The Geometric Structure of the Brain Fiber Pathways

Van J. Wedeen et al.

Science 335, 1628 (2012);
DOI: 10.1126/science.1215280

This copy is for your personal, non-commercial use only.

If you wish to distribute this article to others, you can order high-quality copies for your
colleagues, clients, or customers by clicking here.

Downloaded from www.sciencemag.org on June 10, 2012

Permission to republish or repurpose articles or portions of articles can be obtained by
following the guidelines here.

The following resources related to this article are available online at

www.sciencemag.org (this information is current as of June 10, 2012 ):

A correction has been published for this article at:

http://www.sciencemag.org/content/336/6082/670.full.html
Updated information and services, including high-resolution figures, can be found in the online
version of this article at:
http://www.sciencemag.org/content/335/6076/1628.full.html
Supporting Online Material can be found at:
http://www.sciencemag.org/content/suppl/2012/03/28/335.6076.1628.DC1.html
http://www.sciencemag.org/content/suppl/2012/03/29/335.6076.1628.DC2.html
A list of selected additional articles on the Science Web sites related to this article can be
found at:
http://www.sciencemag.org/content/335/6076/1628.full.html#related
This article cites 29 articles, 5 of which can be accessed free:
http://www.sciencemag.org/content/335/6076/1628.full.html#ref-list-1
This article has been cited by 1 articles hosted by HighWire Press; see:
http://www.sciencemag.org/content/335/6076/1628.full.html#related-urls
This article appears in the following subject collections:
Neuroscience
http://www.sciencemag.org/cgi/collection/neuroscience

Science (print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by the
American Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005. Copyright
2012 by the American Association for the Advancement of Science; all rights reserved. The title Science is a
registered trademark of AAAS.
 CORRECTED 11 MAY 2012; SEE LAST PAGE
REPORTS
(fig. S13), and was strongest when the GoC pop- interneurons in cortical and subcortical structures 21. S. J. Cruikshank et al., Proc. Natl. Acad. Sci. U.S.A. 101,
ulation was sparsely activated by chemical excit- (7) use similar mechanisms. Our results suggest 12364 (2004).
22. M. R. Deans, J. R. Gibson, C. Sellitto, B. W. Connors,
atory synaptic inputs (fig. S14). that interneurons do not operate as fully indepen- D. L. Paul, Neuron 31, 477 (2001).
Our results show that the passive properties dent neuronal units but share charge during chem- 23. P. Gleeson, V. Steuber, R. A. Silver, Neuron 54, 219
of GoC dendrites confer distance-dependent sub- ical synaptic excitation and thus exhibit features (2007).
linear chemical synaptic integration. This weakens of a syncitium. 24. W. Rall, J. Neurophysiol. 30, 1138 (1967).
25. S. Cash, R. Yuste, Neuron 22, 383 (1999).
the impact of distal excitatory inputs. However, 26. S. A. Bloomfield, B. Völgyi, Nat. Rev. Neurosci. 10, 495
the high density of dendritic GJs in the molecular References and Notes (2009).
layer enables PF synaptic charge to flow into the 1. T. F. Freund, G. Buzsáki, Hippocampus 6, 347 (1996). 27. E. Yaksi, R. I. Wilson, Neuron 67, 1034 (2010).
dendrites of neighboring GoCs. This GJ-mediated 2. P. Jonas, J. Bischofberger, D. Fricker, R. Miles, 28. A. M. Kerlin, M. L. Andermann, V. K. Berezovskii,
Trends Neurosci. 27, 30 (2004). R. C. Reid, Neuron 67, 858 (2010).
lateral excitation counteracts the effects of sub- 3. F. Pouille, M. Scanziani, Science 293, 1159 (2001). 29. D. Marr, J. Physiol. 202, 437 (1969).
linear dendritic behavior by enabling distal inputs 4. C. J. McBain, A. Fisahn, Nat. Rev. Neurosci. 2, 11 (2001).
to drive network activity more effectively. Den- 5. H. Hu, M. Martina, P. Jonas, Science 327, 52 (2010). Acknowledgments: Funded by U.K. Biotechnology and
dritic GJs therefore counteract the problem of 6. M. Martina, I. Vida, P. Jonas, Science 287, 295 (2000). Biological Sciences Research Council (BBSRC) (F005490),
7. B. W. Connors, M. A. Long, Annu. Rev. Neurosci. 27, 393 Medical Research Council (G0400598), and Wellcome Trust
dendritic saturation (24) without the need to (2004). (064413). R.A.S. holds a Wellcome Trust Principal Research
boost electrically remote synaptic input with 8. R. T. Kanichay, R. A. Silver, J. Neurosci. 28, 8955 Fellowship (095667) and a European Research Council (ERC)
active dendritic conductances (25). A key role of (2008). Advanced Grant, Z.N. a Wellcome Trust Project Grant and an
interneurons is to counteract and balance network 9. S. Dieudonne, J. Physiol. 510, 845 (1998). ERC Advanced Grant, and A.L. a Janos Bolyai Scholarship.
excitation. The combination of passive dendrites 10. K. Vervaeke et al., Neuron 67, 435 (2010). We thank D. Paul for the Cx36–/– mice; E. Chaigneau,

