Positive selection on NIN, a gene involved in neurogenesis, and primate brain evolution

Authors


Abstract

A long-held dogma in comparative neurobiology has been that the number of neurons under a given area of cortical surface is constant. As such, the attention of those seeking to understand the genetic basis of brain evolution has focused on genes with functions in the lateral expansion of the developing cerebral cortex. However, new data suggest that cortical cytoarchitecture is not constant across primates, raising the possibility that changes in radial cortical development played a role in primate brain evolution. We present the first analysis of a gene with functions relevant to this dimension of brain evolution. We show that NIN, a gene necessary for maintaining asymmetric, neurogenic divisions of radial glial cells (RGCs), evolved adaptively during anthropoid evolution. We explored how this selection relates to neural phenotypes and find a significant association between selection on NIN and neonatal brain size in catarrhines. Our analyses suggest a relationship with prenatal neurogenesis and identify the human data point as an outlier, possibly explained by postnatal changes in development on the human lineage. A similar pattern is found in platyrrhines, but the highly encephalized genus Cebus departs from the general trend. We further show that the evolution of NIN may be associated with variation in neuron number not explained by increases in surface area, a result consistent with NIN's role in neurogenic divisions of RGCs. Our combined results suggest a role for NIN in the evolution of cortical development.

According to the radial unit hypothesis (Rakic 1988, 1995) variation in the production of cortical neurons could arise in two key ways (Fig. 1): a greater proliferation of neural progenitors leading to an increase in the number of ontogenetic columns of neurons (radial units) or a prolonged proliferation of cortical progenitor cells leading to an increase in neurons/radial unit. Radial units are represented in the adult brain by cortical minicolumns (Horton & Adams 2005; Mountcastle 1997), the basic microcircuit in the cortex (Casanova et al. 2009). A long-held dogma in comparative neuroanatomy was that the number of neurons in each minicolumn was constant across mammals (Cheung et al. 2007; Rockel et al. 1980).

Figure 1.

A radial unit lineage model of cortical neurogenesis. Based on a modified figure from Rakic (2005). The figure illustrates how changes in the number of (1) symmetric divisions of neural progenitors and (2) asymmetric divisions of RGCs. Changes in the latter could also increase cortical thickness, but here we illustrate how it may affect neuron density and produce the sort of variation in number of neurons/unit surface area showed by Herculano-Houzel et al. (2008). Neurons are shown as triangles, and the columns represent radial units. (CC, cerebral cortex (layers I–VI); AsD, asymmetrical division; SD, symmetrical division).

However, convincing evidence has recently shown that the number of neurons under a unit area of cortical surface, and therefore the number of neurons/minicolumn, is not constant across primates (Herculano-Houzel et al. 2008) confirming previous reports of variation across mammals (Casanova et al. 2009; Haug 1987; Poth et al. 2005; Stolzenburg et al. 1989). In humans, variation in cortical thickness is associated with neurodevelopmental (Hardan et al. 2006; Narr et al. 2005) and neurodegenerative disorders (Lerch & Evans 2005; Rosas et al. 2005), suggesting that there may be phenotypic consequences to these evolved differences.

The number of neurons/radial unit is determined by the number of times radial glial cells (RGCs), cortical progenitors, undergo successive rounds of asymmetric division self-replicating and producing a neuron (Götz & Huttner 2005). Variation in the duration of this type of division presumably underlies the observed variation in neurons/radial unit (Herculano-Houzel et al. 2008). Although several candidate genes have been proposed to contribute to the evolution of brain size by altering the duration of asymmetric divisions of neural progenitors, leading to a lateral expansion of the cortex (e.g. Fish et al. 2006; Montgomery et al. 2011), no candidate genes have been linked to the radial expansion of neuron number. We propose that NIN (or ninein) may have a function relevant to this cell fate switch.

