Cortical cell and neuron density estimates in onechimpanzee hemisphereChristine E. Collinsa,b, Emily C. Turnera, Eva Kille Sawyerc, Jamie L. Reeda, Nicole A. Younga,d, David K. Flahertye,and Jon H. Kaasa,1

aDepartment of Psychology, Vanderbilt University, Nashville, TN 37240; bEli Lilly and Company, Indianapolis, IN 46285; cNeuroscience Graduate Program,Vanderbilt University, Nashville, TN 37240; dDepartment of Neuroscience, Neuroscience Research Institute, The Ohio State University College of Medicine,Columbus, OH 43210; and eVanderbilt University Medical Center, Flow Cytometry Shared Resource, Nashville, TN 37232

Contributed by Jon H. Kaas, December 8, 2015 (sent for review November 10, 2015; reviewed by Patrick R. Hof and Chet C. Sherwood)

The density of cells and neurons in the neocortex of many mammalsvaries across cortical areas and regions. This variability is, perhaps,most pronounced in primates. Nonuniformity in the composition ofcortex suggests regions of the cortex have different specializations.Specifically, regions with densely packed neurons contain smallerneurons that are activated by relatively few inputs, thereby pre-serving information, whereas regions that are less densely packedhave larger neurons that have more integrative functions. Here wepresent the numbers of cells and neurons for 742 discrete locationsacross the neocortex in a chimpanzee. Using isotropic fractionationand flow fractionation methods for cell and neuron counts, weestimate that neocortex of one hemisphere contains 9.5 billion cellsand 3.7 billion neurons. Primary visual cortex occupies 35 cm2 ofsurface, 10% of the total, and contains 737 million densely packedneurons, 20% of the total neurons contained within the hemisphere.Other areas of high neuron packing include secondary visual areas,somatosensory cortex, and prefrontal granular cortex. Areas of lowlevels of neuron packing density include motor and premotor cortex.These values reflect those obtained from more limited samples ofcortex in humans and other primates.

primate neocortex | neuron density | visual cortex | isotropic fractionator |flow fractionator

The present study is part of our ongoing effort to determinehow human brains are both similar and different from the

brains of our closest living primate relatives (e.g., refs. 1 and 2).Present-day chimpanzees and bonobos diverged from the line ofhominins that led to modern humans some 6 million years ago(3). The modern chimpanzee brain is similar in size to our ear-liest hominin ancestors (4). In contrast, the brains of laterhominins rapidly increased in size about 1.5 million years ago,resulting in modern humans having brains about three times thesize of early hominins and present-day chimpanzees. Similarly,the neocortical sheet of one cerebral hemisphere is almost threetimes larger in humans (average, ∼975 cm2 in humans; e.g., refs.5 and 6) than in chimpanzees (∼341 cm2). This increase alsoappears to be reflected in an increase in the number of areas thatare architectonically and physiologically modified for differentspecialized functions. Such specializations are reflected in a widerange of neuron types and laminar and sublaminar patterns ofregional and areal organization in larger primate brains, espe-cially those of humans (7). It would be surprising if averageneuron size and neuron packing densities did not vary acrossareas, especially in large-brained primates. However, the claimhas been made that neuron densities are uniform across corticalareas and mammalian species, except for a twofold increase inprimary visual cortex (V1) in some primates (8, 9). This con-clusion is contrary to the results of several previous studies (e.g.,ref. 10), especially those of our recent reports on neuron den-sities across the cortical sheet in several primate species (1, 2).The primates examined to date (baboons, macaques, and gala-gos) had much higher neuron densities in V1. Neuron densitieswere also higher than average in other sensory areas, and

considerably lower in primary motor cortex (M1), with more pro-nounced differences appearing in larger-brained primates. Asthere is a relationship between neuron density in cortex andaverage neuron size (1, 11), and larger neurons have largerdendritic arbors with more synaptic contacts (12), larger neuronsare more suitable for integrative functions, whereas smaller neu-rons preserve information and more faithfully reflect the sourcesof their activation (13–15).Here, we extend our observations on areal neuron densities in

strepsirrhine galagos and anthropoid monkeys to the full neo-cortical sheet of a chimpanzee (Pan troglodytes). We provide themost detailed assessment of cortical neuron densities across thecortical sheet for any species, having dissected the chimpanzeecortex into 742 tissue pieces, providing roughly six times moredetail than has ever been reported for the cortex of any primatespecies (1) and nearly 100 times more evaluations than a typicalquantitative study (10). This comprehensive approach allows usto demonstrate significant regional differences in neuron densitiesthat may be small enough to be lost in the noise of more limitedstudies. Although this has not yet been attempted for a humanbrain, direct comparisons will be possible in future studies.

ResultsThe total surface area of the cerebral cortex of the right hemi-sphere of this chimpanzee was 341 cm2, although that could be aslight underestimate resulting from a loss of some neocortexalong the outer margins of the cortical sheet. The neocortexcontained an estimated 9.51 billion cells (count includes all celltypes identified by DAPI; e.g., glial, neuronal, etc.), of whichabout 3.71 billion, or 39%, were neurons (only cells that stainpositive for neuronal nuclear antigen). The cell and neuron

Significance

Chimpanzees are our closest relatives, and understanding theorganization of their brains can help us understand our ownevolution. Here we present a detailed examination of cell andneuron densities across the chimpanzee cortex. We found simi-larities to other mammals, including primary sensory areas withhigh neuron densities and a trend of decreasing neuron densitiesalong the posterior to anterior axis of the cortex. However, wealso found a prefrontal region with anomalously high neurondensity that disrupts the trend of decreased neuron densities infrontal brain regions. The data reported here allow valuablecomparisons among the brains of our close relative and those ofhumans and other primates.

