chimera.
8. Select residues of interest and input “findhbond selRestrict
both distSlop 0.35 angleSlop 60.0 saveFile filename” in the
command line. The hydrogen bonds will be shown on the
screen and details will be saved in the file “filename”. One can
adjust the distance (distSlop) and angle (angleSlop) parameters
in the definition of hydrogen bonds.
9. (Optional) Go to Tools -> Structural Analysis -> FindHBond
to find the hydrogen bonds.
Acknowledgements
This work was supported by the Intramural Research Program of

the National Library of Medicine.
References
1. Abecasis GR, Altshuler D, Auton A, Brooks variation. Nucleic Acids Res 41(Database
LD, Durbin RM, Gibbs RA, Hurles ME, issue):D936–D941
McVean GA, Genomes Project C (2010) A 11. Kibbe WA, Arze C, Felix V, Mitraka E, Bolton
map of human genome variation from E, Fu G, Mungall CJ, Binder JX, Malone J,
population- scale sequencing. Nature Vasant D, Parkinson H, Schriml LM (2015)
467(7319):1061–1073 Disease Ontology 2015 update: an expanded
2. Phillips JC, Braun R, Wang W, Gumbart J, and updated database of human diseases for
Tajkhorshid E, Villa E, Chipot C, Skeel RD, linking biomedical knowledge through disease
Kale L, Schulten K (2005) Scalable molecular data. Nucleic Acids Res 43(Database
dynamics with NAMD. J Comput Chem issue):D1071–D1078
26(16):1781–1802 12. Ramos EM, Hoffman D, Junkins HA, Maglott
3. Brooks BR, Bruccoleri RE, Olafson BD, States D, Phan L, Sherry ST, Feolo M, Hindorff LA
DJ, Swaminathan S, Karplus M (1983) (2014) Phenotype-Genotype Integrator
Charmm – a program for macromolecular (PheGenI): synthesizing genome-wide associ-
energy, minimization, and dynamics calcula- ation study (GWAS) data with existing
tions. J Comput Chem 4(2):187–217 genomic resources. Eur J Hum Genet
4. Humphrey W, Dalke A, Schulten K (1996) 22(1):144–147
VMD: visual molecular dynamics. J Mol Graph 13. Stenson PD, Ball EV, Mort M, Phillips AD,
14(1):33–38, 27-38 Shiel JA, Thomas NS, Abeysinghe S, Krawczak
5. Pettersen EF, Goddard TD, Huang CC, M, Cooper DN (2003) Human Gene Mutation
Couch GS, Greenblatt DM, Meng EC, Ferrin Database (HGMD): 2003 update. Hum Mutat
TE (2004) UCSF Chimera—a visualization 21(6):577–581
system for exploratory research and analysis. 14. Amberger JS, Bocchini CA, Schiettecatte F,
J Comput Chem 25(13):1605–1612 Scott AF, Hamosh A (2015) OMIM.org:
6. Wang Y, Geer LY, Chappey C, Kans JA, Bryant Online Mendelian Inheritance in Man
SH (2000) Cn3D: sequence and structure (OMIM(R)), an online catalog of human genes
views for Entrez. Trends Biochem Sci and genetic disorders. Nucleic Acids Res
25(6):300–302 43(Database issue):D789–D798
7. Mailman MD, Feolo M, Jin Y, Kimura M, 15. Landrum MJ, Lee JM, Riley GR, Jang W,
Tryka K, Bagoutdinov R, Hao L, Kiang A, Rubinstein WS, Church DM, Maglott DR
Paschall J, Phan L, Popova N, Pretel S, Ziyabari (2014) ClinVar: public archive of relationships
L, Lee M, Shao Y, Wang ZY, Sirotkin K, Ward among sequence variation and human pheno-
M, Kholodov M, Zbicz K, Beck J, Kimelman type. Nucleic Acids Res 42(Database
M, Shevelev S, Preuss D, Yaschenko E, Graeff issue):D980–D985
A, Ostell J, Sherry ST (2007) The NCBI 16. Forbes SA, Bindal N, Bamford S, Cole C, Kok
dbGaP database of genotypes and phenotypes. CY, Beare D, Jia M, Shepherd R, Leung K,
Nat Genet 39(10):1181–1186 Menzies A, Teague JW, Campbell PJ, Stratton
8. Rubinstein WS, Maglott DR, Lee JM, Kattman MR, Futreal PA (2011) COSMIC: mining
BL, Malheiro AJ, Ovetsky M, Hem V, complete cancer genomes in the Catalogue of
Gorelenkov V, Song G, Wallin C, Husain N, Somatic Mutations in Cancer. Nucleic Acids
Chitipiralla S, Katz KS, Hoffman D, Jang W, Res 39(suppl 1):D945–D950
Johnson M, Karmanov F, Ukrainchik A, 17. Weinstein JN, Collisson EA, Mills GB, Shaw
Denisenko M, Fomous C, Hudson K, Ostell KR, Ozenberger BA, Ellrott K, Shmulevich I,
JM (2013) The NIH genetic testing registry: a Sander C, Stuart JM, Cancer Genome Atlas
new, centralized database of genetic tests to Research N (2013) The Cancer Genome Atlas
enable access to comprehensive information Pan-Cancer analysis project. Nat Genet
and improve transparency. Nucleic Acids Res 45(10):1113–1120
41(Database issue):D925–D935 18. Singh A, Olowoyeye A, Baenziger PH, Dantzer
9. Sherry ST, Ward M, Sirotkin K (1999) dbSNP- J, Kann MG, Radivojac P, Heiland R, Mooney
database for single nucleotide polymorphisms SD (2008) MutDB: update on development of
and other classes of minor genetic variation. tools for the biochemical analysis of genetic
Genome Res 9(8):677–679 variation. Nucleic Acids Res 36(Database
10. Lappalainen I, Lopez J, Skipper L, Hefferon T, issue):D815–D819
Spalding JD, Garner J, Chen C, Maguire M, 19. Mottaz A, David FP, Veuthey AL, Yip YL
Corbett M, Zhou G, Paschall J, Ananiev V, (2010) Easy retrieval of single amino-acid
Flicek P, Church DM (2013) DbVar and polymorphisms and phenotype information
DGVa: public archives for genomic structural using SwissVar. Bioinformatics 26(6):851–852
20. Li M, Kales SC, Ma K, Shoemaker BA, Crespo- 29. Hernansaiz-Ballesteros RD, Salavert F,
Barreto J, Cangelosi AL, Lipkowitz S, Panchenko Sebastian-Leon P, Aleman A, Medina I,
AR (2015) Balancing protein stability and activ- Dopazo J (2015) Assessing the impact of
ity in cancer: a new approach for identifying mutations found in next generation sequenc-
driver mutations affecting CBL ubiquitin ligase ing data over human signaling pathways.
activation. Cancer Res 76(3):561–571 Nucleic Acids Res 43(W1):W270–W275
21. NR Coordinators (2014) Database resources 30. Choi Y, Sims GE, Murphy S, Miller JR, Chan
of the National Center for Biotechnology AP (2012) Predicting the functional effect of
Information. Nucleic Acids Res 42(Database amino acid substitutions and indels. PLoS One
issue):D7–D17 7(10):e46688
22. Perez B, Mechinaud F, Galambrun C, Ben 31. Adzhubei IA, Schmidt S, Peshkin L, Ramensky
Romdhane N, Isidor B, Philip N, Derain-Court VE, Gerasimova A, Bork P, Kondrashov AS,
J, Cassinat B, Lachenaud J, Kaltenbach S, Sunyaev SR (2010) A method and server for
Salmon A, Desiree C, Pereira S, Menot ML, predicting damaging missense mutations. Nat
Royer N, Fenneteau O, Baruchel A, Chomienne Methods 7(4):248–249
C, Verloes A, Cave H (2010) Germline muta- 32. Thusberg J, Olatubosun A, Vihinen M (2011)
tions of the CBL gene define a new genetic Performance of mutation pathogenicity predic-
syndrome with predisposition to juvenile tion methods on missense variants. Hum Mutat
myelomonocytic leukaemia. J Med Genet 32(4):358–368
47(10):686–691 33. Hashimoto K, Rogozin IB, Panchenko AR
23. Marchler-Bauer A, Derbyshire MK, Gonzales (2012) Oncogenic potential is related to acti-
NR, Lu S, Chitsaz F, Geer LY, Geer RC, He J, vating effect of cancer single and double
Gwadz M, Hurwitz DI, Lanczycki CJ, Lu F, somatic mutations in receptor tyrosine kinases.
Marchler GH, Song JS, Thanki N, Wang Z, Hum Mutat 33(11):1566–1575
Yamashita RA, Zhang D, Zheng C, Bryant SH 34. Schlebach JP, Narayan M, Alford C, Mittendorf
(2015) CDD: NCBI’s conserved domain data- KF, Carter BD, Li J, Sanders CR (2015)
base. Nucleic Acids Res 43(D1):D222–D226 Conformational stability and pathogenic mis-
24. Ashkenazy H, Erez E, Martz E, Pupko T, Ben- folding of the integral membrane protein
Tal N (2010) ConSurf 2010: calculating evolu- PMP22. J Am Chem Soc 137(27):8758–8768
tionary conservation in sequence and structure 35. Potapov V, Cohen M, Schreiber G (2009)
of proteins and nucleic acids. Nucleic Acids Res Assessing computational methods for predict-
38(suppl 2):W529–W533 ing protein stability upon mutation: good on
25. Shoemaker BA, Zhang D, Thangudu RR, average but not in the details. Protein Eng Des
Tyagi M, Fong JH, Marchler-Bauer A, Bryant Sel 22(9):553–560
SH, Madej T, Panchenko AR (2010) Inferred 36. Khan S, Vihinen M (2010) Performance of
Biomolecular Interaction Server – a web server protein stability predictors. Hum Mutat
to analyze and predict protein interacting part- 31(6):675–684
ners and binding sites. Nucleic Acids Res
38(Database issue):D518–D524 37. Zhang Z, Wang L, Gao Y, Zhang J,
Zhenirovskyy M, Alexov E (2012) Predicting
26. Goncearenco A, Shaytan AK, Shoemaker BA, folding free energy changes upon single point
Panchenko AR (2015) Structural perspectives mutations. Bioinformatics 28(5):664–671
on the evolutionary expansion of unique
protein-protein binding sites. Biophys 38. Nishi H, Tyagi M, Teng S, Shoemaker BA,
J 109(6):1295–1306 Hashimoto K, Alexov E, Wuchty S, Panchenko
AR (2013) Cancer missense mutations alter
27. Shoemaker BA, Zhang D, Tyagi M, Thangudu binding properties of proteins and their inter-
RR, Fong JH, Marchler-Bauer A, Bryant SH, action networks. PLoS One 8(6):e66273
Madej T, Panchenko AR (2012) IBIS (Inferred
Biomolecular Interaction Server) reports, pre- 39. Teng S, Madej T, Panchenko A, Alexov E
dicts and integrates multiple types of conserved (2009) Modeling effects of human single
interactions for proteins. Nucleic Acids Res nucleotide polymorphisms on protein-protein
40(Database issue):D834–D840 interactions. Biophys J 96(6):2178–2188
28. Dou H, Buetow L, Hock A, Sibbet GJ, 40. Ghersi D, Singh M (2014) Interaction-based
Vousden KH, Huang DT (2012) Structural discovery of functionally important genes in
basis for autoinhibition and phosphorylation- cancers. Nucleic Acids Res 42(3):e18
dependent activation of c-Cbl. Nat Struct Mol 41. Li M, Petukh M, Alexov E, Panchenko AR
Biol 19(2):184–192 (2014) Predicting the impact of missense
mutations on protein-protein binding affinity. cBio cancer genomics portal: an open platform
J Chem Theory Comput 10(4):1770–1780 for exploring multidimensional cancer genom-
42. Porollo A, Meller J (2007) Prediction-based ics data. Cancer Discov 2(5):401–404
fingerprints of protein–protein interactions. 55. Peri S, Navarro JD, Amanchy R, Kristiansen
Proteins 66(3):630–645 TZ, Jonnalagadda CK, Surendranath V,
43. Qin S, Zhou H-X (2007) meta-PPISP: a meta Niranjan V, Muthusamy B, Gandhi TK,
web server for protein-protein interaction site Gronborg M, Ibarrola N, Deshpande N,
prediction. Bioinformatics 23(24):3386–3387 Shanker K, Shivashankar HN, Rashmi BP,
44. Zhou H-X, Qin S (2007) Interaction-site pre- Ramya MA, Zhao Z, Chandrika KN, Padma N,
diction for protein complexes: a critical assess- Harsha HC, Yatish AJ, Kavitha MP, Menezes
ment. Bioinformatics 23(17):2203–2209 M, Choudhury DR, Suresh S, Ghosh N,
Saravana R, Chandran S, Krishna S, Joy M,
45. Porollo A, Meller J (2012) Computational Anand SK, Madavan V, Joseph A, Wong GW,
methods for prediction of protein-protein Schiemann WP, Constantinescu SN, Huang L,
interaction sites. Protein-Protein Interactions – Khosravi-Far R, Steen H, Tewari M, Ghaffari
Computational and Experimental Tools S, Blobe GC, Dang CV, Garcia JG, Pevsner J,
472:3–26 Jensen ON, Roepstorff P, Deshpande KS,
46. Li M, Simonetti FL, Goncearenco A, Chinnaiyan AM, Hamosh A, Chakravarti A,
Panchenko AR (2016) MutaBind estimates Pandey A (2003) Development of human pro-
and interprets the effects of sequence variants tein reference database as an initial platform for
on protein-protein interactions. Nucleic Acids approaching systems biology in humans.
Res. Jul 8;44(W1):W494-501. Genome Res 13(10):2363–2371
47. Li M, Shoemaker BA, Thangudu RR, Ferraris 56. Kerrien S, Aranda B, Breuza L, Bridge A,
JD, Burg MB, Panchenko AR (2013) Broackes-Carter F, Chen C, Duesbury M,
Mutations in DNA-binding loop of NFAT5 Dumousseau M, Feuermann M, Hinz U,
transcription factor produce unique outcomes Jandrasits C, Jimenez RC, Khadake J,
on protein-DNA binding and dynamics. J Phys Mahadevan U, Masson P, Pedruzzi I,
Chem B 117(42):13226–13234 Pfeiffenberger E, Porras P, Raghunath A,
48. Case DA, Cheatham TE, Darden T, Gohlke H, Roechert B, Orchard S, Hermjakob H (2012)
Luo R, Merz KM, Onufriev A, Simmerling C, The IntAct molecular interaction database in
Wang B, Woods RJ (2005) The Amber biomo- 2012. Nucleic Acids Res 40(Database
lecular simulation programs. J Comput Chem issue):D841–D846
26(16):1668–1688 57. Joshi-Tope G, Gillespie M, Vastrik I,
49. Stefl S, Nishi H, Petukh M, Panchenko AR, D'Eustachio P, Schmidt E, de Bono B, Jassal B,
Alexov E (2013) Molecular mechanisms of Gopinath GR, Wu GR, Matthews L, Lewis S,
disease-causing missense mutations. J Mol Biol Birney E, Stein L (2005) Reactome: a knowl-
425(21):3919–3936 edgebase of biological pathways. Nucleic Acids
50. Mizuguchi K, Deane CM, Blundell TL, Res 33(Database issue):D428–D432
Johnson MS, Overington JP (1998) JOY: pro- 58. Kanehisa M, Goto S, Hattori M, Aoki-
tein sequence-structure representation and Kinoshita KF, Itoh M, Kawashima S, Katayama
analysis. Bioinformatics 14(7):617–623 T, Araki M, Hirakawa M (2006) From genom-
51. Tina KG, Bhadra R, Srinivasan N (2007) PIC: ics to chemical genomics: new developments in
protein interactions calculator. Nucleic Acids KEGG. Nucleic Acids Res 34(Database
Res 35(suppl 2):W473–W476 issue):D354–D357
52. Stenson P, Mort M, Ball E, Shaw K, Phillips A, 59. Niknafs N, Kim D, Kim R, Diekhans M, Ryan
Cooper D (2014) The Human Gene Mutation M, Stenson PD, Cooper DN, Karchin R
Database: building a comprehensive mutation (2013) MuPIT interactive: webserver for map-
repository for clinical and molecular genetics, ping variant positions to annotated, interactive
diagnostic testing and personalized genomic 3D structures. Hum Genet
medicine. Hum Genet 133(1):1–9 132(11):1235–1243
53. Thorn CF, Klein TE, Altman RB (2010) 60. Peterson TA, Adadey A, Santana-Cruz I, Sun
Pharmacogenomics and bioinformatics: Y, Winder A, Kann MG (2010) DMDM:
PharmGKB. Pharmacogenomics domain mapping of disease mutations.
11(4):501–505 Bioinformatics 26(19):2458–2459
54. Cerami E, Gao J, Dogrusoz U, Gross BE, 61. Jegga AG, Gowrisankar S, Chen J, Aronow BJ
Sumer SO, Aksoy BA, Jacobsen A, Byrne CJ, (2007) PolyDoms: a whole genome database
Heuer ML, Larsson E, Antipin Y, Reva B, for the identification of non-synonymous cod-
Goldberg AP, Sander C, Schultz N (2012) The ing SNPs with the potential to impact disease.
Nucleic Acids Res 35(Database evolutionary information. Bioinformatics

issue):D700–D706 22(22):2729–2734
62. Ng PC, Henikoff S (2003) SIFT: predicting 74. Al-Numair NS, Martin AC (2013) The SAAP
amino acid changes that affect protein func- pipeline and database: tools to analyze the
tion. Nucleic Acids Res 31(13):3812–3814 impact and predict the pathogenicity of muta-
63. Yue P, Melamud E, Moult J (2006) SNPs3D: tions. BMC Genomics 14(Suppl 3):S4
candidate gene and SNP selection for associa- 75. Yates CM, Filippis I, Kelley LA, Sternberg MJ
tion studies. BMC Bioinformatics 7:166 (2014) SuSPect: enhanced prediction of single
64. Ferrer-Costa C, Gelpi JL, Zamakola L, Parraga amino acid variant (SAV) phenotype using net-
I, de la Cruz X, Orozco M (2005) PMUT: a work features. J Mol Biol 426(14):2692–2701
web-based tool for the annotation of patho- 76. Simonetti FL, Tornador C, Nabau-Moreto N,
logical mutations on proteins. Bioinformatics Molina-Vila MA, Marino-Buslje C (2014) Kin-
21(14):3176–3178 Driver: a database of driver mutations in pro-
65. Bromberg Y, Rost B (2007) SNAP: predict tein kinases. Database 2014:bau104.
effect of non-synonymous polymorphisms on 77. McSkimming DI, Dastgheib S, Talevich E,
function. Nucleic Acids Res Narayanan A, Katiyar S, Taylor SS, Kochut K,
35(11):3823–3835 Kannan N (2015) ProKinO: a unified resource
66. Shihab HA, Gough J, Cooper DN, Stenson for mining the cancer kinome. Hum Mutat
PD, Barker GL, Edwards KJ, Day IN, Gaunt 36(2):175–186
TR (2013) Predicting the functional, molecu- 78. Guerois R, Nielsen JE, Serrano L (2002)
lar, and phenotypic consequences of amino Predicting changes in the stability of proteins
acid substitutions using hidden Markov mod- and protein complexes: a study of more than
els. Hum Mutat 34(1):57–65 1000 mutations. J Mol Biol 320(2):369–387
67. Reva B, Antipin Y, Sander C (2011) Predicting 79. Dehouck Y, Grosfils A, Folch B, Gilis D,
the functional impact of protein mutations: Bogaerts P, Rooman M (2009) Fast and accu-
application to cancer genomics. Nucleic Acids rate predictions of protein stability changes
Res 39(17):e118 upon mutations using statistical potentials and
68. Carter H, Chen S, Isik L, Tyekucheva S, neural networks: PoPMuSiC-2.0.
Velculescu VE, Kinzler KW, Vogelstein B, Bioinformatics 25(19):2537–2543
Karchin R (2009) Cancer-specific high- 80. Yin S, Ding F, Dokholyan NV (2007) Eris: an
throughput annotation of somatic mutations: automated estimator of protein stability. Nat
computational prediction of driver missense Methods 4(6):466–467
mutations. Cancer Res 69(16):6660–6667 81. Parthiban V, Gromiha MM, Schomburg D
69. Capriotti E, Calabrese R, Fariselli P, Martelli (2006) CUPSAT: prediction of protein stabil-
PL, Altman RB, Casadio R (2013) ity upon point mutations. Nucleic Acids Res
WS-SNPs&GO: a web server for predicting the 34(Web Server Issue):W239–242
deleterious effect of human protein variants 82. Potapov V, Cohen M, Inbar Y, Schreiber G
using functional annotation. BMC Genomics (2010) Protein structure modelling and evalu-
14(Suppl 3):S6 ation based on a 4-distance description of side-
70. Wang M, Zhao XM, Takemoto K, Xu H, Li Y, chain interactions. BMC Bioinformatics
Akutsu T, Song J (2012) FunSAV: predicting 11:374–374
the functional effect of single amino acid vari- 83. Deutsch C, Krishnamoorthy B (2007) Four-
ants using a two-stage random forest model. body scoring function for mutagenesis.
PLoS One 7(8):e43847 Bioinformatics 23(22):3009–3015
71. Bao L, Zhou M, Cui Y (2005) nsSNPAnalyzer: 84. Willard L, Ranjan A, Zhang H, Monzavi H,
identifying disease-associated nonsynonymous Boyko RF, Sykes BD, Wishart DS (2003)
single nucleotide polymorphisms. Nucleic VADAR: a web server for quantitative evalua-
Acids Res 33(Web Server Issue):W480–482 tion of protein structure quality. Nucleic Acids
72. Mi H, Muruganujan A, Thomas PD (2013) Res 31(13):3316–3319
PANTHER in 2013: modeling the evolution 85. Giollo M, Martin AJ, Walsh I, Ferrari C,
of gene function, and other gene attributes, in Tosatto SC (2014) NeEMO: a method using
the context of phylogenetic trees. Nucleic residue interaction networks to improve pre-
Acids Res 41(Database issue):D377–D386 diction of protein stability upon mutation.
73. Capriotti E, Calabrese R, Casadio R (2006) BMC Genomics 15(Suppl 4):S7
Predicting the insurgence of human genetic 86. Pires DE, Ascher DB, Blundell TL (2014)
diseases associated to single point protein DUET: a server for predicting effects of muta-
mutations with support vector machines and tions on protein stability using an integrated
computational approach. Nucleic Acids Res based method for predicting protein stability
42(Web Server Issue):W314–319 changes upon mutations. Bioinformatics
87. Laimer J, Hofer H, Fritz M, Wegenkittl S, 23(10):1292–1293
Lackner P (2015) MAESTRO – multi agent 93.
Dehouck Y, Kwasigroch JM, Rooman
stability prediction upon point mutations. M, Gilis D (2013) BeAtMuSiC: predic-
BMC Bioinformatics 16(1):116 tion of changes in protein–protein bind-
88. Capriotti E, Fariselli P, Rossi I, Casadio R ing affinity on mutations. Nucleic Acids Res
(2008) A three-state prediction of single point 41(W1):W333–W339
mutations on protein stability changes. BMC 94. Berliner N, Teyra J, Çolak R, Garcia Lopez S,
Bioinformatics 9(Suppl 2):S6 Kim PM (2014) Combining structural model-
89. Cheng J, Randall A, Baldi P (2006) Prediction ing with ensemble machine learning to accu-
of protein stability changes for single-site muta- rately predict protein fold stability and binding
tions using support vector machines. Proteins affinity effects upon mutation. PLoS One
62(4):1125–1132 9(9):e107353
90. Chen CW, Lin J, Chu YW (2013) iStable: off- 95. Kruger DM, Gohlke H (2010) DrugScorePPI
the-
shelf predictor integration for predicting webserver: fast and accurate in silico alanine
protein stability changes. BMC Bioinformatics scanning for scoring protein-protein interac-
14(Suppl 2):S5 tions. Nucleic Acids Res 38(Web Server
91. Teng S, Srivastava A, Wang L (2010) Sequence Issue):W480–486
feature-based prediction of protein stability 96. Zhao N, Han JG, Shyu CR, Korkin D (2014)
changes upon amino acid substitutions. BMC Determining effects of non-synonymous SNPs
Genomics 11(Suppl 2):1–8 on protein-protein interactions using super-
92. Huang L-T, Gromiha MM, Ho S-Y (2007) vised and semi-supervised learning. PLoS
iPTREE-STAB: interpretable decision tree Comput Biol 10(5):e1003592
Chapter 18
Protein Micropatterning Assay: Quantitative Analysis

of Protein–Protein Interactions
Gerhard J. Schütz, Julian Weghuber, Peter Lanzerstorfer,
and Eva Sevcsik
Abstract
Characterization, especially quantification, of protein interactions in live cells is usually not an easy
endeavor. Here, we describe a straightforward method to identify and quantify the interaction of a mem-
brane protein (“bait”) and a fluorescently labeled interaction partner (“prey”) (membrane-bound or cyto-
solic) in live cells using Total Internal Reflection Fluorescence microscopy. The bait protein is immobilized
within patterns in the plasma membrane (e.g., via an antibody); the bait–prey interaction strength can be
quantified by determining the prey bulk fluorescence intensity with respect to the bait patterns. This
method is particularly suitable also for the analysis of weak, transient interactions that are not easily acces-
sible with other methods.
Key words Micropatterning, Protein–protein interactions, Soft lithography, TIRF microscopy,

