EFTA01122834.pdf

nature LETTERS physics PUBLISHED ONLINE:1 FEBRUARY 20161DO1: 10.1038/NPHYS3632 On the growth and form of cortical convolutions Tuomas Tallinenl * , Jun Young Chung2'3 ' , Francois Rousseau'', Nadine Girard5.6, Julien Lefevres and L. Mahadevan23.9.'" The rapid growth of the human cortex during development is of axonal tension driving gyriftcation10. At present, the most accompanied by the folding of the brain into a highly convoluted likely hypothesis is also the simplest one: tangential expansion structure' 3. Recent studies have focused on the genetic and of the cortical layer relative to sublayers generates compressive cellular regulation of cortical growth", but understanding stress, leading to the mechanical folding of the cortex'-" 2' -'s. This the formation of the gyral and sulcal convolutions also mechanical folding model produces realistic sizes and shapes of requires consideration of the geometry and physical shaping gyral and sulcal patterns" that are presumably modulated by brain of the growing brain'-'. To study this, we use magnetic geometry26, but the hypothesis has not been tested before with real resonance images to build a 3D-printed layered gel mimic three-dimensional (3D) fetal brain geometries in a developmental of the developing smooth fetal brain; when immersed in a setting. Here we substantiate and quantify this notion using both solvent, the outer layer swells relative to the core, mimicking physical and numerical models of the brain, guided by the use of cortical growth. This relative growth puts the outer layer into 3D magnetic resonance images (MRI) of a smooth fetal brain as a mechanical compression and leads to sulci and gyri similar to starting point. those in fetal brains. Starting with the same initial geometry, We construct a physical simulacrum of brain folding by we also build numerical simulations of the brain modelled taking advantage of the observation that soft physical gels swell as a soft tissue with a growing cortex, and show that this superficially when immersed in solvents. This swelling relative to also produces the characteristic patterns of convolutions over the interior puts the outer layers of the gel into compression, yielding a realistic developmental course. All together, our results surface folding patterns qualitatively similar to sulci and gyri". An show that although many molecular determinants control the MRI image of a smooth fetal brain at gestational week (GW) 22 tangential expansion of the cortex, the size, shape, placement (Fig. lb; see Supplementary Methods) serves as a template for a and orientation of the folds arise through iterations and 3D-printed cast of the brain. A mould of this form allows us to variations of an elementary mechanical instability modulated create a gel-brain (mimicking the white matter) that is then coated by early fetal brain geometry. with a thin layer of elastomer gel (mimicking the cortical grey The convoluted shape of the human cerebral cortex is the result matter layer). When this composite gel is immersed in a solvent of gyrification that begins after mid-gestation" (Fig. la); before the (see Supplementary Methods) it swells starting at the surface; this sixth month of fetal life, the cerebral surface is smooth. The first leads to superficial compression and the progressive formation sulci appear as short isolated lines or triple junctions during the of cusped sulci and smooth gyri in the cortex similar in both sixth month. These primary sulci soon elongate and branch, and morphology and relative timing to those seen in real brains (Fig. Ic secondary and tertiary sulci form, resulting in a complex pattern and Supplementary Movie I). We note that although the mechanical of gyri and sulci at birth. Some new sulci develop after birth, creasing or sulcification instability is due to the swelling-induced further complicating the pattern. Although the course and patterns compression, the effect is convoluted by the complex curvature of of gyrification vary across individuals, the primary gyri and sulci the initial shape. have characteristic locations and orientations". To obtain a more quantitative assessment of this process, we carry Gyrification is, however, not unique to humans, and