Author contributions: C.E.C. and N.A.Y. designed research; C.E.C., E.C.T., E.K.S., N.A.Y.,and D.K.F. performed research; E.C.T., E.K.S., and J.L.R. analyzed data; and C.E.C., N.A.Y.,and J.H.K. wrote the paper.

Reviewers: P.R.H., Mount Sinai School of Medicine; and C.C.S., The GeorgeWashington University.

The authors declare no conflict of interest.1To whom correspondence should be addressed. Email: jon.h.kaas@vanderbilt.edu.

740–745 | PNAS | January 19, 2016 | vol. 113 | no. 3 www.pnas.org/cgi/doi/10.1073/pnas.1524208113

Dow

nloa

ded

by g

uest

on

Mar

ch 1

4, 2

020

counts for individual pieces of tissue varied considerably (Fig. 1).Some of this variation may be random, possibly reflecting biasesbased on tissue distortions during flattening or other processingfactors. Nevertheless, there were large, statistically significantregional and areal differences that were reflected by the com-parisons of mean counts (see Fig. 2 A and B for statistics).

Visual Areas of Cortex. The estimated surface area of V1 from theright hemisphere of the present chimpanzee was 35.04 cm2,about 10% of the total neocortical surface. This estimate wasbased on the piece of cortex that was separated along the marginof the lunate fissure, which marks the rostral border of V1 (16,17). We compensated for a slight error in that cut by adding anarrow margin of tissue from the larger main piece, as shown inred in Fig. 1 B and C. The estimated volume of V1 (5.30 cm3) isclose to previous estimates of 5.52 cm3 (18) and 4.64 cm3 (19), aswell as those from the other hemisphere of our chimpanzeebased on measures from serial brain sections (see ref. 20; 4.8 cm3).Cell and neuron packing densities were greatest in V1. At only10% of the neocortical surface, V1 contained just more than 1.13billion cells, of which 737 million, or 65%, were neurons. Celldensities in V1 averaged about 32 million cells per square cen-timeter of surface (see Fig. 2B, which shows estimated means) or138 million cells per gram of tissue. The average neuron densitywas 21 million neurons per square centimeter of cortical surface(Fig. 2A), or 89 million neurons per gram of cortical tissue. Celland neuron densities varied across the 77 pieces of V1, but mosthad neuron densities near or higher than 20 million per squarecentimeter. The packing densities of neurons in V1 were 1.2, 2.1,3.3, and 3.5 times greater than neuron densities in secondaryvisual cortex (V2) and somatosensory, motor, and premotorcortices, respectively. Total cell densities were less variableacross cortex, but were still 1.3, 1.5, 1.3, and 1.4 timesgreater than in V2, somatosensory, motor, and premotor areas,respectively.The higher cell and neuron packing densities distinguished V1

from all other areas of cortex. Although the full extent of V2 in

chimpanzees is somewhat uncertain, the location shown in Fig. 1conforms to expectations. This estimate of V2 comprises 28.69 cm2

of cortical surface, or 8% of the total neocortical surface. Thetotal cells in V2 were just more than 788 million, of which 396million, or 50%, were neurons. The average cell densities in V2were 27 million per square centimeter of surface (Fig. 2B), or115 million cells per gram of tissue. The average neuron densityin V2 was 14 million neurons per square centimeter of corticalsurface (Fig. 2A), or 59 million neurons per gram of corticaltissue, a drop of 7 million neurons per square centimeter or 30million neurons per gram from V1, but still higher than othercortical regions. Neuron densities in V2 were 1.2, 2.1, and 2.2times greater than those in somatosensory, motor, and premotorcortices, respectively.Cortex just rostral to V2 is expected to contain a number of

visual areas that have been identified in macaque monkeys andother primates, including visual areas V3 and V4 and dorso-medial visual area (21, 22). These and other caudal visual areaswould all have higher than average cell, and especially neuron,packing densities, although not as high as V1 and V2. These datasupport the conclusion that retinotopically organized visual areasin chimpanzees have high neuron packing densities.

Somatosensory Areas. Somatosensory areas had higher than av-erage neuron packing densities relative to nonprimary sensoryareas. Anterior parietal cortex of chimpanzees and other an-thropoid primates includes areas 3a, 3b, 1, and 2 (23), with area3b being the primary tactile area. The region designated as so-matosensory cortex in Fig. 1 likely contains most of areas 3b, 1,and 2. As we separated rostral from caudal cortex along thedepth of the central sulcus, our “somatosensory cortex” likelyincludes the caudal half of area 3a, with the rostral half includedin motor cortex. The most lateral part of somatosensory cortexmay have been excluded from our designated block, which was16.6 cm2 of surface, or 4.86% of neocortex. Cell densities in thesomatosensory block averaged 22 million cells per square cen-timeter of surface (Fig. 2B), or 91 million cells per gram of tissue.