Downloaded from www.sciencemag.org on June 10, 2012

11. G. P. Dugué et al., Neuron 61, 126 (2009). T. Branco, and P. Gleeson for help; D. Attwell, M. Farrant,
and dendritic GJs facilitates this by enabling a 12. A. Losonczy, J. C. Magee, Neuron 50, 291 (2006). H. Hu, D. Kullmann, J. Rothman, and D. Ward for comments on
larger fraction of interneurons to respond to lo- 13. T. Nevian, M. E. Larkum, A. Polsky, J. Schiller, the manuscript; and T. Fernandez-Alfonso, H. Hu, and D. Ruedt
calized patches of synaptic excitation. Our results Nat. Neurosci. 10, 206 (2007). for discussions.
reveal how GJs on inhibitory interneuron den- 14. L. Forti, E. Cesana, J. Mapelli, E. D’Angelo, J. Physiol.
574, 711 (2006). Supporting Online Material
drites could contribute to spatial averaging, which 15. J. C. Magee, E. P. Cook, Nat. Neurosci. 3, 895 (2000). www.sciencemag.org/cgi/content/full/science.1215101/DC1
has been proposed in the retina (26) and excitatory 16. S. R. Williams, G. J. Stuart, Science 295, 1907 (2002). Materials and Methods
olfactory neurons in insects (27), and to the broad 17. H. Ohishi et al., Neuron 13, 55 (1994). Figs. S1 to S15
tuning of inhibitory interneurons in cortex (28). 18. B. P. Vos, A. Volny-Luraghi, E. De Schutter, Eur. J. References (30–45)
Neurosci. 11, 2621 (1999).
These mechanisms are also likely to contribute to 19. P. L. van Kan, A. R. Gibson, J. C. Houk, J. Neurophysiol. 11 October 2011; accepted 21 February 2012
gain control in the granule cell layer through PF- 69, 74 (1993). Published online 8 March 2012;
mediated feedback (29), and it seems likely that 20. S. Solinas et al., Front. Cell Neurosci. 1, 2 (2007). 10.1126/science.1215101

The Geometric Structure of the ing.” Crossing is a pervasive feature of brain

structure and may be essential for efficient con-
nectivity (20, 21). Owing to crossing, the map-
Brain Fiber Pathways ping of connectivity must untangle pathways
from cellular to macroscopic scales simultane-
ously (22, 23). This was accomplished with tract
Van J. Wedeen,1* Douglas L. Rosene,2 Ruopeng Wang,1 Guangping Dai,1 Farzad Mortazavi,2
tracers methods, which are considered a gold stan-
Patric Hagmann,3 Jon H. Kaas,4 Wen-Yih I. Tseng5
dard (18, 19). Tracer studies inject compounds
into the live brain and allow them to disperse by
The structure of the brain as a product of morphogenesis is difficult to reconcile with the means of axonal transport, marking individual
observed complexity of cerebral connectivity. We therefore analyzed relationships of adjacency and axons over large distances. However, these can
crossing between cerebral fiber pathways in four nonhuman primate species and in humans by map only a small fraction of the pathways in any
using diffusion magnetic resonance imaging. The cerebral fiber pathways formed a rectilinear single brain and are not feasible in humans. Thus,
three-dimensional grid continuous with the three principal axes of development. Cortico-cortical the discovery and analysis of the structural re-
pathways formed parallel sheets of interwoven paths in the longitudinal and medio-lateral axes, lationships between pathways—and their context
in which major pathways were local condensations. Cross-species homology was strong and showed within cerebral connectivity—has remained chal-
emergence of complex gyral connectivity by continuous elaboration of this grid structure. This lenging (18, 19, 24).
architecture naturally supports functional spatio-temporal coherence, developmental path-finding, To address these limitations, methods have
and incremental rewiring with correlated adaptation of structure and function in cerebral been developed to map the fiber pathways of the
plasticity and evolution.
1
Department of Radiology, Massachusetts General Hospital
(MGH), Harvard Medical School and the MGH/Massachussetts
he organizing principles of cerebral con- stem (3–5, 7–9), and functional studies (10–13)