Quantitive analysis, Membrane proteins
1 Introduction
Although there are many methods to analyze protein–protein

interactions, quantitative analysis of protein interactions in live
cells is still less than straightforward. Most approaches rely on
immunoprecipitation, affinity purification or chemical crosslinking
and, thus, analysis of cell lysates [1, 2]. In live cells, assays are rather
challenging, laborious, suffer from detection of false positives or
negatives, do not allow for easy quantification, and/or are not
readily accessible for many labs (e.g., bimolecular fluorescence
complementation [3], yeast two-hybrid screen [4], fluorescence
resonance energy transfer [5], or single-molecule methods [6]).
Protein micropatterning is a technique that circumvents many
of these problems: it is simple, inexpensive, does not need elaborate
equipment, can also capture transient interactions, and is performed
in live cells, and data analysis is uncomplicated. The method is based
on the work of several groups who forced membrane proteins into
261
262 Gerhard J. Schütz et al.
specific patterns within the plasma membrane of living cells [7, 8].
We have extended this approach to use it as a tool for characteriza-
tion and quantification of protein interactions: One interaction
partner (bait) is restricted to specific regions (typically regular
micropatterns) in the live cell plasma membrane and the lateral dis-
tribution of a fluorescently labeled interaction partner (prey) is
monitored. In case of an interaction, prey molecules will follow the
bait pattern; homogeneous distribution of prey protein in the
plasma membrane indicates the absence of an interaction (Fig. 1).
Quantification can be achieved by comparing the prey signal inten-
sity within and outside the bait regions: the signal contrast between
these regions provides a measure of the interaction strength.
While patterned surfaces can be generated by different meth-
ods (e.g., photolithography [9] or dip-pen nanolithography [10]),
soft lithography [11] is probably the most convenient: it is fast,
simple, and lends itself to high throughput routines. In this proto-
col, the patterned cell substrate is produced by printing streptavi-
din patterns on a glass coverslip, to which a bait-specific biotinylated
antibody is then attached. We have first used this approach to
Fig. 1 Principle of protein micropatterning in the plasma membrane. (a) Sketch

and (b) TIRF image of a cell grown on a micropatterned substrate. Bait antibody
is arranged in a regular pattern of 3 μm sized dots with 3 μm interspaces. The
bait protein (unlabeled) reorganizes according to the antibody patterns, but the
fluorescently labeled prey protein is distributed homogeneously in the plasma
membrane, indicating no interaction between bait and prey protein. Scale bar is
7 μm. (c, d) As in (a, b), but here the prey protein interacts strongly with the bait
protein and localizes according to the bait patterns. The cell outline is indicated
by a dashed white contour line
Protein Micropatterning Assay 263
c haracterize the interaction of two proteins involved in immuno-

signaling: CD4, a transmembrane protein, and the tyrosine kinase
Lck, a palmitoylated protein that is transiently associated with the
plasma membrane [12]. Since then, it has been applied to charac-
terize various protein–protein interactions in several different cell
types [10, 13–17] and has been used to determine protein binding
curves [18] and dissociation constants [18]. Recently, we have also
used Protein Micropatterning to interrogate lipid-mediated pro-
tein interactions [19]. Versions of the Protein Micropatterning
assay have been reviewed in [20, 21].
2 Materials
Prepare all work solutions fresh each time. Store epoxy-coated cov-
erslips in the desiccator after opening. This protocol is optimized
for PDMS stamps; if a different material is used, conditions may
need to be adjusted for optimal printing results.
1. Polydimethylsiloxane (PDMS) stamps (see Note 1).
2. Epoxy-coated coverslips: NEXTERION® slide E (Schott,
Germany).
3. Streptavidin stock solution: dissolve 0.5 mg/ml streptavidin
(Sigma, USA) in phosphate buffered saline (PBS) pH 7.4.
Store aliquots at −20 °C. Do not freeze and thaw.
4. Streptavidin work solution: dilute streptavidin stock solution
to 50 μg/ml in PBS pH 7.4.
5. Secure SealTM Hybridization chambers (Grace Biolabs, USA).
6. BSA-Cy5 stock solution (see Note 2): dissolve Cy5-labeled
bovine serum albumin (BSA-Cy5; Nanocs, USA) to 1 mg/ml in
PBS pH 7.4. Store aliquots at −20 °C. Do not freeze and thaw.
7. BSA-Cy5 work solution: dilute BSA-Cy5 stock solution to
100 μg/ml in PBS pH 7.4.
8. Antibody work solution: dilute biotinylated antibody to
10 μg/ml in PBS pH 7.4 containing 1 % BSA.
9. Imaging buffer: Hank’s Balanced Salt Solution (HBSS) with
Ca2+ and Mg2+ and 2 % fetal calf serum (FCS) (see Note 3).
10. Cells expressing bait proteins and fluorescent prey proteins (see
Note 4), Accutase (Sigma, USA) (see Note 5).
3 Methods
Carry out all procedures at room temperature unless otherwise

specified.
Fig. 2 Soft lithography and functionalization. (a) Streptavidin work solution is

incubated on a PDMS stamp. (b) After washing and drying of the stamp, strepta-
vidin is printed onto an epoxy-coated coverslip. (c, d) The stamp is removed;
BSA-Cy5 is added to fill the interspaces. (e) When biotinylated antibody is added,
it binds specifically to the streptavidin patterns
3.1 Soft Lithography The workflow of “3.1 Soft lithography and functionalization” is
and Functionalization sketched in Fig. 2.
1. Wash PDMS stamp by rinsing with ethanol (p.a.) and ultra-
pure water. Dry the PDMS stamp under a stream of a dry inert
gas such as nitrogen or argon.
2. Place ~50 μL of streptavidin work solution (50 μg/ml) on the
PDMS stamp (the whole pattern area should be covered). Let
protein adsorb to stamp for 15 min at room temperature (see
Note 6).
3. Wash the PDMS stamp by rinsing carefully with water and dry
under a stream of nitrogen or argon.
4. Place the PDMS stamp face-down under its own weight onto
an epoxy-coated coverslip and incubate for 30 min at room
temperature or overnight at 4 °C in a humidified atmosphere
(e.g., a petri dish with a wet tissue) (see Note 7).
5. Mark the position of the patterned area on the back of the
coverslip with a water-resistant marker and separate the stamp
from the slide using tweezers (see Note 8).
6. Stick a Secure SealTM Hybridization chamber over the marked

area.
7. Add BSA-Cy5 work solution (100 μg/ml) to the hybridization
chamber and incubate for 15 min at room temperature (see
Note 9).
8. Wash with 500 μl PBS by adding the buffer into one port of
the hybridization chamber and removing it at the second port.
9. Add antibody work solution (10 μg/ml) to the hybridization
chamber and incubate for 15 min at room temperature.
10. Wash with 500 μl PBS.
11. Store the micropatterned surfaces with PBS in the dark at
room temperature until seeding of cells (see Note 10).
3.2 Seeding Cells 1. Grow adherent cells expressing bait and prey proteins of inter-
est to 70 % confluency in a 10 cm tissue culture dish.
2. Detach cells with Accutase® solution and centrifuge 4 min at
300 × g. This protocol has been tested for T24, HeLa and
CHO cells (see Note 11).
3. Pellet cells by spinning for 5 min at ~300 × g.
4. Discard the supernatant and resuspend the cell pellet in 1 ml of
the appropriate growth medium. Then, dilute this ~1:10 in
growth medium (see Note 12).
5. Remove the PBS from the hybridization chamber on the
micropatterned coverslip and seed cell suspension.
6. Check cell density on a light microscope. Cells should be sin-
gle but not too sparse.
7. Put coverslips in a petri dish humidity chamber to prevent the
sample from running dry and incubate for 1.5–2 h at 37 °C in
a 5 % CO2 atm.
8. Before analyzing the cells on the microscope, replace the
medium with imaging buffer.
3.3 Total Internal 1. Place the coverslip on a TIRF microscope in a suitable mount
Reflection (see Note 13).
Fluorescence (TIRF) 2. The BSA-Cy5 grid needed for quantitative analysis is recorded
Microscopy at 647 nm.
3. Distribution of fluorescent prey protein (tagged with e.g.,
GFP) is recorded (at e.g., 488 nm (see Note 14)).
3.4 Contrast 1. Export microscopy images as 8-bit TIF image. For contrast
Quantitation quantitation it is necessary to export images of the fluorescent
prey/bait protein (Fig. 3a) and the respective image with the
BSA-Cy5 grid (Fig. 3b). Figure 3c shows the overlaid images.
2. 8-bit TIF images are imported in the semi-automated micropa-

tterning analysis software (“Spotty”, see Note 15).
3. An automatic gridding algorithm is used to calculate the grid-
size and the rotation angle ϕ of the used image. The algorithm
automatically determines the grid parameters that correctly fit
the micropatterned structure (see Note 16). Cells to be ana-
lyzed are detected automatically or can also be selected manu-
ally (Fig. 3d).
4. Based on the correct identification of the grid position with
respect to fluorescent patterns, the fluorescence contrast can
be calculated for each pattern in the image as C = (F+–F−)/
(F+–FBG), where F+ denotes the average intensity of the inner
pixels of the micropatterns, F− the average intensity of the pix-
els surrounding the micropatterns, and FBG the intensity of
the global background (see Note 17) (Fig. 3e).
Fig. 3 Quantitation of protein interactions using “Spotty”. Image recorded of the fluorescently labeled prey
protein (a) and the corresponding BSA-Cy5 grid (b). (c) Overlay. (d) An automatic gridding algorithm automati-
cally optimizes the grid parameters and produces a grid that correctly fits the micropatterns. Yellow lines
denote the cell areas to be used for analysis. (e) The grid subdivides the total image into adjacent squares,
each of which is quantified according to the average signal within a central circle comprising the micropattern
spot (F+) and the signal outside this circle (F−). (f) Statistical analysis of multiple cells is shown in a 2D histo-
gram of the fluorescence brightness and contrast. The color scale corresponds to the number of events (i.e.,
individual analyzed spots)
Fig. 4 Examples of generated 2D histograms (a) T24 cell transiently expressing CD4 and Lck-YFP grown on CD4
antibody patterns. Lck-YFP interacts strongly with the patterned CD4, which is also reflected in the high contrast
values shown in the 2D histogram on the right. The low contrast values at lower fluorescence intensities are
probably a result of low CD4 (and Lck-YFP) expression levels of a cell subpopulation. For calculating the mean
contrast <C>, we only consider data points above a certain intensity threshold (indicated by the yellow line). (b)
T24 cell transiently expressing CD4 and cytosolic YFP grown on CD4 antibody patterns. No copatterning of YFP
with CD4 can be observed, the contrast values fluctuate around zero. Scale bars are 10 μm. The color scale
corresponds to the number of events (i.e., individual analyzed spots). Figure modified from [12]
5. Several fluorescence parameters (e.g., mean brightness, back-

ground fluorescence, contrast,…) as well as graphical descrip-
tions can be extracted from the software for further processing.
For statistical analysis of multiple cells, we find it useful to present
the data in two-dimensional histograms, with the fluorescence
brightness F = F+–FBG on the ordinate against the signal contrast
C on the abscissa (Fig. 3f) (see Note 18). To facilitate compari-
son of two-dimensional histograms, we use the mean contrast
<C>. Figure 4 shows examples of 2D histograms in the presence
and absence of protein–protein interaction, yielding high con-
trast and low contrast values in the 2D histograms, respectively.
4 Notes
1. PDMS is an often-used and reliable material for soft lithogra-

phy, but it is rather soft. Stamp feature sizes need to be above
1 μm. We prefer stamps featuring regularly spaced dots (3 μm
in size, with 3 μm interspaces).
2. BSA can also be labeled with a different fluorophore. Its fluo-

rescence should be spectrally separated from the fluorescence
of the prey protein.
3. Growth medium is exchanged for imaging buffer (a) to reduce
background fluorescence (if Phenol Red-containing medium is
used) and (b) to keep cells at pH 7.4 during measurements.
4. For initial tests, it is convenient to use cells expressing a fluores-
cent bait protein. This way, successful immobilization of bait
protein at the antibody patterns can be evaluated. Alternatively,
this can also be verified by staining patterned bait protein with
a fluorescently labeled antibody targeting a different epitope
than the biotinylated capture antibody. It may be useful, how-
ever, to use antibody Fab-fragments, since full antibodies may
be excluded from very densely populated patterns.
5. We use Accutase to detach cells because it is gentler than tryp-
sin but equally efficient for most cell types. We found that e.g.,
loss of glycosylphosphatidylinositol-anchored proteins from
the cell surface was significantly reduced when using Accutase
instead of trypsin.
6. You can use the pipet tip to spread the streptavidin drop. Do
not touch the stamp surface.
7. Water is needed for the streptavidin binding covalently to the
epoxy-coated coverslips. In their protocol for protein printing
onto Nexterion E coverslips, the manufacturer suggests a
humidity of 75 % during printing. We found that using a wet
tissue in a petri dish gives satisfactory results.
8. Be careful to lift the stamp without dragging it across the
surface.
9. BSA-Cy5 serves two purposes: (a) blocking areas of the cover-
slips not covered with streptavidin (interspaces) and (b) pro-
viding the grid necessary for quantitative analysis.
10. We have found that micropatterned surfaces with the stamps
still attached can be stored at 4 °C for at least 2 days without
losing imprint quality.
11. Other adherent cell types may be suited for micropatterning as
well. For some cell types it may be beneficial to replace
BSA-Cy5 (completely or partially) with fibronectin or polyly-
sine to promote cell adhesion in the interspaces between strep-
tavidin regions.
12. Best results will be obtained when cells are plated to ~30–50 %
confluency. We use growth medium without Phenol Red to
reduce background fluorescence.
13. TIRF microscopy is used to ensure that only membrane-bound
prey protein is detected. Otherwise, detection of cytosolic prey
protein can lead to an apparently reduced contrast.
14. When using this assay for the first time, we recommend using
a fluorescently labeled bait protein as described in Note 4. If
the fluorescence signals of bait, prey and analysis grid are suf-
ficiently spectrally separated, labeled bait protein can be used
for all measurements.
15. “Spotty” can be obtained from www.protein-interaction-lab.at
upon request.
16. Evolutionary computation strategies are used for optimized
grid identification and detection of micropatterns in biological
samples.
17. A relevant factor for the success of contrast evaluation is the
size of the F+ region. It has to be adjusted to fit the actual size
of the printed patterns (as shown in Fig. 3e).
18. Taking into account the fluorescence brightness is especially
useful when dealing with a heterogeneous cell population with
very different expression levels of bait and prey protein (see also
Fig. 4). It may be advantageous to analyze cell subpopulations
of different expression levels separately, or to apply an intensity
threshold as shown in Fig. 4.
Acknowledgments
This work was funded by the Austrian Science Fund (FWF projects
P 26337 and P 25730), the Austrian Research Promotion Agency
(FFG project 842379), the program ‘Regionale
Wettbewerbsfähigkeit OÖ 2007–2013’ with the financial support
of the European Fund for Regional Development, as well as the
Federal State of Upper Austria.
References
1. Barrios-Rodiles M, Brown KR, Ozdamar B, 4. Young KH (1998) Yeast two-hybrid: so many

Bose R, Liu Z, Donovan RS, Shinjo F, Liu Y, interactions, (in) so little time. Biol Reprod
Dembowy J, Taylor IW, Luga V, Przulj N, 58:302–311
Robinson M, Suzuki H, Hayashizaki Y, Jurisica 5. Maurel D, Comps-Agrar L, Brock C, Rives
I, Wrana JL (2005) High-throughput mapping ML, Bourrier E, Ayoub MA, Bazin H, Tinel N,
of a dynamic signaling network in mammalian Durroux T, Prezeau L, Trinquet E, Pin JP
cells. Science 307:1621–1625 (2008) Cell-surface protein-protein interaction
2. Rigaut G, Shevchenko A, Rutz B, Wilm M, analysis with time-resolved FRET and snap-tag
Mann M, Seraphin B (1999) A generic protein technologies: application to GPCR oligomer-
purification method for protein complex char- ization. Nat Methods 5:561–567
acterization and proteome exploration. Nat 6. Suzuki KG, Fujiwara TK, Sanematsu F, Iino R,
Biotechnol 17:1030–1032 Edidin M, Kusumi A (2007) GPI-anchored
3. Kerppola TK (2006) Design and implementation receptor clusters transiently recruit Lyn and Ga
of bimolecular fluorescence complementation for temporary cluster immobilization and Lyn
(BiFC) assays for the visualization of protein inter- activation: single-molecule tracking study 1.
actions in living cells. Nat Protoc 1:1278–1286 J Cell Biol 177:717–730
7. Orth RN, Wu M, Holowka D, Craighead HG, 15. Lanzerstorfer P, Borgmann D, Schutz G,

Baird B (2003) Mast cell activation on pat- Winkler SM, Hoglinger O, Weghuber J (2014)
terned lipid bilayers of subcellular dimensions. Quantification and kinetic analysis of Grb2-
Langmuir 19:1599–1605 EGFR interaction on micro-patterned surfaces
8. Mossman KD, Campi G, Groves JT, Dustin for the characterization of EGFR-modulating
ML (2005) Altered TCR signaling from geo- substances. PLoS One 9:e92151
metrically repatterned immunological synapses. 16. Lanzerstorfer P, Yoneyama Y, Hakuno F,
Science 310:1191–1193 Muller U, Hoglinger O, Takahashi S,
9. Waichman S, You C, Beutel O, Bhagawati M, Weghuber J (2015) Analysis of insulin receptor
Piehler J (2011) Maleimide photolithography substrate signaling dynamics on microstruc-
for single-molecule protein-protein interaction tured surfaces. FEBS J 282:987–1005
analysis in micropatterns. Anal Chem 17. Bashour KT, Gondarenko A, Chen H, Shen K,
83(2):501–508 Liu X, Huse M, Hone JC, Kam LC (2014)
10. Gandor S, Reisewitz S, Venkatachalapathy M, CD28 and CD3 have complementary roles in
Arrabito G, Reibner M, Schröder H, Ruf K, T-cell traction forces. Proc Natl Acad Sci U S A
Niemeyer C, Bastiaens P, Dehmelt L (2013) A 111:2241–2246
protein-interaction array inside a living cell. 18. Sunzenauer S, Zojer V, Brameshuber M, Trols
Angewandte Chemie 52:4790–4794 A, Weghuber J, Stockinger H, Schutz GJ
11. Kane RS, Takayama S, Ostuni E, Ingber DE, (2013) Determination of binding curves via
Whitesides GM (1999) Patterning proteins protein micropatterning in vitro and in living
and cells using soft lithography. Biomaterials cells. Cytometry A 83:847–854
20:2363–2376 19. Sevcsik E, Brameshuber M, Folser M,
12. Schwarzenbacher M, Kaltenbrunner M, Weghuber J, Honigmann A, Schutz GJ (2015)
Brameshuber M, Hesch C, Paster W, Weghuber GPI-anchored proteins do not reside in
J, Heise B, Sonnleitner A, Stockinger H, ordered domains in the live cell plasma mem-
Schütz GJ (2008) Micropatterning for quanti- brane. Nat Commun 6:6969
tative analysis of protein-protein interactions in 20. Weghuber J, Brameshuber M, Sunzenauer S,
living cells. Nat Methods 5:1053–1060 Lehner M, Paar C, Haselgrubler T,
13. Weghuber J, Sunzenauer S, Plochberger B, Schwarzenbacher M, Kaltenbrunner M, Hesch
Brameshuber M, Haselgrubler T, Schutz GJ C, Paster W, Heise B, Sonnleitner A, Stockinger
(2010) Temporal resolution of protein-protein H, Schutz GJ (2010) Detection of protein-
interactions in the live-cell plasma membrane. protein interactions in the live cell plasma
Anal Bioanal Chem 397:3339–3347 membrane by quantifying prey redistribution
14. Alexander RA, Prager GW, Mihaly-Bison J, upon bait micropatterning. Methods Enzymol
Uhrin P, Sunzenauer S, Binder BR, Schutz GJ, 472:133–151
Freissmuth M, Breuss JM (2012) VEGF- 21. Weghuber J, Sunzenauer S, Brameshuber M,
induced endothelial cell migration requires Plochberger B, Hesch C, Schutz G. J (2010)
urokinase receptor (uPAR)-dependent integrin In-vivo detection of protein-protein interac-
redistribution. Cardiovasc Res 94:125–135 tions on micro-patterned surfaces. J Vis Exp 37
Chapter 19
Designing Successful Proteomics Experiments

Daniel Ruderman
Abstract
Because proteomics experiments are so complex they can readily fail, and do so without clear cause. Using
standard experimental design techniques and incorporating quality control can greatly increase the chances
of success. This chapter introduces the relevant concepts and provides examples specific to proteomic
workflows. Applying these notions to design successful proteomics experiments is straightforward. It can
help identify failure causes and greatly increase the likelihood of inter-laboratory reproducibility.
Key words Design of experiments, Randomization, Bias, Variance
1 Introduction
This chapter’s goal is to help researchers design proteomics experi-

ments that succeed. I will present the concepts of experimental
design in the context of proteomics workflows. The particular
combination of experimental factors that impact proteomic offers
specific challenges for their design. Since each experiment is
unique, some extrapolation to particular circumstances will be
needed. I hope to provide the tools to do so. Additionally, there
are many good experimental design texts (e.g., [1–3]) and online
resources (e.g., [4]), which I encourage the reader to explore.
1.1 Why Proteomics Proteomics aims to quantify thousands of proteins in complex bio-
Experiments Benefit logical backgrounds, such as tissue and plasma. It is typically done
from Design using sensitive instruments like mass spectrometers following mul-
tiple preparation steps, including protein extraction, enrichment,
fractionation, and digestion. Because these complex laboratory
processes can lack stability across a full study, they can negatively
impact the outcome unless the experiment is designed to take this
into account.
In their 2005 paper [5], Hu et al. describe three proteomics
studies that failed due to poor experimental design. In one study,
cancer sample data showed strong dissimilarity between run dates.
271
272 Daniel Ruderman
Because one cancer subtype’s samples were all run on a single day,
any biological signal was confounded with any technical changes
that day. In another study, samples were found to group in similarity
not by cancer subtype, as expected, but instead by the date the col-
lection protocol was changed. Their third example was the victim of
erroneous mass spectrometer calibration and sample degradation.
These studies would likely have succeeded with proper experimental
design and inclusion of quality controls. Poor experimental design is
cited as a major impediment to translational proteomics [6].
Science is ultimately held to the standard of reproducibility. So
it is surprising that attempts to reproduce the results of high-profile
biomedical studies mostly failed [7, 8]. Poor experimental design
has been highlighted as a cause reproducibility failure [9]. The US
National Institutes of Health has recently emphasized reproduc-
ibility, stating that grant applications will be reviewed specifically
for “rigorous experimental design for robust and unbiased results”
[10]. Thus experimental design is not only important in executing
research, but in securing the funds to do so in the first place.
2 Materials
As best you can, gather and enumerate the following:

1. The scientific question you are addressing with the study.
2. A list of available samples and their annotations (including
experimental factors such as treatment group and “nuisance”
factors like sample processing date and technician).
3. Response variables you will measure (e.g., peptide abundance,
protein identity, retention time).
4. A list of experimental factors that will be varied to answer the
question (e.g., disease vs. control, drug dose, time point).
5. A list of nuisance factors which need to vary across the experi-
ment (but ideally would not) and may affect response variables
(e.g., run date, reagent batches, LC columns).
6. Experimental protocols.
7. Ground truth analytes and quality control samples.
8. Experimental design and data analysis software (see Note 1).
9. Estimates of the effect size and biological/technical variability
in the response variable (typically through a pilot
experiment).
10. Statistician consultant/collaborator.
Proteomic Experimental Design 273
3 Methods: Seven Steps to Executing a Well-Designed Proteomics Experiment
A successful experiment is one that can be replicated by others.

This means that all the factors specific to one laboratory—like
equipment, reagents and technique—must not influence its out-
come. The experimental result must accurately reflect a sample
weighing micrograms, and yet be unaffected by any variability in
the tons of lab equipment used to measure it.
The techniques of experimental design have been recognized as
an important component of proteomics research for over a decade
[11–16]. An in-depth treatment can be found in the seminal work
of Vitek and colleagues [11, 12]. The basic steps outlined below
should help researchers achieve unbiased, efficient proteomics
experiments. A visual outline of the approach is shown in Fig. 1.
3.1 Define A clear research goal determines the experiment you’ll want to
the Research Goal run. It embodies questions like: “What effect will you measure?”
and Relevant and “Which sample groups will you compare?” These in turn iden-
Response Variables tify the response variables you will measure and the experimental
factors you will change to determine their impact.
Response variables are the experimental outputs (numbers) used
to answer the research question. In quantitative proteomics they
may be the abundances of peptides or proteins. In proteome surveys
they may instead be the number of distinct proteins found across
samples. Core to designing an experiment is choosing the experi-
mental methodology. Other chapters in this volume can aid in the
decision, for example, between MRM (Multiple Reaction
Monitoring), DDA (Data-Dependent Acquisition), DIA (Data-
Independent Acquisition), or other methods [17]. The best choice
will optimize the ability to accurately measure the response variables,
possibly at the expense of reduced performance by other metrics.
For example, in setting a mass isolation window there is a trade-off
between analyte selectivity and quantitation sensitivity [18].
It is important to have an estimate of the effect size, as it will
help determine how many samples will need to be run (see below).
For early stage biomarker screening any significant difference may be
of interest, no matter how small. In other cases, such as toxicology
studies, only physiologically meaningful differences are of value.
Fig. 1 Four major activities and their component steps to designing an unbiased and efficient experiment
274 Daniel Ruderman
3.2 List Relevant Factors are those things that potentially affect response variables.
Factors Ideally the response would only be impacted by the physical prop-
erty the experimental system is designed to measure. But this is
almost never the case. For example, the measured peptide peak
area depends not just on that peptide’s abundance but also on elec-
trospray stability, chromatography column carryover, and the effi-
ciency of proteolytic digestion. Digestion efficiency may in turn
depend on the trypsin reagent lot number and which lab tech per-
forms the digestion.
Those factors we change as part of the experimental investiga-
tion are called experimental factors. For example, when comparing
the proteome between subcellular fractions, the proteomic fraction
(e.g., cytoplasm or nucleus) is such a factor. If we are instead test-
ing the effect of chromatography on peak height, then flow rate
might be an experimental factor. Once the experimental factors
have been chosen to answer the research question, all those remain-
ing in the list are called nuisance factors. These are conditions that
can affect response variables but do not remain fixed for all samples
or runs. Examples include reagent batches, intra-day LC-MS run
order, and personnel. The goal of identifying nuisance factors is to
minimize their effect on response variables while allowing experi-
mental factors to systematically alter those responses.
3.3 Develop Control Two kinds of control samples are important for verifying that the
Samples and Quality experimental platform actually works. Experimental controls ensure
Control Procedures accuracy: Quantitation should reflect what is in the sample; actual
differences should be detected; lack of change between sample groups
should not. Quality control samples (QCs) ensure that processes and
equipment perform within defined specifications [19, 20].
Experimental controls for accuracy are used to ensure that
quantities like retention time, m/z, and protein/peptide abun-
dance are well measured. They are typically created by spiking
compounds into the experimental samples. Controls can also be
run through the mass spectrometer at different times than the
experimental samples (e.g., lock masses during alternate MS scans,
separate QC samples). Krokhin and Spicer [21] describe a set of
spike-in peptides for normalizing reverse phase HPLC retention
times. The peptides enable both retention time normalization of
data between runs and hydrophobicity index prediction of sample
peptides [22] for improved identification. For protein abundance
normalization between runs of complex proteomic samples, the
“super-SILAC” approach of Geiger et al. can be used [23]. Here a
proteomic standard sample is derived from species-specific cells
cultured in SILAC media, providing mass shifts to tryptic peptides
so they do not overlap sample peptides. This approach has two
important benefits. First the proteins are matched to the sample, so
most peptides will have a corresponding quantitation standard.
Second, because the standard can be spiked into the sample prior
to fractionation and digestion, the ratio of sample to standard pep-

tide abundance is unaffected by any variability from those upstream
processes.
Positive and negative experimental controls can be used to nest
an “orthogonal” experiment within the main experiment. In these
controls protein abundances differ by a known amount, which
enables testing of the experiment’s ability to detect them. These
control samples should closely resemble the experimental samples.
For example, if studying the plasma proteome, the control samples
are ideally based on a single plasma sample that is available in suf-
ficient quantity to run periodically during the entire experiment
(see Note 2). A positive control for detecting protein differences
between two samples might contrast a human plasma sample with
and without spiked-in proteins [24]. The corresponding negative
control would be multiple runs of the same sample to ensure that
few proteins are found that differ (false discoveries). To instead
control an experiment quantifying peak area differences between
sample groups, proteins can be spiked into control samples at two
different concentrations (which vary from protein to protein, e.g.,
UPS2 from Sigma-Aldrich). The concentration differences might
cover a range from multiple fold-changes down to a few percent in
order to detect the assay’s lower limit of change detection. Running
both control samples multiple times during the course of the
experiment will mimic an experiment on two separate sample
groups (though without biological variability within each group).
The purpose of quality control samples is to indicate when
processes or equipment have failed, or, better, reassure us that they
have not. Measurements on quality control samples should indi-
cate quantitatively how well processes and equipment perform.
Multiple samples can be used to QC different experimental steps.
For example, a protein fractionation workflow can be periodically
assessed using a cell lysate or other complex mixture, with quanti-
tative readout given by UV absorption across fractions. LC-MS
performance can be measured using a simple mixture of peptides,
quantifying the peak areas across peptides and charge states as well
as other peaks reflecting contamination or carry-over. Alternatively,
a QC sample derived from biological samples of interest (or a
vendor-provided reference sample [25]) can be periodically run
through the entire workflow. Sets of performance metrics and anal-
ysis software are available from the NIST [26], the QuaMeter tool
[27, 28], and SProCoP [29]. Metrics specific to isobaric labeling
can be found in [30]. Finally, QCs should be implemented to
detect any previously identified failure modes.
Accurate peptide/protein quantitation is a general goal of pro-
teomics. In unlabeled experiments, this hinges on chromatographic
reproducibility between runs, for two reasons. First, quantitation
itself is impacted by variations in elution peak width. Second, the
ability of software to reproducibly identify the same peak across
276 Daniel Ruderman
runs is degraded by retention time shifts, which can lead to missing

values and artificially increased unique peptide counts. Use of per-
formance metrics can assist in optimizing chromatography [26, 27,
29]. For isobarically labeled experiments (e.g., iTRAQ, TMT),
peptide/protein abundances are compared to a known standard
that is labeled and present in all runs. To reduce variability, it is best
to label the standard once as a large batch. Although including a
standard in this way reduces experimental efficiency, it both enables
LC-MS quality control and data normalization between runs. For
some experiments there are alternative computational normaliza-
tion methods that do not require a control standard [31].
3.4 Optimize Process Protocols such as found in this volume and in vendor application
notes give detailed roadmaps for key parts of the experiment. But
there remain many parameters to adjust in making the experimen-
tal system work well as a whole. For example, Agilent’s Jet Stream
electrospray ionization source has six adjustments, such as drying
gas temperature, nebulizer voltage, and capillary voltage [32].
Additionally, MS data analysis packages have their own settings.
The OMSSA search engine [33], for example, has 9 of them, not
including PTM selection. An experiment’s success may hinge on
getting them right.
Fortunately, there are systematic methods for optimizing these
settings. The statistics sub-field of Response Surface Methodology
(RSM) [34] is used to design and analyze an experiment to esti-
mate the settings that maximize a performance metric. Although
process optimization is not strictly part of experimental design (it
need not be done for each individual experiment), it is equally
important and often overlooked. The basic idea is to perform a
designed experiment across a set of parameter adjustments. RSM
then fits the performance metric to a (usually) quadratic surface in
the space of parameters, and estimates the parameter settings that
optimize performance. To optimize data analysis parameters no
additional experiments need be performed. Instead, the analysis
software can be run on a single representative data set under differ-
ent parameter settings. A number of examples demonstrating
proteomics performance gains through process optimization have
been published [35–42].
It is not always obvious how to choose the performance met-
ric. For example, when trying to discover biomarkers, is it best to
maximize the total number of spectral “features” detected or
instead to maximize the number of features with high signal-to-
noise ratio? Is the number of identified proteins also important? A
standard approach in such situations is to mathematically combine
a number of relevant metrics using Derringer’s desirability metric
[34] (see Note 1). It tends to compromise well between competing
criteria. One such tradeoff is found during label-free MS1 quanti-
tation in tandem mass spectrometry, where higher accuracy can be
achieved by dedicating more scans per eluted peak to MS1 while

sacrificing MS2-based identification scans. The desirability metric
can be used to find a principled compromise between quantitation
accuracy and depth of protein identification.
3.5 Estimate Underpowered studies do not have sufficient independent samples

the Number or technical replicates to answer research questions with high prob-
of Samples Needed ability. They can lead to either false positive or false negative con-
clusions [43]. The probability of detecting an effect when it is
actually present is called the power. Researchers typically aim for
80–90 % power. To create a study that is likely to find the effect we
are looking for, we must estimate how many samples will be needed
and design the experiment accordingly. This involves knowing two
things: how big an effect we are looking for (the signal) and how
much the measured response varies between samples in the same
group (the noise).
Measuring the noise level through pilot experiments is a
critical step in the design. I recommend running at least ten inde-
pendent biological samples from within the same group (e.g., dis-
ease, control) to assess “all-in” variability. This includes noise both
from the technical experimental aspects and those from biological
variation between replicate samples. Additionally, one biological
sample should be processed and measured at least five times to
assess the technical variability magnitude. While these experiments
may seem to be a costly investment, the knowledge they provide is
key to running a successful experiment the first time around.
In designing experiments, we quantify noise as the statistical
variance in a measurement when the “signal” (in this case, the sam-
ple group or the sample itself) does not change. In proteomics
there is often more than one signal of interest and thus many such
variances to measure [44]. For MS1 data there may be more than
105 individual peaks quantified, each with its signal-to-noise ratio
(see Note 3). In such cases a good shortcut is estimate a single typi-
cal noise variance (say, the 75th percentile across all peaks) and use
it to compute the number of samples needed to detect a range of
biologically relevant peak area differences (see below). Note that
for abundance measurements noise processes are usually multipli-
cative, and logarithmic transforms of the data are typically used
instead of raw abundance values [11].
In general, a study’s power calculation method depends on the
actual statistical analysis to be performed. However, a simple and
useful method for estimating sample sizes uses the Gaussian noise
approximation. This assumes the response variable is either con-
tinuous (e.g., a peptide’s abundance) or even a large integer (e.g.,
the number of proteins identified), and that noise due to both
intra-group and technical variability is approximately Gaussian and
additive. In this case the well-known sample size estimate
278 Daniel Ruderman
N = éë( z1-a / 2 + z1- b ) s / d ùû

2

holds, where N is the minimum number of samples needed per
group, zq is the qth quantile of the normal distribution, α is the
desired significance level (typically 0.05), 1-β is the power, σ is the
estimated total noise standard deviation (biological and technical),
and δ is the expected effect size. At a significance level of 0.05 and
power of 80 %, z1-a / 2 = 1.96 and z1- b = 0.84 , so N = 7.84 (s / d ) .
2
Thus for a signal-to-noise ratio (σ/δ) of 1, at least 8 samples per

group are needed to power such a study. For a more complete dis-
cussion on choosing the number of technical and biological repli-
cates, see Ref. 11. As noted above, running pilot experiments will
provide the biological and technical noise levels needed to make
these important decisions. Software such as JMP and Stat-Ease
provide methods for determining sample size (see Note 1).
Often the study goal is to estimate a number, like a peptide’s
average abundance across samples, to some precision. Assuming
independent Gaussian variability (combined biological and techni-
cal) in each sample’s measurement with standard deviation σ, the
number of samples to achieve a precision of τ is approximately
N = (s / t ) .
2
Although outside the scope of this chapter, determining the

sample size while controlling the false discovery rate instead of the
false positive rate (significance) can also be achieved [45].
3.6 Configure An experiment’s design maps its idealized set of measurements

the Experiment onto laboratory reality. Which samples should be analyzed? In
what order should they be run? How many QC samples should be
included? A good experimental design will provide enough statisti-
cal power to answer the research question and do so without bias-
ing the result. In what follows I will assume there are two sample
groups being compared. While many of the same ideas apply to
more complex (e.g., multi-factorial) investigations, I direct the
reader to standard texts for more complete treatments [1, 2].
Modern design of experiments goes back to R.A. Fisher’s
work, first published in 1926 [46]. With the joint goals of effi-
ciency and validity, he identified three cornerstones of experimen-
tal design: blocking, randomization, and replication. Blocking and
randomization serve to eliminate biased answers. Replication
increases the signal-to-noise ratio to improve precision of the
experiment’s result. The concepts of blocking, randomization
and replication are the most important takeaways of this chap-
ter. I present them and other design considerations below.
3.6.1 Blocking Many factors change discretely over the course of an experiment.
Some examples are reagent batches, personnel, date, and run order.
As one of these factors changes it can alter the response of the
experimental system. Lab members may use different techniques.