19 - 21 22 -14

25 - 27 > 40 31 - 33

28 - 30 < 18 34 - 36

37 - 39

Million cells/cm2

BV1

V2Somatosensory

Motor Premotor Frontal

DP

A

6 - 88 -10

10 - 12 > 20 14 - 16

12 - 14 < 6 16 - 18

18 - 20

Million neurons/cm 2

C

Fig. 1. (A) Lateral view of intact, adult female chimpanzee brain. (B) Complete dissection map, illustrating the 742 pieces of tissue dissected from thechimpanzee cortex. Total cell density in millions of cells per square centimeter of cortical surface is illustrated on the cortex flat map, with the darkest shadingindicated areas of high cell density and lighter shading indicating low cell density. The total cell density includes all types of cells in cortex; that is, neurons,glial cells, and epithelial cells. Areal boundaries are estimated for V1, V2, somatosensory (areas 3b, 3a, 1, and 2), motor (areas 4 and 6), and premotor cortex.(C) Complete dissection map illustrating the 742 pieces of tissue dissected from the chimpanzee cortex. Total neuron density in millions of cells per squarecentimeter of cortical surface is illustrated on the cortex flat map, with the darkest shading indicating areas of high neuron density and lighter shadingindicating low neuron density.

Collins et al. PNAS | January 19, 2016 | vol. 113 | no. 3 | 741

NEU

ROSC

IENCE

Dow

nloa

ded

by g

uest

on

Mar

ch 1

4, 2

020

Neuron numbers averaged 10 million per square centimeter(Fig. 2A), or 41 million neurons per gram of tissue. The so-matosensory section of cortex contained 362 million cells, ofwhich 164 million, or 45%, were neurons.

Motor and Premotor Areas. Our block of tissue pieces designatedas motor in Fig. 1 likely contains most of M1, including thecortex of the rostral bank of the central sulcus and cortex of thecaudal half of the precentral gyrus (17, 23–25). The more rostralblock of pieces likely includes dorsal and ventral divisions ofpremotor cortex. Our motor cortex block consisted of 24.97 cm2

of cortical surface with 625 million cells, of which 163 million, or27%, were neurons. Cell packing densities averaged 25 millioncells per square centimeter of cortical surface (Fig. 2B), or 91million cells per gram of tissue. Neuron packing densities aver-aged 7 million neurons per square centimeter of surface (Fig.2A), or 24 million neurons per gram. Nearly all of the individualpieces of M1 had low neuron densities, in the 6–7 million range.Thus, M1 was characterized by low neuron packing densities andmoderate cell packing densities.Our premotor block included 24.46 cm2 of cortical surface,

with 636 million cells and 172 million neurons, for a compositionof 26% neurons. Cell densities in premotor cortex averaged 25million per square centimeter of cortical surface (Fig. 2B), or 88million per gram of tissue. The average neuron packing densitywas 6 million neurons per square centimeter of cortical surface(Fig. 2A), or 23 million neurons per gram of cortical tissue. Thus,cortex in the premotor block had low levels of neuron packing,with little variability across tissue pieces, and closely matched thelow neuron packing densities across motor areas.

Prefrontal Cortex. Prefrontal cortex generally consists of the cortexrostral to premotor cortex, and it has several subdivisions includinga large dorsolateral region of granular frontal cortex and adjoiningregions of orbital frontal and medial frontal cortex (26–28). Arealvariations in neuron densities across prefrontal cortex have pre-viously been shown in macaques (29). Overall, the prefrontal blockof tissue denoted in Fig. 1 B and C has pieces of tissue with highercell packing densities and neuron packing densities than those ineither motor or premotor cortex. Most of this cortex with highercell and neuron packing densities would be considered granularfrontal cortex. Pieces of cortex with lower values were located alongthe margins of frontal cortex, including medial frontal and orbitalfrontal regions. However, we distinguished a dorsomedial block ofcortex within the presumptive region of granular frontal cortex (Fig.1C) as having higher neuron densities than other prefrontal regions.Pieces of cortex in this block had average neuron packing densitiesof 11 million per square centimeter. Total cell densities averaged 30million per square centimeter, or 106 million per gram, with 36%neurons. This high-density block of tissue covered 13.11 cm2 with515 cells and 181 million neurons.

The Anterior to Posterior Pattern. To examine whether the chim-panzee brain showed uniform cell and neuron packing densitiesacross cortex (or a linear decrease in density beyond V1 fromposterior to anterior), we examined densities in pieces of cortexacross the anterior-to-posterior (A-P) dimension, using linearregression and curve estimation, testing multiple model func-tions. Recently, there have been a number of descriptions ofneuron packing densities across the cortical sheet in monkeysand other mammals that revealed a trend from low packingdensities to high neuron packing densities from anterior to poste-rior cortex (11, 30–33). When we assessed how cell and neurondensities varied in the selected cortical regions across the A-P di-mension, using generalized linear modeling with robust estimators,most cortical areas showed significant differences in estimatedmeans of neuron and cell density (Fig. 2 A and B). In addition, thelocations of samples within cortical regions were better predictorsthan A-P coordinates alone of cell density (cortical area, P = 6.4 ×10−10; A-P, P = 0.818) and of neuron density (cortical area, P =1.0 × 10−4; A-P, P = 0.835).When the density values for individual pieces were plotted by A-P