T nectivity remain unclear. In the brainstem

and spinal cord, fiber pathways are or-
ganized as parallel families derived from the three
also suggest that connectivity is geometrically or-
ganized. Several leading theories of cerebral func-
tion (14–17) propose geometric organization at
Institute of Technology Athinoula A. Martinos Center for
Biomedical Imaging, Building 129, 13th Street, 2nd Floor,
Charlestown, MA 02129, USA. 2Department of Anatomy and
Neurobiology, Boston University, Medical Campus, 700 Albany
Street, W701, Boston, MA 02118, USA. 3Department of
principal axes of embryonic development: the multiple scales. However, high-resolution studies Radiology, University Hospital Center–University of Lausanne,
rostro-caudal, the medio-lateral (or proximo-distal), of cerebral connectivity with tract tracers have Rue du Bugnon, 46, CH-1011 Lausanne, Switzerland. 4Depart-
and the dorso-ventral (1–6). In the forebrain of given only limited evidence of geometric orga- ment of Psychology, College of Arts and Sciences, Vanderbilt
advanced species, however, corresponding pat- nization (10–12, 18, 19). University, 301 Wilson Hall, 111 21st Avenue South, Nashville,
TN 37240, USA. 5Department of Radiology, Center for Opto-
terns of connectivity have yet to be established. A challenge in the investigation of cerebral electronic Biomedicine, National Taiwan University College of
Many studies of evolution, development, and gene structure and connectivity can be traced to the Medicine, 1 Jen-Ai Rd, Taipei, Sec 1, 100 Taiwan.
expression point to a geometric organization of common occurrence of distinct pathways within *To whom correspondence should be addressed. E-mail:
cerebral fiber pathways similar to that of the brain- the same small volumes of tissue, or “path cross- [email protected]

1628 30 MARCH 2012 VOL 335 SCIENCE www.sciencemag.org

 REPORTS
brain through use of diffusion magnetic resonance ways. These correlations represent the mesoscale was acquired in whole-brain specimens ex vivo
imaging (MRI). Diffusion MRI creates multi- structure of connectivity, within the scope of which in rhesus, owl monkey, marmoset, and the pro-
dimensional contrast that is representative of the are the questions of whether cerebral pathways are simian galago, and in vivo in subjects (515 direc-
distribution of fiber orientations at each location discrete versus continuous and the detailed char- tions; pathways were computed with deterministic
in the tissue (21). Though of lower resolution than acter of the spatial organization of connectivity. streamline integration) (21). To demonstrate path-
tract tracing, diffusion MRI is noninvasive, appli- To probe the spatial relations between the path- ways’ structural relationships, we augmented in-
cable to humans and synoptic, and able to map the ways of the brain, we analyzed path-adjacency and teractive software to compute for any path the set
connectional anatomy of a single brain in its entire- path-crossing in four nonhuman primate species of all paths with which it shares one or more
ty, including spatial correlations between path- and in humans. Diffusion spectrum MRI (DSI) voxels, termed its path neighborhood.

assoc.

SLF3

Downloaded from www.sciencemag.org on June 10, 2012

seed

SLF3

A B

SLF1 SLF1

CC CC

CB

CB

C D

Fig. 1. Neighborhood structure of cerebral pathways. (A) MRI analog of tracer A region (red sphere, arrow) is selected in the white matter deep to the cin-
injection. A major association pathway of the rhesus forebrain SLF3 (blue) was gulate gyrus. The paths incident on this region (white) are near-orthogonal
identified by its termination in a selected region (yellow sphere in the parietal paths within the corpus callosum and SLF1. (D) The set of all paths incident on
lobule). A co-terminus association pathway (red) is noted. These pathways these paths in (C) forms a curved sheet of interwoven orthogonal pathways [(E)
appear as isolated structures. (B) The path neighborhood of SLF3 is identified: shows detail], including mutually parallel transverse paths of the callosum (red
a curved sheet of parallel paths (orange) that cross SLF3 nearly orthogonally. to green) and longitudinal paths (blue) within SLF1 and the cingulum bundle
(C to E) A path neighborhood in the rhesus frontal lobe (right lateral view). (C) (CB). No other path orientations were evident.

www.sciencemag.org SCIENCE VOL 335 30 MARCH 2012 1629

 REPORTS
The character of a typical cerebral path neigh- entirely consists of a single curved two-dimensional formed in the rhesus frontal lobe, as illustrated in
borhood is illustrated in Fig. 1. In Fig. 1A, an (2D) sheet of paths, all mutually parallel, trans- Fig. 1, C to E. The path neighborhood comprises
association pathway—rhesus monkey superior versely oriented, and all crossing SLF3 at nearly two sets—transverse callosal paths and longitudinal
longitudinal fasciculus-3 (SLF3)—is identified by right angles. paths of the cingulum and SLF1—that crossed like
its termination within the parietal lobe; this is the To investigate neighborhood structure indepen- the warp and weft of a fabric as a near-orthogonal
MRI analog of a tracer injection (25). So defined, dent of path identifications, we adopted the follow- grid. Thus, these paths formed a single biaxial sys-
this pathway appeared as an almost isolated struc- ing procedure: select a small region, identify its tem. This pattern was typical of cerebral white matter.
ture. In Fig. 1B, the path neighborhood of SLF3 incident paths, and compute the paths incident on The 3D structure of cerebral pathways was
was identified and was astonishingly simple. It these paths, their neighborhood. This was per- demonstrated through analysis of several path