If two sample groups are isolated by levels of such a factor (e.g., all
responding patients’ sample run on one day and all non-responders
on another), then a significant technical change may be mistaken
for a biological effect.
The solution is to use blocking. For each level of a nuisance fac-
tor (e.g., date, personnel) a small balanced experiment is run. As an
example, if only eight samples per day can be run, then four from
each of two sample groups would be run per day. Any day-to-day
variation in the response variable would then affect both groups
nearly equally, preventing bias which might lead to a false discovery.
When there are multiple such factors, it can be challenging to solve
the design in a balanced way, particularly if there are interactions
between those factors. When a balanced design is not possible (e.g.,
sample counts are not multiples of the block size), randomization
should be used instead (see below). George Box offered the sound
advice: “Block what you can, randomize what you cannot.”
If a nuisance factor is known to have little or no effect on the
response variable, then it is safe to forego blocking it. One such
factor might be pipette tip lot numbers; it is simply very unlikely
that it makes any difference. The magnitude of a nuisance factor’s
impact can be assessed by running a control sample across that fac-
tor’s levels, followed by appropriate data analysis (e.g., linear fixed-
effects or mixed-effects modeling). It is a good idea to run QC
samples periodically to monitor longitudinal performance and
detect any such effects. Furthermore, once a balanced experiment
has been run its data can be analyzed to determine the magnitude
of variation and temporal trends due to nuisance factors. This can
be a good way to detect process issues even though they may not
directly impact the experiment’s results.
3.6.2 Randomization There may be nuisance factors with many levels where it is not
obvious how to perform blocking, for example the assignment of
samples to wells in a 96-well plate. Although it is known that plate-
based experiments can have systematic changes toward plate edges,
it is very effective simply to randomize sample placement. Similarly,
the order in which a day’s samples are processed can be random-
ized in case of changing equipment or personnel attentiveness.
Temporal randomization (both within and across days) in pro-
teomics is particularly important since there are many potential
sources of drift, including LC column degradation, electrospray
instability, transfer tube build-up, and tuning loss.
Randomization has the added advantage of counteracting bias
due to unrecognized influences. For example if a contiguous sub-
set of vials gets contaminated, and they were randomized relative
to sample groups, then the experimental result will show increased
variability but without a biased effect on the conclusion (although
with increased confidence intervals). Randomization is particularly
280 Daniel Ruderman
Fig. 2 A randomized block design for a proteomics LC-MS workflow. Two sample groups (A and B) are being
measured for proteomic differences. The experimental day is a blocking factor, and morning (AM) versus after-
noon (PM) is a blocking factor nested within the day. Each such block contains six samples run (balanced
between three samples of each group). The 6-sample run order is randomized to avoid confounding temporal
drift with the A/B proteomic difference, with laboratory personnel blinded to the ordering. At the middle of each
LC-MS run a QC sample is analyzed to monitor process. QC and sample data should be analyzed to ensure no
systematic AM versus PM differences or large day-to-day variation impact the signals of interest
important in the case of longitudinally collected samples. They

should be processed in random order to ensure that any procedural
changes are not mistaken for biological trends. Many experiments
randomize samples within blocks, which is called a block-randomized
design. An example is shown in Fig. 2.
Multi-run isobaric labeling (e.g., iTRAQ, TMT) experiments
[47] are cases where blocking and randomization should be used.
An example using 4-plex iTRAQ to contrast two sample groups of
eight samples each in a completely balanced design is shown in
Fig. 3a. Here each of the four labels is applied twice per sample
group (Fig. 3b). Also, each pooled MS run contains two samples
from each group (Fig. 3c). Thus bias due to either a particular
label or run will have limited impact on the inter-group compari-
son. When sample counts are unbalanced between groups or not
divisible by the number of labels, assignment of sample to label
and/or run should be made randomly. Although many have
reported only minimal label bias (e.g., [30, 48]), it is best to start
with label randomization and discontinue only if statistical testing
demonstrates no bias. When using randomization, there is a
trade-off between reducing bias and increasing both experi-
mental effort and the risk of introducing procedural error.
3.6.3 Replication Unlike blocking and randomization, which seek to remove system-
atic bias that could be mistaken for or mask a real effect, the goal
of replication is to reduce the impact of random measurement vari-
ation. This leads to more precise estimates of the effects that answer
Fig. 3 (a) Blocked and balanced design for a 4-plex iTRAQ experiment contrasting two sample groups (A and
B) with eight samples per group (A1 through A8, B1 through B8) in four runs/pools. (b) Frequency table for
labels across sample groups showing balanced design. (c) Frequency table for runs across sample groups
showing balanced design
the research question. Replication comes in two types. By repeat-

edly drawing independent random samples from a population we
achieve more and more accurate estimates of that population’s
mean response value (e.g., a protein expression level). This is bio-
logical replication. Additionally, any given sample can be processed
multiple times in the laboratory and the results averaged to reduce
the impact of process variability. This is technical replication.
Biological replication serves to accurately reflect a population
of interest using only a subset of samples. The population may, for
example, be patients who responded to a therapy or cell cultures
grown under specific conditions. Because there is inherent variabil-
ity among patients and the cells that happen to grow best in cul-
ture, a single sample from those populations may deviate greatly
from the population average of interest. The number of biological
replicates to employ in a study depends in large part on the desired
power to detect an estimated effect, as mentioned above. However,
there are other considerations.
Some sample populations are particularly heterogeneous, hav-
ing “long tails” that provide sample values far from the mean more
frequently than expected from Gaussian statistics [49]. In my expe-
rience clinical samples often have this property. In such cases it is
important to run enough samples to include some from the tails,
otherwise replication will fail when they show up in validation sets.
I thus recommend at least 10 samples from each group when het-
erogeneity is expected. In contrast, replicates of laboratory sam-
ples, like cell culture, are more likely to have Gaussian statistics. In
this case the estimates from power calculations are likely to work
well, even if fewer than 10 samples.
Technical replication reduces the impact of process noise. It is
useful particularly when laboratory procedures introduce variabil-
ity which is on the scale of or larger than the population variability.
This may particularly be the case for low concentration analytes
282 Daniel Ruderman
that are inherently difficult to measure precisely. Protein identifica-

tion in complex samples is often improved by replicating data-
dependent LC-MS/MS runs and accumulating the list of proteins
identified across them. Liu et al. [50] found that 10 such runs are
needed to identify 95 % of the proteins in yeast lysate. To replicate
a process the sample must be split and processed multiple times,
meaning sufficient sample must be available to rerun the protocol
in question. To determine which steps in a workflow cause the
most variability, an analysis of variance can be performed [51].
Highly variable steps can be remedied through process improve-
ment or technical replication [13].
3.6.4 Blinding Modern clinical trials—arguably some of the most carefully

designed experiments anywhere—are blinded for good reason.
When patients, clinicians, or data analysts know which subjects
belong to which group, results become biased [52]. There is no
reason to expect scientific studies to be any different. Indeed, many
biological results were found not to be reproducible by the original
investigators once they are blinded [53]. Data analysts also need to
be blinded [54], which suggests that confirmation bias has strong
impact on biomedical research.
3.6.5 Control Samples An experiment’s design should include control samples to ensure
process. This is most crucial in complex experiments involving
many samples measured across multiple days or instruments.
However, even a simple experiment on just a few samples will ben-
efit, particularly when a negative result is buoyed by knowledge
that the measurements are reliable. The ~15 % additional cost in
time and resources is well worth the payoff.
Quality control samples measure whether technology is func-
tioning consistently and within specifications. Since most pro-
teomic workflows involve multiple steps (e.g., capture, fractionation,
LC-MS) it is best to insert the QC samples as early in the workflow
as possible to test them all. A daily series of sample preparations can
be “bookended” by QC samples at either end to measure perfor-
mance changes. If intra-day changes are unlikely then a single QC
sample could instead be placed in the middle of a day’s experimen-
tal samples. In either case, if a QC sample demonstrates the equip-
ment has failed, those runs can be rejected and the samples possibly
rerun. Additionally, the response measures on the QC samples can
be used to normalize experimental data for possible drift.
Figure 4 shows a design exemplifying blocking, randomiza-
tion, and QCs. This design was motivated by a set of blood samples
collected from eight subjects in pairs, one prior to and one after a
seizure. The research goal was to determine whether there are
detectable changes in the plasma proteome following seizure. The
design specifies how samples are run on LC-MS on a system with
two LC columns (LC-1 and LC-2) alternately driving a single
Fig. 4 Example of blocking by LC system and inclusion of QCs in a paired sample (pre- [no asterisk] and post-
seizure [asterisk]) workflow. See Subheading 3.6.5 for details
MS. It includes technical QC samples on every fifth elution from

each column to monitor performance. Because samples are evalu-
ated pairwise for differences, it is important that they are run in a
manner that minimizes any technical changes. Thus each LC forms
a block that contains all samples from four patients so that no
patient’s samples are subject to inter-column differences.
Furthermore, to mitigate temporal drift, both samples from a
patient are run as close in time as possible. Finally, to ensure there
is no systematic effect of temporal drift on the pre- vs post-seizure
effect, half of the patients have the pre-seizure sample (no asterisk)
run first and half had the post-seizure sample (asterisk) run first.
As previously mentioned, positive and negative experimental
control samples enable researchers to demonstrate the capability of
the experiment to correctly discover positive results while ignoring
negatives. More complex controls may instead be designed to
quantify the sensitivity to detect effects of various sizes against a
fixed specificity (e.g., p < 0.05). To best mimic the measurement
conditions of experimental samples, these controls should be ran-
domized into the general sample population and their data ana-
lyzed separately.
3.7 Choose In addition to vendor-specific software, there are now many freely
the Statistical Analysis available analysis platforms for proteomics data [55–58]. Analysis
steps include protein identification, quantification, and statistical
testing. The most common methods are t-tests and ANOVA. Since
these tests make certain assumptions about the data statistics (e.g.,
Gaussian noise, homogeneity of variance, independence across
samples) one should be sure these preconditions are met (particu-
larly for small data sets) [3]; if not, other methods such as nonpara-
metric tests (e.g., Mann–Whitney U) can be used. Quantitation
through spectral counting, for example, can often have non-
Gaussian noise statistics. With blocked designs one must ensure
blocking is accounted for in the analysis (so that the possibly large
variation between blocks is ignored when estimating the noise
level). This can typically be accomplished using a blocking factor in
ANOVA [59] or mixed effects models [60]. As a general rule the
284 Daniel Ruderman
statistical analysis must match both the research question and the
experimental design. It must not be altered after the collection of
data (“HARKing” [61]). I strongly recommend engaging a
statistician early in the experiment design process to address
these questions. After data are collected is often too late.
Control samples also must be analyzed. QC samples should be
checked for instrumentation and process failure. These QCs can
also be used to detect performance variation between blocks (e.g.,
across days) to characterize changes in equipment performance.
Analysis of positive and negative control samples should validate
that the entire experimental workflow functions properly.
4 Conclusions
Oberg and Vitek state that “…all quantitative investigations fail to

deliver reproducible and accurate results if proper attention is not
devoted to the experimental design” [11]. A well-designed experi-
ment does not take much more effort to execute than a poorly
designed one. Yet, it can be crucial to the experiment’s success or
understanding why it failed. R. A. Fisher’s basic prescription was
for blocking, randomization, and replication. Additionally, by
instituting appropriate QCs, not only can outlier samples caused
by technical failures be recognized and removed, but long-term
changes in process performance can also be detected before they
negatively impact experiments. The goal is results that others can
consistently reproduce. Finally, I encourage the sharing of
proteomic data via public repositories (e.g., proteomexchange.org).
It not only reduces unneeded replication of research, but also
enables the comparison of similar data sets to evaluate quality and
accuracy. Although standards for LC-MS quality control runs do
not currently exist, sharing this information together the experi-
mental data could one day greatly ease the process of debugging
and optimizing proteomics experiments. This would help make
efficient, reproducible proteomics accessible to a far wider group of
researchers.
5 Notes
1. For select tasks, the following software are useful:

The MSStats R package for statistical analysis of quantitative
mass spectrometry data [60].
Using Bioconductor in R to analyze mass spectrometry data
[55].
The R package desirability [62] aids in formulating and optimiz-
ing multiple experiment performance read-outs (e.g., mass
accuracy and quantitation noise), using the methods of

Derringer and Suich [63].
JMP (www.jmp.com) and Stat-Ease (www.statease.com) offer
tools for designing and powering experiments.
2. To ensure sufficient quantity, a pool of multiple plasma samples
could also be used. But its characteristics may differ in unknown
but important ways from an actual plasma sample. Thus I
(weakly) prefer using a single patient’s sample.
3. Many of these peaks may be redundant since they correspond
to different peptides from the same protein.
Acknowledgments
I thank Dr. Parag Mallick for introducing me to the field of experi-

mental design and Dr. Nicholas Graham for comments on the
manuscript.
References
1. Montgomery DC (2013) Design and analysis of papers3://publication/uuid/0C526949-

experiments, 8th edn. John Wiley, Hoboken, NJ 4171-407F-834A-A82A37275234
2. Box GEP, Hunter JS, Hunter WG (2005) 9. Begley CG, Ioannidis JPA (2015)
Statistics for experimenters : design, innova- Reproducibility in science: improving the stan-
tion, and discovery. Wiley series in probabil- dard for basic and preclinical research. Circ Res
ity and statistics, 2nd edn. Wiley-Interscience, 116(1):116–126
Hoboken, NJ 10. The National Institutes of Health (2015)
3. Quinn GP, Keough MJ (2002) Experimental Enhancing Reproducibility through Rigor and
design and data analysis for biologists. Transparency. http://grants.nih.gov/grants/
Cambridge University Press, Cambridge, UK guide/notice-files/NOT-OD-15-103.html
4. NIST/SEMATECH e-Handbook of statistical 11. Oberg AL, Vitek O (2009) Statistical design
methods. http://www.itl.nist.gov/div898/ of quantitative mass spectrometry-based
handbook/index.htm proteomic experiments. J Proteome Res
5. Hu J, Coombes KR, Morris JS (2005) 8(5):2144–2156
The importance of experimental design in 12. Riter LS, Vitek O, Gooding KM, Hodge BD,
proteomic mass spectrometry experiments:
Julian RK (2005) Statistical design of experi-
some cautionary tales. Brief Funct Genomic ments as a tool in mass spectrometry. J Mass
Proteomic 3(4):322–331 Spectrom 40(5):565–579
6. Maes E, Cho WC, Baggerman G (2015) 13. Karp NA, Lilley KS (2007) Design and analy-
Translating clinical proteomics: the impor- sis issues in quantitative proteomics studies.
tance of study design. Expert Rev Proteomics Proteomics 7(Suppl 1):42–50
12(3):217–219 14. Leek JT, Scharpf RB, Bravo HC, Simcha
7. Prinz F, Schlange T, Asadullah K (2011) D, Langmead B, Johnson WE, Geman D,
Believe it or not: how much can we rely on Baggerly K, Irizarry RA (2010) Tackling the
published data on potential drug targets? widespread and critical impact of batch effects
Nature: pp 1–2. doi: papers3://publication/ in high- throughput data. Nat Rev Genet
doi/10.1038/nrd3439-c1 11(10):733–739
8. Begley CG, Ellis LM (2012) Drug develop- 15. Cairns DA (2011) Statistical issues in quality
ment: raise standards for preclinical cancer control of proteomic analyses: good experi-
research. Nature 483 (7391).[Online]. doi: mental design and planning. Proteomics
11(6):1037–1048
286 Daniel Ruderman
16. Cairns DA (2014) Statistical issues in the Zimmerman LJ, Carr SA, Fisher SJ, Gibson
design and planning of proteomic profiling BW, Paulovich AG, Regnier FE, Rodriguez
experiments. Clin Proteomics 18:223–236 H, Spiegelman C, Tempst P, Liebler DC,
17. Leitner A, Aebersold R (2013) SnapShot: mass Stein SE (2010) Performance metrics for liq-
spectrometry for protein and proteome analy- uid chromatography- tandem mass spectrom-
ses. Cell 154(1):252–252e251. doi:10.1016/j. etry systems in proteomics analyses. Mol Cell
cell.2013.06.025 Proteomics 9(2):225–241
18. Gallien S, Bourmaud A, Kim SY, Domon B 27. Ma Z-Q, Polzin KO, Dasari S, Chambers
(2014) Technical considerations for large- MC, Schilling B, Gibson BW, Tran BQ, Vega-
scale parallel reaction monitoring analysis. Montoto L, Liebler DC, Tabb DL (2012)
J Proteomics 100:147–159. doi:10.1016/j. QuaMeter: multivendor performance metrics
jprot.2013.10.029 for LC–MS/MS proteomics instrumentation.
19. Montgomery DC (2013) Introduction to Anal Chem 84(14):5845–5850
statistical quality control, 7th edn. Wiley,
2 8. Wang X, Chambers MC, Vega-Montoto LJ,
Hoboken, NJ Bunk DM, Stein SE, Tabb DL (2014) QC
20. Bramwell D (2013) An introduction to statis- metrics from CPTAC raw LC-MS/MS data
tical process control in research proteomics. interpreted through multivariate statistics. Anal
J Proteomics 95(C):3–21, doi: papers3://pub- Chem 86(5):2497–2509
lication/doi/10.1016/j.jprot.2013.06.010
2 9. Bereman MS, Johnson R, Bollinger J, Boss
21. Krokhin OV, Spicer V (2009) Peptide reten- Y, Shulman N, MacLean B, Hoofnagle
tion standards and hydrophobicity indexes AN, MacCoss MJ (2014) Implementation
in reversed-phase high-performance liq- of statistical process control for proteomic
uid chromatography of peptides. Anal experiments via LC MS/MS. J Am Soc Mass
Chem 81(22):9522–9530, doi: papers3:// Spectrom 25(4):581–587. doi:10.1007/
publication/doi/10.1021/ac9016693 s13361-013-0824-5
22. Krokhin OV (2006) Sequence-specific retention
3 0. Burkhart JM, Vaudel M, Zahedi RP, Martens
calculator. algorithm for peptide retention predic- L, Sickmann A (2011) iTRAQ protein quan-
tion in ion-pair RP-HPLC: application to 300- tification: a quality-controlled workflow.
and 100-Å pore size C18 sorbents. Anal Chem Proteomics 11(6):1125–1134, doi: papers3://
78(22):7785–7795. doi:10.1021/ac060777w publication/doi/10.1002/pmic.201000711
23. Geiger T, Wisniewski JR, Cox J, Zanivan S,
3 1. Herbrich SM, Cole RN, West KP Jr, Schulze
Kruger M, Ishihama Y, Mann M (2011) Use K, Yager JD, Groopman JD, Christian P, Wu
of stable isotope labeling by amino acids in L, O'Meally RN, May DH, McIntosh MW,
cell culture as a spike-in standard in quantita- Ruczinski I (2013) Statistical inference from
tive proteomics. Nat Protoc 6(2):147–157, multiple iTRAQ experiments without using
doi: papers3://publication/doi/10.1038/ common reference standards. J Proteome Res
nprot.2010.192 12(2):594–604. doi:10.1021/pr300624g
24. Levin Y, Hradetzky E, Bahn S (2011)
3 2. Greco G, Boltner A, Letzel T (2014)
Quantification of proteins using data- Optimization of jet stream ESI parameters
independent analysis (MSE) in simple and- when coupling agilent 1260 infinity analytical
complex samples: a systematic evaluation. SFC system with agilent 6230 TOF LC/MS.,
Proteomics 11(16):3273–3287 Agilent Technologies, https://www.agilent.
com/cs/library/applications/5991-4510EN.
25. Ivanov AR, Colangelo CM, Dufresne CP, pdf
Friedman DB, Lilley KS, Mechtler K, Phinney
BS, Rose KL, Rudnick PA, Searle BC, Shaffer
3 3. Geer LY, Markey SP, Kowalak JA, Wagner L,
SA, Weintraub ST (2013) Interlaboratory Xu M, Maynard DM, Yang X, Shi W, Bryant
studies and initiatives developing standards SH (2004) Open mass spectrometry search
for proteomics. Proteomics 13(6):904–909, algorithm. J Proteome Res 3(5):958–964.
doi: papers3://publication/doi/10.1002/ doi:10.1021/pr0499491
pmic.201200532
3 4. Bezerra MA, Santelli RE, Oliveira EP, Villar
26. Rudnick PA, Clauser KR, Kilpatrick LE, LS, Escaleira LA (2008) Response surface
Tchekhovskoi DV, Neta P, Blonder N, methodology (RSM) as a tool for optimization
Billheimer DD, Blackman RK, Bunk DM, in analytical chemistry. Talanta 76(5):965–977
Cardasis HL, Ham A-JL, Jaffe JD, Kinsinger 35. Coscollà C, Navarro-Olivares S, Martí P, Yusà V
CR, Mesri M, Neubert TA, Schilling B, Tabb (2014) Application of the experimental design
DL, Tegeler TJ, Vega-Montoto L, Variyath of experiments (DoE) for the determination
AM, Wang M, Wang P, Whiteaker JR, of organotin compounds in water samples
using HS-SPME and GC–MS/MS. Talanta RR, Ioannidis JPA, Jankowski J, Julian BA,
119:544–552 Klein JB, Kolch W, Luider T, Massy Z, Mattes
36. Eliasson M, Rännar S, Madsen R, Donten MA, WB, Molina F, Monsarrat B, Novak J, Peter
Marsden-Edwards E, Moritz T, Shockcor JP, K, Rossing P, Sánchez-Carbayo M, Schanstra
Johansson E, Trygg J (2012) Strategy for opti- JP, Semmes OJ, Spasovski G, Theodorescu D,
mizing LC-MS data processing in metabolo- Thongboonkerd V, Vanholder R, Veenstra TD,
mics: a design of experiments approach. Anal Weissinger E, Yamamoto T, Vlahou A (2010)
Chem 84(15):6869–76 Recommendations for biomarker identification
37. Székely G, Henriques B, Gil M, Ramos A, and qualification in clinical proteomics. Science
Alvarez C (2012) Design of experiments as a Transl Med 2 (46): 46ps42
tool for LC–MS/MS method development for 44. Anderle M, Roy S, Lin H, Becker C, Joho K
the trace analysis of the potentially genotoxic (2004) Quantifying reproducibility for differen-
4-dimethylaminopyridine impurity in glucocor- tial proteomics: noise analysis for protein liquid
ticoids. J Pharmaceut Biomed Anal 70:251–8 chromatography-mass spectrometry of human
38. Maes K, Van Liefferinge J, Viaene J, Van serum. Bioinformatics 20(18):3575–3582
Schoors J, Van Wanseele Y, Béchade G, 45. Liu P, Hwang JTG (2007) Quick calculation
Chambers EE, Morren H, Michotte Y, for sample size while controlling false discov-
Vander Heyden Y, Claereboudt J, Smolders ery rate with application to microarray analysis.
I, Van Eeckhaut A (2014) Improved sensitiv- Bioinformatics 23(6):739–746
ity of the nano ultra-high performance liquid 46. Fisher RA, Others (1926) The arrangement of field
chromatography- tandem mass spectrometric experiments. doi: citeulike-article-id:10709753
analysis of low-concentrated neuropeptides 47. Westbrook JA, Noirel J, Brown JE, Wright
by reducing aspecific adsorption and optimiz- PC, Evans CA (2015) Quantitation with
ing the injection solvent. J Chromatogr A chemical tagging reagents in biomarker stud-
1360:217–228 ies. Proteomics Clin Appl 9(3-4):295–300.
39. Passeport E, Guenne A, Culhaoglu T, Moreau doi:10.1002/prca.201400120
S, Bouyé J-M, Tournebize J (2010) Design of 48. Oberg AL, Mahoney DW, Eckel-Passow JE,
experiments and detailed uncertainty analysis Malone CJ, Wolfinger RD, Hill EG, Cooper
to develop and validate a solid-phase micro- LT, Onuma OK, Spiro C, Therneau TM,
extraction/gas chromatography–mass spec- Bergen Iii HR (2008) Statistical analysis of rela-
trometry method for the simultaneous analysis tive labeled mass spectrometry data from com-
of 16 pesticides in water. J Chromatograph A plex samples using ANOVA. J Proteome Res
1217(33):5317–5327 7(1):225–233, doi: papers3://publication/
40. Raji MA, Schug KA (2009) Chemometric study doi/10.1021/pr700734f
of the influence of instrumental parameters on 49. Niepel M, Spencer SL, Sorger PK (2009) Non-
ESI-MS analyte response using full factorial genetic cell-to-cell variability and the conse-
design. Int J Mass Spectrom 279:100–106 quences for pharmacology. Curr Opin Chem
41. Zhou Y, Song J-Z, Choi FF-K, Wu H-F, Qiao Biol 13(5–6):556–561, doi: http://dx.doi.
C-F, Ding L-S, Gesang S-L, Xu H-X (2009) An org/10.1016/j.cbpa.2009.09.015
experimental design approach using response 50. Liu H, Sadygov RG, Yates JR 3rd (2004)
surface techniques to obtain optimal liquid A model for random sampling and estima-
chromatography and mass spectrometry condition of relative protein abundance in shotgun
tions to determine the alkaloids in Meconopsi proteomics. Anal Chem 76(14):4193–4201.
species. J Chromatogr A 1216(42):7013–7023 doi:10.1021/ac0498563
42. Switzar L, Giera M, Lingeman H, Irth H, 51. Mercier C, Truntzer C, Pecqueur D, Gimeno
Niessen WMA (2011) Protein digestion opti- J-P, Belz G, Roy P (2009) Mixed-model
mization for characterization of drug–protein of ANOVA for measurement reproducibil-
adducts using response surface modeling. ity in proteomics. J Proteomics 72(6):974–
J Chromatogr A 1218(13):1715–1723, doi: 981, doi: http://dx.doi.org/10.1016/j.
papers3://publication/doi/10.1016/j. jprot.2009.05.005
chroma.2010.12.043 52. Day SJ, Altman DG (2000) Blinding in clinical
43. Mischak H, Allmaier G, Apweiler R, Attwood T, trials and other studies. BMJ 321(7259):504
Baumann M, Benigni A, Bennett SE, Bischoff 53. Begley CG (2013) Six red flags for suspect
R, Bongcam-Rudloff E, Capasso G, Coon JJ, work. Nature 497(7450):433–4
D'Haese P, Dominiczak AF, Dakna M,
Dihazi H, Ehrich JH, Fernandez-Llama P, Fliser 54. MacCoun R, Perlmutter S (2015) Blind analy-
D, Frokiaer J, Garin J, Girolami M, Hancock sis: hide results to seek the truth. Nature
WS, Haubitz M, Hochstrasser D, Holman 526(7572):187–189
288 Daniel Ruderman
55. Gatto L, Christoforou A (2014) Using R and 59. Kleinbaum DG, Kleinbaum DG (1998)
Bioconductor for proteomics data analysis. Applied regression analysis and other multivari-
Biochim Biophys Acta 1844(1 Pt A):42–51 able methods, 3rd edn. Duxbury, Pacific Grove
56. Colangelo CM, Chung L, Bruce C, Cheung 60. Choi M, Chang CY, Clough T, Broudy D,
K-H (2013) Review of software tools for Killeen T, MacLean B, Vitek O (2014)
design and analysis of large scale MRM pro- MSstats: an R package for statistical analysis of
teomic datasets. Methods 61(3):287–298 quantitative mass spectrometry-based pro-
57. Schwacke JH, Hill EG, Krug EL, Comte- teomic experiments. Bioinformatics
Walters S, Schey KL (2009) iQuantitator: a 30(17):2524–2526
tool for protein expression inference using 61. Kerr NL (1998) HARKing: hypothesizing
iTRAQ. BMC Bioinformatics 10:342. after the results are known. Pers Soc Psychol
doi:10.1186/1471-2105-10-342 Rev 2(3):196–217
58. Gonzalez-Galarza FF, Lawless C, Hubbard SJ, 62. Kuhn M. Desirability: desirability function
Fan J, Bessant C, Hermjakob H, Jones AR (2012) optimization and ranking. https://github.
A critical appraisal of techniques, software pack- com/topepo/desirability
ages, and standards for quantitative proteomic 63. Derringer G, Suich R (1980) Simultaneous
analysis. OMICS 16(9):431–442, doi: papers3:// optimization of several response variables.
publication/doi/10.1089/omi.2012.0022 J Qual Technol 12:214–219
Chapter 20
Automated SWATH Data Analysis Using Targeted