also exists out a numerical simulation of the developing brain constructed in a range of primates and other species". It has evolved as an using the same 3D fetal brain MRI (Fig. Id) as an initial condition efficient way of packing a large cortex into a relatively small skull for the growth of a soft elastic tissue model of the brain. The model with natural advantages for information processing"•". Thus, assumes that a cortical layer of thickness h is perfectly adhered to a although the functional rationale for gyrification is clear, the white matter core and grows with a prescribed tangential expansion physiological mechanism behind gyrification has been unclear. ratio g. with both tissues assumed to be soft neo-Hookean elastic Hypotheses include gyrogenetic theoriesc° proposing that solids with similar elastic moduli (see Supplementary Methods). biochemical prepatterning of the cortex controls the rise of gyri, Combining these facts with the known overall isometric growth and the axonal tension hypothesisn proposing that axons in white of the brain' yields a differential-strain-based elastic model of matter beneath the cortex draw together densely interconnected brain growth that we solve numerically using custom finite-element cortical regions to form gyri. There is, however, no evidence of methods". For problem parameters, we note that from GW 22 prepatterning that matches gyral patterns, nor is there evidence to adulthood (Supplementary Fig. I) there is an approximately NATURE PHYSICS I ADVANCE ONLINE PUBLICATION I www.naluce.ccovinalurephy5ccs 2016 Macmillan Publishers Limited. All tights reserved. EFTA01122834 LETTERS NATURE PHYSICS DOI: 10.10313/NPHYS3632 a b to GW 22-23 :eepliedtec .br.: t. ev.e.ed gel-brain earth than later GW 33-34 r GW 36-37 3D-printed brain model NI ter moulds t=0 lcm / t I=0. 41 1 Nicie r seis, GW 22 I, d Simulation mesh from MRI image * at GW 22 7 .- Tangential expansion White matter- GW 34 Figure 11Physical mimic and numerical simulation of tangential cortical expansion. a, Gyrification of the human brain during the latter half of gestation (photographs from ref. 1, adapted with permission from Elsevier). b. A 3D-printed model of the brain is produced from a 3D MRI image of a smooth fetal brain and then used to create a pair of negative silicone moulds for casting. To mimic the constrained growth of the cortex, a replicated gel.brain (white matter) is coated with a thin layer of gel (cortex) that swells by absorbing a solvent (hexanes) over time / (h 4 min, 12 s9 min, b 2--16 min). C. The layered gel progressively evolves into a complex pattern of sulci and gyri during the swelling process. d. A simulation starting from a smooth fetal brain shows gyrification as a result of uniform tangential expansion of the cortical layer. The brain is modelled as a soft elastic solid and a relative tangential expansion is imposed on the conical layer as shown at left, and the system allowed to relax to its elastic equilibrium. 20-fold increase in brain volume (approximately 60 ml to 1,200 ml), Physically, the similar stiffness of the cortex and sublayers implies and a 30-fold increase in cortical area (approximately 80 cm' to that gyrification arises as a non-trivial combination of a smooth 2,400 cm2), whereas the expanding cortical layer changes little, with linear instability" and a nonlinear sukification instabilitr". a typical thickness of 2.5 mm in the undeformed reference state Sections of the physically and numerically simulated brains (the deformed thickness is about 3 mm). In physiological terms, shown in Fig. 2a,b exhibit a bulging of gyri and deepening of sulci we thus assume that tangential expansion during the fetal stage in a sequence resembling the observations from MRI sections. Our extends through the cortical plate (which has a thickness of about simulations of gyrification driven by constrained cortical expansion 1-1.5 mm at GW 22) and decays rapidly in the subplate (Fig. Id, allow us to also measure the gyrification index (GI, defined as left). The subplate diminishes during gyrification while the cortical the ratio of the surface contour length to that of the convex hull, plate thickens and develops into the cerebral cortex", so that in determined here