Downloaded from www.sciencemag.org on June 10, 2012

A

C B

µm µm
D

Fig. 2. Grid structure of cerebral pathways. The grid structure of the pathways each depth [(B), arrows]. The paths within each sheet were orthogonal: of the
of the sagittal stratum in the rhesus occipital lobe was demonstrated by means longitudinal fronto-occipital fasciculus (FOF) (blue) or the transverse callsum
of neighborhood analysis. (A and B) Four seed regions were selected (spheres (red) and association system (green). Paths of the third mutually orthogonal
are indicated as superficial by red and green and as deep by yellow and direction (arrowhead), perpendicular to the local cortical surface, were noted.
orange), their paths identified (white), and neighborhoods computed (fronto- (D) Confocal microscopy of a sagittal slice oblique cut parallel to the lateral
occipital fasciculus, blue; callosal paths, red and orange). The interior view face of the sagittal stratum showed in-plane crossing of FOF (horizontal) and
along the axis of the structure (C) illustrates its character as an orthogonal callosal paths (vertical oblique), and the 2D autocorrelation map of this
grid. These neighborhoods comprised 2D sheets of closed quadrilaterals at microscopy, representative of the fiber orientations.

1630 30 MARCH 2012 VOL 335 SCIENCE www.sciencemag.org

 REPORTS
neighborhoods in a single region. Deep white mat- that parallel grid structure of pathways was not system [supporting online material (SOM) text
ter of the occipital lobe—the sagittal stratum—of limited to particular 2D surfaces but extended and fig. S1].
the rhesus monkey is shown in Fig. 2. Here, four throughout entire 3D volumes. Paths in the third In all present studies, cerebral path crossings
seed regions were selected whose incident paths nearly orthogonal axis were also seen and no formed well-defined 2D sheets. For example, in
and their neighborhoods formed the edges and diagonal paths observed. Similar 3D organization Fig. 2B paths crossed (arrows) so as to form closed
faces of a curved rectangular box, demonstrating was demonstrated in the highly curved midline quadrilaterals that are elements of well-defined

Fig. 3. Continuous grid struc-

ture of rhesus frontal lobes. (A)
The grid structure of subcortical
pathways was continuous with- A
in and between path neighbor-
hoods. Three neighborhoods were
constructed of the left arcuate
and right central sulci and the
midline. (Insets) All cortico-cortical
pathways were highly curved el-
ements in a sheet of interwoven

Downloaded from www.sciencemag.org on June 10, 2012

paths in two nearly perpendic-
ular orientations. The grid struc-
ture was continuous within and
between neighborhoods. Path
orientation was aligned with
gyral topography in the acuate
sulcus and but oblique in cen-
tral sulcus, with spiral trajecto-
ries (violet). (B) Major pathways
in deep white matter in the rhesus
frontal lobe (from top left), in-
cluding SLF 1-3, and the cingulum
bundle (CB) (blue) were compo-
nents of a single path-grid. Their
intersecting pathways (green, or-
ange, and red) are transverse,
parallel at all scales, and cross
orthogonally.

www.sciencemag.org SCIENCE VOL 335 30 MARCH 2012 1631

 REPORTS
2D surfaces. Geometrically, this configuration is struction, and tractography are purely local, limited of Fig. 2, A to C, that showed axons interweaving
highly exceptional (26, 27). Just as it is exceed- to single or to adjacent voxels, whereas the spa- and crossing in two axes, which was confirmed
ingly unlikely that several points fall on a straight tial correlations entailed in this pattern were long- by the orientation distribution of their spatial
line, it is similarly improbable that two families range and nonlinear, this structure could not be autocorrelation (28).
of curves in 3D have any 2D surfaces in com- attributed to technical artifacts related to the im- Grid structure of cerebral pathways was per-
mon. This sheet structure was found throughout aging of diffusion (SOM text and fig. S2). A vasive, coherent, and continuous with the three
cerebral white matter and in all species, orienta- histological counterpart of the crossing structure principal axes of development. In each region
tions, and curvatures. Moreover, no brain pathways of fiber pathways was observed with confocal in Fig. 3A, continuous grid structure is demon-
were observed without sheet structure. Further, be- microscopy: a tissue section through the rhesus strated across several regions in the frontal lobes
cause the processes of diffusion encoding, recon- monkey sagittal stratum cut parallel to the sheets in rhesus monkey, including left arcuate sulcus,

Downloaded from www.sciencemag.org on June 10, 2012

Fig. 4. Homologous cerebral grid structure in (A) galago, (B) marmoset, (C) of the anterior commissure and fornix (AC/FX) was resolved in all ex vivo studies
owl monkey, (D) rhesus monkey, and (E) human, left lateral views. Homologous [(A) to (D)]. In the rhesus monkey, grid structure is shown in gyri and sulci, in-
grid structures including those of the corpus callosum/cingulum bundle (CC/CB); cluding the principal, arcuate, and central sulci and the superior temporal
sagittal stratum and supra-Sylvian region were identified in all species, and that gyrus, continuous as grids with those of the adjacent deep white matter.