NIN encodes a centrosome maturation factor (Chen et al. 2003; Mogensen et al. 2000), which localizes primarily on the mother centrioles in post-mitotic cells (Piel et al. 2000). During the proliferation of RGCs, mother and daughter centrioles show differences in NIN concentration (Ohama & Hayashi 2009). This intrinsic asymmetry contributes to cell fate with the differentiated neuron inheriting the daughter centrioles and the renewed RGC inheriting the mother centrioles (Wang et al. 2009). Suppression of NIN function is sufficient to disrupt this pattern of inheritance and results in depletion of RGCs as they prematurely terminate their cycle of proliferative, asymmetric division (Wang et al. 2009). This suggests that NIN is critical for maintaining neurogenic divisions of RGCs (Wang et al. 2009). Here, we describe an analysis of the molecular evolution of NIN across 22 species of anthropoid primates designed to test whether this gene has been targeted by positive selection during anthropoid evolution, and to ask what the phenotypic relevance of this selection might be.

Materials and methods

Laboratory methods

Genomic DNA samples had previously been extracted from tissue samples using DNeasy kits (QIAGEN UK, Crawley, UK). A preliminary analysis performed on the full NIN coding sequences for six anthropoids is presented in Fig. S1, Supporting Information. To obtain a greater sample that encapsulates the diversity of primate brain sizes, the two largest exons of NIN, exon 18 (503 bp) and exon 19 (2141 bp), which together account for 41.3% of the total coding sequence of the longest human transcript predicted on Ensembl, were targeted for resequencing. This region includes the important functional coiled-coil domain. Sequences were obtained for 5 apes, 6 Old World monkeys and 11 New World monkeys representing all major clades of anthropoid primates. The phylogeny of these species was taken from Montgomery et al. (2010) and is shown in Fig. 2. Primer sequences, PCR and sequencing reaction conditions are presented in Table S1, Supporting Information. Sequences were edited in SEQMAN v. 5.05 (DNASTAR Inc., Madison, WI, USA); the two exons were concatenated and aligned using MUSCLE in MEGA 5.0 (Tamura et al. 2011).

Figure 2.

A phylogeny of the 22 species considered in the study. All species have data for adult brain mass and newly collected sequence data for NIN exons 18 and 19. Y (yes) indicates when data are also available for the full coding sequence (Full CDS), neonatal brain mass (NBM) and neuron number, cortical area and thickness (NAT). Stem-C is the catarrhine stem and Stem-P is the platyrrhine stem.

Sequence analysis

We implemented a sliding window analysis of dN/dS across the alignment of the full coding sequence from six anthropoids using SWAAP (Pride 2000), and the Nei and Gojobori (1986) method with a window size of 150 codons and a step size of 15 codons, to visualize the distribution of selection pressures and identify any stark differences between exons targeted for resequencing and other exons. These results are presented in Supporting Information. Estimation of dN/dS ratios (ω), a common measure used to infer selection pressures acting on coding sequence, was carried out using a codon-based maximum likelihood method (CODEML in PAML version 4; Yang 2007). To detect positive selection across primates, we implemented the site models. These allow ω to vary among sites but not across lineages (Nielsen & Yang 1998; Yang et al. 2000). The first test compares model M1a and model M2a (Wong et al. 2004; Yang et al. 2005). Model M1a (nearly neutral) allows sites to fall into two categories with ω < 1 (purifying selection) and ω = 1 (neutral evolution), whereas model M2a (positive selection) allows sites to fall into three categories with ω < 1, ω = 1 and ω > 1 (positive selection) (Yang et al. 2005). The second test compares model 8a and model 8 (Swanson et al. 2003; Wong et al. 2004), which use the beta distribution to describe the numbers of sites across different categories of ω. Site models implement a Bayes Empirical Bayes method to identify specific codons under positive selection by calculating the posterior probability for site classes (Yang et al. 2005). In addition, the branch models were used to calculate the root-to-tip ω for the lineage leading from the last common ancestor of a clade to each terminal species (see below; Montgomery et al. 2011).