location and color-coded for tissue block of origin, the average curve

2 ) mc /snoilli

m( aera ecafrus yb sllec fo rebmun nae

M

* * * *

****

** ** ** *

Mea

n nu

mbe

r of n

eur o

ns b

y su

rfac

e ar

ea (m

illio

ns/ c

m )2

**** *

* **** ***

*** **** *

** *

B

A

Fig. 2. Pairwise comparison results of estimated marginal means based onthe original scale of the dependent variable, either neuron (A) or cell (B)density in square centimeter, with mean differences significant at P < 0.05and robust SE bars shown. All significant pairwise comparisons are high-lighted by the significance bars. (A) V1 estimated marginal means of neurondensity are significantly higher than all other cortical regions, whereas V2 issignificantly higher than motor, premotor, and somatosensory blocks. So-matosensory predicted values are also significantly higher than motor andpremotor blocks. Motor and premotor cortices do not significantly differfrom one another, but are both significantly lower than all other areas ofcortex. (B) V1 and the frontal region estimates show no significant differ-ences in cell density, but each region independently contains higher meansthan every other area. Somatosensory, premotor, and motor cortex do notsignificantly differ from one another.

742 | www.pnas.org/cgi/doi/10.1073/pnas.1524208113 Collins et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 1

4, 2

020

for neuron density had higher values in posterior cortex and adownward slope to lower values in anterior cortex (Fig. 3A). Al-though an A-P gradient is apparent in the array of values for indi-vidual pieces, it clearly does not correspond to a simple linear A-Pgradient. This type of plotting allows V2 values to mix with V1 valuesto contribute to a steep rise in the posterior slope that obscures theclear difference between V1 and V2. In addition, the prefrontalgranular region contains high neuron packing values with a fringe oflower values. However, motor and premotor regions contain lowvalues, the somatosensory region has moderate values, and much oftemporal and posterior parietal regions have moderate values. Asimilar pattern, with less variation, is apparent for all cells (Fig. 3B).For the densities of cells and neurons, we found that cubic modelsprovided the best correlations (Fig. 3 A and B), all with P values<0.0001. These results, including increased densities in frontal re-gions of cortex, support deviations from linear or uniform patternsacross the A-P dimension.

DiscussionIn the present study, we flattened the neocortex of one cerebralhemisphere of a chimpanzee into a sheet, divided the sheet intothree main parts, and then further divided the large pieces into 742small pieces of tissue. Tissue blocks were individually processed forestimates of total neuron numbers using the rapid and accurate flowfractionation method (34, 35), and estimates of total cell numberwere obtained using the isotropic fractionation method (36). Theresults are shown in cell or neuron number per square centimeter ofcortical surface because the number of neurons in the verticallydefined cortical columns that extend across the depth of the cortexhas been considered basic to cortical function (37). The results in-dicate that areas of neocortex in chimpanzees differ greatly inpacking density, such that visual areas V1 and V2 have the highestneuron density, and the motor and premotor areas are among thosethat have the lowest densities. Perhaps surprisingly, a region ofgranular prefrontal cortex had higher neuron densities than sur-rounding cortical regions. These clear differences in cell and neurondensities, when considered together with the rapidly accumulatingevidence from other primates, and even nonprimate species (38, 39),should dispel any notion that the neocortex is uniform in this respect.The present results have functional implications for the neocortex inchimpanzees and invite comparisons with the neocortex of otherprimates, especially with humans, as the closest biological extantrelative of chimpanzees and bonobos.

Comparisons with Other Primates. Our results allow detailed compar-isons with similar maps of flattened neocortex in macaques andbaboons (1, 2, 40). In these primates, V1 had the highest neurondensities, as much as three to six times that of most cortical regions.This is unsurprising, as larger numbers of neurons per cortical col-umn have been previously reported in V1 of monkeys (8), and morerecently reproduced (9). However, those authors argued for the“basic uniformity” in neuron numbers across the depth of cortex forother areas of cortex, and across mammalian species. Here, we showthat neuron numbers vary across the chimpanzee cortical sheet, withhigh values also in V2, somatosensory cortex, and part of frontalgranular cortex, and low values in motor and premotor cortex. Mostnotably, neocortex in macaques and baboons also reflects this gen-eral pattern. In a more limited study of flattened cortex of a galago,V1 also had much higher neuron densities than other areas (1), andM1 of owl monkeys, squirrel monkeys, and galagos has been shownto have low neuron densities in comparison with other cortical re-gions when studied with our current methods (2). Such a detailedcomparison with neocortex of humans is not yet possible, butmeaningful comparisons of neuron densities using another approachindicate impressive similarities with our present results. By usingthick frontal sections of a human brain and comparing neuronaldensities across anterior to posterior slices of cortex, Ribeiro et al.(30) reported very high neuron densities for posterior slices including

A

B

Mea

n nu

mbe

r of n

euro

ns b

y su

rfac

e ar

ea (m

illio

ns/c

m )2

Mea

n nu

mbe

r of c

ells

by

surf

ace

area

(mill

ions

/cm

)2

Fig. 3. The cell and neuron densities of all 742 pieces are plotted based on theiranterior–posterior position in the flattened cortex, and designated to a corticalarea (shown in small map). Neuron (A) and cell (B) packing densities follow thesame pattern, in which densities are highest in the most posterior positions andlower in more anterior positions, with the exception of a block of frontal cortexthat contains higher densities. For both cell and neurons, although the correla-tion is low, cubic models provide a better correlation (neurons: R2 = 0.424; cells:R2 = 0.066) compared with linear (neurons: R2 = 0.303; cells: R2 = 0.011) orquadratic (neurons: R2 = 0.414; cells: R2 = 0.035) models. Most outliers are cell- orneuron-dense pieces in the posterior part of the brain.