1632 30 MARCH 2012 VOL 335 SCIENCE www.sciencemag.org

 REPORTS
midline callosal region, and right central sulcus. We have found that the fiber pathways of the simpler species, suggesting that the evolutionary
The grid orientations in central sulcus were not forebrain are organized as a highly curved 3D emergence of discrete pathways parallels the in-
parallel to gyral topography but were oblique, grid derived from the principal axes of develop- creasing cerebral complexity in the primate lineage.
following spiral trajectories. This contrasted with ment. This structure has a natural interpretation. Functionally, pervasive cerebral organization with
the temporal lobes, where path and gyral ori- By the Frobenius theorem, any three families of parallel paths and similar lengths would naturally
entations were closely aligned (SOM text and curves in 3D mutually cross in sheets if and only support neural coding via spatial and temporal
fig. S3A). if they represent the gradients of three corre- coherence (16).
In rhesus deep white matter, the major frontal sponding scalar functions (26, 27). Accordingly, The grid structure of cerebral pathways has
association pathways were parallel elements of we hypothesize that the pathways of the brain implications for brain mapping: It suggests a sim-
a single grid (Fig. 3B, SOM text, and fig. S4). follow a base-plan established by the three chemo- plifying framework and natural coordinate sys-
These pathways were not highly distinct but con- tactic gradients of early embryogenesis (30). Thus, tem (35) for the description of brain structure, its
tinuous with their neighbors via sparse interposed the pathways of the mature brain presents an im- pathways, and connectivity; simplifies and con-
parallel paths. This was consistent in analysis at age of these three primordial gradients, plastically strains models of cerebral white matter; and indi-
multiple depths (SOM text and fig. S4). Thus, deformed by development. This could be tested cates that topographic organization is characteristic
major longitudinal pathways of the frontal lobe by obtaining diffusion MRI during cerebral em- of cerebral connectivity and not limited to a few
as well as longitudinal U-fiber pathways could be bryogenesis and assessing the derivation of path- major pathways. Of concern, present findings sug-
considered local condensations of a single system ways from the primordial directions. Grid structure gest that existing MRI tractography may under-
that spans the frontal lobe. The grid structure of should restrict and simplify axonal path-finding estimate sharp turning (36). It also provides a
pathways of the cerebral mantle was continuous compared with models that allow less constrained means to validate MRI tractography through con-