Phenotypic data and tests for gene–phenotype associations

To explore the potential role of NIN in brain size, we first conduct a broad analysis using brain mass. Although whole brain mass is a relatively crude phenotype, brain size correlates with a number of structural and cellular phenotypes (Herculano-Houzel et al. 2007; Stephan et al. 1981), and data are more readily available for brain mass. Brain size, therefore, provides a means to uncover genes with a general role in brain evolution (Montgomery et al. 2011). Data for adult body and brain mass (Bauchot & Stephan 1969; Stephan et al. 1981; Zilles & Rehkemper 1988) and neonatal brain and body size (Capellini et al. 2011) were obtained from previously published sources (Table S2). Next, we focused on specific structural changes in the cerebral cortex to test a specific hypothesis regarding NINs function (see Introduction section). Data for cortical phenotypes (cortical thickness, cortical surface area and cortical neuron number) are available for eight anthropoids (Herculano-Houzel et al. 2008): Callithrix jacchus, Aotus trivirgatus, Callimico goeldii, Saimiri sciureus, Cebus apella, Papio sp., Macaca fascicularis and Macaca radiata (Table S3). Herculano-Houzel et al. (2008) defined the cerebral cortex as ‘all cortical regions lateral to the olfactory tract, including the hippocampus and piriform cortex’. Our sequence data come from Macaca mulatta, so in order to maximize sample size we took the average values of the two Macaca species; this is likely to introduce additional noise to the analysis but should not bias the results.

To test for associations between selection on NIN and brain phenotypes, we followed the method presented in Montgomery et al. (2011). Briefly, branch models implemented in CODEML were used to estimate the average dN/dS ratio from the last common ancestor of all the species in each dataset to each terminal species tip. These values were then set as species data and used in a phylogenetic generalized least squares (PGLS) regression with phenotypic measures in BayesTraits as explained above (Organ et al. 2007; Pagel 1999; Pagel et al. 2004). In PGLS phylogenetic non-independence is corrected for by converting the phylogeny into a variance–covariance matrix, where the diagonal of the matrix gives information on the path length from root to tips (the ‘variance’) and the off-diagonal values of the matrix provide information on the shared evolutionary history of any pair of species (the ‘covariance’). The variance–covariance matrix is included into the error term of the regression model, and the resulting estimated regression parameters (i.e. slopes and intercepts) are ‘phylogenetically controlled’ (Pagel 1999). All results presented are therefore account phylogenetically controlled. The significance of the regression coefficient was determined using a one-tailed t test as we specifically hypothesize an association between dN/dS and the phenotype in the direction that the phenotype is evolving as higher dN/dS values should be associated with greater phenotypic changes where that gene contributes to the genetic basis of that phenotype (Montgomery et al. 2011). To test for associations between the evolution of NIN and radial expansion, we first performed a regression analysis between neuron number and cortical surface area using PGLS in BayesTraits (Pagel 1999; Pagel et al. 2004). The regression was significant (t6 = 12.716, P < 0.001, R2 = 0.970) as previously reported (Herculano-Houzel et al. 2008). The residuals from the regression equation

display math

were used as a measure of the variation in neuron number not explained by lateral expansion, which must therefore be explained by radial expansion; these values are termed ‘radial neuron number’.

Results

NIN bears a signature of positive selection

We found strong evidence that NIN evolved under positive selection during anthropoid primate evolution, with a significant proportion of sites evolving with a dN/dS above one (likelihood ratio statistic (LRS) = 18.014, P < 0.001; Table S4). The average dN/dS ratio does not vary significantly between the three major clades (LRS = 1.586, P = 0.452; Table S5), indicating that the strength of selection is fairly uniform across anthropoids. However, root-to-tip dN/dS ratios calculated for each species (Table S2) showed significantly higher values in platyrrhines than catarrhines (t20 = 3.716, P = 0.013, R2 = 0.269). This is due to a significantly higher rate of evolution along the stem lineage leading to platyrrhines (dN/dS = 0.627; Fig. 2) than on the stem lineage leading to catarrhines (dN/dS = 0.168) (LRS = 5.105, P = 0.024). The average dN/dS in each clade excluding these branches is not significantly different (LRS = 0.073, P = 0.964). Hence, the selection pressure acting on NIN differed during the origin of platyrrhines and catarrhines but was similar during their diversification. Whether this difference relates to some difference in phenotypic evolution at the origin of these two clades is difficult to resolve but an enticing possibility. Of potential relevance is the observation that platyrrhines have higher neuron densities and larger numbers of neurons under 1 mm2 of cortical surface than catarrhines (Herculano-Houzel et al. 2008).