from coronal sections as described in ref. 3). We the simulated adult brain the expanding layer corresponds to the see that there is a clear increase in the GI with developing brain cerebral cortex (which is about 3 mm thick in adults). volume in agreement with observations (Fig. 2c). The GI arising Our simple parametrization of brain growth leads to emergence from our numerical simulation reaches 2.5, matching observations of gyrification in space and time along a course similar to real of adult brains. A different measure of the GI based on the cortical brains (Fig. Id and Supplementary Movie 2): gyrification is initiated surface area rather than that of sections shows that the simulated through the formation of isolated line-like sulci (GW 26), which adult brain has a cortical area that is approximately four times the elongate and branch, establishing most of the patterns before birth exposed cortical area (Supplementary Fig. 1). For comparison, we (GW 40). After birth, brain volume still increases nearly threefold, also section our physical gel simulacrum that swells from an initial and during this time our model shows that the gyral patterns are unpatterned state (GI = 1.07, GW 22) and see that as a function modified mainly by the addition of some new bends to existing of swelling, the GI increases to about 1.55, a modest increase gyri in agreement with longitudinal morphological analyses". The associated with an approximately twofold increase in brain volume, characteristic spatiotemporal appearance of these convolutions— the latter state corresponding to roughly GW 30-34 (Fig. 2c and rounded gyri between sharply cusped sulci in a mixture of threefold Supplementary Fig. 2). The ultimate limiting factor in our physical junctions and S-shaped bends"—is a direct consequence of the experiments is the inability for our gel to swell and increase its mechanical instability induced by constrained cortical expansion. volume 20-fold like in fetal brains. 2 NATURE PHYSICS I ADVANCE ONUNE PUBLICATION I whewnaturecorninatutephymes K1 2016 Macmillan Publishers Limited. All rights reserved. EFTA01122835 NATURE PHYSICS DOI: 10.1038/NPHYS3632 LETTERS c 3.0 • Falai/children brains 4,1S2 2 :41 • Averageadult Imam • Numerical Imam 5606. • -,:y 44 • 23 • Gel-brain • • S6 /2 •e • --* 4040 t.5 31$ • • 40% w "a • 2.0 my 40 3•5 31 4 t_ 3436, 51 IS 30 29A Cf•29 22 Z25 'or8 29 1.0 - - , 2 50 100 200 400 800 1.600 Brain volume (m0 22 m GW 22 Adult Figure 2 [Sectional views of model brains during convolutional development. a. Planform and cross-sectional images of a physical gel-brain showing convolutional development during the swelling (folding) process that starts from an initially smooth shape (left panels) to a moderately convoluted shape (right panels). b, The coronal sections of the simulated brain (top panels) with comparisons to corresponding MRI sections3539. c, Gyrification index as a function of brain size for real brains (data from refs 3,40), a numerically simulated brain, and a physical gel.brain. Gestational week is indicated for fetal and children brains. The initial volume of the gel.brain ('34 ml) is scaled to match that of the simulated brain (-4=57 ml). Note that in the gel experiments only the outer layer swells and therefore the volume grows less than in real brains. Having seen that our physical and numerical experiments can that in highly curved regions the maximum compressive stress is capture the overall qualitative picture ofhow gyri form, we now turn perpendicular to the highest (convex) curvature. However, this does to the question of the role ofbrain geometry and mechanical stresses not hold at the ellipsoidal surfaces of the frontal and temporal in controlling the placement and orientations of the major and lobes; these lobes elongate and bend towards each other as a result minor sulci and gyri. In Fig. 3a, we show the field of the simulated of cortical growth (Supplementary Fig. 3; Supplementary Movie I compressive stress just before the primary sulci form. Although shows the analogue in our gel experiments). This reduces the cortical growth in our model is relatively uniform in space, the compressive stress in the direction ofelongation and bending, which curvature of the surface