Downloaded from www.sciencemag.org on June 10, 2012

with those of the limbic system and basal ganglia and less correlated connectivity within and be- sistency with grid structure.
(SOM text and fig. S3B). tween cerebral areas. If grid structure guides con-
To investigate how pathways may change nectivity similar to the lane markers in a highway, References and Notes
course, we analyzed a known turn in a major ce- then navigation would be reduced from a general 1. W. His, Die Abh. Kon. Sacuh. Ges. Wiss. Math. Phys. Kl.
rebral pathway identified in the tracer studies of 3D problem to a far simpler question of when 15, 675 (1889).
2. S. Ramon y Cajal, Histologie du Systeme Nerveux de
Schmahmann and Pandya in rhesus monkey (17). to exit. In the brain, fibers growing in any axis l’Homme et des Vertebres (A. Malone, Paris, 1909).
Tracer and DSI showed bifurcation of a frontal would have a choice at each moment of just the 3. B. F. Kingsbury, J. Comp. Neurol. 32, 113 (1920).
projection pathway into transverse and dorso- four orthogonal directions perpendicular to their 4. H. Bergquist, B. J. Kallen, Comp. Neurol. 100, 627 (1954).
ventral components. In lieu of diagonals, path- course. Grid structure would increase the efficacy 5. M. N. Bernside, A. G. Jacobsen, Dev. Biol. 18, 537 (1968).
6. C. Nicholson, in The Central Nervous System of
ways were found to bifurcate and turn between of path orientation as a mechanism of axonal path- Vertebrates, R. Niewenhuys, H. J. ten Donkelaar, Eds.
axes (SOM text and fig. S5). finding (31, 32). Simultaneously, this structure (Springer-Verlag, Berlin, 1998), vol. 1, chap 5.
To investigate grid continuity in the cerebral supports incremental modification of connectivity 7. R. Nieuwenhuys, Brain Res. Bull. 57, 257 (2002).
hemisphere, widely separated path neighborhoods by geometric modification within broad continu- 8. M. J. Katz, R. J. Lasek, H. J. W. Nauta, Neuroscience 5,
821 (1980).
were constructed in owl monkey. Grid structure ous families of parallel paths. Thus, the grid or-
9. L. Puelles, Brain Behav. Evol. 46, 319 (1995).
was maintained at all scales, from the single ganization of cerebral pathways may represent a 10. E. A. DeYoe, D. J. Felleman, D. C. Van Essen,
voxel, to the lobe, to the hemisphere (SOM text “default connectivity,” on which adaptation of E. McClendon, Nature 371, 151 (1994).
and fig. S6). Continuity of grid structure between structure and function can both occur incremen- 11. J. H. Kaas, Anat. Rec. A Discov. Mol. Cell. Evol. Biol.
superficial gyral and deep cerebral pathways was tally in evolution and development, plasticity, 281A, 1148 (2004).
12. L. W. Swanson, Brain Res. Brain Res. Rev. 55, 356 (2007).
demonstrated through analysis of neighborhoods and function. 13. D. Badre, M. D’Esposito, Nat. Rev. Neurosci. 10, 659
at sequential depths (SOM text and fig. S4). Grid A grid organization of cerebral pathways has (2009).
structure in all three orthogonal axes was ob- been suggested in several contexts. Katz et al. (8) 14. M. M. Mesulam, Brain 121, 1013 (1998).
served in the centrum semiovale, including the hypothesized that a large-scale substrate of grid or- 15. K. Friston, Nat. Rev. Neurosci. 11, 127 (2010).
16. K. E. Stephan et al., Philos. Trans. R. Soc. Lond. B
longitudinal and transverse paths and dorso-ventral ganization is applicable to forebrain development. Biol. Sci. 356, 1159 (2001).
projection paths. This was observed in the rhesus Checkerboard organization of cortical Brodmann 17. J. D. Schmahmann, D. N. Pandya, Fiber Pathways of the
and in the human, in vivo and ex vivo (SOM text fields has been noted in visual areas of the tem- Brain (Oxford Univ. Press, Oxford, 2006).
and fig. S7). Further validation of grid structure poral lobe (1) and frontal motor areas (3) in mon- 18. M. Abeles, Corticonics (Cambridge Univ. Press,
Cambridge, 1991).
of cerebral pathways was obtained with an alter- keys. In humans, Badre and D’Esposito (13) have 19. D. Mumford, Biol. Cybern. 66, 241 (1992).
native MRI contrast mechanism of circular diffu- suggested hierarchic organization of the frontal 20. V. J. Wedeen, P. Hagmann, W. Y. Tseng, T. G. Reese,
sion contrast (SOM text and fig. S8) (29), and the cortex along a rostro-caudal axis. R. M. Weisskoff, Magn. Reson. Med. 54, 1377 (2005).
recognized triaxial structure of pathways of the The correlation between grid and topographic 21. V. J. Wedeen et al., Neuroimage 41, 1267 (2008).
22. J. Klingler, Schweiz. Arch. Neurol. Psychiatr. 36, 247
brainstem was shown with present methods orientations observed in the rhesus temporal lobe
(1935).
(SOM text and fig. S9). supports Van Essen’s hypothesis relating cerebral 23. W. J. S. Krieg, Connections of the Cerebral Cortex
Strong homology of deep cerebral grid struc- folding to fiber tension (33); however, the var- (Brain Books, Chicago, IL, 1963).
ture was found across all species studied (Fig. 4). iable relation of fiber structure and cortical fold- 24. J. Livet et al., Nature 450, 56 (2007).
These included the grid systems of the callosum, ing observed in the frontal lobe merits further 25. J. D. Schmahmann et al., Brain 130, 630 (2007).
26. W. P. Thurston, Three Dimensional Tand Geometry,
sagittal stratum, and supra-Sylvian pathways, as investigation. More recently, Clochoux et al. have S. Levy, Ed. (Princeton Univ. Press, Princeton, NJ, 1997)
well as the crossing of the fornix and anterior derived from its patterns of folding a spherical ch 3.7.
commissure in all species studied ex vivo at high coordinate system for the longitude and latitude 27. C. W. Meisner, K. S. Thorn, J. A. Wheeler, Gravitation
resolution. In the rhesus monkey, central and sub- of the human cerebral cortex (34). This coordi- (W. H. Freeman, San Francisco, 1973).
28. W.-Y. I. Tseng, V. J. Wedeen, T. G. Reese, R. N. Smith,
cortical grid structures (Fig. 4D), including those nate system seems generally congruent with that E. F. Halpern, J. Magn. Reson. Imaging 17, 31 (2003).
of the major frontal sulci (principal, arcuate, implied by the present grid structure of pathways, 29. V. J. Wedeen, J. G. Dai, W.-Y. I. Tseng, R. Wang, T. Benner,
central), fit together continuously like a jigsaw and investigation of their relationship may shed Proc. Intl. Soc. Mag. Reson. Med. 14, 851 (2006).
puzzle (SOM text and figs. S3A, S4, and S6, owl light on the linkage in structure and development 30. S. N. Sanson, F. J. Livesey, Cold Spring Harb. Perspect.
Biol. 1, a002519 (2009).
monkey). Thus, we hypothesize that the complex between cerebral gray matter and white matter. 31. L. Vasung et al., J. Anat. 217, 400 (2010).
connectivity of the cerebral mantle represents a Greater differentiation between pathways was 32. A. Chédotal, L. J. Richards, Cold Spring Harb.
continuous elaboration of the simpler core. observed in the more complex rhesus brain than in Perspect. Biol. 2, a001917 (2010).