A role for NIN in the evolution of brain size

We next sought to explicitly test the hypothesis that positive selection acting on NIN is related to evolutionary shifts in brain size using comparative methods to test for gene–phenotype associations. The shift in dN/dS between platyrrhines and catarrhines precludes any significant association between adult brain mass and NIN root-to-tip dN/dS across all anthropoids (Table 1), but within each clade evidence is found supporting an association between the evolution of NIN and brain mass (Fig. 3). Within catarrhines there is a strong significant association between neonatal brain mass and root-to-tip dN/dS (P = 0.001; Table 1). When adult brain mass is considered the significance of the association is lost but a trend remains (P = 0.055). The loss of significance can be partially attributed to the data point for Homo: when removed the association regains significance, albeit narrowly (P = 0.044). Within the platyrrhines a similar effect is seen for adult brain mass with Cebus dramatically reducing the significance of the association (from P = 0.005 to P = 0.086; Table 1). The Cebus data point also affects the results of the neonatal regression analysis, but here the lower sample size limits the power to detect gene–phenotype associations. Hence, in each clade, there is evidence linking the evolution of NIN to the evolution of brain size but both include a single data point that appears to be an outlier: Homo for catarrhines and Cebus for platyrrhines.

Table 1. Phylogenetically controlled regression analysis between root-to-tip dN/dS and brain phenotypes
CladenAdult whole brainnNeonate whole brain
t-statPR2t-statPR2
All anthropoids210.7940.2180.032130.9660.1770.072
Catarrhines101.7990.0550.28885.0540.0010.810
Catarrhines without Homo91.9780.0440.35975.4300.0010.855
Platyrrhines111.4860.0860.19751.3690.1320.385
Platyrrhines without Cebus103.3680.0050.58642.7410.0560.790
Figure 3.

Phylogenetically controlled regressions between root-to-tip dN/dS and brain mass. Catarrhines are shown in light gray circles, and platyrrhines in dark gray diamonds. Data points are raw species values. Solid lines are phylogenetically controlled regressions performed separately in BayesTraits for the two clades without Homo or Cebus, apparent outliers, dashed lines are including them.

We have argued elsewhere that a stronger association with neonatal brain mass than adult brain mass is consistent with a locus having a role in neurogenesis (Montgomery et al. 2011). This is because primate neocortical neurogenesis is restricted to prenatal development (Bhardwaj et al. 2006; Rakic 1988). Postnatal brain growth is largely driven by gliagenesis, axon growth and myelination (Low & Cheng 2006; Sauvageot & Stiles 2002; Sowell et al. 2001), rather than by production of new neurons, and apoptosis eliminates large numbers of neurons after birth (Buss et al. 2006). Variation in these non-neurogenic processes will weaken any association between the molecular evolution of genes under selection in relation to prenatal neurogenesis. Anthropoid primates are all considered highly precocial (Capellini et al. 2011; Derrickson 1992), but any slight variation in maturity at birth will likely reduce the association with neonatal brain size and will not bias the results.

A role for NIN in the evolution of radial neuron number?

Finally, we attempted to test the more developmentally specific hypothesis that NIN contributes to the evolution of brain size by altering the duration of asymmetric, neurogenic divisions of RGCs. This analysis is complementary to the analysis of whole brain size but, because of the limited data on cortical phenotypes, is necessarily preliminary. We tested for an association between root-to-tip dN/dS and ‘radial neuron number’, the variation in neuron number not explained by increases in surface area using a phylogenetically controlled regression. This showed a trend (t5 = 1.732, P = 0.072, R2 = 0.373; Fig. 4), which may be suggestive of an underlying relationship. In a multiple regression with dN and dS as separate variables, dN shows a positive trend with variation in radial neuron number but dS shows no association (dN: t3 = 1.957, P = 0.073; dS: t3 = −0.376, P = 0.366; R2 = 0.493), which is consistent with the action of positive selection driving the trend (Montgomery et al. 2011). Multiple regressions between root-to-tip dN/dS, neuron number and cortical surface area also confirm that the trend is driven by variation in neuron number, controlled for surface area, rather than vice versa (Table S6).

Figure 4.

Phylogenetically controlled regressions between root-to-tip dN/dS and radial neuron number. Data points are raw species values and are labeled with their genus. The phylogenetically controlled regression line was estimated in BayesTraits and superimposed on top of raw species data.