is not. This yields a non-uniform stress field in turn is reflected in the dominant orientations of the frontal and in the cortical layer. Thus, compressive stresses are reduced in the temporal gyri. vicinity ofhighly curved convex regions, so that the first sulci appear Although the global brain shape directs the orientations of at weakly curved or concave regions in our simulations, consistent the primary gyri, the finer details of the gyrification patterns are with observations in fetal brainstm. Furthermore, compression- sensitive to variations in the initial geometry. In Fig. 3c we see induced sulci should favour their alignment perpendicular to the that the patterns of gyri and sulci on a physical gel-brain exhibit largest compressive stress, and indeed directions of the largest some deviations from perfect bilateral symmetry. The hemispheres compressive stress in our model are perpendicular to the general are not identical in real brains either, but we note that in our orientations ofprimary gyri and sulci (Fig. 3a). Figure 3b shows that models artefacts from imaging, surface segmentation, and sample the first generations of sulci form perpendicular to the maximum preparation can cause the two hemispheres to differ more than in compressive stress in real and simulated model brains; this reality. The sequential patterns emerging from the folding (swelling) correlation is particularly clear for the primary sulci in real brains. and unfolding (drying) process also show some degree of variations Although the shape of the initially smooth fetal brain is (Supplementary Fig. 4), but this process is highly repeatable; indeed, described by the curvature of its surface, cortical growth eventually the resulting gyral patterns are found to be robust and reproducible couples the curvature and mechanical stress in non-trivial ways. on multiple repetition of the experiment with the same gel-brain In Supplementary Fig. 3 we compare the stress field and curvature sample (lower left of Fig. 3c). The folding patterns vary in detail at the cortical surface just before the first sulci form, and see across samples (Supplementary Fig. 4), but they share general NATURE PHYSICS I ADVANCE ONLINE PUBLICATION I www.n.stute.cominaturephysics 3 Kt 2016 Macmillan Publishers Limited. All rights reserved. EFTA01122836 LETTERS NATURE PHYSICS DOI: 10.1038/NPH 4' *IL ... Lel; b Left hemisphere Right hemisphere GW 34 29 GW 22_,>\4 GW it° I 1145_ \ ( ttfr AN.INItior- 4.51 4/ kf t, . 72% ment Apeoment Perpendicular to max. sires A A Parallel to mex. st ess Figure 31Mechanical stress orients convolutions. a, simulated stress field just before the onset of gyrification. Lines indicate the direction of maximum numerical simulation, in real brains(MRI,25 22,2934), compressive stress and colour indicates the magnitude of the stress (red, high). General directions of primary sulci are indicated using solid arrows. simplified scheme of the human brain" indicating the general orientations of sulci is shown at the top right. b, View of the first generations of sulci in a different brains for GW and and a physical gel brain on both hemispheres, with the former two dotted lines are perpendicular (angle>4S°)A <45°) showing the inner surfaces of cortical plates. The sulcal lines (blue, simulation; green, real brain; red, experiment) are superposed on the simulated vector fields of maximum compressive stress at GW and show that most sulci form perpendicular to the direction of maximum compressive stress (solid and and parallel (angle 4). to the stress vector field, respectively). The percentage of sulcal length perpendicular to the stress vectors is indicated in each case. c, folded gel-brain showing a certain lack of symmetry between the right and left hemispheres. The panel on the lower left shows nearly overlapping sets of sulcal lines obtained from three repeated tests with the same gel-brain (see Supplementary Fig. 5 similarities in terms of the alignment of sulci and gyri and the inter- sulcal spacing. These experimentally derived findings corroborate numerical results; in Supplementary Fig. we see the extent of variations associated with imperfections, by comparing the left and right hemispheres of the simulated brain when the