Extraction of Ion Chromatograms
Hannes L. Röst, Ruedi Aebersold, and Olga T. Schubert
Abstract
Targeted mass spectrometry comprises a set of methods able to quantify protein analytes in complex mixtures
with high accuracy and sensitivity. These methods, e.g., Selected Reaction Monitoring (SRM) and SWATH
MS, use specific mass spectrometric coordinates (assays) for reproducible detection and quantification of
proteins. In this protocol, we describe how to analyze, in a targeted manner, data from a SWATH MS experi-
ment aimed at monitoring thousands of proteins reproducibly over many samples. We present a standard
SWATH MS analysis workflow, including manual data analysis for quality control (based on Skyline) as well
as automated data analysis with appropriate control of error rates (based on the OpenSWATH workflow). We
also discuss considerations to ensure maximal coverage, reproducibility, and quantitative accuracy.
Key words Targeted proteomics, SWATH, SWATH MS, SWATH acquisition, Data-independent
acquisition, DIA, OpenSWATH, pyProphet, TRIC aligner, Skyline
1 Introduction
Over the past decade, protein analysis strategies based on mass

spectrometry (MS) have steadily gained popularity. Untargeted
shotgun proteomics is currently the most widely used technique
for qualitative and quantitative measurements of proteins on a
large scale. It allows the detection of hundreds to several thousands
of proteins in a single run. However, its focus on high proteome
coverage leads to some curtailments in reproducibility, quantitative
accuracy, and sample throughput [1]. The targeted proteomic
technique selected/multiple reaction monitoring (SRM) alleviates
these limitations by focusing the mass spectrometer on a defined
set of proteins of interest [2]. SRM excels at consistent and accu-
rate quantification of proteins over large sample sets with coeffi-
cients of variation (CV) below 15 % and offers the largest dynamic
range of all MS-based techniques available today [3]. However,
SRM measurements are limited to a few dozens of target proteins
per sample injection. Novel MS-based proteomic methods, such as
289
290 Hannes L. Röst et al.
SWATH MS, utilize untargeted, data-independent acquisition

(DIA) with targeted data extraction to improve the throughput of
targeted proteomics by massively multiplexing the targeted detec-
tion of peptides. They thereby combine the comprehensiveness of
the shotgun method with the quantitative accuracy and reproduc-
ibility of SRM. In terms of quantification reproducibility, SWATH
MS performs similarly to SRM and offers a dynamic range of at
least three orders of magnitude, positioning it as an ideal technol-
ogy for large-scale and high-quality proteome measurements.
1.1 SWATH MS In the SWATH MS workflow, proteins are enzymatically cleaved to

produce a mixture of homogeneous peptides and then separated
by online liquid chromatography (LC) coupled to tandem mass
spectrometry (MS/MS). Similar to other DIA methods [4, 5], the
mass spectrometer recursively cycles through a large m/z range and
co-fragments all peptide ions in relatively large isolation windows
(or “swathes”) of several m/z (Fig. 1a) [6]. For each isolation win-
dow, the resulting fragment ions are recorded in a high-resolution
composite mass spectrum. Thus, the instrument does not explicitly
target single precursors as in shotgun or SRM but rather fragments
all precursor ions falling into one isolation window simultaneously.
Window size and accumulation time per window are chosen such
that the instrument can cycle through all windows relatively
quickly, allowing every peptide to be fragmented 8–10 times dur-
ing its chromatographic elution (Fig. 1b). In SWATH acquisition,
all ionized species of the sample are thus fragmented and recorded
in a systematic, unbiased fashion, independently of their abun-
dance. Compared to shotgun approaches, which record fragment
ion spectra based on precursor ion intensity, SWATH acquisition
systematically records fragment ion spectra every few seconds,
thereby allowing reconstruction of the LC elution profile of each
fragment ion (Fig. 1b). Similar to SRM, the deterministic acquisi-
tion strategy makes SWATH acquisition highly reproducible. The
acquisition of full, high-resolution fragment ion data provides
additional information compared to SRM, similar to parallel reac-
tion monitoring [7, 8].
What distinguishes SWATH MS from other DIA methods is
the way the data, that is, the highly multiplexed fragment ion spec-
tra, are analyzed [6, 9]. In SWATH MS, intensities of specific frag-
ment ions are extracted from the highly multiplexed fragment ion
spectra in a targeted manner to produce extracted ion chromato-
grams (XICs) (Fig. 1b). These XICs are similar to SRM traces and
can be analyzed analogously. This approach facilitates data analysis
by reducing the complexity of the data significantly, but requires
prior knowledge on the peptides and proteins to be analyzed. This
prior knowledge consists of a set of MS coordinates uniquely
describing the protein of interest (also called an “assay”). The
complete set of coordinates describes (1) which peptides of a
SWATH Data Analysis 291
A
1200
}
SWATH
isolation
1175 window
500
475 MS2
S2
m/z
MS2
S2
MS2
S2
MS2
MS2
450
425
400
{
Time
Cycle
time
B
MS2
MS2
MS2
MS2
Intensity
MS2
Intensity
m/z
Tim
e
Time
Fig. 1 SWATH MS. (a) While peptides elute from the LC, the mass spectrometer cycles through a large m/z
range (here from 400 to 1200 m/z) and co-fragments all peptide ions in relatively large isolation windows (or
“swathes”) of several m/z (here 25 m/z) [6]. For each isolation window, the resulting fragment ions are
recorded in a high-resolution composite mass spectrum (MS2). For every isolation window, the instrument
accumulates fragment ions for 100 ms, for 32 isolation windows covering 400–1200 m/z this results in a cycle
time of 3.2 s. At the beginning of each cycle a survey scan (MS1) is recorded (not shown). (b) In SWATH MS,
intensities of specific fragment ions are extracted from the highly multiplexed fragment ion spectra in a tar-
geted manner to produce extracted ion chromatogram peak groups
protein are most representative, i.e., unique and well detectable,

(2) the elution time of these peptides from the LC, (3) their pre-
dominant charge state, (4) the most abundant fragment ions
formed during fragmentation, and (5) their relative intensity.
Comprehensive, high-quality assay libraries are a critical prerequi-
site for the success of a SWATH MS analysis [10].
1.2 Automated While small-scale analysis of SWATH MS data can be performed

SWATH MS Analysis manually (similarly to SRM data analysis), this strategy is not fea-
sible any more for proteome-wide analysis. Automated software
packages provide fast, unbiased and comprehensive data analysis
solutions using advanced machine learning and data processing
algorithms. These software packages are able to automatically
import assay libraries, extract chromatographic traces from the raw
data and identify peak groups resulting from co-eluting fragment
ions (Fig. 2). These peak groups are then scored and statistically
evaluated using a noise model (negative distribution of scores).
This step allows the software to assign each peak group a quality
score, which reflects the probability of making a correct identifica-
tion. Researchers can then use this score to apply an experiment-
wide false discovery rate (FDR) cutoff based on how many false
quantifications they are willing to tolerate in the final dataset. If
multiple runs are present, a cross-run alignment can be performed
that integrates all available data to further increase the consistency
of the final data matrix (Fig. 2).
In this chapter we provide a detailed workflow for automated
analysis of SWATH MS data using the software suite OpenSWATH,
including pyProphet and TRIC Aligner [9, 11]. Because data qual-
ity is key for a successful automated SWATH MS analysis, we start
Data conversion
to open format
Raw data files

control
Quality
MSConvert
or qtofpeakpicker
(ProteoWizard)
Converted data files
OpenSWATH
(OpenMS)
Peak group identification
SWATH assay
SWATH assay library

and statistical scoring
for target proteins

library
Identified
peak groups
SWATH assay library
for iRT peptides
pyProphet
Scored
peak groups
TRIC Alignment
Cross-run
alignment
(msproteomicstools)
Aligned output
Fig. 2 Overview of the workflow for automated SWATH data analysis

with a brief description on how to load SWATH MS data into

Skyline [12] to visually inspect a small number of peptides in every
run for quality control.
2 Materials
Any files described in this chapter, including example files to test the
protocol, can be downloaded from the following website: http://
www.peptideatlas.org/PASS/PASS00779. Most of the example files
are related to a recent study in Mycobacterium tuberculosis that used
SWATH MS [13]. The installation of software tools is described for a
64-bit Microsoft Windows 7 system. However, many of the tools can
be installed on other platforms, such as Mac OS X and Linux, as well
(OpenMS, Python, pyProphet, TRIC aligner). Specific software ver-
sions used to develop this protocol are noted in the SoftwareVersions.
txt file available at the above mentioned website. It is important to
have enough free disk space available as the data files and the interme-
diary analysis files tend to be very large (dozens of GB).
1. SWATH data files. A detailed protocol of how to obtain high-
quality SWATH MS data on a TripleTOF instrument is pro-
vided in Chapter 16 by Hunter et al., but other instrument
types can be used as well (see Note 1). See Note 2 for a discus-
sion of parameters that are important during acquisition, such
as LC gradient, SWATH window size, and acquisition time.
Importantly, the samples have to contain retention time calibra-
tion (iRT) peptides at the time they are measured (see Note 3).
In this protocol, we assume that the 11 synthetic iRT peptides
provided in the iRT-kit by Biognosys have been spiked into
each sample [14]. Please note that the AB SCIEX TripleTOF
instrument produces two data files per SWATH run, a .wiff and
a .wiff.scan file, which should always be stored together (see
three .wiff and corresponding .wiff.scan example files).
2. File describing SWATH window setup. The selection of appro-
priate SWATH windows is an important part of a SWATH MS
experiment (Chapter 16 by Hunter et al.). For the current pro-
tocol, we assume that a fixed window setup (32 windows of
26 m/z each, 1 m/z overlap: 400–425 m/z, 424–450 m/z, 449–
475 m/z, … 1149–1175 m/z, 1174–1200 m/z) has been used
for data acquisition (see example file SWATHwindows_acquisi-
tion.tsv). In contrast to the SWATH windows used for data
acquisition, the SWATH windows used for data analysis should
not be overlapping. Therefore, the 1 m/z overlap that was used
for acquisition among the neighboring SWATH windows is
split, such that each window is only 25 m/z (first and last win-
dows are slightly different): 400–424.5 m/z, 424.5–449.5 m/z,
449.5–474.5 m/z, … 1149.5–1174.5 m/z, 1174.5–1200 m/z
(see example file SWATHwindows_analysis.tsv).
3. SWATH assay library. In addition to the SWATH data files, the

following workflow requires as an input a SWATH assay library,
containing precomputed decoy transition groups as well as
assays for the retention time peptides. These assays can be built
from experimental data (see Ref. 10 for a detailed protocol) or
downloaded from the SWATHAtlas database (see Note 4). As
an example library for this protocol we use a comprehensive
SWATH assay library of M. tuberculosis (see example file Mtb_
TubercuList-R27_iRT_UPS_decoy.tsv).
4. iRT peptide SWATH assays. In order to perform targeted
extraction, the retention time of each SWATH MS injection
needs to be transformed into a normalized retention time
space, which is also used by the assay library (see Note 3). The
protocol we describe here assumes that the 11 synthetic iRT
peptides from the iRT-Kit by Biognosys have been spiked into
every sample. For this protocol we provide an iRT peptide
SWATH assay library in TraML format for OpenSWATH anal-
ysis (see example file iRTassays.TraML) and a reduced table for
import into Skyline (see example file iRTassays_Skyline.tsv).
5. Skyline. Skyline is a free, open-source software for targeted
data analysis of various types of proteomics data [12] and pro-
vides great visualization options. The software can be down-
loaded from the website: http://skyline.maccosslab.org. It is
only available for Microsoft Windows operating systems.
6. Proteowizard. To convert vendor raw data files into an open
format, such as mzML or mzXML, we use the msconvert tool,
which is part of the Proteowizard software suite (Chapter 23
by Mallick et al.). Download Proteowizard from http://pro-
teowizard.sourceforge.net/downloads.shtml and install it on
your machine. Due to vendor licensing constraints, the data
conversion functionality of ProteoWizard is only available for
Microsoft Windows operating systems.
7. OpenMS. OpenMS is a cross-platform and open-source analy-
sis package for mass spectrometric data [15–18]. It is specifi-
cally suited for automated, large-scale analysis and can be run
on any machine, from a Desktop computer to a high-
performance Linux cluster [19, 20]. Download the appropri-
ate version of OpenMS for your operating system from
https://sourceforge.net/projects/open-ms/files/OpenMS/
OpenMS-2.0/. Install OpenMS into a local folder, for exam-
ple C:\Program Files\OpenMS-2.0. On newer Windows ver-
sions (Windows Vista, 7, and higher) some of the.NET
dependencies will already be installed, which will be indicated
during the install and removes the need to download and
install them. (The warning during.NET installation “Turn
Windows features on or off” can be ignored.)
8. Python. Several scripts used in this protocol require Python

2.7. On Windows, the easiest way to install Python is through
Anaconda, which can be downloaded from https://store.con-
tinuum.io/cshop/anaconda/. Choose the “Graphical
Installer” under the “PYTHON 2.7” tab appropriate for your
system (64-bit or 32-bit). Install Anaconda as Administrator
on your system (right-click, “Run as” and then select the
Administrator user). Select the installation for all users and
choose C:\Anaconda2\ as install path (should be the default).
Unless you have another Python installation on your system,
allow Anaconda to be added to your PATH and register it as
the default Python 2.7 installation (these options become
available during the installation process).
9. pyProphet. pyProphet is a tool to calculate a discriminating
score separating target from decoy assays and to compute a false
discovery rate (FDR) cutoff based on this score [21]. Open the
“Anaconda Prompt” by opening the Start Menu (Windows icon
in lower left corner) and type “Anaconda” in the search field.
Right-click the “Anaconda Prompt” entry and select “Run as
administrator”. To install the pyProphet package, enter:
pip install scikit-learn==0.15.2
pip install pyprophet==0.13.3
If running a 32-bit system, there are no pre-built packages and
a working C++ compiler is required on your machine. To get
such a compiler, go to http://aka.ms/vcpython27 and down-
load the file called VCForPython27.msi. After installing this
file, use the command above to install pyProphet.

10.
TRIC feature aligner. The TRansfer of Identification
Confidence (TRIC) software tool integrates information from
multiple OpenSWATH runs using cross-run alignment and
retention time correction [11]. Open the “Anaconda Prompt”
by opening the Start Menu (Windows icon in lower left cor-
ner) and type “Anaconda” in the search field. Right-click the
“Anaconda Prompt” entry and select “Run as administrator”.
To install the TRIC aligner package, enter:
conda install biopython
pip install msproteomicstools==0.3.2
3 Methods
3.1 Visualization Before proceeding with automated SWATH data analysis, it is cru-
of Selected Peptides cial to ensure adequate quality of the raw data. Skyline offers great
for Quality Control visualization options to assess SWATH data quality and we recom-
mend loading all data files to be subjected to automated analysis by
OpenSWATH first into Skyline to manually inspect a few quality
control peptides, such as the spiked-in iRT peptides. Optimally,
every SWATH data file is loaded into Skyline right after acquisition
to monitor LC stability as well as performance of the mass
spectrometer.
1. Open a blank Skyline document and go to Settings → Peptide
Settings → “Filter” tab. Uncheck the option to “Auto-select all
matching peptides” and click “OK”.
2. Go to Settings → Transition Settings → “Filter” tab. Uncheck the
option to “Auto-select all matching transitions” and click “OK”.
3. Go to Settings → Transition Settings → “Full-Scan” tab. Set the
following parameters: Acquisition method: DIA; Product mass
analyzer: TOF; Isolation scheme: From the drop-down menu,
select “Add…” and fill the “Edit Isolation Scheme” window as
shown in Fig. 3a (for this protocol, the window boundaries can
be copy-pasted from the file SWATHwindows_analysis.tsv).
Resolving power: 15,000 (depends on the state of the instru-
ment during SWATH data acquisition); Retention time filter-
ing: Include all matching scans. Click “OK”.
4. Go to Edit → Insert → Transition list. In the opening window,
paste assays of peptides you would like to monitor. For this
protocol, the assays for the iRT peptides can be copy-pasted
from the file iRTassays_Skyline.tsv. Click “Insert”.
5. Save the Skyline file.
6. Go to File → Import → Results and select “Add single-injection
replicates in files” and click “OK”.
7. Select the raw SWATH data .wiff files (the .wiff.scan files are not
showing up here, but need to be located in the same folder) and
click “Open”. When asked about removing the common prefix
of the file name, click “Remove”. Now it will take a while to
import all the data depending on file size and number of files.
8. After the data is loaded, arrange the Skyline file such that it is
convenient to inspect the runs (Fig. 3b). First, get the peak
area and retention time overview plots: View → Retention
Times → Click “Replicate Comparison” and View → Peak
Areas → Click “Replicate Comparison”. Next, arrange the
panels by drag-and-drop to the desired location. Then go to
“Settings” and click “Integrate all”. Finally, right-click on a
chromatogram plot → Auto-zoom X-axis → Best Peak.
9. Inspect the SWATH peak groups to judge performance of LC
(peak shape, retention time stability) and mass spectrometer
(signal intensity and signal-to-noise ratio for low abundant
peaks) over time (Fig. 3b). Note that Skyline uses a scoring
system to determine the correct peak group but it can still hap-
pen that it does not pick the correct one. In these cases, peak
boundaries can be changed manually by click-and-drag in the
retention time axis of the chromatogram plot. A confidently
Fig. 3 SWATH data visualization in Skyline. (a) Screen shot showing how SWATH windows are defined in Skyline.
This is essential for the software to know from which SWATH window the specified fragment ions should be
extracted. (b) Example of how to organize the Skyline window for fast and efficient monitoring of large numbers
of SWATH runs. The example here shows just three runs, but Skyline can easily handle dozens of runs
identified peptide is characterized by a well identifiable peak

group near the expected retention time with co-eluting frag-
ment ions that match the relative intensities given in the assay
library. More criteria that are typically used to identify correct
peak groups are discussed elsewhere [9, 22]. For detailed
Skyline tutorials please consult the Skyline website (http://
skyline.maccosslab.org).
3.2 Raw Data MSConvert, provided through the ProteoWizard software suite
Conversion Into mzML (Chapter 23 by Mallick et al.), enables conversion of proprietary
raw data files (.wiff for AB SCIEX, .raw files for Thermo Fisher
instruments) into an open format, such as mzML or mzXML.
OpenSWATH can handle both file types, but when running on a
Windows PC, the input files need to be in mzML format.
1. Start the software by opening the Start Menu (Windows icon
in lower left corner), type “MSConvert” in the search field and
click on “MSConvert”.
2. In the MSConvert window, select “List of Files” and use the
“Browse” button to locate the raw files to be converted (the
.wiff and .wiff.scan files need to be located in the same folder).
Click the “Add” button. Choose an output directory using the
second “Browse” button. Then set the following parameters:
Output format: mzML; Extension: mzML; Binary encoding
precision: 64-bit; Write index: yes; Use zlib compression: yes.
All other checkboxes should be unchecked and no filters
should be set (see Fig. 4). While OpenSWATH produces best
results on profile data, it is possible to run it on centroided data
to reduce disk space and execution time (see Note 5).
3.3 Automated OpenSWATH can be run as a single command, which executes all
Analysis steps of the OpenSWATH data analysis pipeline automatically. The
with OpenSWATH individual steps can be controlled through flags passed on the com-
mand line. Use the --helphelp command to see all options and
to learn about particular speed and memory optimizations avail-
able. Table 1 summarizes the most important options.
1. Start the software by opening the Start Menu (Windows icon
in lower left corner), type “TOPP” in the search field and click
on “TOPP command line” which will start a command prompt
preloaded with the necessary paths to execute OpenSWATH.
2. OpenSWATH requires four input files: (1) The actual data in
mzXML or mzML format (on Windows PC only mzML for-
mat is accepted). (2) An assay library containing assays for all
target peptides. OpenSWATH can take assay libraries in both
tab-separated table (tsv) and TraML format. For more details
see Notes 4 and 6. (3) An assay library containing assays for all
iRT peptides (see Note 3) in TraML format. 4) A file with a
Fig. 4 SWATH data conversion with MSConvert. Conversion of raw data to mzML
format using the MSConvert GUI, showing the individual options that should be
selected during the conversion step for profile-mode conversion
small table specifying the SWATH windows out of which the

fragment ion traces should be extracted. Note that the extrac-
tion windows are slightly different from the windows specified
for data acquisition, i.e., they should not contain overlaps (see
example files described in the Materials section). A full com-
mand for running OpenSWATH looks like this (enter entire
command on a single line):
OpenSwathWorkflow.exe
– in data.mzML
– tr library.tsv
– sort_swath_maps
– readOptions cache
– tempDirectory C:\Temp
– batchSize 1000
– tr_irt iRT_assays.TraML
– swath_windows_file SWATHwindows_analysis.tsv
– out_tsv osw_output.tsv
Table 1
OpenSWATH parameters
Parameter Explanation
--helphelp To see all options
-out_chrom Additionally also output the generated XICs. This option is
required if the OpenSWATH results are to be inspected
manually with TAPIR [23] or for other downstream
processing not discussed here
-swath_windows_file Text file specifying the SWATH windows (see example file and
main text for more details)
-rt_extraction_window The size of the retention time extraction window in seconds
(centered around the theoretical retention time calculated
using the iRT alignment), usually 300–600 s are reasonable,
but may be adjusted according to the chromatography and the
accuracy of expected retention times in the assay library
-mz_extraction_window Size of the extraction window in Dalton. The default value of
0.05 may need to be adjusted depending on the instrument
resolution
-ppm Indicates that -mz_extraction_window is given in ppm
instead of Dalton
-Scoring:TransitionGroup Minimal peak width in seconds preventing spurious peaks from
Picker:min_peak_width appearing in the results. Defaults to 14 s
-TransitionGroupPicker: Indicates whether background subtraction should be performed
background_subtraction during quantification. This option may increase the accuracy
of the quantification
-Scoring:TransitionGroup Smoothing parameter for Savitzky-Golay smoothing (expressed
Picker:PeakPicker MRM: in number of scans). Increasing this parameter leads to
sgolay_frame_length stronger smoothing. Alternatively, Gaussian smoothing may be
used by -Scoring:TransitionGroupPicker:Peak
PickerMRM:use_gauss
-Scoring:stop_report_ Determines the number of peak groups per assay that are
after_feature reported in the result. A smaller number produces less output;
generally 5–10 is appropriate
-batchSize Determines the size of the batch of chromatograms that are
loaded into memory and analyzed at once. The smaller this
number, the less memory is required
-sort_swath_maps This parameter ensures that the SWATH maps in the input are
sorted by m/z (this assumes that the swath window file given
in -swath_windows_file also contains windows sorted by
m/z)
-readOptions Setting this option to cache instructs the software to not load
all data into memory at once but to create a local cache of the
data. Make sure to specify a directory for the temporary files
using the -tempDirectory parameter
-tempDirectory Defines the directory to store temporary files (required if
-readOptions cache is used)
The output of the OpenSWATH command is a large table contain-

ing one scored peak group per row (usually more than one peak
group is scored per assay). The properties of each peak group are
given in columns (i.e., the assay used to generate the peak group,
the retention time, the individual scores etc.).
3.4 Error Rate After running OpenSWATH, q-values corresponding to the FDR
Estimation of peak identification can be estimated with the pyProphet soft-
with pyProphet ware tool. The only input for pyProphet is the OpenSWATH result
table file (osw_output.tsv).
1. Use the same command line window as above and enter (on a
single line):
C:\Anaconda2\Scripts\pyprophet.exe
--ignore.invalid_score_columns
--d_score.cutoff=0.5
osw_output.tsv
This command will output a set of files, of which osw_out-
put_with_dscore_filtered.csv will be used in the next step.
2. pyProphet will also output a table containing the number of
true and false positives for different q-value cutoffs (_sum-
mary_stat.csv). A q-value cutoff (or FDR) between 1 and 10 %
is appropriate for most studies. For further information, the
pdf report file should be consulted, which contains plots of the
fitted distributions as well as ROC curves. This step provides
an opportunity to identify common errors, such as using an
assay library without decoys or using an assay library unsuited
for the measured sample (e.g., from another organism). In a
successful run, target and decoy distributions should be clearly
separated as shown in Fig. 5.
Fig. 5 Peak group scoring with pyProphet. To ensure that the FDR estimation step using pyProphet worked
properly, the output should be manually inspected to identify potential problems. The plotted distributions
should be bimodal and the decoy distribution should be close in shape to the false positive distribution (left part
of the bimodal distribution)
3.5 Cross-Run If multiple SWATH MS runs were analyzed with OpenSWATH,

Alignment with TRIC the resulting quantitative data matrix may contain a substantial
Aligner number of missing values. One way to reduce these missing values
is to use the TRIC algorithm to perform cross-run alignment. This
algorithm aligns peak groups across all runs using either a reference-
based alignment or a tree-based alignment strategy and is imple-
mented in the feature_alignment.py tool from the
msproteomicstools package. Table 2 describes the individual
parameters. For further information you may also consider the
online documentation at https://github.com/msproteomics-
tools/msproteomicstools.
Table 2
TRIC alignment parameters
Parameter Explanation
--help To see all options
--method Alignment method, either reference-based alignment (best_overall or
global_best_overall) or reference-free, tree-based alignment
(LocalMST). The method best_overall is most conservative as it
does not change peak groups with a good q-value, thereby avoiding errors
introduced during alignment. Alternatively, global_best_overall
may be chosen, which, however, requires that the maximal shift in
retention time is low, otherwise even good peak groups may be removed
from the output if their retention time is too far off. The method
LocalMST is recommended for large-scale SWATH MS analysis
--realign_method Describes how the alignment between individual runs is performed and
may be either linear (linear or diRT) or nonlinear (lowess or
CVSpline). For small datasets, linear alignment methods typically
perform sufficiently well and are faster than the nonlinear methods
--max_rt_diff Maximal shift in retention time that is tolerated after alignment
(in seconds). The software will automatically estimate a sensible value if
it is set to auto_3medianstdev. Adjust this parameter to your
chromatography. Nonlinear alignment methods typically allow for lower
values than linear alignment methods
--target_fdr Desired q-value (or FDR) on assay level (estimated using decoys)
--max_fdr_quality q-value (or FDR) cutoff to still consider a feature for quantitation.
Typically, a value of 0.01 (1 %) is appropriate. Higher values (e.g., 0.05)
may be appropriate for small to medium size datasets as they increase
completeness of the data matrix, but may also introduce lower quality
peaks groups
1. Use the same command line window as above and enter (on a
single line):
C:\Anaconda2\python.exe
C:\Anaconda2\Scripts\feature_alignment.py
--in file1_with_dscore.csv file2_with_dscore.
csv file3_with_dscore.csv
--out aligned.tsv
--method best_overall
--realign_method diRT
--max_rt_diff 90
--target_fdr 0.01
--max_fdr_quality 0.01
The command exemplified here will run alignment on three files
using linear iRT alignment and pick an appropriate peak group in
each run within the aligned window using a reference-based align-
ment. Only peptides that pass the identification threshold esti-
mated by the TRIC algorithm will be reported in the final output
(set to 1 % FDR using the --target_fdr parameter in the
example above). Note that recent results suggest substantial
improvements when using reference-free, nonlinear retention time
alignment, which can be enabled using the “LocalMST” tree-
based alignment option and the “lowess” RT correction option.
4 Notes
1. While SWATH MS was originally developed on an AB SCIEX