We performed additional checks to test the specificity of the result. When a multiple regression was performed replacing neuron number with cortical thickness no underlying trend with cortical thickness was found (Table S6). In addition, there is no trend with total neuron number (t5 = −0.635, P = 1.000, R2 = 0.075) or surface area (t5 = −0.932, P = 1.000, R2 = 0.148) when considered alone, suggesting that the evolution of NIN is not associated with lateral expansion of the primate cortex. This implies specificity with variation in the number of neurons under a unit area of cortical surface rather than a general association with cortical thickness, which may also be influenced by differences in development unrelated to changes in neuron number.

As the raw data values of radial neuron number for platyrrhines and catarrhines are clumped (Fig. 4), it is of interest to know whether the trend is driven by a difference between these clades. A phylogenetically controlled t test indicates that platyrrhines have higher ‘radial neuron numbers’ than catarrhines (t5 = 3.348, P = 0.020) and a narrowly non-significantly higher dN/dS (2.552, P = 0.051). It is therefore possible that this clade effect drives the association across anthropoids. However, it should be noted that if the clade difference was the sole driver of the association this would be accounted for in the PGLS regression. The fact that a trend is found after correcting for phylogeny indicates that variation within platyrrhines and catarrhines may contribute to the association. If selection on NIN is related to variation in ‘radial neuron number’ and platyrrhines have evolved a greater ‘radial neuron number’, we would expect that there would be a significant, positive shift in the root-to-tip dN/dS values in platyrrhines as periods of increased positive selection on genes underpinning the evolution of a phenotype should co-occur with changes in that phenotype. This is in contrast to the association reported in Fig. 3, where a shift in dN/dS is not associated by a positive increase in the phenotype, i.e. there is a significant increase in dN/dS on the stem platyrrhine branch without a comparable increase in brain size, indicating that the evolution of NIN and brain mass is decoupled at this point in the phylogeny. Unfortunately, the data are insufficient to test for associations within platyrrhine or catarrhines, which would provide a more robust confirmation that NIN contributes to the evolution of this axis of neuron number variation both between and within the major anthropoid clades. Hence, while the data available provide suggestive results, additional data on neural phenotypes are required to confirm the hypothesized developmental role of NIN in primate brain evolution.

Discussion

We present robust evidence that NIN, a gene that plays a key role in maintaining neurogenic divisions of RGCs (Wang et al. 2009), evolved adaptively during anthropoid evolution. We explored the phenotypic relevance of this positive selection using tests to detect associations between genetic and phenotypic evolution. We found a complex but significant relationship between gross brain mass and the selection pressure acting on NIN, with a strongly significant relationship between NIN and neonatal brain size in catarrhines. The combined results indicate that the adaptive evolution of NIN has phenotypic relevance to brain evolution. We suggest that this relevance could relate to its contribution to cell fate during asymmetric division of neural progenitors. Variation in the duration of asymmetric division of RGCs or radial glial-like progenitors in the outer subventricular zone could make substantial contributions to the evolution of neuron number (Fietz & Huttner 2010; Fietz et al. 2010; Götz & Huttner 2005) and knockdown experiments directly implicate NIN in having a role in this type of division (Wang et al. 2009). Changes in the function of centrosomal genes, such as NIN, could therefore influence total neuron number and brain size (Fig. 1).

We sought to test this hypothesis directly using cortical neuron data and, although the results do not reach significance, they are intriguing. The hypothesis that NIN may have a role in the evolution of asymmetric division of RGCs and contributes to the evolution of neuron number per radial unit is supported by a trend between the molecular evolution of NIN and variation in neuron number that is not explained by lateral expansion, as measured by cortical surface area. There is no trend with either raw neuron number or surface area, suggesting no link with lateral cortical expansion through the addition of extra columns. Furthermore, the higher rate of NIN evolution at the origin of platyrrhines is consistent with the suggestion that this clade has evolved significant higher ‘radial neuron numbers’. While we cannot delineate a continuous role for NIN in producing variation in this phenotype from a discrete shift at the origin of platyrrhines with the available phenotypic data, both scenarios are consistent with the patterns of selection pressures shaping the evolution of NIN. It may be the case that the relationship between NIN and brain mass within each clade reflects covariation between brain expansion and the addition of extra neurons to radial units, whereas the ‘grade shift’ between platyrrhines and catarrhines is due to selection on NIN to alter ‘radial neuron number’ independently from brain size.