Collins et al. PNAS | January 19, 2016 | vol. 113 | no. 3 | 743

NEU

ROSC

IENCE

Dow

nloa

ded

by g

uest

on

Mar

ch 1

4, 2

020

V1, and very low values for anterior slices in frontal and prefrontalcortex. It is not clear yet from these results whether V1 and V2differ, whether M1 and premotor cortex are specifically low inneuron number, or whether a granular region of prefrontal cortexwith higher neuron densities exists. It is also uncertain whether pri-mary somatosensory cortex is higher in primates in general, asexpected from its well-developed layer of granular cells, or whethergranular primary auditory cortex values are higher in any primate.As for the nonprimates that have been studied, areal differences inmean densities may be less pronounced, but they vary considerablyacross cortical areas with high densities in V1 and primary so-matosensory cortex (S1) in mice (39) and rats (41). Others havereported higher neuron densities in S1 than M1 and V2 in rats (42),and higher values in visual areas and S1 than M1 in cats (10).

Does Neuron Packing in Chimpanzee Cortex Reflect a DevelopmentalPattern of Cortical Neurogenesis? In a series of publications, Finlayand colleagues have presented evidence that an anterior to posteriorpattern of cortical development that is seen in primates and othermammals results in a matching gradient of neuron densities from lowto high across the cortex (e.g., refs. 11 and 31–33). Our chimpanzeeresults roughly reflect such a pattern, with at least four exceptions.These exceptions include the sharp increase in neuron densities in V1from V2 at the V1–V2 border, the very low densities in motor andpremotor cortex, the increased neuron densities in anterior so-matosensory cortex, and the higher than expected neuron densities indorsofrontal cortex of “granular” frontal cortex. These exceptions donot argue against the developmental gradient having an importantrole in creating neuron density differences across the cortex, but theexceptions do indicate that other factors are also involved.Such additional factors may include areal differences in neuron

death during development (31). Direct evidence for this possibilitycomes from V1 of macaques after loss of visual inputs to the brainduring fetal development. In such monkeys, parts of V1 fail to de-velop normally histologically (43, 44), and these abnormal parts hadcell densities reduced by about 25% (43). In view of the de-velopmental gradient theory of regional differences in neuron den-sities, it may just be a fortunate circumstance for primates that V1evolved in caudal neocortex. Higher neuron densities in V1 functionto precisely preserve visual information, whereas motor and pre-motor cortex, as well as parts of prefrontal cortex, evolved withinmore anterior cortex to contain extremely low neuron densities as aresult of larger neurons. This makes neurons in these anterior re-gions better suited to integrate information from many sources ofactivation. Alternatively, more modular features of cortical devel-opment could have played a prominent role in the evolution of suchstriking differences in neuron densities across cortex.

Functional Implications. Differences across cortical areas in neu-ron packing densities imply there is an inverse relationship withaverage neuron size (45). Larger neurons take up more spaceand require more glial and other support cells that vary in size.Smaller neurons have smaller dendritic arbors and are connectedby fewer inputs (46). Overall, small cortical neurons are betterdesigned for preserving information from a small number ofactivating inputs, whereas large neurons are better suited forintegrating information from a larger number of activating inputs(47, 48). Thus, V1 has densely packed small neurons (granularcells) in layer 4 that are activated by just a few neurons in thedorsal lateral geniculate nucleus, and they activate other neuronsin V1 that are, with few exceptions, small pyramidal neurons withsmall apical arbors contacted by relatively few inputs (46, 48).The high neuron densities for V1 of primates have been postu-lated as a mechanism for preserving the high visual acuity ofprimates (49). Motor cortex is known for its large pyramidal cellsand lack of small layer 4 granular cells, which promotes integrationfrom more sources of information. Primary sensory areas generallyhave smaller neurons with smaller dendritic arbors, whereas higher-

order sensory areas have larger neurons with larger arbors. Theneuron density values in cortex illuminate this hypothesis in detailby indicating average levels of information preservation and in-tegration for areas across the cortical sheet for chimpanzee.However, neurons of quite different sizes may occur in the samearea and play different functional roles. For example, V1 containsboth small granular cells in layer 4 and large Meynert cells inlayers 5 and 6. However, the average neuron size should suggest adominant role for a region or area.One of the important findings of the present study was that a

dorsal part of granular prefrontal cortex in the chimpanzee had aregion of higher neuron density. As the term “granular frontal cortex”implies, this cortex contains small neurons in layer 4. However, largepyramidal neurons have also been described in prefrontal cortex,suggesting such neurons receive and sum many inputs (12). Althoughall parts of granular prefrontal cortex would seem suitable for pre-serving information, the dorsomedial part seems more specialized forthis function. Notably, granular frontal cortex is thought to be aspecialization of the primate brain (26) that appears to be importantin working memory (50). Our present results are consistent with theconclusion that granular frontal cortex is not uniform in function (28),and the region of particularly dense neuron packing in frontal cortexmay be a specialization of primates that occurs to a lesser extent inOld World macaques (1) and baboons (40), is especially marked inchimpanzees, and is likely in humans.