TripleTOF instrument, SWATH-like data-independent acqui-
sition can also be obtained from other types of instruments.
One of the major considerations is the acquisition speed, which
needs to allow sufficient sampling during elution of an analyte
from the LC column. The software tools described here, i.e.,
Skyline and OpenSWATH, can both analyze data from multi-
ple vendors, including data acquired on Waters Synapt TOF
instruments, Thermo Fisher Q Exactive instruments and AB
SCIEX TripleTOF instruments [9]. Multiple groups have
reported SWATH-like acquisition on Thermo Q Exactive
instruments and successfully used OpenSWATH and other
SWATH MS software for data analysis [24–26].
2. SWATH acquisition requires careful optimization of multiple
parameters including chromatographic peak width, m/z win-
dow width, acquisition time per window and overall m/z
coverage. In the initial implementation of SWATH MS,
these parameters were set to ca. 30 s for the chromatographic
peak width, 25 m/z for the window width, 100 ms for the
acquisition time per window and a precursor mass range of
400–1200 m/z was chosen, resulting in 32 SWATH windows.
Since then, multiple improvements to the scheme have been

proposed (Chapter 16 by Hunter et al.). It is important to
remember that one of the main constraints is to retain suffi-
cient sampling of each peptide precursor during its chromato-
graphic elution. In the original implementation, no more than
100 ms acquisition time could be allowed in order to complete
a 32-window cycle within 3.3 s. Assuming an average chro-
matographic peak width of 30 s, this would enable the sam-
pling of on average 9 data points across the peak which is
sufficient for the OpenSWATH algorithm to reconstruct a
peak and perform fragment ion trace cross-correlation. If more
powerful chromatographic separation were employed (as
offered by UPLC, for example), the other parameters would
have to be adjusted accordingly. For example, the acquisition
time of each SWATH window could be reduced in order to
retain sufficient sampling. Also, the precursor mass range could
be decreased or the size of each window increased, both result-
ing in a smaller number of SWATH windows allowing the
instrument to complete each cycle faster. Also, working with
flexible window sizes has been shown to improve performance
(Chapter 16 by Hunter et al.).
3. The presence of iRT retention time reference peptides, the so-
called iRT peptides, is crucial to calibrate the retention time
information present in the SWATH assay library to the reten-
tion times recorded in the SWATH data. In principle, any set
of well-detectable peptides spanning a wide retention time
range can be used [14]. Even endogenous peptides can serve
as iRT peptides, making it unnecessary to purchase and spike
synthetic peptides into every sample. A set of conserved endog-
enous peptides found in most eukaryotic samples has recently
been described [27]. These Common internal Retention Time
standards (CiRT) can be used in the same way as the iRT pep-
tides without the need to spike in additional peptides. However,
for some specialized applications, i.e., blood plasma analysis,
the CiRT approach may be suboptimal and the use of the com-
mercial iRT peptides is recommended.
4. Comprehensive, ready-to-use SWATH assay libraries have
been published for a number of organisms, including human
[28] and yeast [29], and can be downloaded from http://
www.SWATHAtlas.org. Schubert et al. provide an extensive
discussion on various aspects of SWATH assay libraries and
provide a detailed protocol on how to build high-quality assay
libraries from shotgun data [10].
5. The OpenSWATH software is designed and tested on profile
data and produces optimal results on profile data. However, it
is possible to run OpenSWATH on centroided data, reducing
the required disk space and execution time of the whole work-
flow substantially. As centroiding reduces the number of peaks
in the data and sometimes may remove low-intensity peaks, the
XICs generated by OpenSWATH become less smooth and peak
detection becomes less sensitive. If centroiding is performed, it
is crucial to compare the results of OpenSWATH to those
obtained on profile data on the same dataset to obtain an accu-
rate estimation of how centroiding affects the identification
rate. Multiple centroiding algorithms are freely available,
including the internal msconvert centroiding (enabled during
conversion by the “Filter” settings) as well as the separate exe-
cutable qtofpeakpicker.exe (found in the same folder as
msconvert.exe after installation of ProteoWizard). The
qtofpeakpicker has been specifically designed for data derived
from QTOF instruments and we recommend to use it to cen-
troid TripleTOF data [10]. The qtofpeakpicker is run directly
on the .wiff files and we recommend using the same parame-
ters as Schubert et al. to generate centroided mzML files
[10]: --resolution=20000 --area=1 --thresh-
old=1 --smoothwidth=1.1.
6. A high-quality SWATH assay library is an essential feature for any
targeted analysis of DIA or SWATH MS data. The OpenSWATH
software uses this information to extract XICs for the targeted
peptide from the appropriate fragment ion spectra and to score
the peaks according to how well they match the information in
the assay library (expected retention time, fragment ion m/z,
etc.). Alongside the assays for the proteins of interest (target
assays), the library needs to contain a set of “decoy assays” which
represent peptides not present in the sample (usually shuffled tar-
get peptide sequences). These assays are scored in the same fash-
ion as the targeted assays and used by pyProphet to estimate the
null distribution of different OpenSWATH scores and compute
q-values (for FDR control). Having decoys in the assay library is
essential for this step to work. Repositories such as http://www.
SWATHAtlas.org provide assay libraries that already contain
decoys. Please refer to Röst et al. and Schubert et al. for more
information on the scoring algorithm and the importance of
decoys and FDR control [9, 10]. OpenSWATH accepts assay
libraries in both tab-separated table (tsv) and TraML format. For
large assay libraries, the tsv format is more memory efficient than
the TraML format. OpenMS comes with the tools
ConvertTraMLToTSV and ConvertTSVToTraML, which
enable conversion between the two formats (regardless of the
names of these conversion tools, the file ending for the table for-
mat has to be.csv, while the actual data format is tab-separated).
References
1. Domon B (2012) Considerations on selected for creating and analyzing targeted proteomics
reaction monitoring experiments: implications experiments. Bioinformatics 26:966–968.
for the selectivity and accuracy of measure- doi:10.1093/bioinformatics/btq054
ments. Proteomics Clin Appl 6:609–614. 13. Schubert OT, Ludwig C, Kogadeeva M et al
doi:10.1002/prca.201200111 (2015) Absolute proteome composition and
2. Picotti P, Aebersold R (2012) Selected reaction dynamics during dormancy and resuscitation of
monitoring-based proteomics: workflows, Mycobacterium tuberculosis. Cell Host
potential, pitfalls and future directions. Nat Microbe. doi:10.1016/j.chom.2015.06.001
Methods 9:555–566. doi:10.1038/ 14. Escher C, Reiter L, MacLean B et al (2012)
nmeth.2015 Using iRT, a normalized retention time for
3. Picotti P, Bodenmiller B, Mueller LN et al more targeted measurement of peptides.
(2009) Full dynamic range proteome analysis of Proteomics 12:1111–1121. doi:10.1002/
S. cerevisiae by targeted proteomics. Cell pmic.201100463
138:795–806. doi:10.1016/j.cell.2009.05.051 15. Kohlbacher O, Reinert K, Gröpl C et al (2007)
4. Venable JD, Dong M-Q, Wohlschlegel J et al TOPP--the OpenMS proteomics pipeline.
(2004) Automated approach for quantitative Bioinformatics 23:e191–7. doi:10.1093/bio-
analysis of complex peptide mixtures from tan- informatics/btl299
dem mass spectra. Nat Methods 1:39–45. 16. Sturm M, Bertsch A, Gröpl C et al (2008)
doi:10.1038/nmeth705 OpenMS – an open-source software framework
5. Chapman JD, Goodlett DR, Masselon CD for mass spectrometry. BMC Bioinformatics
(2013) Multiplexed and data-independent tan- 9:163. doi:10.1186/1471-2105-9-163
dem mass spectrometry for global proteome 17. Röst HL, Schmitt U, Aebersold R, Malmström
profiling. Mass Spectrom Rev. doi:10.1002/ L (2014) pyOpenMS: a Python-based interface
mas.21400 to the OpenMS mass-spectrometry algorithm
6. Gillet LC, Navarro P, Tate S et al (2012) Targeted library. Proteomics 14:74–77. doi:10.1002/
data extraction of the MS/MS spectra generated pmic.201300246
by data-independent acquisition: a new concept 18. Röst HL, Schmitt U, Aebersold R, Malmström
for consistent and accurate proteome analysis. L (2015) Fast and efficient XML data access
Mol Cell Proteomics 11:O111.016717. for next-generation mass spectrometry. PLoS
doi:10.1074/mcp.O111.016717 One 10:e0125108. doi:10.1371/journal.
7. Gallien S, Duriez E, Crone C et al (2012) pone.0125108
Targeted proteomic quantification on 19. Junker J, Bielow C, Bertsch A et al (2012)
quadrupole- orbitrap mass spectrometer. Mol TOPPAS: a graphical workflow editor for the
Cell Proteomics 11:1709–1723. doi:10.1074/ analysis of high-throughput proteomics data.
mcp.O112.019802 J Proteome Res 11:3914–3920. doi:10.1021/
8. Peterson AC, Russell JD, Bailey DJ et al (2012) pr300187f
Parallel reaction monitoring for high resolu- 20. Aiche S, Sachsenberg T, Kenar E et al (2015)
tion and high mass accuracy quantitative, Workflows for automated downstream data
targeted proteomics. Mol Cell Proteomics.
analysis and visualization in large-scale computa-
doi:10.1074/mcp.O112.020131 tional mass spectrometry. Proteomics 15:1443–
9. Röst HL, Rosenberger G, Navarro P et al 1447. doi:10.1002/pmic.201400391
(2014) OpenSWATH enables automated, tar- 21. Teleman J, Röst HL, Rosenberger G et al
geted analysis of data-independent acquisition (2014) DIANA-algorithmic improvements for
MS data. Nat Biotechnol 32:219–223. analysis of data-independent acquisition MS
doi:10.1038/nbt.2841 data. Bioinformatics. Oxford, England.
10. Schubert OT, Gillet LC, Collins BC et al (2015) doi:10.1093/bioinformatics/btu686
Building high-quality assay libraries for targeted 22. Reiter L, Rinner O, Picotti P et al (2011)
analysis of SWATH MS data. Nat Protoc mProphet: automated data processing and sta-
10:426–441. doi:10.1038/nprot.2015.015 tistical validation for large-scale SRM experi-
11. Röst HL, Liu Y, D’Agostino G et al (2016) ments. Nat Methods 8:430–435. doi:10.1038/
TRIC: an automated alignment strategy for nmeth.1584
reproducible protein quantification in targeted 23. Röst HL, Rosenberger G, Aebersold R,
proteomics. Nat Methods 13:777–783. Malmström L (2015) Efficient visualization of
doi:10.1038/nmeth.3954 high-throughput targeted proteomics experi-
12. MacLean B, Tomazela DM, Shulman N et al ments: TAPIR. Bioinformatics. Oxford,
(2010) Skyline: an open source document editor England. doi:10.1093/bioinformatics/btv152
24. Malmström L, Bakochi A, Svensson G et al 27. Parker SJ, Röst HL, Rosenberger G et al
(2015) Quantitative proteogenomics of human (2015) Identification of a set of conserved
pathogens using DIA-MS. Proteomics 129:98– eukaryotic internal retention time standards for
107. doi:10.1016/j.jprot.2015.09.012 data-independent acquisition mass spectrome-
25. Bruderer R, Bernhardt OM, Gandhi T et al try. Mol Cell Proteomics 14:2800–2813.
(2015) Extending the limits of quantitative doi:10.1074/mcp.O114.042267
proteome profiling with data-independent 28. Rosenberger G, Koh CC, Guo T et al (2014) A
acquisition and application to acetamino- repository of assays to quantify 10,000 human
phen treated 3D liver microtissues. Mol proteins by SWATH-MS. Sci Data 1:140031.
Cell Proteomics. doi:10.1074/mcp.M114. doi:10.1038/sdata.2014.31
044305 29. Selevsek N, Chang C-Y, Gillet LC et al (2015)
26. Egertson JD, Kuehn A, Merrihew GE et al Reproducible and consistent quantification of
(2013) Multiplexed MS/MS for improved the Saccharomyces cerevisiae proteome by
data-independent acquisition. Nat Methods SWATH-MS. Mol Cell Proteomics 14:739–
10:744–746. doi:10.1038/nmeth.2528 749. doi:10.1074/mcp.M113.035550
Chapter 21
Virtualization of Legacy Instrumentation Control

Computers for Improved Reliability, Operational Life,
and Management
Jonathan E. Katz
Abstract
Laboratories tend to be amenable environments for long-term reliable operation of scientific measurement
equipment. Indeed, it is not uncommon to find equipment 5, 10, or even 20+ years old still being rou-
tinely used in labs. Unfortunately, the Achilles heel for many of these devices is the control/data acquisi-
tion computer. Often these computers run older operating systems (e.g., Windows XP) and, while they
might only use standard network, USB or serial ports, they require proprietary software to be installed.
Even if the original installation disks can be found, it is a burdensome process to reinstall and is fraught
with “gotchas” that can derail the process—lost license keys, incompatible hardware, forgotten configura-
tion settings, etc. If you have running legacy instrumentation, the computer is the ticking time bomb
waiting to put a halt to your operation.
In this chapter, I describe how to virtualize your currently running control computer. This virtualized
computer “image” is easy to maintain, easy to back up and easy to redeploy. I have used this multiple times
in my own lab to greatly improve the robustness of my legacy devices.
After completing the steps in this chapter, you will have your original control computer as well as a
virtual instance of that computer with all the software installed ready to control your hardware should your
original computer ever be decommissioned.
Key words Legacy hardware, Virtual computers, System reliability, Mass Spectrometry, VirtualBox,
Cloning, Systems management
1 Introduction
Scientific measurement devices with attached control computers

often have a viable life that greatly exceeds the supported life of the
operating system that hosts the vendor software. It is also often the
case that the vendors have either gone out of business or have dis-
continued support for their older devices in this same time frame.
Even for the vendors still in business, the process of, for example,
reinstalling and reconfiguring a mass spectrometer, an LC system
and its associated autosampler is not a fast nor simple process.
309
310 Jonathan E. Katz
Many labs often have multiple deployments built on (now) legacy

operating systems, such as Windows XP, for which there is no
driver support under more current versions of Windows. In spite
of being well past their amortization, these devices still provide
invaluable daily service. A failed computer alone (all too common
an occurrence) should not be deciding factor forcing procurement
of replacement funds.
In this chapter, I show you how to take the required steps to
virtualize a legacy control computer. A virtual computer is com-
puter that is created in software that runs on a host computer. In
its most common deployment, there is a definition file that
describes the features of the computer (the memory, the processor
type, what interface cards are available, etc.) and there is an image
file that contains the contents of the hard drive. Modern comput-
ers are more than capable of storing, simultaneously, the definition
and image files of multiple virtual computers. While there is a
minor performance penalty for running the virtual equivalent of a
real computer, this deficit is greatly outweighed by the administra-
tive and logistic advantages. Effectively, a virtualized computer is
entirely recapitulated by a single file; a file as easily copied, moved,
or otherwise manipulated as any other. Physical computers are
often different from one another, but, from the perspective of the
operating system and the applications, every virtual computer with
the same definition file looks the same.
When deploying a new physical control computer (say to
replace one with a failed motherboard), it requires installation of
an operating system, all the applications and then configuring all
the components—hours of work. On the other hand, to deploy a
duplicate of a configured virtual machine requires copying a file
and starting the virtual machine—minutes of work.
Creating a new virtual computer instance is fairly straightfor-
ward and not unlike deploying a new physical computer. It requires
installation of an operating system, installation of applications and
then configuration of these components. However, for legacy con-
trol, this has the same problems as installing a new physical com-
puter—copies of operating systems need to be found, original
control software and drivers needs to be installed and configured,
etc. The scope of this chapter, however, is the virtualization of an
existing installation.
Your currently running machine is already installed and config-
ured, so, all the installation steps are already done! This is the
inherent advantage to virtualizing a currently running
machine. However, virtualizing existing Windows installations has
some other complications as compared to creating a new virtual
instance. The primary complication comes from the fact that
Windows has a concept of the underlying hardware and virtualiza-
tion will change the hardware that the operating system is e xpecting
to see—Windows interprets these changes as errors. There are two
primary forms these errors take. One is licensing—if Windows sees
Legacy Instrumentation Virtualization 311
too many changes to what it perceives as the underlying hardware,

it will assume that it has been copied and generate an error;
Windows will boot, but it will require reactivation. The other error
is that if the underlying hardware changes, Windows will assume it
has been corrupted and will not successfully boot. Both of these
issues will have to be addressed. The final complication to address
here is that of virtualization peripherals—the cards and ports on
the computer that the applications need to use. The serial ports,
the network interfaces, etc. all need to be made available to the
virtual computer. There is one final error that can occur, the origi-
nal computer may not have appropriate drivers installed to match
what is presented by the virtual hardware; this tends to be rare for
XP virtualizations.
As an example, I will be using a 2007 install of an LTQ-Orbitrap
with a CTC-Leap autosampler and an Eksigent 2D LC system. The
control computer is running Windows XP SP3. This configuration
presents some nice challenges in that it requires the use of 2 network
interface cards as well as several RS-232 serial connections. I have
used very similar steps to virtualize spectrophotometers, DNA spot-
ters, and gel documentation systems. While the presented process
focuses on the virtualization of Microsoft Windows based installa-
tions (primarily XP) using VirtualBox virtualization software, the
essential steps remain unchanged for different operating systems and
are as follows:
1. Prepare the live machine to be virtualized.
2. Cleanly shutdown the live machine.
3. Duplicate the raw hard drive data to an image file contains the
complete image of the hard drive.
(a) This will usually be done by booting an alternate operat-
ing system on the live machine, attaching an external hard
drive and executing a duplication job. Sometimes it will be
required (or desired) to remove the current live hard drive
and duplicate it on a different machine.
4. Prepare the host computer to run the VirtualBox Software
(a) Adjust BIOS settings.
(b) Install VirtualBox.
5. Convert the hard drive image file into an image readable by
VirtualBox.
6. Create an initial hardware definition and attempt booting the
virtualized machine.
7. Finish configuration and perform required post-virtualization
processes.
(a) Re-validate windows.
(b) Configure peripherals.
(c) Check application configurations.
2 Materials
1. Software—Within this chapter, there are many referenced

tools that are to be downloaded from the Internet. Because of
the uncontrollable volatility of Internet links, we have made
available an uncontrollably volatile mirror site for these appli-
cations. The mirror link is:
https://drive.google.com/folderview?id=0B41aLALlbwPW
X2FQRjVqMjJoeFU
Conveniently accessible by its volatile short link:
https://goo.gl/3MDo5Y
This mirror should only be used as a last resort when the offi-
cial channels are unavailable.
2. Host computer.
(a) Hardware—Most modern computers should be suffi-

ciently powerful to virtualize a computer from a couple of
years or more in the past. There are two important consid-
erations. The first is that the available RAM should be at
least 1–2 GB more than the RAM that is installed on the
machine you are virtualizing. The second is that the com-
puter should have free hard drive space at least the size of
the hard drive of the computer you are virtualizing. For
example, if I am virtualizing a machine with a 500 GB
hard drive, I find a host computer with a 750 GB hard
drive installed usually more than sufficient.
(b) Operating system—My preference in virtualization soft-
ware is VirtualBox from Oracle. Any operating system
that can run VirtualBox will be sufficient, this includes
Linux, Windows, and Mac OS.
(c) Virtualization software—VirtualBox is available for free
download from [1]; for Linux distributions there is often
an easy install process, see Note 1 for an Ubuntu
example.
3. External hard drive—An external USB hard drive that is at
least the size of the hard drive you wish to virtualize. Note, if
you are virtualizing a 32-bit machine, it is probably best to use
a 2 TB or smaller external USB drive.
4. External drive reader (optional)—For some people and for
some significantly older computers, it is difficult to create the
image of the hard drive from the machine that you wish to
virtualize. This can happen for a number of reasons—the
computer might not be able to boot external media, the com-
puter might have severely limited resources, etc. In these cases,
it is often easier to remove the hard drive and duplicate the
contents on another machine. To do this, it is recommended to
obtain a multi-format USB hard drive reader. I have had good
success with the SATA/IDE to USB converters made by

Sabrent, however, other models should work similarly well.
One that mounts SATA, PATA, and IDE will fulfill most needs.
5. External boot media (e.g., CD-R, USB flash drive)—If you are
going to create your disk image on the existing physical com-
puter, you will need to boot into an alternate operating system
to allow for the duplication of hard drive contents without cor-
ruption. Depending on your physical computer, this can be
either a blank CD-R or USB stick prepared with a bootable
copy of an operating system suitable for making the image
copy. Typically a 2 GB USB drive is more than sufficient.
6. External USB->Serial adapters (possibly required). It was very
common for older devices to use RS-232 serial ports as their
means of connection to a host computer. Sometimes the live
computer would use “multiple port serial cards” to increase the
available RS-232 ports. It is common for your new host
machines to have less of these ports available. Further, it is
sometimes difficult to virtualize the multiple-port RS-232
cards and even the onboard RS-232 port. These issues are
readily solved with USB based RS-232 interfaces. I have had
much success with the TRENDnet TU-S9 USB to RS-232
interfaces, however, other models should work similarly well.
3 Methods
1. Prepare the physical computer to be virtualized. Much of this

section mirrors the unsupported protocols referenced from
VirtualBox [2]:
(a) The primary hardware incompatibility that stops successful
Windows XP virtualization is that the Windows hard drive
driver does a check of the attached hard drive controller; if
the controller has changed, Windows will stop the boot
process. The easiest remedy is to relax these checks before
creating a hard drive image; VirtualBox provides a tool on
their website (MergeIDE) that greatly simplifies this pro-
cess. This is not a problem with Windows98.
●● Obtain MergeIDE (noting capitalization) from:
https://www.virtualbox.org/attachment/wiki/
Migrate_Windows/MergeIDE.zip
●● Extract the contents of MergeIDE.zip.
●● Execute MergeIDE.bat on the live computer (“dou-
ble click” works fine for this. Be sure to double click
“MergeIDB.bat” and not “MergeIDE.reg”).
(b) Cleanly shutdown the live computer.
2. Create a raw (“dd”) copy (i.e., “image”) of the live machine’s
hard drive. In this example, we will boot the existing live com-
puter with either a CD or USB containing a copy of OSFClone.

This is the least invasive process and, in practice, perhaps the
most useful technique but it does require the live computer to
be able to boot off of secondary media such as a CD-ROM or
a USB flash drive. While we give the OSFClone example here,
this can also be performed from any live Linux distribution as
described in Note 2. While least optimal, if the computer can-
not boot from secondary media, it will be required to physi-
cally take the hard drive out of the live machine and then, on
a different machine, use an external drive interface to create
the image. This is also the required process if the live machine
is having problems booting off of CD-R’s and USB sticks. If
needed, the process of removing a hard drive to image it is
described in Note 3.
Below is the method for creating a full hard drive copy (i.e.,
creating a hard drive image) on a live system using OSFClone.
(a) Determine which bootable media type will work best on
your current live computer. While more convenient, older
machines often have difficulty booting from USB media
and are often more successful at booting from CD-ROM.
(b) Go to http://www.osforensics.com/tools/create-disk-
images.html and download either the.iso (to create a
bootable CD-ROM) or the.zip (to create a bootable USB).
(c) Follow the instructions on the osforensics website to burn
the image to your chosen media. For example, in Windows
7, for the.iso, right-click and select burn image, for the.zip,
unpack it then run the ImageUSB tool. In Linux, one might
chose to use “brasero” to write the.iso to a CD-R disk.
(d) Power off the live computer if not already.
(e) Attach the USB external hard drive you will use to hold
your hard drive image.
(f) Boot OSFClone on your live computer using the CD-R or
USB you created. For this to be successful, you may need
to enter the BIOS to change the boot order (see Fig. 1 and
Note 4 for some further description of this).
(g) Figure 2 shows the boot process for OSFClone. You will
be presented with a prompt “boot:” at which point you
should press the enter key.
(h) You may be presented with a prompt to “see video modes
available”, it is ok to just press the space key here to
continue.
(i) Once OSFClone is started, select the option to “image” a
complete drive (type the number 2 then press the enter
key; please note that this interface does not use the mouse
or the arrow keys. You will need to type the values you are
Fig. 1 Booting from secondary media. Panel (a) shows a representative example of a splash screen when a
computer is powered on. Pressing F12, as prompted by this screen brings up the menu shown in panel (b).
From this menu, the system can be booted from prepared USB flash drives, removable CD drives, or other
media
Fig. 2 Booting into OSFClone. OSFClone is built on an operating system with very rudimentary drivers. As such,
it is entirely keyboard controlled. In this screen shot, the user was prompted to press enter at the “boot:”
prompt and then asked to press the space bar to boot without further defining the video card
prompted for and press enter). Image will copy the entire
hard drive into a file (in contrast to “clone” which would
duplicate one device onto another device). You may be
presented with options of which “format you wish to
use” and you should select “dd” to perform this imaging
operation (i.e., type 1 and press the enter key).
(j) You will now need to select a source and destination. See
Fig. 3 as an example.
●● Type 1 and press enter to select the source; most
likely it will be “/dev/hda"—confirm this on your
system; you will be able to confirm this based on the
reported size (it will match the size of your “c:” drive).
Return to the option menu.
●● Type 2 and press enter to select the destination; most
likely it will be “/dev/sda1"—confirm this on your
system; you will be able to confim this based on the
reported size (it will most likely be the largest size
available). Note that for the destination of an image
operation, the partition will be mounted and a file will
be created in that partition.
Fig. 3 Setting Source and Destination in OSFClone. OSFClone uses a command line interface to set all the
parameters. Initially, to set the source, type “1” and press enter (as shown in panel (a)). You will be presented
with a screen similar to panel (b). Most likely you will select the option corresponding to /dev/hda—confirm
this based on the device size. Then you will be back at panel (a) and you can select option 2 for the destination.
Presented with a screen similar to Panel (c), you will probably select the option for /dev/sda1 or /dev/sdb1;
confirm this by matching the size of your external hard drive to the option you wish to select
(k) Start the dd. After it is complete, shutdown the computer

and remove the external drive containing your image.
3. Prepare the host computer to run the virtual computer
(a) It may be required to enable virtualization extensions

within the BIOS; this is often indicated by the error mes-
sage “VT-x is not available.” The process to enable Intel
virtualization extensions is described in Note 5.
(b) Boot the host machine and install VirtualBox which can
be obtained from http://www.virtualbox.org/ (Ubuntu
install described in Note 1).
(c) If you have USB based peripherals, you will also want to
download and install the VirtualBox extension pack from
virtualbox.org; the downloaded file is then installed from
within VirtualBox -- File->Preferences->Extensions.
(Note 1 also describes how to match the extension pack
version in Ubuntu.)
(d) Attach your external hard drive that contains the dd image
file.
(e) Convert the hard drive image file to a VirtualBox image.
This is best done from the command line. In Windows,
this process is described below (see Note 6 for the Linux
equivalent commands).
●● Press the “Start Button”, type “cmd” and press enter
to open a command line interface (Fig. 4a).
●● Execute the VBoxManage command to perform the
conversion. The concept is:
●● VBoxManage convertfromraw image.img out-drive.vdi
●● where “VBoxManage” is the name of a command
installed when you installed VirtualBox, “convert-
fromraw” is the action you want the command to per-
form “image.img” is the name of the image that you
created and out-drive.vdi is the name of the VirtualBox
compatible hard drive. All of the commands and file
names will need to reference the full path to the folder
where the commands or image files reside. It is easy to
get the full path by navigating to the directory within
the Windows file explorer and then right clicking and
selecting “Select address as text” (Fig. 4b). Below is
an example of the command as it might be typed.
Note that the VBoxManage portion had to be in
quotes because the path had a space in it. At the C:\>
prompt type:
●● “C:\Program Files\Oracle\VirtualBox\VBoxManage”
convertfromraw E:\diskcopy.img D:\HD.vdi
Fig. 4 Converting the dd image file into a VirtualBox image file. Panel (a) demonstrates how to open a com-
mand line to type commands. Press the “Start” button in the lower left, type “cmd” and press enter. Panel (b)
demonstrates how to copy the complete path for the command line from within Windows Explorer. Panel (c)
shows a sample of how the command window looks when executing the VBoxManage command under
Windows 7
●● replacing the paths with your paths; this assumes that

the removable hard drive with your disk image was
mounted as the E: drive and the path where you want
to store the VirtualBox hard drive image is just D:\
and that VirtualBox was installed in C:\Program Files\
Oracle (Fig. 4c).
●● It is recommended that the.vdi image be placed on a
physically installed hard drive, or, if it is to be on an
external USB hard drive that USB 3.0 be used.
4. Start VirtualBox and select the option to create a new virtual
machine. Select the appropriate type (e.g., “Microsoft
Windows”) and version (e.g., “Windows XP (32-bit)”). For
memory, a reasonable starting setting is to match the setting
on the machine you are virtualize—all things being equal, 2 or
3 GB is often a good starting point for XP machines. Note
that for Windows 98, a maximum 512 MB memory setting
avoids some compatibility issues. When asked for hard drive
options, select the .vdi file that you created above.
5. Before starting the new virtual computer, connect all USB

peripherals you want to control. In the VirtualBox list of vir-
tual machines, select the machine you created, click on set-
tings, click on USB. On the right hand side, click the icon of
the USB connector with the “+ “symbol. Add all the devices
that should be attached to that computer (e.g., USB based LC
Pumps). If you are using USB to RS-232 interfaces, do not
forget to connect these as well. You do not need to add mice
or keyboards as they are handled in a separate fashion by
VirtualBox.
6. Set up the network interfaces. For most machines, the default
settings (e.g., one attached interface set as “Attached to:
NAT”) is sufficient. This will allow your virtual computer to
share the host network connection in a similar fashion to con-
necting a computer to a router that is attached to your home
internet feed. In our example, however, the Thermo-Fisher
mass spectrometer uses a second network interface for control.
To make this work properly in the virtual machine, select
Network->Adapter 2 and then set it to be “Attached to:
Bridged Adapter” and then select the network interface
attached to the mass spectrometer router. Note that for
Windows 98 installations, selecting the network interface type
as “PCNet-PCI II” will usually work better than the default
option.
7. Start the Virtual Machine; if Windows hangs on a black screen,
power off the virtual machine and try enabling I/O APIC. In
the VirtualBox list of virtual machines, select the machine you
created, click on settings, I/O APIC will be found by navigat-
ing to System->Motherboard->Extended Features.
8. If Windows asks you to reactivate, use the license key that is
associated with your installation. Note 7 provides some point-
ers if Windows does require activation and your license key
does not work.
9. Install the guest additions. The guest additions allow your vir-
tual machine to more gracefully handle the graphics display as
well as sharing keyboards, mice, and files with the host com-
puter. Installation is performed on the running guest machine
by choosing the “Devices” option from the menu bar, select-
ing to insert the “Guest Additions CD” (see Fig. 5) and then
installing the software from within your virtual computer. You
may be prompted to do this automatically after you “insert”
the CD, or, you may have to navigate to the CD folder to
execute the install script.
10. Check your system application settings to make sure they are
interfacing to the appropriate network cards and RS-232
COM ports.
Fig. 5 Installing Guest Additions in Your Virtual Machine. Within the started virtual machine, select Devices-
>Insert Guest Additions CD. Then install the guest additions as prompted
11. Enjoy your easy to manage new virtual instrumentation con-

trol computer (Note 8 describes how to export your new vir-
tual machine suitable for rapid redeployment)
4 Notes
1. Ubuntu install of VirtualBox.