www.sciencemag.org SCIENCE VOL 335 30 MARCH 2012 1633

 REPORTS
33. D. C. Van Essen, Nature 385, 313 (1997). Science Foundation, the National Institute Of Mental Health, or has a major competing interest to declare. All human studies
34. C. Clouchoux et al., Neuroimage 50, 552 (2010). the National Institutes of Health. This work is directly were obtained after signed informed consent, with review and
35. R. Nieuwenhuys, Brain Struct. Funct. 214, 79 supported by grants NSF PHY-0855161, NIH R01-MH652456, approval by the Institutional Review Board. Tissue specimens
(2009). P41 RR-023953, P41 RR-14075, and The Human Connectome were studied as discarded materials, with review and approval
36. P. J. Basser, S. Pajevic, C. Pierpaoli, J. Duda, A. Aldroubi, Project U01 MH093765. V.J.W. designed the methods of by the Institutional Subcommittee on Animal Care. Human
Magn. Reson. Med. 44, 625 (2000). MRI acquisition, reconstruction, and analysis; acquired and tissue specimens were studied as deidentified discarded
analyzed the data; discovered the grid structure; and wrote the materials, with review and approval by the Institutional
Acknowledgments: The authors thank L. C. Abbate, paper. W.-Y.I.T. collaborated in the development of theoretical, Review Board.
J. W. Belliveau, D. A. Feinberg, B. R. Rosen, S. A. Wedeen, and imaging, and anatomic ideas and methods. G.D. implemented
M. W. Weiner for reviewing this manuscript; J. D. Schmahmann, MRI acquisition techniques, including hardware and pulse Supporting Online Material
B. I. Shraiman, R. Turner, H. E. Stanley, and T. J. Brady for sequences; optimized protocols; and acquired ex vivo MRI. www.sciencemag.org/cgi/content/full/335/6076/1628/DC1
their comments; K. Mansfield of the New England Primate R.W. implemented analysis algorithms and created their user Materials and Methods
Center and L. Worthylake of the Oregon National Primate interface. J.H.K. and D.L.R. obtained and prepared specimens, SOM Text
Research Center for primate specimens; C. Devitt for myelin participated in analyses, and contributed to this manuscript. Figs. S1 to S9
stains; and M. P. Frosch for human specimens. The content F.M. optimized and obtained immunofluorescent microscopy.
is solely the responsibility of the authors and does not P.H. provided in vivo human studies. All authors discussed 13 October 2011; accepted 10 February 2012
necessarily represent the official views of the National the results and commented on the manuscript. No author 10.1126/science.1215280

Hierarchical Genetic Organization of 1000-fold increase in surface area relative to the

mouse brain (11). In effect, we sought to develop

Human Cortical Surface Area

a brain atlas of human cortical surface area that

Downloaded from www.sciencemag.org on June 10, 2012

was based entirely on genetic correlations, rather
than a priori structural or functional information.
Chi-Hua Chen,1 E. D. Gutierrez,2 Wes Thompson,1 Matthew S. Panizzon,1 Terry L. Jernigan,1,2 To delineate the genetic patterning of the cor-
Lisa T. Eyler,1,3 Christine Fennema-Notestine,1,4 Amy J. Jak,1,5 Michael C. Neale,6 tical area, we measured relative surface areal ex-
Carol E. Franz,1,7 Michael J. Lyons,8 Michael D. Grant,8 Bruce Fischl,9 Larry J. Seidman,10 pansion using cortical surface reconstruction and
Ming T. Tsuang,1,5,6 William S. Kremen,1,5,6*† Anders M. Dale1,4,11* spherical atlas mapping developed by Dale and
colleagues (12–14). We divided the area mea-
Surface area of the cerebral cortex is a highly heritable trait, yet little is known about genetic influences sured at each location by the total surface area in
on regional cortical differentiation in humans. Using a data-driven, fuzzy clustering technique with order to account for global effects. Using the twin
magnetic resonance imaging data from 406 twins, we parceled cortical surface area into genetic design, which compares monozygotic and dizy-
subdivisions, creating a human brain atlas based solely on genetically informative data. Boundaries gotic twins, we then estimated genetic correlations
of the genetic divisions corresponded largely to meaningful structural and functional regions; between different points on the cortical surface.
however, the divisions represented previously undescribed phenotypes different from conventional These genetic correlations represent shared ge-
(non–genetically based) parcellation systems. The genetic organization of cortical area was hierarchical, netic influences on relative areal expansion be-
modular, and predominantly bilaterally symmetric across hemispheres. We also found that the results tween cortical regions (15). Details of these methods
were consistent with human-specific regions being subdivisions of previously described, genetically have been previously described (8, 16). After
based lobar regionalization patterns. computing pairwise genetic correlations, we used
an unsupervised pattern recognition method—
s early as the 1950s, Bergquist and Kallen control in regional differentiation (2, 3). This fuzzy cluster analysis (17)—to demarcate the ge-