Materials and MethodsExperimental procedures were all approved by the Vanderbilt InstitutionalAnimal Care and Use Committee. One adult female chimpanzee brain wasobtained for this study from the Texas Biomedical Research Institute. The ageof the chimpanzee was estimated to be 53 y. The animal was humanelykilled because of myocarditis (heart failure). The neocortex is expected tohave few age-related changes in a chimpanzee of this age (51, 52). Shortlybefore death, her body weight was 34.8 kg. The brain was flushed with 0.9%PBS, removed from the skull, and shipped overnight in the same solution. Onarrival, the brain weight was 344 g. The brain was bisected into right and lefthemispheres. The right neocortex was flattened and used in this report, and theleft hemisphere was sectioned and histologically processed for other studies.

Todetermineneuronand cell number per unit of cortical surface area across allparts of the cortical surface, we first manually flattened the neocortex into asheet. The unfixed cortex of the right hemisphere was separated from the un-derlying structures andmost of thewhitematter. A cut wasmade in the depth ofthe central sulcus to separate themost anterior region, and another cutwasmadeto separate primary visual cortex from the rest of the caudal cortex, creating threeseparate cortical pieces for flattening. This was done to preserve the integrity ofthe tissue during flattening. Sulci were carefully opened, and the cortical sheetwas then unfolded under gentle pressure (Fig. 1A). Excess white matter wasremoved. After postfixing in 4% (wt/vol) paraformaldehyde for 2 wk, each sheetwas cut into small pieces, ∼5 mm2 in surface area, resulting in 742 pieces thatwere photographed, weighed, numbered, and assigned to a cortical area whenpossible (Fig. 1 B and C). Pieces were assigned to cortical areas based on theirexpected relation to sulcal landmarks and their myelin content, as primarysensory areas appear dark relative to the surrounding cortex when viewed on alight box because of their myelin-dense composition. The surface area of eachpiece of cortex was measured from the photograph, using ImageJ (NIH). Eachpiece of cortex was individually disassociated to a solution, where cell mem-branes were ruptured but cell nuclei remained intact, as previously described(34, 36). We determined the number of cells in each tissue piece, using the iso-tropic fractionator method (36). The flow fractionator method was used to esti-mate of the total number of neuron nuclei labeled with the anti-NeuN antibodyand DAPI in each tissue piece, as previously described in detail (1, 2, 34, 35, 40).

IBM-SPSS software (version 22) was used to test for deviations from thenormal distribution (Kolmogorov-Smirnov test), to test for nonlinearities(linear regression), and to compare cell and neuron densities between se-lected cortical areas (generalized linear modeling). Significance was con-sidered for P < 0.05. Statistical analysis included using Huber-White robustSEs through the generalized linear modeling procedures specifying “robust”covariance matrix estimators (also called Huber-White or sandwich estima-tors) in SPSS software. Robust SEs are corrections to help account for viola-tions of the assumption of independence between cell counts on tissuepieces from a single chimpanzee brain. The resulting P values can be used to

744 | www.pnas.org/cgi/doi/10.1073/pnas.1524208113 Collins et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 1

4, 2

020

help describe differences within the one brain, but they do not directly allowinferences about the population of chimpanzees.

Cell and neuron numbers per square centimeter of cortical surface wereobtained for each piece of cortex and color coded for cells (Fig. 1B) or neurons(Fig. 1C). In addition, each dissected piece of tissue was assigned coordinatesalong the A-P axis by generating centroid measures for each piece, using NIHImageJ software. Cell and neuron numbers per square centimeter of surfacearea versus the A-P locations were plotted and used in linear regressioncurve-fitting estimations to assess whether cell and neuron densities wereuniform or increased or decreased in the A-P dimension with linear,quadratic, and cubic regression models.

The results were tested for deviations from the normal distribution(Kolmogorov-Smirnov test), and generalized linear regression modelingwas then used to compare cell and neuron densities between selected corticalareas. Selected cortical areas defined by experts were the components of one

fixed factor (see Fig. 1C for regional bin map), and A-P coordinates (x-valuefrom the centroid calculation) were binned for use as the components of thesecond fixed factor. Model main effects were tested and evaluated for therelative contribution of each factor to the variance in mean density. Multiplecomparisons were adjusted by the Bonferroni method, and robust estimators,as addressed earlier, were used in this analysis to account for violations ofindependence between the samples from a single chimpanzee.

ACKNOWLEDGMENTS. We thank Laura Trice, Kallie Yeoman, and FeyiAworunse for laboratory assistance. Flow cytometry experiments wereperformed in the Vanderbilt University Medical Center (VUMC) Flow CytometryShared Resource. The VUMC Flow Cytometry Shared Resource is supportedby the Vanderbilt Ingram Cancer Center (P30 CA68485) and the VanderbiltDigestive Disease Research Center (DK058404). This work was supportedby a grant from the G. Harold and Leila Y. Mathers Foundation (to J.H.K.).