(a) VirtualBox exists in the software repository for Ubuntu.
While this facilitates installation, this is often not the most
current copy of VirtualBox that is available; this will usu-
ally not present a problem, but, it does mean that you will
have to look through the download archives to find the
appropriately version-matched extension pack.
(b) First, open a terminal window. Clicking on the top left
icon to open the Dash, type terminal and click on the
resulting icon. Or use the hotkey shortcut of “ctrl+alt+t”.
(c) Type “sudo apt-get install virtualbox” and enter your
password when prompted.
(d) Start up virtual box (VirtualBox), check the version,

download the extensions pack from http://www.virtual-
box.org/ that matches the version you installed from the
repository.
2. Imaging a removed hard drive on a Linux computer.
(a) Remove the hard drive from the computer to be

virtualized.
(b) Open a terminal window on your Linux machine (see

Note 1(b)).
(c) Become “root” (e.g., type “sudo /bin/bash”).
(d) Using your USB based external drive reader, attach the
hard drive to the Linux machine.
(e) Figure out which device id was assigned (type dmesg and
look for the last message, you will see something like “/
dev/sde” referenced—sda is the first hard drive, sdb the
second, and so forth. In this example, we will assume that
this drive was assigned as /dev/sde.
●● If you are booting Linux on the machine to be virtual-
ized, this is often /dev/sda.
(f) Type mount and see if any partitions from the live drive
were auto-mounted, if so, unmount them—e.g., umount
/dev/sde1 repeat for all mounted sde partitions.
(g) Connect the external USB drive that you acquired to hold
the copy of the hard drive of the live computer.
(h) Figure out which device id was assigned (type dmesg and
look for the last message, you will see something like “/
dev/sdf” referenced. In this example, we will assume that
this drive was assigned as /dev/sdf.
(i) Type mount and note the partitions from sdf (in this exam-
ple) that have been mounted and where. Often this will take
the form of /media/user/xxxx-xxxx-xxxx-xxxx/. If sdf has
not automounted its partition, you can do this by hand:
mkdir /tmp/out; mount /dev/sdf1 /tmp/out This
assumes that the first partition on the external USB drive is
the one you want to use and it was assigned as “sdf”.
(j) This is dangerous. If you are unsure of the device and path
names, you can accidentally overwrite data! Proceed with
caution. From the root command prompt type the imag-
ing command:
dd if=/dev/sde of=/tmp/out/file.dd bs=4M
conv=noerror,sync
where “/tmp/out” should be replaced by the path identi-
fied in step 2.9. The “conv” options are meant to intel-
ligently handle read errors if they were to arise; noerror
means the copy will continue and sync means to replace
error reads with the appropriate number of 0’s to match
the amount of data that was not correctly read.
3. Imaging a removed hard drive on a Windows computer.
(a) Remove the hard drive from the computer to be

virtualized
(b) Using your USB based external drive reader, attach the
hard drive to another Windows machine
(c) Connect the external USB drive that you acquired to hold
the copy of the hard drive of the live computer
(d) Download and install the HDD raw copy tool from HDD
Guru (http://hddguru.com/software/HDD-Raw-
Copy-Tool/). I use the “portable Windows executable”
that does not require installation.
(e) Run HDD Raw Copy, select the source hard drive—this
will easily be findable because it will be a “USB” device
and have the size you are expecting. For destination, select
“FILE” and navigate where you want to store the image.
For file type, be sure to select “Raw Image (dd image)”
versus compressed image.
(f) Initiate the copy.
(g) Cleanly unmount the external USB drive and the mounted
hard drive. If you are uncertain of how, you can just shut-
down your computer.
(h) Remove the USB drive containing the hard drive image
and remove the mounted hard drive. The drive with the
disk image does not need to be removed if it is on the
machine you are going to use as the host computer.
(i) If desired, reinstall the hard drive in the original physical
computer.
(j) Windows does provide a tool to convert a physical device
to a .vhd image file (“disk2vhd”), however, .vhd files are
not the best to use with VirtualBox so conversion to. vdi-
will still be required.
4. Configuring BIOS to boot from CD or USB before the hard
drive.
(a) For most instances, the example show in Fig. 1 will be suf-
ficient—press the prompted for key and then select either
the USB or CD as appropriate.
(b) For some versions of the BIOS, you will need to find the
options for the boot order and then set the CD or USB to
boot before the hard drive.
5. Configuring BIOS settings to enable virtualization extensions
(a) If you are running an AMD CPU, hardware virtualization
is usually turned on by default. Intel machines tend to be
a mix of default enabled and default disabled.
(b) To check and change settings, you need to be in the BIOS
configuration screen. Turn on the host computer and, on
the initial splash screen, select the option to activate the

“Setup”/BIOS menu. This is often the “F2”, “DEL” or
“ESC” key—look at the splash screen to see if they tell
you the appropriate key; you can also perform an Internet
search for the make and model of your computer. This can
vary considerably, for e xample, for some Lenovo laptops,
there is a little physical button that needs to be pressed
during the boot process.
(c) BIOS menus also vary considerably, you are looking to
enable “Intel VT-x” aka “Intel Virtualization Extensions“
aka “Intel Virtual Technology”.
6. Converting a dd image into a vdi file in Linux.
(a) Open a terminal window.
(b) Become “root” (e.g., type sudo /bin/bash).
(c) Type VBoxManage convertfromraw /media/user/
xx/image.dd ~user/out-drive.vdi replacing the paths
to the input and output file with the appropriate fully
qualified pathnames.
7. Windows reactivation
(a) Unfortunately, the new virtualized Windows will some-
times require reauthentication. This is complicated
because Windows XP and earlier are no longer supported
by Microsoft. Often the reauthentication can be accom-
plished by using the original license key. If, however, the
original key will not validate Windows, there is no sup-
ported reactivation option. Many people have found suc-
cess in a work-around that involves editing the Windows
registry, resetting the countdown timer and then prevent-
ing the operating system from updating the countdown
timer value. There are many sites on the internet that
describe this process [3–5].
8. Comments about managing your new virtualized computer.
(a) Some functions may work better if you install the

VirtualBox guest additions. In the running virtualized
computer select Devices->Insert Guest Additions CD
Image. The image will be downloaded and inserted into
the virtual CD drive. At this point, you will either be
prompted to install the software, or, you can select the
CD drive from within the virtualized computer.
(b) Backups and redeployment are easy. With the virtual com-
puter shut down, in the VirtualBox main console, select
“File->Export Appliance”. This exported appliance (i.e., a
single file that contains everything required by the virtual
image) is easily imported into other VirtualBox installa-
tions. The exported files can be archived and instantiated

as needed.
(c) Recovery from configuration mistakes and from viruses is
now trivial! After your machine is running, shut down the
machine then, from the VirtualBox main console, select
“Machine->Clone”. To save disk space, select a “linked
clone”. Then, always use your new clone instance of your
virtual machine. If a configuration gets corrupted, just
start up the original instance and all the changes are
undone.
Acknowledgments
I thank David B. Agus, The Lawrence J. Ellison Institute for

Transformative Medicine at USC and the USC Center for Applied
Molecular Medicine for generous institutional support. Also, I
thank Lara Bideyan for her critical review of this document and the
testing of the protocols described within.
References
1. VirtualBox download site: https://www.virtu- 4. http://www.americancomputerenterprises.

albox.org/wiki/Download com/downloads/how_to_activate_windows_
2. https://www.virtualbox.org/wiki/Migrate_ xp_witho.htm
Windows 5. http://www.technotrait.com/2012/02/27/
3. http://community.spiceworks.com/how_to/ activate-windows-xp-without-genuine-key/
show/3381-how-to-fix-windows-xp-
activation-after-a-windows-xp-repair
Chapter 22
Statistical Assessment of QC Metrics on Raw

LC-MS/MS Data
Xia Wang
Abstract
Data quality assessment is important for reproducibility of proteomics experiments and reusability of pro-
teomics data. We describe a set of statistical tools to routinely visualize and examine the quality control
(QC) metrics obtained for raw LC-MS/MS data on different instrument types and mass spectrometers.
The QC metrics used here are the identification free QuaMeter metrics. Statistical assessments introduced
include (a) principal component analysis, (b) dissimilarity measures, (c) T2-chart for quality control, and
(d) change point analysis. We demonstrate the workflow by a step-by-step assessment of a subset of Study
5 for the Clinical Proteomics Technology Assessment for Cancer (CPTAC) using our R functions.
Key words Abnormal experiment, Change point, Dissimilarity, Euclidean distance, MS/MS pro-
teomics, Principal component analysis, Quality control
1 Introduction
Shotgun proteomics experiments are subject to multiple sources of

variations, from the sample preparation, to the internal spectrom-
eter operation and external environmental factors, to the data
acquisition methods. Large variations inhibit the reproducibility of
the experiments across labs and across time [1, 2]. Timely quality
assessment is thus important in lab practice. Also, with massive
data available in the field of proteomics, the data quality is an
important problem to be addressed, especially for decision making
by combing multiple data sets from different labs, such as the pub-
lished data. When multiple labs have investigated the same condi-
tion, many challenges may hinder meta-analysis. The data must be
scrutinized to determine whether they have been consistently col-
lected and if outliers or batch effects exist. Both evaluation of
experiment reproducibility and examination of data reusability
help improve the effectiveness and accuracy of biological and clini-
cal decisions based on the experimental proteomics data.
325
326 Xia Wang
This note describes a toolkit of multivariate statistical proce-

dures and its R implementation we developed for quality assessment
on raw data from a wide range of shotgun proteomics experiments.
The quality assessment is based on the identification free, multidi-
mensional quality metrics produced by QuaMeter [3]. We empha-
size that the multidimensional features of a shotgun proteomics
experiment should be kept integrated in its quality assessment and
thus employ multivariate statistical procedures in all the assess-
ments. The objective is to produce visualization of the multidi-
mensional quality metrics, to detect potential abnormal
experimental runs, to test for “out-of-control” experiments, and to
examine batch effects.
The major statistical procedures include (a) principal compo-
nent analysis (PCA). This facilitates visualization of multiple qual-
ity control (QC) metrics and generates the principal component
(PC) scores for further quality assessment. (b) Dissimilarity mea-
sures. This measure is based on the Euclidean distance between the
robust PCA coordinates for each LC-MS/MS experiment. It helps
to detect potential abnormal experimental runs. (c) T2-chart for
quality control. This QC method provides a statistical test to detect
those “out-of-control” experiments for further investigation with
a cutoff significance level. The difference and connection between
procedures (b) and (c) are discussed in Subheading 3. (d) Change
point analysis. Change points are caused by sustained changes in
the experimental conditions and thus are particularly importance
in investigating longitudinal studies and examining batch effects.
We include a step-by-step demonstration of the above analysis in R
functions using CPTAC Study 5 data (see Notes 1 and 3).
2 Materials
Identification-free QC metrics on raw LC-MS/MS data are gener-

ated by QuaMeter. QuaMeter software [3] is a tool built on the
ProteoWizard library [4] to produce quality metrics from LC-MS/
MS data. It has a special “IDFree” mode to produce metrics that
are independent of identification success rates for MS/MS scans
(see Note 2). As only the raw data need to be accessed, the process
time of QuaMeter is very short. It typically takes less than 5 min
per file for Thermo Orbitrap Velos raw data files, with a large por-
tion of time consumed in the extraction of ion chromatograms [5].
Thus, QuaMeter is well-positioned to give timely feedback in sup-
port of go/no-go decisions during experiments.
For each data file obtained from an LC-MS/MS experiment,
QuaMeter IDFree generates 44 numerical measures in four cate-
gories: XIC (extracted ion chromatograms, 7 metrics), RT (reten-
tion times, 13 metrics), MS1 (mass spectrometry, 11 metrics), and
MS2 (tandem mass spectrometry, 13 metrics). It aims to provide a
Statistical Assessment of QC Metrics 327
full evaluation of multidimensional aspects of the experiment.

Some of the 44 numerical metrics need to be omitted in the mul-
tivariate analysis due to low variance in the metric itself or high
linear correlation with other metrics (see Note 3 initial data exami-
nation). Besides these numerical measures, QuaMeter also records
the name of the raw LC-MS/MS data file and the start time of the
experiment, which are very useful for data file identification and for
examining the temporal evolution of the experiments. Table 1 con-
tains a full list of all 46 metrics along with their computations.
Table 1
46 metrics produced in the “IDFree” mode of QuaMeter
Filename What is the name of the file from which the metrics were computed?
StartTimeStamp At what time did acquisition begin for this experiment?
XIC-WideFrac What fraction of precursor ions account for the top half of all peak width?
XIC-FWHM-Q1 What is the 25%ile of peak widths for the wide XICs?
XIC-Height-Q2 The log ratio for 50%ile of wide XIC heights over 25%ile of heights.
XIC-Height-Q3 The log ratio for 75%ile of wide XIC heights over 50%ile of heights.
XIC-Height-Q4 The log ratio for maximum of wide XIC heights over 75%ile of heights.
RT-Duration What is the highest scan time observed minus the lowest scan time
observed?
RT-TIC-Q1 The interval when the first 25 % of TIC accumulates divided by
RT-Duration.
RT-TIC-Q2 The interval when the second 25 % of TIC accumulates divided by
RT-Duration.
RT-TIC-Q3 The interval when the third 25 % of TIC accumulates divided by
RT-Duration.
RT-TIC-Q4 The interval when the fourth 25 % of TIC accumulates divided by
RT-Duration.
RT-MS-Q1 The interval for the first 25 % of all MS events divided by RT-Duration.
RT-MS-Q2 The interval for the second 25 % of all MS events divided by RT-Duration.
RT-MS-Q3 The interval for the third 25 % of all MS events divided by RT-Duration.
RT-MS-Q4 The interval for the fourth 25 % of all MS events divided by RT-Duration.
RT-MSMS-Q1 The interval for the first 25 % of all MS/MS events divided by
RT-Duration.
RT-MSMS-Q2 The interval for the second 25 % of all MS/MS events divided by
RT-Duration.
(continued)
328 Xia Wang
Table 1
(continued)
RT-MSMS-Q3 The interval for the third 25 % of all MS/MS events divided by
RT-Duration.
RT-MSMS-Q4 The interval for the fourth 25 % of all MS/MS events divided by
RT-Duration.
MS1-TIC-Change-Q2 The log ratio for 50%ile of TIC changes over 25%ile of TIC changes.
MS1-TIC-Change-Q3 The log ratio for 75%ile of TIC changes over 50%ile of TIC changes.
MS1-TIC-Change-Q4 The log ratio for largest TIC change over 75%ile of TIC changes.
MS1-TIC-Q2 The log ratio for 50%ile of TIC over 25%ile of TIC.
MS1-TIC-Q3 The log ratio for 75%ile of TIC over 50%ile of TIC.
MS1-TIC-Q4 The log ratio for largest TIC over 75%ile TIC.
MS1-Count How many MS scans were collected?
MS1-Freq-Max What was the fastest frequency for MS collection in any minute? (Hz)
MS1-Density-Q1 What was the 25%ile of MS scan peak counts?
MS2-Count How many MS/MS scans were collected?
MS2-Freq-Max What was the fastest frequency for MS/MS collection in any minute? (Hz)
MS2-Density-Q1 What was the 25%ile of MS/MS scan peak counts?
MS2-PrecZ-1 What fraction of MS/MS precursors is singly charged?
MS2-PrecZ-2 What fraction of MS/MS precursors is doubly charged?
MS2-PrecZ-3 What fraction of MS/MS precursors is triply charged?
MS2-PrecZ-4 What fraction of MS/MS precursors is quadruply charged?
MS2-PrecZ-5 What fraction of MS/MS precursors is quintuply charged?
MS2-PrecZ-more What fraction of MS/MS precursors is charged higher than +5?
MS2-PrecZ-likely-1 What fraction of MS/MS precursors lack known charge but look like +1 s?
MS2-PrecZ-likely-multi What fraction of MS/MS precursors lack known charge but look like >+1 s?
3 Methods
All the statistical procedures are done using R software [6].
3.1 Initial Data A timeline plot of the data files, grouped by participating laborato-
Examination ries, individual instruments, sample types, or other features, helps
visualize the data files in the order that they were produced in the
study. The metric “StartTimeStamp” is used to extract starting
time of the experiment for a raw data file.
Many of the QuaMeter metrics are based on quartiles of a
quantity. For example, XIC-FWHM-Q1, XIC-FWHM-Q2, and
XIC-FWHM-Q3 are the 25%ile, 50%ile, and 75%ile of peak widths
for the wide XICs. This helps to provide a rough approximation of
the distribution of the peak widths for the wide XICs. However, it
also brings in potential high correlation among these measures. In
practice, we usually remove any metric that shows correlation
greater than 0.98 or less than −0.98 with the other metric(s). The
analysis loses little information from this removal because one can
predict the values for these removed variables almost perfectly
from the retained information. Also, metrics that are related to
directly configurable options, such as MS2-PrecZ-* and
RT-Duration, are also omitted in the statistical analysis (see Note 3
initial data examination).
3.2 Principal Principal component analysis (PCA) helps to visualize the multidi-
Component Analysis mensional quality control (QC) metrics in lower dimension. It
makes the interpretation easier by transform correlated metrics to
uncorrelated linear combinations of the metrics. To reduce impact
by any extreme data files, we use robust PCA method [7]: the vari-
ance–covariance matrix of the QC metrics as well as the center of
each metrics are estimated by cov.rob function in R and are used in
the PCA analysis (cov.rob function in R). The PCA analysis is car-
ried out by princomp R function using the QC metrics normalized
by the individual metric’s variance. The results contain PC scores
matrix (linear combinations of original metrics), whose columns
corresponding to the dimension of the QC metrics that are ana-
lyzed and whose rows are the newly transformed QC metrics for
each data file (see Note 3 principal component analysis for the
example). The data structure can then be visually examined by sys-
tematically exploring two- or three-dimensional plots of different
combinations of PC scores [8]. The one used most frequently is
the plot based on the first two PCs, which account for the largest
proportion of variance in the data. A more complete examination
is needed for a comprehensive view of the data structure.
3.3 Dissimilarity This step is to exploratorily examine how similar two experiments
Measures are in terms of the QC metrics. If a spectrometer has been operat-
ing stably, it is expected that QC metrics obtained from all the
experiments on this spectrometer should be randomly distributed
around a “true” stable status, and thus the distance between each
other is approximately equal. If an experiment is “far away” from
the rest of the experiments based on the QC metrics, it reflects the
overall difference in the QC metrics and thus implies this experi-
ment may have been operated under quite different conditions
330 Xia Wang
compared to others. We call an experiment that is “far away” from

the other experiments “an abnormal experimental run.”
To detect if an experiment is “far away,” we design the dissimi-
larity measure. Suppose the p-dimensional PC scores for the two
experiments are x1 and x2, the dissimilarity between these two
p-dimensional coordinates is defined as
+ ( x12 - x 22 ) +  ( x1 p - x 2 p ) .
2
( x11 - x 21 )
2 2
Thus, this measure is based on the Euclidean distance between

the robust PCA coordinates for each LC-MS/MS experiment. The
larger the dissimilarity values, the lower the similarity between the
two experimental data files. Compared to T2-chart in
Subheading 3.4, this measure has a wider application as it does not
require that most of experiments are similar to each other. Actually,
since these is no benchmark profile required in comparison, this
distance measure is designed to be automatically outlier-proof and
can be used in almost any situation. An abnormal experiments can
be easily identified by its large distance (higher dissimilarity) from
a few other experiments.
3.4 T2-Chart For experiment i with the p-dimensional PC scores xi, a T2 statistic
for Quality Control is computed as
Ti 2 = ( xi 1 - x1 ) / l1 + ( xi 2 - x 2 ) / l2 +  + ( xip - x p ) / lp ,
2 2 2

where the p-dimensional vector x is the median of the p QC met-
rics and λj is variance of the jth PC component. The statistic Ti2 is
a normalized Euclidean distance. Since xi’s are PC scores and the
variance–covariance matrix is a diagonal matrix with λj, j = 1, …, p ,
as the diagonal element, it is assumed that Ti2 approximately fol-
lows a χ2-distribution with p degrees of freedom [9]. The upper
and lower control limits can be obtained at a given significance
level α as cutoff between in-control and out-of-control experi-
ments. While the comparison based on the dissimilarity measure is
more exploratory, this statistic provides a more rigorous evaluation
of the abnormal experiments with a cutoff significance level. It may
be impacted, however, by data files from extreme experiments.
3.5 Change Point Change point analysis is a tool to study the potential batch effects
Analysis in experiments. The multidimensional QC metrics is summarized
in the normalized p-dimensional Euclidean distance (Ti2). The
batch effects may be caused by the changes in the mean of Ti2, or
in the variability of Ti2, or both. With a smaller number of experi-
mental runs on each spectrometer, we only assume changes in the
mean. The R function cpt.mean in the R library changepoint is
used in studying the batch effects in the mean level [10]. When
variance or mean and variance changes are suspected, cpt.var or

cpt.meanvar functions can be used.
4 Notes
1. Overview of CPTAC Study 5. Study 5 for the Clinical

Proteomics Technology Assessment for Cancer (CPTAC) had
multiple-site participants with standard operating procedures
(SOPs) implementation [11]. It focused on Thermo LTQ and
Orbitrap instruments. The six spectrometers included are
LTQ73, LTQ295, LTQc65, Orbi86, OrbiP65, and OrbiW56,
where the first three to five letters and number denote the
instrument type and the last two numbers are the anonymous
code for the laboratory that participated. The samples included
the digested NCI-20 (labeled “1B”), the yeast reference mate-
rial (“3A”), and the yeast reference material with bovine serum
albumin spiked at 10 fmol/μL in 60 ng/μL yeast lysate (“3B”).
The run order in Study 5 repeated the following block twice
1B, 3A, 3A, 3A, 1B, 3B, 3B, 3B, followed by an additional
1B. This leads to a total of 103 experimental runs, including
30 on Sample 1B (five runs for each spectrometer). Here we
only study the data collected from experiments on Sample
1B. Table 1 in the Supporting Information of the CPTAC
Repeatability article [6] lists the commercial sources of the
proteins in the NCI-20 mixture and the detailed description
on the creation of NCI-20 reference materials can be found in
[4]. Data were collected on three LTQ and three Orbitrap
mass spectrometers between October of 2007 and January of
2008. Raw data files for Study 5 can be found at the CPTAC
Public Portal: https://cptac-data-portal.georgetown.edu/
cptacPublic/. QuaMeter metrics for Study 5 data and the R
functions are available on the author’s website (http://homep-
ages.uc.edu/~wang2x7/).
2. QuaMeter IDFree execution. The QuaMeter software was
executed with this command line:
quameter.exe *.raw -MetricsType idfree
-OutputFilepath metrics.tsv
The quameter.cfg files used for each type of instrument
included the following:
Orbitrap configuration:
ChromatogramMzLowerOffset = "0.01mz"
ChromatogramMzUpperOffset = "0.01mz"
Instrument = "orbi"
Ion Trap configuration:
332 Xia Wang
Instrument = "LTQ"
QqTOF configuration:
Instrument = "orbi"
#Here, “Orbi” means
“can resolve isotopes.”
3. Demonstration using CPTAC Study 5.
Read in the QuaMeter data. There were a total six .tsv files for
QuaMeter metrics with the 46 metrics. The sample type, the
spectrometer instrument type, and a label for the spectrometer
number were added. Thus the “QuaMeter.csv” file had a total
of 50 columns with the first column as the index number. In
this demonstration, we only examined 30 runs from Sample
1B, 5 runs for each spectrometer.
Initial data examination. Figure 1 shows the timeline plot for
all the data files in CPTAC Study 5. We first removed
RT-Duration and MS2.PrecZ.* as they are operating condi-
tions set by the operator. We then used cor R function to detect
and removed metrics with high correlations with other met-
rics. At the end, 27 numeric metrics, excluding Filename and
StartTimeStamp, were used in the following statistical analysis.
These 27 numeric metrics were:
XIC (5): XIC.WideFrac, XIC.FWHM.Q2, XIC.Height.Q2,
XIC.Height.Q3, XIC.Height.Q4
RT (11): RT.TIC.Q1, RT.TIC.Q2, RT.TIC.Q3, RT.TIC.Q4,
RT.MS.Q1, RT.MS.Q2, RT.MS.Q3, RT.MS.Q4, RT.
MSMS.Q1, RT.MSMS.Q2, RT.MSMS.Q3
MS1 (9): MS1.TIC.Change.Q2, MS1.TIC.Change.Q3, MS1.
TIC.Change.Q4, MS1.TIC.Q2, MS1.TIC.Q3, MS1.TIC.
Q4, MS1.Count, MS1.Freq.Max, MS1.Density.Q1, MS1.
Density.Q2, MS1.Density.Q3
MS2 (2): MS2.Freq.Max, MS2.Density.Q1
Principal component analysis. The R object metrics.pca con-
tained the 27 metrics for the 30 experiment runs. The follow-
ing R codes carried out robust PCA analysis.
metrics.pca<-metrics.1[,5:(ncol(metrics.1)-3)]
robust.cov<-cov.rob(metrics.pca)
robust.cor<-cov2cor(robust.cov$cov)
robust.cor.1<-robust.cov
robust.cor.1$cov<-robust.cor
metrics.pca.1<-metrics.pca
for ( i in 1:ncol(metrics.pca)){
CPTAC Study 5: Experimental Time and Sample Types for Participants
1B
3A
OrbiW56
3B
OrbiP65
Orbi86
LTQc65
LTQ295
LTQ73
Oct 13 2007 Nov 2007 Dec 2007 Jan 14 2008
StartTimeStamp
Fig. 1 Timeline for experiments in CPTAC Study 5. It illustrates the run times for each LC-MS/MS experiment
in the course of Study 5 across six spectrometers. The run order was prescribed by the SOPs under which
Study 5 was conducted. Sample 1B: digested NCI-20 mixture; Samples 3A: yeast; 3B: yeast + BSA
robust.cor.1$center[i]<-robust.cov$center[i]/sqrt(robust.
cov$cov[i,i])
metrics.pca.1[,i]<-metrics.pca[,i]/sqrt(robust.cov$cov[i,i])
}
print(summary(pc.cr <- princomp(metrics.pca.1, covmat =
robust.cor.1, scores=T)))
Figure 2 shows the PC plot for the first two principal compo-
nents. The experimental runs from the same spectrometer
tended to cluster together, and LTQ and Orbi instruments
tended to cluster together, along with some potential abnormal
experimental runs, such as the one in LTQ73.
Dissimilarity measures. For pairwise comparison, we used run1
to represent the p-dimensional PC coordinate for one experi-
ment and run2 to represent that of the other experiment. The
dissimilarity measures were then calculated in R as distance
<-sqrt(sum((run1-run2)^2)). Figure 3 shows all the values of
dissimilarity measure grouped by spectrometers. Clearly, there
was a cluster of large distances on LTQ73, which were all
334 Xia Wang
The Second Principal Component