simulation has gel model shows that we can capture the qualitative features of the folding patterns that are driven by a combination of surface swelling and surface geometry, setting the stage for how biology can build on this simple physical pattern-forming instability. Our numerical model is fully quantitative, based on parameters that been run forwards (folding) or backwards (unfolding) in time. In are known by simple measurements: cortical thickness, white and Supplementary Fig. 6 we compare the simulated folding of two grey matter tissue stiffness, brain growth and relative tangential different brains (starting from MRI images of different fetal brains) expansion of the cortex. Together, they show that gyrification and see that the patterns in both brains are consistent with the is an inevitable mechanical consequence of constrained cortical scheme depicted in Fig. 3a. 4 We quantify our qualitative comparison of the similarities between real and simulated brains in Fig. using morphometric techniques. Conformal mapping of the curvature vector fields (see Supplementary Methods) allows for a visual (Fig. 4a) as well as expansion; patterned gyri and sulci that are consistent with observations arise even in the absence of any patterning of cortical growth. Furthermore, although cortical expansion and brain shape couple to guide placement and orientation of gyri, the finer details of the patterns, on scales comparable to the cortical thickness, quantitative comparison (Fig. 4b), which indicates similar average are sensitive to geometrical and mechanical perturbations. Real alignments of curvature fields, and thus folds, in both the simulated brains are likely to have small inter-individual variations in shape, and real brains. As a final comparison, in Fig. 4c,d we identify tissue properties and growth rates, and the sensitivity of mechanical visually many of the named primary gyri' from the simulated brains folding to such small variations could explain the variability and compare to a real fetal brain at a similar stage, and note that our of gyrification patterns, although the primary convolutions are simulations can capture many of the trends quantitatively. consistently reproducible in their location and timing. All together, by combining physical experiments and numerical Our organ-level approach complements the recent emphases on simulations originating with initially smooth 3D fetal brain the many molecular determinants underlying neuronal cell size geometries, we have quantified a simple mechanical folding scenario and shape, as well patterns in their proliferation and migrationtn. for the development of convolutions in the fetal brain. Our physical 4 I These variables are functions of time and space in the cortex NATURE PHYSICS ADVANCE ONLINE PUBLICATION I wwvanaturexaminaturephysics al 2016 Macmillan Publishers Limited. All rights reserved. EFTA01122837 NATURE PHYSICS DOI: 10.1038/NPHYS3632 LETTERS a b 90° 80° 70° 60° 50° 40° 30' 20° 10° 0° Map of angles between the vectors of least curvatite Average Pearson angle correlation GW 29 (Ieft/right) 23°/26° 0.84/0.81 Guy 34 (left/right) 32°/30° 0.62/0.65 Precentral d Superior Precentral frontalgyrus gyros Postcentral Superior Postcentral 8Yrus gyrus frontal gyros gyros Middle , — frontal gyms Middle \ Suplartiarginal % SJoramarginal Mous frontal gv.L. gyros Angular / gyrus Aular • u < I rUS Inlet c frontal gy: \Middle Superior temporal gyros Inferior temporal gyros frontal gyros Superior Inferior Inferior temporal gyros temporal gyros Middle temporal gyros temporal gyrus Figure 4 I Comparison of real and simulated folding patterns. a, Conformal spherical mappings (see Supplementary Methods) of inner cortical surfaces of real and simulated fetal brains. b, A map of angles between the vectors of the least absolute curvature of the real and simulated brain surfaces. The angle field is computed in the spherical domain and mapped back on the real brain. In grey regions the curvature directions match and in green regions they are mismatched. Average angles and Pearson correlations between real and simulated curvature fields at GW 29 and 34 are given in the table (details in Supplementary Methods). c, All notable gyri have formed by GW 37 as identified on a real fetal brain (adapted from ref. 1 with permission from Elsevier). d, Analogous regions shown in a simulated brain driven by constrained cortical expansion. In both cases, the colouring is based on visual identification of the major gyri. and the sub-ventricular zone, but ultimately feed into just two References physical parameters for a given initial brain shape: relative cortical I. Griffiths. P. D., Reeves. M.. Morris. P. & Larroche. J. Atlas ofFetal andNeonatal expansion and the ratio of cortical thickness/brain size. By Brain MR Imaging (Mosby. 