1. Collins CE, Airey DC, Young NA, Leitch DB, Kaas JH (2010) Neuron densities vary acrossand within cortical areas in primates. Proc Natl Acad Sci USA 107(36):15927–15932.

2. Young NA, Collins CE, Kaas JH (2013) Cell and neuron densities in the primary motorcortex of primates. Front Neural Circuits 7:30.

3. Goodman M, et al. (1989) Molecular phylogeny of the family of apes and humans.Genome 31(1):316–335.

4. Le Gros Clark WE (1959) The antecedents of man (Edinburgh University Press,Edinburgh).

5. Tramo MJ, et al. (1995) Surface area of human cerebral cortex and its gross mor-phological subdivisions: In vivo measurements in monozygotic twins suggest differ-ential hemisphere effects of genetic factors. J Cogn Neurosci 7(2):292–302.

6. Van Essen DC, Glasser MF, Dierker DL, Harwell J (2012) Cortical parcellations of themacaque monkey analyzed on surface-based atlases. Cereb Cortex 22(10):2227–2240.

7. Sherwood CC, Hof PR (2007) The evolution of neuron types and cortical histology inapes and humans. The Evolution of Nervous Systems, eds Preuss TM, Kaas JH (Aca-demic Press, Oxford), Vol 4, pp 355–378.

8. Rockel AJ, Hiorns RW, Powell TP (1980) The basic uniformity in structure of theneocortex. Brain 103(2):221–244.

9. Carlo CN, Stevens CF (2013) Structural uniformity of neocortex, revisited. Proc NatlAcad Sci USA 110(4):1488–1493.

10. Beaulieu C, Colonnier M (1989) Number of neurons in individual laminae of areas 3B,4 gamma, and 6a alpha of the cat cerebral cortex: A comparison with major visualareas. J Comp Neurol 279(2):228–234.

11. Cahalane DJ, Charvet CJ, Finlay BL (2012) Systematic, balancing gradients in neurondensity and number across the primate isocortex. Front Neuroanat 6:28.

12. Elston GN, et al. (2006) Specializations of the granular prefrontal cortex of primates:Implications for cognitive processing. Anat Rec A Discov Mol Cell Evol Biol 288(1):26–35.

13. Kaas JH (2000) Why is brain size so important: Design problems and solutions asneocortex gets bigger or smaller. Brain Mind 1(1):7–23.

14. Elston GN (2007) Specialization of the neocortical pyramidal cell during primateevolution. Evolution of Nervous Systems 4:191–242.

15. Collins CE (2011) Variability in neuron densities across the cortical sheet in primates.Brain Behav Evol 78(1):37–50.

16. Holloway RL, Broadfield DC, Yuan MS (2003) Morphology and histology of chimpanzeeprimary visual striate cortex indicate that brain reorganization predated brain expansionin early hominid evolution. Anat Rec A Discov Mol Cell Evol Biol 273(1):594–602.

17. Bailey P, Von Bonin G, McCulloch WS (1950) The isocortex of the chimpanzee (Uni-versity of Illinois Press, Urbana).

18. Bush EC, Allman JM (2004) The scaling of frontal cortex in primates and carnivores.Proc Natl Acad Sci USA 101(11):3962–3966.

19. de Sousa AA, et al. (2010) Comparative cytoarchitectural analyses of striate and ex-trastriate areas in hominoids. Cereb Cortex 20(4):966–981.

20. Miller DJ, Balaram P, Young NA, Kaas JH (2014) Three counting methods agree on celland neuron number in chimpanzee primary visual cortex. Front Neuroanat 8:36.

21. Lyon DC, Kaas JH (2002) Evidence for a modified V3 with dorsal and ventral halves inmacaque monkeys. Neuron 33(3):453–461.

22. Lyon DC, Kaas JH (2002) Evidence from V1 connections for both dorsal and ventral sub-divisions of V3 in three species of New World monkeys. J Comp Neurol 449(3):281–297.

23. Qi HX, Preuss TM, Kaas JH (2008) Somatosensory areas of the cerebral cortex: Ar-chitectonic characteristics and modular organization. The Senses: A ComprehensiveReference, eds Gardner EP, Kaas JH (Elsevier, London), Vol 6, pp 143–170.

24. Leyton SSF, Sherrington CS (1917) Observations on the excitable cortex of the chim-panzee, orangutan and gorilla. Q J Exp Physiol 11(2):135–222.

25. Hines M (1940) Movements elicited from the precentral gyrus of adult chimpanzeesby stimulation with sine wave currents. J Neurophysiol 3(5):442–466.

26. Preuss TM, Goldman-Rakic PS (1991) Myelo- and cytoarchitecture of the granularfrontal cortex and surrounding regions in the strepsirhine primate Galago and theanthropoid primate Macaca. J Comp Neurol 310(4):429–474.

27. Semendeferi K, Armstrong E, Schleicher A, Zilles K, Van Hoesen GW (2001) Prefrontalcortex in humans and apes: A comparative study of area 10. Am J Phys Anthropol114(3):224–241.