2
LTQ73
0
LTQ295
LTQc65
Orbi86
OrbiP65
−2
OrbiW56
−4
−6
−6 −4 −2 0 2 4
The First Principal Component
Fig. 2 PCA plots based on the first two principal components of quality metrics
for experiments on Sample 1B on the six spectrometers
6
Euclidean distance
LTQ295 LTQ73 LTQc65 Orbi86 OrbiP65 OrbiW56
Fig. 3 Dissimilarity measures for experiments on Sample 1B on the six

spectrometers
istances of the first run compared to the other runs on LTQ73.

d
This implied that the first experiment on LTQ73 was poten-
tially abnormal/different compared to the other runs and
needed further investigation. When there is a large number of
experiments to compare, the number of dissimilarity measures
increases fast (in the order of n2). A boxplot of the dissimilarity
measures is a helpful visualization tool.
T2-chart for quality control. For each spectrometer, the T2 sta-
tistics for each data file was calculated using the following R
codes:
temp.dist$T2[j] <- t(temp)%*%solve(pca.var)%*%temp,
where temp.dist was the R object containing the PC scores on
experiment runs for the given spectrometer, temp was the met-
rics vector for the experiment centralized by the median vector
and pca.var was a diagonal matrix with the diagonal elements
equal to the variance of each PC component. The median vec-
tor of a given spectrometer was calculated by R function l1me-
dian and saved as L1median. The T2’s were assumed following
a χ2-distribution with 27 degrees of freedom. We used an over-
all familywise error rate a = 0.01 and used Bonferroni correc-
tion for the multiple testing on each spectrometer (five files per
spectrometer).
Figure 4 shows the T2 statistics along the experiment starting
time, grouped by the six spectrometers. The blue dots repre-
sent normal experiment runs, while the pink dot represents the
“out-of-control” one detected by the χ2 test and the “*” sym-
bol represents the experiment with large dissimilarity measures.
As shown in Fig. 4, the first experiment in LTQ73 (labeled as
“*”) was an experiment quite different from the rest of experi-
ments on LTQ73 and the last experiment in OrbiW56 (pink
dot) was detected as an out-of-control experiment. Further
investigation was needed to identify the cause of the
abnormality.
Change point analysis. We first removed any isolated out-of-
control experiments detected by the T2 statistic because change
point analysis studies sustained changes. Change point detec-
tion was then carried out upon the remaining experiment runs.
In Fig. 4, the last experiment in OrbiW56 was removed from
change point analysis as it is an isolated out-of-control experi-
ment (black circled). Except LTQc65, the other five spectrom-
eters all had one change point, which divided the experiments
on Sample 1B into two batches (first batch: red squares; sec-
ond batch: green diamonds). Particularly, it turned out that
the first run in LTQ73 was different from the rest of the runs
and was detected as a change point location.
336 Xia Wang
1 2 3 4 5
LTQc65 OrbiW56
40
200 400 600 800 1000

30
20
10
0
0
LTQ73 OrbiP65
40
*
30
30
20
T2
20
10
10
0
LTQ295 Orbi86
40
30
30
20
20
10
10
0
1 2 3 4 5
date
Fig. 4 T2 statistic and change point analysis results for the six spectrometers. The dots represent the values of
T2 statistic with blue dots for “in-control” experiments, pink dots for “out-of-control” experiments, and “*” for
those with large dissimilarity measures. If a pink dot is circled by a black circle, then it is not included in the
change point study (as the one for OrbiW56). The red squares and green diamonds distinguish the two batches
of the experiments runs, if there is a change point
References
1. Bell AW, Deutsch EW, Au CE et al (2009) A 4. Chambers MC, Maclean B, Burke R et al (2012)
HUPO test sample study reveals common A cross-platform toolkit for mass spectrometry
problems in mass spectrometry-based pro- and proteomics. Nat Biotechnol 30:918–920
teomics. Nat Methods 6:423–430 5. Wang X, Chambers MC, Vega-Montoto LJ
2. Mann M (2009) Comparative analysis to guide et al (2014) QC metrics from CPTA raw
quality improvements in proteomics. Nat LC-MS/MS data interpreted through multi-
Methods 6:717–719 variate statistics. Anal Chem 86(5):2497–2509
3. Ma ZQ, Polzin KO, Dasari S (2012) QuaMeter: 6. R Core Team (2015) R: a language and envi-
multivendor performance metrics for LC–MS/ ronment for statistical computing. R
MS proteomics instrumentation. Anal Chem Foundation for Statistical Computing, Vienna,
84(14):5845–5850 Austria. http://www.R-project.org/
7. Rousseeuw PJ, van Zomeren BC (1990) 10. Killick R, Eckley IA (2014) changepoint: an R
Unmasking multivariate outliers and leverage package for changepoint analysis. J Stat Softw
points. J Am Stat Assoc 85:633–639 58(3):1–19
8. Ringnér M (2008) What is principal compo- 11. Tabb DL, Vega-Montoto L, Rudnick PA et al
nent analysis? Nat Biotechnol 26(3):303–304 (2010) Repeatability and reproducibility in
9. Johnson RA, Wichern DW (2007) Applied proteomic identifications by liquid
multivariate statistical analysis. Pearson Prentice chromatography- tandem mass spectrometry.
Hall, Upper Saddle River J Proteome Res 9:761–776
Chapter 23
Data Conversion with ProteoWizard msConvert

Ravali Adusumilli and Parag Mallick
Abstract
Recent advances in proteome informatics have led to an explosion in tools to analyze mass spectrometry
data. These tools operate across the analysis pipeline doing everything from assessing quality control to
matching peptides to spectra to quantification. Unfortunately, the vast majority of these tools are not able
to operate directly on the proprietary formats generated by the diverse mass spectrometers. Consequently,
the first step in many protocols is the conversion of data from vendor-specific binary files to open-format
files. This protocol details the use of ProteoWizard’s msConvert and msConvertGUI software for this
conversion, taking format features, coding options, and vendor particularities into account. We specifically
describe the various options available when doing conversions and the implications of each option.
Key words Proteomics, Proteome informatics, Data conversion, Open formats, mzML, mzXML,
ProteoWizard
1 Introduction
Mass-spectrometry-based proteomics has become an important

component of biological research. Numerous proteomics methods
have been developed to identify and quantify the proteins in bio-
logical and clinical samples, identify pathways affected by endoge-
nous and exogenous perturbations, and characterize protein
complexes. However, before it is possible to explore any of these
important biological questions, a battery of computational tools is
required to convert mass spectrometry data into knowledge. Each
vendor has developed its own ecosystem of tools, such as
ProteomeDiscoverer (Thermo) and Analyst (SCIEX). to help
interpret the complex data produced by their instruments.
However, the proteome informatics research community has
developed thousands of alternate tools with a wide range of
capabilities. Unfortunately, the vast majority of these tools are not
able to directly read vendors’ binary formats and instead are depen-
dent upon “open” formats such as mzML and mzXML. Translating
from closed binary formats to open formats requires specialized
339
340 Ravali Adusumilli and Parag Mallick
software. Initially a set of tools that specifically converted data from

one vendor format into an open format (e.g., ReAdW) emerged.
More recently, the ProteoWizard toolkit [1, 2] has provided a
swiss-army knife tool in msConvert that is able to convert all popu-
lar vendor formats into diverse open formats.
Beyond performing a direct translation of file contents, from a
closed binary format to an openly readable format, conversion is
also a data processing step. In this step, researchers may perform
lossy data compressions (such as centroiding) or fundamentally
alter the values in files through recalibration (such as in mzRefin-
ery). Here we describe a variety of protocols for data conversion
and particularly detail all the diverse options available to users dur-
ing conversion.
The msConvert tool described below is available as user-
friendly graphical user interface (GUI) program (msConvertGUI)
for Windows and as cross-platform (Windows, Mac, Linux) com-
mand line tool (msConvert). We note that there are more options
available in the command line tool than in the GUI. MS Convert
offers the flexibility of processing multiple input file formats with
options for several filters and transformations. The details of the
filters and transformations available for msConvertGUI and
msconvert command line will be further discussed below. In addi-
tion to converting from vendor formats to open formats, msCon-
vert and msConvertGUI can also convert among open formats,
such as from mzXML to mzML. Conversion from vendor formats
is only achievable on Windows. However, conversion among open
formats can be performed on any platform.
2 Materials
2.1 Installation 1. Make sure you have installed Microsoft .NET Framework 3.5
of ProteoWizard SP1 and 4.0.
(Includes Both (a) MS .NET Framework 3.5 SP1 can be found at: http://
msConvert, www.microsoft.com/en-us/download/details.aspx?id=22
msConvertGUI and (b) MS .NET Framework 4 can be found at: http://www.
msPicture): microsoft.com/en-us/download/details.aspx?id=17851
2. Download the latest version of proteoWizard from: http://
proteowizard.sourceforge.net/
(a) We typically recommend the Windows 64-bit installer for
most users.
(b) Upon download, double-click the .msi file to install.
(c) proteoWizard will typically be installed into a folder ‘C:\
Program Files\ProteoWizard\ProteoWizard 3.0.DDDD’
PWIZ Conversion 341
3. (For Command Line Use) Add “proteoWizard” folder to the

path variable:
(a)
Go to control panel > System and secu-
rity > System > Advanced settings > Environment Variables
(b) You should see a table/list of variables and values. Double
click on “Path”.
(c) You should see a bunch of values separated by “;” in the
Variable value field.
(d) Add your proteowizard location to the list separating with
a semicolon.
For example: “C:\Windows; C:\Program Files\ProteoWizard
3.0.8708\”
(e) Restart Windows.
3 Methods
3.1 Using 1. Navigate to the proteowizard directory (Fig. 1) (For ex: C:\
msConvertGUI Program Files\ProteoWizard\ProteoWizard 3.0.8708)
to Convert the File 2. Start the msConvertGUI application
Format of Data Files Double-click on “msConvertGUI” application to open. This
should open the msconvert GUI application. Visually,
msconvertGUI is divided into four boxes (Fig. 2), which are
described in greater detail below.
(a)
Select input files and output files directory location
(Top left).
Fig. 1 Open from the folder

Fig. 2 MSConvertGUI
(b) Bottom left: Options—to select the output format, extension,

binary encoding precision, write index, use zlib compression,
TPP compatibility, package in gzip, numpress compression
types (linear, short logged float, short positive integer).
(c) Top Right: Filters—Type of filters and filtering levels/
options to add or remove.
(d) Bottom Right: Box to view the selected filters in step c.
3. Choosing files to convert (top left, Figs. 3 and 4):
Type: radio button
Default: List of Files
Select the file import type (list of files or file of filenames)—
further explained below.
(a) List of Files:
Type: Button (“Browse”)
Additional types: Button (“Add” and “Remove”)
PWIZ Conversion 343
Fig. 3 Click on Browse and select input files from MSConvertGUI
i. Select the “List of Files” option on top-left of the

msconvert GUI.
ii. Click on the “Browse” button-this should open a
desktop browser window to select the input files.
iii. Go to the directory with the input formatted (*.RAW
files) from the browser window in step ii.
iv. Select one input file or multiple files by holding down
the shift/control button on the keyboard.
v. Click “open”—the selected files should now be visi-
ble in the msconvert GUI (in the box right below the
“List of Files” and “File of filenames” options.
vi. To remove some files after adding them to the GUI,
select the file/files to remove in the box from step v.
Click “Remove” right above the box with the list of
files.
(b) File of file names:
Additional types: Button (“Add” and “Remove”)
Input files to be converted can be imported from a text file
with input file names and locations.
Fig. 4 MSConvertGUI list of selected files
i. Select the “File of File names” option on top-left of

the msconvert GUI.
ii. Click on the “Browse” button-this should open a
desktop browser window to select the input files.
iii. Go to the folder containing the text file with “input
filename and paths” to be converted.
iv.
Select the text file/files with “input filename and
paths” to be converted.
v. Click “open”.
The files to be imported (selected in either step a or b) can
be viewed in step 2.
PWIZ Conversion 345
4. Select an Output Directory (top left):

Default: Folder path with the input files
(a) The default output location is set to the directory with the
input formatted files selected. If input files are selected
across multiple directories, the directory containing the
first selected input file is selected.
(b) To choose an output location other than the default, click
on “Browse”. This should open a browser window.
(c) Select an already existing folder from the desktop or create
a new folder to save the output files.
(d) To create new output folder, click on “Make New Folder”—
this will create a new folder. Type in the output name for
the folder (For Example: “Project1_output”)
(e) Click “OK” to select the folder or “Cancel” to cancel the
folder selection.
5. Select Conversion Options (Fig. 5):
(a) Output format:
Type: Drop down list
Default: mzXML
Select the format of the converted output file. The follow-
ing output formats can be selected:
i. mzXML (default)
ii. mz5
iii. mgf (Mascot generic format)
iv. text (Proteowizard internal text format)
v. ms1
vi. cms1
vii. ms2
viii. cms2
(b) Extension:
Type: Text box
Default: Blank
To add an extension to the output formatted file other
than the option chosen above. For example: mzXML can be
saved as .fred.mzXML if needed
(c) Binary encoding precision:
Type: radio button
Default: 32-bit
Fig. 5 MSConvertGUI output format options
Select the output file encoding type. The output file can be
encoded as 32-bit (default—smaller file size) or 64-bit.
(d) Write index:
Type: Checkbox
Default: Checked
Check this box to write index to the output file (see Note 1).
(e) Use zlib compression:
Type: Checkbox
Default: Checked
Check this box to “zlib” compress the output file (see Note 2).
(f) TPP compatibility:
Type: Checkbox
PWIZ Conversion 347
Default: Checked
Check this box to make the output file “TPP compatible.”
(g) Package in gzip:
Type: Checkbox
Default: NOT Checked
Check this box to “gzip” the output file (see Note 3).
6. Select Filter Options (Figs. 5, 6 and 7) (see Note 4)
Filters:
Type: Dropdown list
Default: MS Level
Fig. 6 MSConvertGUI filtering options

Fig. 7 MSConvertGUI list of selected filters
Select the filter to apply to the output file. The following filters
can be selected:
(a) MS Level (Default)
(b) Peak Picking (see Note 5)
( c) Zero Samples
(d) ETD Peak Filter
(e) Threshold Peak Filter
( f) Charge State Predictor
(g) Activation
(h) Subset
Click “Add” to add the settings selected from step a and/
or b. Alternately, remove the settings added by clicking
“Remove”. View the selected filters and parameters in the box.
PWIZ Conversion 349
7. Save Settings. Click “Use these settings next time I start

MSConvert” (bottom left) to set the settings from steps 1 to
10 as defaults so they do not need to be set the next time you
run the program.
8. Double-check all the settings. Once all options have been filled
click “Start.” A new window will appear with a list of files to be
converted and current progress. Details on the conversion of the
currently selected file will be shown in the text box at the bot-
tom of the new menu. Once all files have finished converting it
is safe to close the progress window and use the resulting files
3.2 Using 1. Make sure the latest version of proteowizard is downloaded

the Command Line and setup as instructed above.
Version of msConvert 2. Go to command prompt/windows power shell (for windows 8+).
to Convert the File If the path variable is set under “SETUP PROTEOWIZARD”
Format of Data Files instructions, all the proteowizard tools can be called without
typing the absolute path. For example: msconvert.exe can be
called directly if path variable is set. If not, C:\Program Files\
ProteoWizard 3.0.8708\msconvert.exe will be required to call
the tool.
For this protocol, the assumption is that path has been set
and so all the tools will be called directly without the absolute
path.
3. To view all options under the msConvert commandline, type
msconvert.exe (or C:\Program Files\ProteoWizard 3.0.8708\
msconvert.exe) and hit ENTER/return.
(a) This should show all the options available under msCon-
vert commandline (screenshot attached).
(b) All Command line usage terms for msconvert can be sum-
marized into four sections:
1. Usage: Generic command format to run the msconvert tool
2. Options: List of all available options to run msconvert (
●● For example: -f for “text file with input filenames” (like
File of file names in msconvertGUI)
1. -o for “output directory ”
2. -c for configuration file
3. FILTER OPTIONS: This section lists all the options to filter
the data in the output file
●● For example: index (similar to write index on the GUI);
1. msLevel to select the level of ms data (same as MS
Levels in the GUI)
4. Examples: This section list few examples with
explanations to run msconvert commandline options.
Fig. 8 msconvert—default RAW (to mzML) file conversion from commandline
Fig. 9 msconvert—RAW to mzXML file conversion from commandline
4. Execute msConvert (Figs. 8, 9 and 10)

The simplest way to run msConvert is as:
msconvert C:\Users\Administrator\Desktop\pwizard_pub\
inputFile01.raw
The above is the most basic command to run a single file
on one input *.RAW file. The “inputFile01.RAW” in the above
command is converted to “inputFile01.mzML” formatted file
by default. The “inputFile01.mzML” file will be located in
the current working directory (i.e., the directory from which
the command is executed).
However, msconvert provides the user with flexible options
to choose the output conversion format, set the output direc-
tory location, set file format extensions, provide configuration
file choice to set all the conversion properties together. These
options are elaborated in Note 6.
PWIZ Conversion 351
Fig. 10 msconvert commandline using -f (or --filelist) to convert multiple files together
5. Execute msConvert with variations in parameters and filters.

(a) For each filter to be included (see Note 7), add text struc-
tured like the following: --filter “peakPicking true 1-”.
The first piece signals that a new data filter is being added.
The section in double quotes supplies the name of the fil-
ter (see below) and the configuration options for it.
(b) Specify the output format options, such as specifying the
overall format by one of the following flags: “--mzML”,
“--mzXML”, “--mz5”, “--mgf”. If subsequent software
allows it, specify “zip” encoding for data within the files
through use of the “-z” option. The full list of these
options can be found by running the msConvert execut-
able without any parameters.
(c) Specify which files are to be converted. Both absolute and
relative paths are permitted, and wildcard characters such as
‘*’ and ‘?’ can be used to process multiple files in a single pass.
(d) Combine these elements into a single command line, such
as the following:
msconvert.exe --filter "peakPicking
true 1-" --mz5 -z *.raw
4 Notes
1. A file index allows for more rapid scan level access of the file.
The index itself makes files slightly larger. However, instead of
having to read a file from beginning to end, it can now be read
from random positions.
2. zlib compression is a seemingly invisible change to the file.
Essentially, the ms spectra themselves are zlib compressed prior
to writing the file. This greatly compresses the file, but also can
slow down its reading. Furthermore, some readers are unable
to handle zlib compressed files.
3. Unlike zlib compression, gzip actually fully gzips the entire
file. This shrink files by about ~20–30 %. Once gzipped, the
files are no longer directly human readable without being ung-
zipped first. ProteWizard and some other tools are able to read
files, even when gzipped. However, they are no longer able to
provide random access. The vast majority of open source tools
are not able to read gzipped files directly.
4. Filters are operations that are performed on spectra to alter
them prior to writing out a converted file. While some filters
are simple (e.g., only writing out MS/MS data), others can be
quite complex and substantially alter your data. The com-
mandLine msConvert has access to many more filters than are
available in the GUI. Please note, that as filters can fundamen-
tally change your data, they may substantially impact down-
stream operations, such as quantification.
5. Peak Picking, is the msConvert nomenclature for “centroid-
ing” which is a very common procedure for shrinking file sizes.
However, this is an irreversible operation that discards a sub-
stantial amount of data. It should be used with caution.
6. msConvert has a large number of available options (Fig. 8):
All the available options are listed on running the command
“msconvert” from the command prompt.
Parameter Flag Explanation

-f [--filelist] arg Specify text file containing filenames
-o [--outdir] arg (=.) Set output directory (‘-’ for stdout) [.]
-c [--config] arg Configuration file (optionName = value)
--outfile arg Override the name of output file
-e [--ext] arg Set extension for output files [mzML|mzXML|mgf|txt|mz5]
--mzML Write mzML format [default]
--mzXML Write mzXML format
--mz5 Write mz5 format
--mgf Write Mascot generic format
--text Write ProteoWizard internal text format
--ms1 Write MS1 format
--cms1 Write CMS1 format
--ms2 Write MS2 format
--cms2 Write CMS2 format
(continued)
PWIZ Conversion 353
Parameter Flag Explanation

-v [--verbose] Display detailed progress information
--64 Set default binary encoding to 64-bit precision [default]
--32 Set default binary encoding to 32-bit precision
--mz64 Encode m/z values in 64-bit precision [default]
--mz32 Encode m/z values in 32-bit precision
--inten64 Encode intensity values in 64-bit precision
--inten32 Encode intensity values in 32-bit precision [default]
--noindex Do not write index
-i [--contactInfo] arg Filename for contact info
-z [--zlib] Use zlib compression for binary data
--numpressLinear [=arg(=2e-009)] Use numpress linear prediction compression for binary mz
and rt data (relative accuracy loss will not exceed given
tolerance arg, unless set to 0)
--numpressPic intensities Use numpress positive integer compression for binary
intensities (absolute accuracy loss will not exceed 0.5)
--numpressSlof [=arg(=0.0002)] Use numpress short logged float compression for binary
intensities (relative accuracy loss will not exceed given
tolerance arg, unless set to 0)
-n [--numpressAll] Same as --numpressLinear --numpressSlof (see https://
github.com/fickludd/ms-numpress for more info)
-g [--gzip] gzip entire output file (adds .gz to filename)
--filter arg Add a spectrum list filter
--merge level Create a single output file from multiple input files by merging
file-level metadata and concatenating spectrum lists
--simAsSpectra Write selected ion monitoring as spectra, not chromatograms
--srmAsSpectra chromatograms Write selected reaction monitoring as spectra, not
chromatograms
--combineIonMobilitySpectra Write all drift bins/scans in a frame/block as one spectrum
instead of individual spectra
--acceptZeroLengthSpectra Some vendor readers have an efficient way of filtering out
empty spectra, but ittakes more time to open the file
--ignoreUnknownInstrumentError If true, if an instrument cannot be determined from a vendor
file, it will not be an error
--help Show this message, with extra detail on filter options
-f or --filelist (Fig. 10)

●● The -f or --filelist flag is used in case the user would like to
convert multiple input files together.
●● For instance, consider a text file containing the filenames
of three *.RAW files which need to be converted to *.
mzML format.
●● Sample text file (pwizRawFiles.txt) with list of filenames
and locations:
C :\Users\Administrator\Desktop\pwizard_pub\input-
File01.raw
C:\Users\Administrator\Desktop\pwizard_pub\input-
File02.raw
C:\Users\Administrator\Desktop\pwizard_pub\input-
File03.raw
●● Using the pwizRawFiles.txt file above with the flag -f, the
command should look like:
●● msconvert -f C:\Users\Administrator\Desktop\pwizard_
pub\pwizRawFiles.txt
●● Additional flags can be added to convert the files to differ-
ent formats as discussed further.
●● NOTE: One line inside the pwizRawFiles.txt file should
have only one input filename. Each input filename in the
list should be present in a new line
-o or --outdir
●● The flags “-o” or “--outdir” can be used to specify the
directory to write the output file
●● For example:
●● mconvert -o C:\Users\Administrator\Desktop\pwizard_pub\
●● Default: Current working directory
-c or --config
●● This option provides the user to enter all file conversion
properties together on one or more files.
●● Sample configuration (msconvert.config) file:
●● mzXML=true
●● zlib=true
●● filter=“index [3,7]”
●● filter=“precursorRecalculation”
●● IMPORTANT: DO NOT use spaces before and after “=”
inside the configuration file
PWIZ Conversion 355
--outfile
●● The --outfile argument lets the user set the name of the
output file. If missing, the output filename will be based on
the given input filename.
-e or --ext
●● This argument can be used to select the extension of the
output filename.
●● Options available: mzML, mzXML, mgf, txt, mz5
●● If none selected, the extension will be based on the output
file conversion format (For example: using argument
--mzXML with msconvert will result in the <output_file-
name> .mzXML).
●● If no output file conversion is selected, then the default as
mentioned earlier will be “mzML”
--mzML
●● The output converted file will be “.mzML” format.
●● This is the default if no other format is selected
--mzXML
●● Very similar to the above “--mzML” format
●● The output file will be created as <output_filename> .
mzXML
--mz5
●● Similar to --mzML and --mzXML options
●● Writes the output file to “mz5” format
--64
●● The “--64” option is used to set the binary encoding to
64-bit precision for the converted file (same as 64bit preci-
sion option on the msconvertGUI)
●● Default if no other encoding flags are selected
--32
●● Like above, this is used to set the binary encoding.
●● As the flag suggests, this option lets us set the output file
conversion to 32-bit encoding precision
●● The m/z and intensity values can also be encoded inde-
pendently using the --mz23 and --inten64 flags.
--mz64
●● The “mz” prefix to the flag “--mz64” refers to m/z values
This flag along with the following “--mz32” flag below
offers the flexibility of setting the m/z values only encoding

to 64-bit precision in case the user would like to specify the
“intensity” coding separately.
●● --mz64 is the default for the m/z encoding.
--mz32
●● Similar to the “--mz64” option above.
●● In case of mz and intensity independent encoding, We can
select “--mz32” for 32-bit encoding of “m/z” values
--inten64
●● The “inten” prefix to the flag “--inten64” refers to inten-
sity values.
●● This and the below given “--inten32” flags can be used for
encoding intensity values independently.
●● --inten64 is the default for encoding the intensity values.
--noindex
●● Setting this “--noindex” flag will tell the msconvert com-
mand line program NOT to write index to the output file.
-i or --contactInfo
●● Provide the msconvert program with contact information
in a filename and path by using this flag.
-z or --zlib
●● Use zlib compression for binary data by setting this flag.
This option is used for data compression.
--numpressLinear
●● Use numpress linear prediction lossy compression for
binary mz and rt data (relative error guaranteed less than
given tolerance)
--numpressPic
●● Use numpress positive integer lossy compression for binary
intensities (maximum 0.5 absolute error guaranteed)
--numpressSlof
●● Use numpress short logged float lossy compression for
binary intensities (relative error guaranteed less than given
tolerance)
-n or --numpressAll
●● Same as --numpressLinear --numpressSlof
●● See https://github.com/fickludd/ms-numpress for more
info
PWIZ Conversion 357
-g or --gzip
●● Use “-g” or “--gzip” option to gzip the entire output file.
●● Selecting this option reduces the file size and adds “.gz” to
the output filename
--filter
●● IMPORTANT: Adding this argument provides the option
to add/set filters before writing the data to output file.
●● Several options are available for filtering the data. All the
available filter types/options are explained in greater detail
under the next section C (FILTER OPTIONS).
●● Each type of filter mentioned in section C should be pre-
ceded by the “--filter” flag.
●● For Example:
–– msconvert C:\Users\Administrator\Desktop\pwizard_
pub\inputFile01.raw
–– --filter “peakPicking true 1-” --filter “threshold bpi-
relative .5 most-intense”
--merge
●● IMPORTANT: This argument/flag provides us the option
to merge the output of multiple input files into a single
output files. This option can be very useful if needed to
merge fractions with each fraction having an individual *.
RAW file.
--simAsSpectra
●● “sim” refers to selected ion-monitoring. This is a useful
option available in case the selected ion-monitoring is
required as spectra and NOT as chromatograms. This is
often used to represent MRM data in mzXML files, which
do not natively support chromatogram data.
--srmAsSpectra
●● “srm” refers to selected reaction monitoring. Like the
“--simAsSpectra” option, the “--srmAsSpectra” flag can be
used to write the selected reaction monitoring as sepctra
and NOT chromatograms. This is often used to represent
MRM data in mzXML files, which do not natively support
chromatogram data.
--combineIonMobilitySpectra
●● Usually, ion mobility spectra are written individually for
each scan. This option can be used to write all drift bins/
scans in a frame/block as one spectrum instead of indi-
vidual spectra.
--acceptZeroLengthSpectra
●● This option lets you accept zero length spectra in case the
file contains empty spectra.
●● Some (but not all) vendor readers have an efficient way of
filtering out empty spectra.
●● NOTE: Takes more time to open the file.
--ignoreUnknownInstrumentError
●● The instrument information should be present in the ven-
dor file usually but in case it is missing, this option can be
useful.
●● If true, if an instrument cannot be determined from a ven-
dor file, it will not be an error
--help
●● Use this option to display all filtering options in detailed
7. The msconvert commandline has additional filtering options
when compared to the msConvertGUI.
For example, msLevel is present in both commandline and
msConvertGUI (as MS Levels on GUI). This filter selects only
spectra with indicated mslevels (ms level 1). Similarly, the ana-
lyzer option in the commandline allows the user to retain the
spectra with the specified mass analyzer (“FTMS” or “ITMS”)
type only.
Note: Filters are applied sequentially in the order that you
list them, and the sequence order can make a large difference
in your output. In particular, the peakPicking filter must be
first in line if you wish to use the vendor-supplied centroiding
algorithms since these use the vendor DLLs, which only oper-
ate on raw untransformed data.
Filter Flag Explanation

index <index_value_set>
msLevel <mslevels>
chargeState <charge_states>
precursorRecalculation
mzRefiner mzRefinerinput1
precursorRefine
peakPicking [<PickerType>[snr=<minimumsignal-to-noiseratio>]
[peakSpace=<minimumpeakspacing>][msLevel=<ms_levels>]]
scanNumber <scan_numbers>
(continued)
PWIZ Conversion 359
(continued)
Filter Flag Explanation