2010). 2. Girard, N. & GambarellL D. Normal Fetal Brain: Magnetic Resonance Imaging. grounding our study on observations of fetal brain development, An Atlas with Anatomic Correlations (Label Production. 2001). where these parameters are known, our work provides a quantitative 3. Armstrong. E.. Sclelcher. A.. Omran, H.. Curtis. M. & Zilles, K. The ontogeny basis for the proximal biophysical cause of gyrification. Our of human syrifkation. Cereb. Cortex 5.56-63 (1995). results can serve as a template for physically based morphometric 4. Sun. 1'. & Hevner. R. F. Growth and folding of the mammalian cerebral cortex: studies that characterize the developing geometry of the brain from molecules to malformations. Nature Rev. Neurosei. 15,217-232 (2014). surface"-" in terms of variations in initial brain shape and cortical S. Lui. J. H.. Hansen. 1). V. & Kriegstein. A. R. Development and evolution of the expansion. Additionally, our study sets the biophysical basis for human neocortex. Ctrl 146,18-36 (2011). 6. Rae. B. et al. Evolutionarily dynamic alternative splicing of GPRS6 regulates an understanding of a range of functional neurological disorders regional cerebral cortical patterning. Science 343.764-768 (2014). linked to structural cortical malformations"6 that are associated 7. Rash, B. G..Tomasi. S.. Lim, H. D., Suh. C. Y. & Vaccarino. F. M. Conical with neurogenesis and brain size (for example, ASPM and CENPJ gyrificatlon Induced by fibroblast growth factor 2 in the mouse brain. for microcephalr) and cell migration and cortex thickness (for 1. Neuroscl. 33,10602-10814(2013). example, GPR56 for polymicrogyria° and RELN for lissencephaly1tl). & ReIllo. L. de Juan Romero, C.. Garcia-Cabezas. M. A. & Borrell. V. A role of intermediate radial glia In the tangential expansion of the mammalian cerebral With this information, a natural next step is to link the molecular cortex. Cad.. Cortex 21,1674-1694(2011). determinants to the macroscopic mechanics quantitatively, and ask 9. Richman. D. P. Sienna. IL M.. Hutchinson. J. W. & Civilness. V. S. Jr how brain shape and function are mutually co-regulated. Mechanical model ofbrain convolutional development. Science 189. 18-21 (1975). Received 27 September 2015; accepted 9 December 2015; 10. Xu, G. et at. Axons pull on the brain. but tension does not drive cortical folding. published online 1February 2016 1. Biornerh. Eng. 132.071013 (2010). NATURE PHYSICS I ADVANCE ONLINE PUBLICATION I www.nalure.ccen/nalurephysics 5 2016 Macmillan Publishers Limited. All rights refereed. EFTA01122838 LETTERS NATURE PHYSICS DOI: 10303B/NPHYS3632 11. Toro. R. & Burned, Y. A morphogenetic model for the development of cortical 32. Tallinen. T.. Biggins. ). & Mahadevan. L Surface suki in squeezed soft solids. convolutions. Cents. Cortex 15.1900-1913 (2005). Phys. Rev. Lett 110.024302 (2013). 12. &ie.,. et at A computational model of cerebral cortex folding./ Theor. Biol. 33. Toro. Let at. Brain size and folding of the human cerebral cortex. Crash Cortex 261,467-478 (2010). 18.2352-2357(2008). 13. Budday, S.. Raybaud. C. & Kuhl, E. A mechanical model predicts morphological 34. Gernunaud, D. et at Larger is [wisher: spectral analysis of gyrification abnormalities in the developing human brain. Sta. Rep. 4,5644 (2014). (SPANGY) applied to adult brain size polymorphism. Neurobnage 63, 14. Bayly. P. V. Okamoto. It ).. Xu. G., ShL Y. & Taber, L. A. A cortical folding 1257-1272 (2012). model incorporating stress-dependent growth explains gyral wavelengths and 35. Lefivre. ). ri al. Are developmental trajectories of cortical folding comparable stress patterns in the developing brain. Phys. Biol. 10,016005 (2013). between cross-sectional datasets of fetuses and preterm newborns? Cereb. 