A postulated that the entire embryonic

brain is divisible into an anteroposterior
series of segmented neuromeres, each forming a
important finding prompted a search for similar
genetic regulatory organization in other regions
of the developing vertebrate brain (4). In par-
netic topography of cortical surface area based on
the genetic correlations of relative surface area
measures. To determine the appropriate number
complete ring around the brain’s longitudinal ticular, in the past decade the cerebral cortex has of clusters, we computed the widely used sil-
axis (1). Almost 40 years later, experimental data received substantial attention. Studies have shown, houette coefficient.
showed that many gene expression domains re- for example, that several signaling molecules and On the basis of the peak of the silhouette
spect segment boundaries in the embryonic ver- transcription factors are involved in establishing coefficients (fig. S1), we identified 12 natural
tebrate hindbrain, suggesting a role of genetic boundaries between mouse cortical regions (5, 6). clusters. These clusters correspond closely to
Animal data demonstrate that the regional or meaningful structural and functional regions
positional identity of cortical regions is defined (Fig. 1), even though the registration procedure
1
Department of Psychiatry, University of California, San Diego, by the combinatorial expression pattern of var- did not rely on prespecified anatomical landmarks;
La Jolla, CA 92093, USA. 2Department of Cognitive Science, ious genes controlling for regional differentia- rather, it makes use of the continuous pattern of
University of California, San Diego, La Jolla, CA 92093, USA. tion, each of which is expressed in a graded and surface curvature (13). In describing the sub-
3
Veterans Administration (VA) San Diego Healthcare System, restricted pattern with distinct spatiotemporal char- divisions, we use conventional labels, but these
San Diego, CA 92161, USA. 4Department of Radiology, Uni-
versity of California, San Diego, La Jolla, CA 92093, USA. 5VA
acteristics (7). Little is known, however, about only approximate the observed clusters. Subdi-
Center of Excellence for Stress and Mental Health, San Diego, the genetic patterning underlying the human cor- visions of the frontal cortex include the motor-
CA 92093, USA. 6Departments of Psychiatry and Human and tex. In our previous work (8), we showed that premotor, dorsolateral prefrontal cortex extending
Molecular Genetics, Virginia Commonwealth University, Rich- genetic patterning underlying the anteroposterior to the anterior and superior parts, dorsomedial
mond, VA 23219, USA. 7Center for Behavioral Genomics,
gradient and four basic cortical divisions of cor- frontal, and orbitofrontal (Fig. 1, clusters 1 to 4).
University of California, San Diego, La Jolla, CA 92093, USA.
8
Department of Psychology, Boston University, Boston, MA tical surface area demonstrated in mouse models Another cluster is found between the frontal and
02215, USA. 9Department of Radiology, Harvard Medical (7) also existed in the human cortex. Further- parietal cortices, extending from pars opercularis
School and Massachusetts General Hospital, Boston, MA 02115, more, region-specific cortical areal expansion in to the subcentral region, including the inferior
USA. 10Department of Psychiatry, Harvard Medical School, humans has been linked to specific genetic poly- pre- and post-central gyri (Fig. 1, cluster 5). The
Boston, MA 02215, USA. 11Department of Neurosciences, Uni-
versity of California, San Diego, La Jolla, CA 92093, USA. morphisms (9, 10). We sought to go beyond the temporal cortex includes the superior temporal,
*These authors contributed equally to this work.
fundamental commonalities that humans share posterolateral temporal cortex extending to tem-
†To whom correspondence should be addressed. E-mail: with other species and to investigate the genetic poral and parietal junction, and anteromedial tem-
[email protected] patterning specific to the human cortex with its poral cortex (Fig. 1, clusters 6 to 8). The parietal

1634 30 MARCH 2012 VOL 335 SCIENCE www.sciencemag.org

 Corrections & CLarifications

Erratum Post date 11 May 2012

Reports: “The geometric structure of the brain fiber pathways” by V. J. Wedeen et al.
(30 March, p. 1628). In (26), the book title should be Three Dimensional Geometry and
Topology. In (27), the first two authors should be spelled “C. W. Misner, K. S. Thorne.”

Downloaded from www.sciencemag.org on June 10, 2012

www.sciencemag.org SCIENCE erratum post date 11 may 2012 1