28. Passingham RE, Wise SP (2012) The Neurobiology of the Prefrontal Cortex. Anatomy,Evolution, and the Origin of Insight (Oxford University Press, Oxford).

29. Dombrowski SM, Hilgetag CC, Barbas H (2001) Quantitative architecture distinguishesprefrontal cortical systems in the rhesus monkey. Cereb Cortex 11(10):975–988.

30. Ribeiro PF, et al. (2013) The human cerebral cortex is neither one nor many: Neuronaldistribution reveals two quantitatively different zones in the gray matter, three in thewhite matter, and explains local variations in cortical folding. Front Neuroanat 7:28.

31. Cahalane DJ, Charvet CJ, Finlay BL (2014) Modeling local and cross-species neuronnumber variations in the cerebral cortex as arising from a common mechanism. ProcNatl Acad Sci USA 111(49):17642–17647.

32. Charvet CJ, Cahalane DJ, Finlay BL (2015) Systematic, cross-cortex variation in neuronnumbers in rodents and primates. Cereb Cortex 25(1):147–160.

33. Charvet CJ, Finlay BL (2014) Evo-devo and the primate isocortex: The central orga-nizing role of intrinsic gradients of neurogenesis. Brain Behav Evol 84(2):81–92.

34. Collins CE, Young NA, Flaherty DK, Airey DC, Kaas JH (2010) A rapid and reliablemethod of counting neurons and other cells in brain tissue: A comparison of flowcytometry and manual counting methods. Front Neuroanat 4:5.

35. Young NA, et al. (2012) Use of flow cytometry for high-throughput cell populationestimates in brain tissue. Front Neuroanat 6:27.

36. Herculano-Houzel S, Lent R (2005) Isotropic fractionator: A simple, rapid method for thequantification of total cell and neuron numbers in the brain. J Neurosci 25(10):2518–2521.

37. Kaas JH (2012) Evolution of columns, modules, and domains in the neocortex ofprimates. Proc Natl Acad Sci USA 109(Suppl 1):10655–10660.

38. Campi KL, Collins CE, Todd WD, Kaas J, Krubitzer L (2011) Comparison of area 17cellular composition in laboratory and wild-caught rats including diurnal and noc-turnal species. Brain Behav Evol 77(2):116–130.

39. Herculano-Houzel S, Watson C, Paxinos G (2013) Distribution of neurons in functionalareas of the mouse cerebral cortex reveals quantitatively different cortical zones.Front Neuroanat 7:35.

40. Young NA, et al. (2013) Epileptic baboons have lower numbers of neurons in specificareas of cortex. Proc Natl Acad Sci USA 110(47):19107–19112.

41. Campi KL, Krubitzer L (2010) Comparative studies of diurnal and nocturnal rodents:Differences in lifestyle result in alterations in cortical field size and number. J CompNeurol 518(22):4491–4512.

42. Skoglund TS, Pascher R, Berthold CH (1996) Heterogeneity in the columnar number ofneurons in different neocortical areas in the rat. Neurosci Lett 208(2):97–100.

43. Rakic P, Suñer I, Williams RW (1991) A novel cytoarchitectonic area induced experi-mentally within the primate visual cortex. Proc Natl Acad Sci USA 88(6):2083–2087.

44. Dehay C, Savatier P, Cortay V, Kennedy H (2001) Cell-cycle kinetics of neocorticalprecursors are influenced by embryonic thalamic axons. J Neurosci 21(1):201–214.

45. Herculano-Houzel S, Manger PR, Kaas JH (2014) Brain scaling in mammalian evolutionas a consequence of concerted and mosaic changes in numbers of neurons and av-erage neuronal cell size. Front Neuroanat 8:77.

46. Elston GN (2002) Cortical heterogeneity: Implications for visual processing and poly-sensory integration. J Neurocytol 31(3-5):317–335.

47. Kaas J (2002) Convergences in themodular and areal organization of the forebrain ofmammals:Implications for the reconstruction of forebrain evolution. Brain Behav Evol 59(5-6):262–272.

48. Elston GN, Rosa MG, Calford MB (1996) Comparison of dendritic fields of layer IIIpyramidal neurons in striate and extrastriate visual areas of the marmoset: A Luciferyellow intracellular injection. Cereb Cortex 6(6):807–813.

49. Srinivasan S, Carlo CN, Stevens CF (2015) Predicting visual acuity from the structure ofvisual cortex. Proc Natl Acad Sci USA 112(25):7815–7820.

50. Funahashi S, Bruce CJ, Goldman-Rakic PS (1989) Mnemonic coding of visual space inthe monkey’s dorsolateral prefrontal cortex. J Neurophysiol 61(2):331–349.

51. Sherwood CC, et al. (2011) Aging of the cerebral cortex differs between humans andchimpanzees. Proc Natl Acad Sci USA 108(32):13029–13034.

52. Autrey MM, et al. (2014) Age-related effects in the neocortical organization ofchimpanzees: Gray and white matter volume, cortical thickness, and gyrification.Neuroimage 101:59–67.

Collins et al. PNAS | January 19, 2016 | vol. 113 | no. 3 | 745

NEU

ROSC

IENCE

Dow

nloa

ded

by g

uest

on

Mar

ch 1

4, 2

020