15. Tallinen. T.. Chung. I. Y.. Biggins. I. S. & Mahadevan. L Gyrification from Cortex h((p:lfdx.doi.org/1o.1093/cercor/bhv123 (2015). constrained cortical expansion. Prot. Nati Mad. Sci. USA 111. 36. Aronka. F.. Becker. A. I. & Spreafico. R. Malformations of cortical 12667-12672 (2014). development. Brain Pathol 22,380-401 (2012). 16. Ono. ht. Kubik. S. & Abernathey. C D. Atlas of the Cerebral Suter (Georg 37. Bond. I. et al. A centrosomal mechanism involving CDK5RAP2 and CENP) Thkme Verlag. 1990). controls brain size. Nature Genet. 37.353-355 (2005). 17. Striedter, G. F. Principles ofBrain Evolution (Sinauer Associates. 2005). 38. Hong. S. E. a al. Autosomal recessive lissencephaly with cerebellar hypoplasia 18. Meta. B. & Herculano-Houzel. S. Cortical folding scales universally with is associated with human RELN mutations. Nature Genet 26.93-96 (2000). surface area and thickness. not number of neurons. Science 349,74-77 (2015). 39. Serag, A. et at. Construction of a consistent high-definition spatio-temporal 19. Wes, K.. Palomero-Gallagher. N. & Amunts. K. Development of cortical atlas of the developing brain using adaptive kernel regression. &tura:A:age 59, folding during evolution and ontogeny. Trends Neurinci. 36,275-284 (2013). 2255-2265(2012). 20. Welker. W. Why does cerebral cortex fissure and fold: a review of determinants 40. Zilles. K.. Armstrong, E. Sclekher. A. & Kretschmann. H. The human pattern of gyri and thick Cereb. Cortex 8,3-136 (1990). in gyrification in the cerebral cortex. Anat. Embryo!. 170,173-179 (1988). 21. van Essen. D. C. A tension based theory of morphogenesls and compact wiring in the central nervous system. Nature 385,313-318 (1997). Acknowledgements 22. Ronan. Let at Differential tangential expansion as a mechanism for cortical We thank CSC— Center for Science. Finland. for computational resources and gyrifkation. Cereb. Cortex 24,2219-2228 (2014). J. C Weaver for help with 3D priming, This work was suppotted by the. cademy of 23. Holland. M. A.. Miller. K. E. & Kuhl, E. Emerging brain morphologies from Finland (T.T.). Agence Nationale de la Recherche (AMR-124503-0Di -0l."Modegy") axonal elongation. Ann. Bionsed. Eng. 43,1640-1653 (2015). (N.C. and IL). the Wyss Institute for Biologically Inspired Engineering (INC. . and LAL). 24. Striedter, G.E. Srinkasan. S. & Monuki. E. S. Cortical folding: when. where. and fellowships from the MacArthur Foundation and the Radcliffe Institute (L.M.). how and why? Anna.. Rm. Neurosci. 38,291-307 (2015). 25. Bayly. P. V. Taber. L A. & Kroenke. C. D. Mechanical forces In cerebral cortical folding: a review of measurements and models.!. Aka. Behar: Biommf. 29, Author contributions T.T.. !ICI:. and L.M. conceived the model and wrote the paper. T.T. developed and 568-581 (2013). performed the numerical ',mutations. developed and performed the physical 26. Todd, P. H. A geometric model for the cortical folding pattern of simple folded experiments. LL developed and performed the moiphometric analyses. F.R.. N.G. and brains. I. Theor. Biol. 97,529-538 (1982). It.. provided Mitt images and provided feedback on the manuscript. T.T. and LM. 27. Meng, Y., Li. C.. Lin. W. Gilmore. ). H. & Shen. D. Spatial distribution and coordinated the project longitudinal development of deep cortical sulcal landmarks in infants. Neurobnage 100.206-218 (2014). 28. Li. K. et at Gyral folding pattern analysis via surface profiling. Neurolmage 52, Additional information Supplementary information is available in the online version of the paper. Reprints and 1202-1214 (2010). permissions information is avadable online at wwwnaturc.conarreprints. 29. Blot. M. A. Mechanics ofburensental Deformations (lohn Wiley. 1965). Correspondence and requests far materials should be addressed to T.T. or L.51. 30. Hohlfeld. E. & Mahadevan, L Unfolding the sulcus. Phu. Reif Lett 106, 105702 (2011). 31. Hohlfeld. E. & Mahadevan. L Scale and nature of sulcification patterns. Phys. Competing financial interests Rest Left. 109,025701 (2012). The authors declare no competing financial interests. 6 NATURE PHYSICS I ADVANCE ONUNE PUBLICATION I vi 2016 Macmillan Pulettsliet> Limited. All tights relayed EFTA01122839