scanEvent <scan_event_set>
scanTime <scan_time_range>
sortByScanTime
stripIT
metadataFixer
titleMaker <format_string>
threshold <type><threshold><orientation>[<mslevels>]
mzWindow <mzrange>
mzPrecursors <precursor_mz_list>
defaultArrayLength <peak_count_range>
zeroSamples <mode>[<MS_levels>]
mzPresent <tolerance><type><threshold><orientation>
<mz_list>[<include_or_exclude>]
scanSumming [precursorTol=<precursortolerance>][scanTimeTol=<scantimetolerance>]
MS2Denoise [<peaks_in_window>[<window_width_Da>[multicharge_fragment_
relaxation]]]
MS2Deisotope [hi_res[mzTol=<mzTol>]][Poisson[minCharge=<minCharge>]
[maxCharge=<maxCharge>]]
ETDFilter [<removePrecursor>[<removeChargeReduced>[<removeNeutralLoss>[<bl
anketRemoval>[<matchingTolerance>]]]]]
chargeStatePredictor [overrideExistingCharge=<true|false(false)>]
[maxMultipleCharge=<int(3)>][minMultipleCharge=<int(2)>][singleC
hargeFractionTIC=<real(0
turbocharger [minCharge=<minCharge>][maxCharge=<maxCharge>]
[precursorsBefore=<before>][precursorsAfter = <after>]
[halfIsoWidth=<half-widthofisolationwindow>][defaultMinCharge=<def
aultMinCharge>][defaultMaxCharge=<defaultMaxCharge>][useVendor
Peaks=<useVendorPeaks>]
activation <precursor_activation_type>
analyzer <analyzer>
analyzerType <analyzer>
polarity <polarity>
●● msLevel
This filter selects only spectra with the indicated <mslevels>.
–– The value set to this argument needs to be an integer.
–– For Example:
# extract MS1 scans only
msconvert C:\Users\Administrator\Desktop\pwizard_pub\
inputFile01.raw
--filter “msLevel 1”
# Alternatively, inside the configuration file from the
msconvert.config file (input as in section 3.2, step 2),
add the filter for “msLevel 1” as:
mzXML=true
zlib=true
filter=“index [3,7]”
filter=“msLevel 1” #For ms1 scans only
●● chargeState
–– This filter keeps spectra that match the listed charge state(s).
–– The value set to this argument needs to be an integer.
–– Both known/single and possible/multiple charge states
are tested. Use 0 to include spectra with no charge state
at all.
●● precursorRecalculation
–– This filter recalculates the precursor m/z and charge for
MS2 spectra.
–– Looks at the prior MS1 scan to better infer the parent mass.
–– Works only on orbitrap and FT data and does not use any
third party (vendor DLL) code.
–– ONLY need to use as a flag. For example:
# Add this line if using config file
filter=“precursorRecalculation”
# If using as a flag to the command line parameter, add the
following to the command:
--filter “precursorRecalculation”
●● mzRefiner
–– This filter recalculates the m/z and charges, adjusting precur-
sors for MS2 spectra and spectra masses for MS1 spectra.
–– It uses an ident file with a threshold field and value to calcu-
late the error and will then choose a shifting mechanism to
correct masses throughout the file. It only works on Orbitrap,
FT, and TOF data. It is designed to work on mzML files.
PWIZ Conversion 361
●● precursorRefine
–– Similar to precursorRecalculation
–– This filter refines the precursor m/z and charge for MS2
spectra. It looks at the prior MS1 scan to better infer the
parent mass. It only works on orbitrap, FT, and TOF data.
It does not use any third party (vendor DLL) code.
●● peakPicking
–– This filter performs centroiding on spectrawith the selected
ms_level argument
peakPicking [<PickerType>[snr=<minimum signal-to-
noise ratio>] [peakSpace=<minimum peak spacing>]
[msLevel=<ms_levels>]]
●● scanNumber
–– This filter takes input as an integer and lets the user select
spectra by scan number
–– Scan numbers are 1-based and not contiguous
●● scanEvent
–– scanEvent filters spectra by scan event and takes input in
the form of a set.
–– This filter offers the flexibility of excluding the selected
scan events as well.
–– For Example:
# All scan event EXCEPT 5 can be included using-
--filter “scanEvent 1-4 6- ”
●● scanTime
–– Spectra can be selected using a specific time range using
this filter.
●● sortByScanTime
–– As the name suggests, the spectra can be a sorted in ascend-
ing order by scan start time.
●● stripIT
–– “IT” in the filter name refers to “ion trap”.
–– This filter removes out the ion trap data spectra with MS
level 1.
●● metadataFixer
–– As the name suggests, this filter can be used to “fix” metadata—
meaning, it adds/replaces a spectra’s TIC/BPI metadata.
–– This filter traverses the m/z intensity arrays to find the sum
and max.
–– Usually used after peak-picking, for Example:

--filter “peakPicking true 1-” --filter metadataFixer
●● titleMaker
–– Uses a string format as the input
–– For Example:
--filter “titleMaker<RunId>.<ScanNumber>.<ScanNumb
er>.<ChargeState>”
●● Threshold
–– This filter can be used to set several thresholds.
–– Data in the input file meeting all the thresholds is retained
and everything else is filtered out.
–– Threshold filter can be set to “most or least” (i.e., either
upper or lower limit can be set) for all give threshold types.
This can be set using the orientation in the threshold for-
mat as given below.
–– Format for threshold-
threshold<type><threshold><orientation>[<mslevels>]
–– Various threshold types with their input format and expla-
nation are given below:
Threshold Threshold
type format Explanation
count Integer (n) Retains intensity values for the selected number of “n” data points. Data
points where intensity is equal to nth intense data point are removed
count-after- Integer (n) Same as the count but retains data points where intensity = nth data
ties point
absolute Flag (no Keep data whose absolute intensity matches threshold
argument)
bpi-relative Flag (no Data is retained if the intensity matches the percentage of base-peak
argument) intensity. 0.75 = 75 %
tic-relative Flag (no Similar to above but for retains data for individual intensities greater
argument) than or less than the percentage of total ion current for the scan.
1 = 100 %
tic-cutoff Flag (no As the name suggests, this is a cutoff. Data is retained UPTO the
argument) percentage of the total ion current for the scan
ms_levels Integer (n) OPTIONAL.
If selected, only scans with MS level = n will be selected
PWIZ Conversion 363
●● mzWindow
–– As the name suggest, this filter is used to provide a range
to select the m/z within the selected window.
–– The m/z values falling ONLY within the selected range
will be retained.
–– Input type: range [mzLow, mzHigh]
–– For Example:
--filter “mzWindow [100.1,307.5]”
●● mzPrecursors
–– This filter allows us the flexibility of selecting spectra within
a given list of precursor m/z values.
–– The input type is a list of m/z precursor value
–– For Example:
--filter “mzPrecursors [123.4,567.8]”
●● defaultArrayLength
–– For this filter, the default array length refers to the range of
peak counts.
–– The name is derived from mzML format file where
“defaultArrayLength” refers to peak list.
–– This can be specified as range of integers
–– For Example:
# Retain only peakcounts >=100:
--filter “defaultArrayLength 100-”
●● zeroSamples
–– Usage: zeroSamples <mode> [<MS_levels>]
–– This is a useful filter to deal with zero values in the spec-
trum. This can be used one of two ways—either remove
spectra with zero values or add the sero value incase the
spectra is missing.
–– Mode options-
i. removeExtra: consecutive zero intensity peaks are
removed from spectra
“100.1,1000 100.2,0 100.3,0 100.4,0 100.5,0 100.6,1030”
would become
“100.1,1000 100.2,0 100.5,0 100.6,1030”
and a peak list
“100.1,0 100.2,0 100.3,0 100.4,0 100.5,0 100.6,1030
100.7,0 100.8,1020 100.9,0 101.0,0”
would become
“100.5,0 100.6,1030 100.7,0 100.8,1020 100.9,0”
ii. addMissing: When <mode> is “addMissing”, each spec-

trum’s sample rate is automatically determined and
flanking zeros are inserted around non-zero data points.
The optional [=<flankingZeroCount>] value can be used
to limit the number of flanking zeros, otherwise the spec-
trum is completely populated between nonzero points.
For example, to make sure spectra have at least five
flanking zeros around runs on nonzero points,
use filter-
“addMissing = 5”.
–– The <MS levels> is optional but when used, retains spectra
with only the ms levels specified using the option. Format:
integer set (For example: 1–5)
●● mzPresent
–– Usage: mzPresent <tolerance> <type> <threshold> <orientation>
<mz_list> [<include_or_exclude>]
–– This filter is very similar to threshold filter
i. tolerance: Specified as a number and units (PPM or MZ).
For example, “5 PPM” or “2.1 MZ”.
ii. <type>, <threshold>, and <orientation> operate as in
the “threshold” filter (refer to step 14).
iii. <mz_list>: List of mz - [mz1,mz2, … mzn]
#Data points within <tolerance> of any of these values
will be kept. For example,
“[100, 300, 405.6]”
iv. <include_or_exclude>: as the terms suggest, using
include (default) retains all the values that match the
criteria where as exclude drops data points matching
the criteria.
●● scanSumming
–– Usage: scanSumming [precursorTol=<precursor toler-
ance>] [scanTimeTol=<scan time tolerance>]
–– Used to sum “MS2201D sub-scans whose precursors are within
the given <precursor tolerance> and <scan time tolerance>
–– Defaults:
i. Precursor tolerance: 0.05
ii. Scan time tolerance: 10 s
–– Used where sub-scans need to be summed to increase the
signal-to-noise ratio.
–– IMPORTANT: Only tested on waters data so far.
PWIZ Conversion 365
●● MS2Denoise
–– Usage: MS2Denoise [<peaks_in_window> [<window_
width_Da> [multicharge_fragment_relaxation]]]
–– Use to denoise, i.e., remove peaks with noise for spectra
with precursor ions.
i. <peaks_in_window>: the number peaks to select in
moving window, default is 6.
ii. <window_width_Da>: the width of the window in
Da, default is 30.
iii. <multicharge_fragment_relaxation>—if “true” (the
default), allows more data below multiply charged
precursors.
●● MS2Deisotope
–– MS2Deisotope [hi_res [mzTol=<mzTol>]] [Poisson
[minCharge=<minCharge>] [maxCharge=<maxCharge>]]
–– Uses the Markey method or a Poisson model to deisotope
the ms2 spectra.
i. hi_res: set to “false” (default) or “true”
●● mzTol: Input format is a decimal value. Sets the
m/z tolerances .Regular default is 0.5, high resolu-
tion default is 0.01
ii. Poisson: Used to define the search range for the
charge. Default is 1
1. minCharge: Default is 1.
2. maxCharge: Default is 3.
–– For Example:
--filter “MS2Deisotope true mzTol=0.4 1 minCharge=1
maxCharge=3”
●● ETDFilter
–– Usage: ETDFilter [<removePrecursor> [<rem-
oveChargeReduced> [<removeNeutralLoss> [<blanket-
Removal> [<matchingTolerance>]]]]]
i. <removePrecursor>—specify “true” to remove unre-
acted precursor (default is “false”)
ii. <removeChargeReduced>—specify “true” to remove
charge reduced precursor (default is “false”)
iii. <removeNeutralLoss>—specify “true” to remove neu-
tral loss species from charge reduced precursor (default
is “false”)
iv. <matchingTolerance>—specify matching tolerance in
MZ or PPM (examples: “3.1 MZ” (the default) or
“2.2 PPM”)
●● chargeStatePredictor
–– Usage: chargeStatePredictor [overrideExistingCharge=
<true|false (false)>] [maxMultipleCharge=<int (3)>]
[minMultipleCharge=<int (2)>] [singleChargeFractionTIC=
<real (0.9)>] [maxKnownCharge=<int (0)>] [makeMS2=
<true|false (false)>]
i. <overrideExistingCharge> : always override existing
charge information (default:“false”)
ii. <maxMultipleCharge> (default 3) and <minMulti-
pleCharge> (default 2): range of values to add to the
spectrum’s existing “MS_possible_charge_state” val-
ues.If these are the same values, the spectrum’s MS_
possible_charge_state values are removed and replaced
with this single value.
iii. <singleChargeFractionTIC>: is a percentage expressed
as a value between 0 and 1 (the default is 0.9, or 90 %).
This is the value used as the previously mentioned ratio
of intensity above and below the precursor m/z.
iv. <maxKnownCharge> (default is 0, meaning no maxi-
mum): the maximum charge allowed for “known”
charges even if override existing charge is false. This
allows overriding junk charge calls like +15 peptides.
v. <algorithmMakeMS2>: default is “false”, when set to
“true” the “makeMS2” algorithm is used instead of
the one described above.
●● turbocharger
–– turbocharger [minCharge=<minCharge>] [maxCharge=
<maxCharge>] [precursorsBefore=<before>] [precursors
After=<after>] [halfIsoWidth=<half-width of isolation
window>] [defaultMinCharge=<defaultMinCharge>]
[defaultMaxCharge=<defaultMaxCharge>] [useVendorPe
aks=<useVendorPeaks>]
–– <maxCharge> (default: 8) and <minCharge> (default 1):
defines range of possible precursor charge states.
–– <before> (default: 2) and <after> (default 0): number of
survey (MS1) scans to check for precursor isotopes, before
and after a MS/MS in retention time.
–– <half-width of isolation window> (default: 1.25): half-
width of the isolation window from which precursor is
derived. Window is centered at target m/z with a total size
of ± the value entered.
–– <defaultMinCharge> (default: 0) and <defaultMaxCharge>
(default: 0): in the event that no isotope is found in the iso-
PWIZ Conversion 367
lation window, a range of charges between these two values

will be assigned to the spectrum. If both values are left at
zero, no charge will be assigned to the spectrum.
●● activation
–– Usage: activation <precursor_activation_type>
–– Filter to retain only spectra of a specified type of activation.
DOES NOT effect non-MS spectra.
–– Input options:
i. ETD
ii. CID
iii. SA
iv. HCD
v. HECID
vi. BIRD
vii. ECD
viii. IRMPD
ix. PD
x. PSD
xi. PQD
xii. SID
xiii. SORI
●● analyzer
–– Usage: analyzer <analyzer>
–– Filter to retain only spectra of a specified type of mass analyzer
–– Four options for input:
i. “quad”
ii. “orbi”
iii. “FT”
iv. “IT”
v. “TOF”
●● analyzerType
–– Currently deprecated but accepted two options are:
“FTMS” or “ITMS”
●● Polarity
–– This filter allows the user to retain spectra with scan of the
selected polarity.
–– Available Options:
i. “negative”
ii. “positive”
iii. “+”
iv. “−”
References
1. Chambers MC, Maclean B, Burke R, Amodei D, MacCoss M, Tabb DL, Mallick P (2012) A
Ruderman DL, Neumann S, Gatto L, Fischer B, cross-platform toolkit for mass spectrometry and
Pratt B, Egertson J, Hoff K, Kessner D, Tasman proteomics. Nat Biotechnol 30(10):918–920.
N, Shulman N, Frewen B, Baker TA, Brusniak doi:10.1038/nbt.2377
MY, Paulse C, Creasy D, Flashner L, Kani K, 2. Kessner D, Chambers M, Burke R, Agus D,
Moulding C, Seymour SL, Nuwaysir LM, Mallick P (2008) ProteoWizard: open source
Lefebvre B, Kuhlmann F, Roark J, Rainer P, software for rapid proteomics tools develop-
Detlev S, Hemenway T, Huhmer A, Langridge ment. Bioinformatics 24(21):2534–2536.
J, Connolly B, Chadick T, Holly K, Eckels J, doi:10.1093/bioinformatics/btn323
Deutsch EW, Moritz RL, Katz JE, Agus DB,
Printed on acid-free paper
Index
A Enhanced Filter Aided Sample Preparation

(eFASP)��11–17
Abnormal experiment�� 326, 330, 333, 335 Enzymatic solution digest��62
Accutase�� 263, 265, 268 Epoxy-coated coverslips�� 263, 264, 268
Acetone precipitation��2–6, 8 Euclidean distance�� 326, 330
Affinity purification��20, 115–129, 131, 261
Algorithm��109, 165, 224, 229, 243, F
249–251, 254, 266, 292, 302–305, 358, 366
FLAG tag�� 117, 123, 139,
Ammonium deoxycholate��12, 13
140, 143–146
Antibiotic�� 122, 140, 146
Fluorescence�� 116, 138, 152,
Antibody��138, 139, 141–144, 147,
261, 265–269
154–160, 163, 262–265, 267, 268
Fluorophore�� 152, 268
B Formaldehyde�� 20, 22–24, 30, 63
Fractionation��36–39, 43, 58, 65–66,
Bias�� 3, 8, 49, 250, 279, 280, 282 70–81, 84, 85, 87–90, 92–97, 100, 101, 108, 116, 171,
Binary complex�� 138, 339 187–192, 197, 271, 275, 282
BioID�� 115–129, 131
Biotin�� 20, 22, 117–120, 123–127, 129, 131, 132 G
C Genotype�� 236, 237, 240
Glycoproteomics��100
Calcium�� 49, 140, 146
Cell line��23, 30, 35–45, 49, 58, 70, 71, 79, 80, 85, H
118–119, 122–123, 138, 150, 171–173, 178, 179, 182
Heterochromatin��19
Change point��326, 330–331, 335, 336
Hybridization�� 19, 20, 25–26, 31, 32, 263, 265
Chromatin��7, 8, 19–33, 117, 124
Cloning�� 117–119, 121–122, 130, 141, 315, 324 I
D Immunoassay��152
Immunoprecipitation��19, 30, 115, 116,
Databases��44, 51, 56, 67, 73, 76, 78,
138–140, 144, 145, 261
88, 92, 94, 133, 199, 212–214, 221, 224, 236–238, 240,
Informatics��339
244, 245, 249, 294
Internal Reflection Fluorescence (TIRF)
Data conversion�� 294, 298, 299, 340
Microscopy��265
Data-dependent acquisition (DDA)�� 64, 76, 91,
Isotope labeling��23
128, 208–209, 224, 225, 273
Data-independent acquisition (DIA)�� 116, 128, L
223, 229, 273, 290, 296, 303, 305
Laser Capture Microdissection�� 150, 151
Design of experiments��278
Legacy hardware��310
Detergent��2, 11, 36, 61, 62, 80, 116, 124, 133
Liquid chromatography-mass spectrometry
2D Histograms�� 266, 267
(LC-MS)��1–4, 7, 8, 36, 41, 43,
Dimethylation��64–65
51, 52, 55–57, 70, 73, 76–79, 81, 87–88, 90–92,
Dissimilarity��271, 326, 329–330, 333–335
95, 97, 100, 101, 105, 109–110, 128, 129, 173, 174,
E 177–179, 224–226, 228, 232, 274–276, 280, 282,
284, 325, 326, 328–333, 335, 336
Electrospray-ionization��66, 76, 91, 109, 276 Luciferase�� 138–141, 143–146
ELISA�� 140–142, 145–147, 182 LUMIER�� 137–142, 144–146
DOI 10.1007/978-1-4939-6747-6, © Springer Science+Business Media LLC 2017
369
Proteomics: Methods and Protocols
370 Index

M Proteins�� 1, 11, 20, 35, 48, 61, 70, 84, 99, 115, 137,
149, 171, 185, 199, 223, 236, 261, 271, 289, 331, 339
Magnesium�� 49, 140 Protein extraction�� 1–9, 11, 37, 41, 71, 102, 271
Mammalian cells�� 122, 139 Protein glycosylation�� 100, 101
Mass spectrometry��3, 7, 17, 20, 27–31, Protein identification�� 33, 35, 36, 43, 44, 67,
33, 36, 38, 41, 48, 51, 61–67, 70, 71, 76, 84, 85, 87, 91, 81, 102, 108, 109, 193, 277, 282, 283
93, 94, 100, 105, 109, 115–129, 131, 133, 171, Protein interactions�� 28, 116, 138, 235–243, 245–247,
185–197, 199–221, 223–229, 231, 232, 272, 274, 276, 249–255, 261–269
284, 289–291, 294, 296, 309, 319, 326, 339 Protein localization��35, 36
Membrane proteins��12, 61–66, 116, 132, 261 Protein microarray��149–168
Microarray��149–168 Protein network��137–147
Microarray analysis��146–168 Protein phosphorylation��48
Micropatterning�� 261–263, 266, 268 Protein-protein interactions (PPIs)�� 116–119, 132,
Microscopy��265 138–140, 235–243, 245–247, 249–255, 261–269
Multicolor detection��156 Protein purification��11
Multi-lectin affinity chromatography Protein quantitation��99–102, 104–111, 173,
(M-LAC)��99–111 214, 266, 274, 275
Mutations�� 235, 236, 243–254 Proteome�� 2, 36, 59, 61, 70, 84, 151, 171,
MzML��294, 298, 299, 305, 339, 193, 232, 273, 289, 339
340, 352, 355, 360, 363 Proteome informatics��339
MzXML��294, 298, 339, 340, 345, Proteomics�� 1, 19, 36, 56, 61, 70, 84, 109,
350, 352, 354, 355, 357, 360 120, 150, 171, 199, 223, 271, 289, 325, 339
ProteoWizard��179, 294, 298, 305, 326,
N
340–341, 345, 349, 352
Nanocrystals�� 152, 153, 155, 156 Proximity labeling��115
Nuclear fractionation��37, 39 pyProphet�� 292, 293, 295,
301–302, 305
O
Q
Open formats�� 294, 298, 339, 340
OpenSWATH��224, 229, 292, 294, Qdot nanocrystal�� 152, 153, 155–160
295, 298–305 Q-TOF��208–211, 223, 305, 332
Quadrupole��97, 193, 199, 208, 224
P Quality control (QC)��195, 215, 220, 225, 229,
Peptide��1, 12, 30, 38, 50, 61, 70, 84, 109, 272, 274–276, 278–280, 282–284, 293, 295–298, 325,
123, 139, 172, 185, 201, 223, 247, 272, 290, 366 326, 328–333, 335, 336
Pericentromeric repeats��20 Quantification�� 6, 8, 29, 35, 37, 41, 45, 56,
Phase-transfer�� 12, 15, 17, 62 70, 71, 76, 84, 92, 97, 123, 140, 151, 152, 154, 173,
Phenotype�� 236–238, 240 185–197, 261, 283, 289, 290, 292, 300, 352
Phosphate-buffered saline (PBS)��22–24, 30, 37, Quantitation�� 5, 99–102, 104–111, 171, 173,
39, 49, 51, 58, 64, 79, 104, 111, 123, 124, 142, 144–146, 174, 176, 179, 180, 205–206, 214, 231, 232, 265–267,
175–177, 182, 263, 265 273–276, 283, 285, 302
Phosphocapture��47 Quantitative analysis�� 31, 41, 186,
Phosphorylation��48, 58, 117, 123, 139, 261–269
151, 173, 183, 249 Quantitative proteomics�� 11–17, 61–66,
Plasma proteomics�� 275, 282 171–173, 175–182, 223–229, 231, 232, 273
Polylysine��268
R
Post-translational modification��35, 48, 61, 99,
100, 123, 132, 238, 239 Randomization�� 129, 229,
Precision medicine��149–168 278–280, 282, 284
Principal component analysis (PCA)�� 116, 215, 326, Receptor tyrosine kinase��179
329, 330, 332, 334
Proteomics: Methods and Protocols
371
Index

Reverse phase protein microarray (RPPA)�� 149–152, T
154–163
RNase A�� 24, 29, 31 Tandem mass tags (TMT)�� 84, 85, 87–90,
92–97, 185–197, 276, 280
S Targeted proteomics�� 199–221, 289, 290
Telomeres��20, 22, 28, 29, 32
Sample preparation�� 2, 37–40, 49, 62, 70, 71,
Ternary complex�� 139, 140
85, 88, 150, 154, 191, 197, 220, 282, 325
Time-of-flight (TOF)��128, 186, 199, 223,
Serum�� 29, 37, 100, 110, 130, 133, 140,
296, 303, 360, 367
142, 144, 150, 154, 155, 174, 205, 263, 331
TIRF microscopy��265
Signal transduction�� 36, 48, 149
Transfection�� 119, 122, 130, 140, 143, 144, 146
Signaling pathways�� 137, 138, 247, 249
Transmembrane proteins�� 139, 263
Single nucleotide polymorphism
TRIC aligner��292, 293, 295, 302–303
(SNP)�� 235, 237, 238
Triple quadrupole (QQQ)�� 199, 204–207, 216–221
Skyline��201, 206, 216, 221, 224,
Trypsin�� 3, 6, 7, 9, 12, 13, 15, 16, 38–40, 44,
229, 293–298, 303
50, 52, 62–64, 67, 72, 73, 78, 94, 101, 105, 109, 121,
Smad2 and Smad4��138
124–128, 130, 133, 142, 174–178, 188, 190, 193, 195,
Soft lithography�� 262, 264–265, 267
201, 207, 212, 220, 268, 274
Sonication�� 6, 8, 14, 16, 20, 21, 24,
30, 31, 39, 74, 80, 176 V
STAGE-Tip��17, 63, 65, 75, 81, 191
Streptavidin��20, 22, 25, 26, 31, Variable windows�� 227, 228, 231
118–120, 124–127, 132, 262–264, 268 Variance�� 194, 277, 282, 283, 327, 329–331, 335
Subcellular proteomic��43 Virtual computers��310, 311, 317, 319, 323
SWATH�� 116, 223, 290 VirtualBox��311–313, 317–320, 322–324
SWATH acquisition��223–229, 231, 232, 290, 303
W
SWATH MS�� 224, 232, 290–294, 302, 303, 305
System reliability��309 Western-blot�� 20, 27, 28, 125, 139, 164, 181, 182
Systems management��122 Wnt signaling��138

Proteomics: Methods and Protocols

Uploaded by

Copyright:

Available Formats

Proteomics: Methods and Protocols

Uploaded by

Document Information

Original Title

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

Proteomics: Methods and Protocols

Uploaded by

Copyright:

Available Formats

What is the book about?

What is the book about?

What techniques are discussed in the book?

What techniques are discussed in the book?

Methods in

Molecular Biology 1550

Lucio Comai · Jonathan E. Katz

For further volumes:

Methods and Protocols

ISSN 1064-3745 ISSN 1940-6029 (electronic)

Library of Congress Control Number: 2017930810

© Springer Science+Business Media LLC 2017

Printed on acid-free paper

This Humana Press imprint is published by Springer Nature

Los Angeles, CA Lucio Comai

1 A Robust Protocol for Protein Extraction and Digestion . . . . . . . . . . . . . . . . . 1

12 Dual-Color, Multiplex Analysis of Protein Microarrays

Ravali Adusumilli • Department of Radiology, Canary Center at Stanford for Cancer

Andy Gieschen • Agilent Technologies, Inc., Santa Clara, CA, USA

Parag Mallick • Department of Radiology, Canary Center at Stanford for Cancer Early

Dongxue Wang • Chair of Proteomics and Bioanalytics, Technische Universität München,

A Robust Protocol for Protein Extraction and Digestion

Analysis of proteins, the key effectors of most cellular processes, is

their several orders of magnitude abundance variation in cells, has

provided, and critical variables for each method are emphasized so

1. Cell Lysis and Acetone Precipitation.

2.1 Solutions 1. 4 % w/v sodium dodecyl sulfate (SDS).

2.2 Equipment 1. Probe tip sonicator (alternatively, a Diagenode or Covaris

2.3 Peptide 1. 1 M TCEP: Dissolve 287 mg in 1 mL MS water. Store at

2.5 Equipment 1. Thermo Pierce C18 tips (Thermo product 87784).

the tubes in water is the best way to accomplish this. Secondly,

4. Repeat equilibration step and discard solvent.

1. The sonicator is used to shear the chromatin in the sample,

6. Volume of lysis buffer stated is for probe tip sonicator; vol-

17. Again, make sure this is at least a fourfold volume addition of

Improving Proteome Coverage and Sample Recovery

The integrity of proteomic experiments hinges on consistent and

Prepare all solutions fresh using ultrapure water and MS-grade

2.1 Solutions 1. Passivation Solution: 5 % (v/v) TWEEN® 20.

7. Peptide Recovery Buffer: 50 mM ABC, pH 8.

2.2 Solutions 1. Passivation Solution: 5 % (v/v) TWEEN® 20.

2.3 Equipment 1. Microcon® ultrafiltration units (YM-30 30 kDa cutoff limit;

Two eFASP protocols are described. The first is the standard

Carry out all procedures at room temperature or as specified

6. Dispense 100 μl Alkylation Buffer to the filter unit and trans-

11. Dispense 50 μl of Peptide Recovery Buffer to the filter unit

1. This protocol employs deoxycholic acid rather than sodium

could substitute for deoxycholic acid, with STAGE tip cleanup

Proteome Characterization of a Chromatin Locus Using

Key words Proteomics, Chromatin, Protein Purification, Repeated Sequences, Heterochromatin,

Each cell of an organism harbors the same genetic information. Yet

The PICh method is powerful because it provides a comprehensive

Formaldehyde Chromatin Chromatin

Fig. 1 Outline of the PICh protocol

provided the abundance of that locus is comparable in the two sam-

Prepare all solutions using ultrapure water and analytical grade

2.1 Special 1. Sonicator with high power sonication probe (Qsonica).

2.2 Buffers 1. 100 mg/mL RNaseA.

Scramble: Desthiobiotin-TEG-4XC18 spacers –

Los Angeles, CA Lucio Comai

3.4 NanoLC- 5 μL of each peptide fraction is analyzed using 180 min data-