These hypotheses require confirmation from larger datasets that can only be provided through the collection of data on cortical thickness, surface area and neuron number in a larger number of species, and we do not rule out the possibility that selection on NIN relates to other cell fate switches during neurogenesis or to its functions in other cells. As well as being localized at the centrosome NIN is found in cytoplasmic speckles (Moss et al. 2007). These speckles are particularly prominent in neurons (Baird et al. 2004) and may have a role in the rapid reorganization of microtubules in dendrites and epithelial cells and therefore to cell migration and morphology (Baird et al. 2004; Matsumoto et al. 2008; Moss et al. 2007). The possibility that selection on NIN relates to changes in neural morphology is also of potential interest but difficult to test using a comparative approach.

Independent genetic mechanisms for different aspects of brain evolution?

Our results indicate that cortical surface area and cortical thickness, which together contribute to cortical volume, are under the control of different genetic mechanisms. Two genes associated with primary microcephaly, ASPM and CDK5RAP2, have functions, clinical features and patterns of evolution that implicate them as having a role in the lateral expansion of the cortex by influencing the duration of symmetric, proliferative divisions of neural progenitor cells, but do not affect the development of cortical thickness (Fish et al. 2006; Mochida & Walsh 2001; Montgomery & Mundy 2010, 2012; Montgomery et al. 2011; Rimol et al. 2010). In contrast, as we show here, NIN may be involved in the evolution of radial expansion and cortical thickness. The suggestion that the evolution of cortical thickness and surface area has independent genetic bases is consistent with the radial unit hypothesis (Rakic 1988, 1995) and associated hypotheses (Fietz & Huttner 2010; Fish et al. 2008; Kriegstein et al. 2006; Pontious et al. 2008). It is also in agreement with quantitative genetics studies in humans which show that variation in cortical thickness and surface area shows distinctive patterns of heritability and no genetic correlations (Panizzon et al. 2009; Rimol et al. 2010; Winkler et al. 2010).

The potential effects of variation in minicolumn spacing

Semendeferi et al. (2011) have shown that the spacing between cortical minicolumns is wider in the human prefrontal cortex than in other apes, which have relatively constant spacing, and suggested that this may be a general feature of the human frontal cortex. This wider spacing could explain why the human data point is an outlier compared with other catarrhines. The human root-to-tip dN/dS is lower than expected for a catarrhine brain of its mass, as would be predicted if cortical expansion was associated with an increase in cortical column spacing and therefore a smaller increase in the density of these columns. If this is the case, the large residual for Cebus may suggest similar changes in cortical structure. These changes would presumably occur postnatally as cortical spacing is determined by axon and dendrite growth, the number of interneurons and blood vessels (Buxhoeveden & Casanova 2002). The smaller effect on Homo and Cebus data points in neonatal than adult regression analyses and the observation that Cebus is not an outlier when neuron number is considered rather than brain mass are consistent with this effect. An alternative explanation of why these species are outliers may be that the relationship between NIN and brain evolution is not conserved in these species, with selection targeting other loci to bring about similar phenotypic changes. However, although we cannot rule out this explanation, it would not explain why these species are greater outliers for adult brain size than neonatal brain size.

Platyrrhines are rarely considered in comparative studies of cortical cytoarchitecture that generally seek to identify differences between humans and other apes, so unfortunately no data exist to explore potential differences between Cebus and other platyrrhines. Such a comparison would be of great interest. The increased spacing between cortical minicolumns in the human prefrontal cortex (Casanova et al. 2009; Schenker et al. 2008; Semendeferi et al. 2011) has been proposed to have some relevance to cognitive evolution in humans (Elston 2007; Gustafsson 1997, 2004; Semendeferi et al. 2011). Wider spacing between minicolumns is thought to be associated with greater connections between parts of the cortex (Douglas et al. 1995) and variation in the amount of information convergence on a neuron (Williams & Jacobs 1997) to produce more generalized processors (Gustafsson 1997). If the results presented here do indeed represent an evolutionary difference in cortical cytoarchitecture between Cebus and other platyrrhines, it is tempting to relate this to the evolution of increased general cognitive ability in this genus (Deaner et al. 2006; Fragaszy et al. 2004; Reader et al. 2011). Cebus does at least have a lower cortical neuron density and approximately the same number of neurons per unit cortical surface area as Saimiri, its sister genus, despite having a cortical surface area approximately 50% larger (Herculano-Houzel et al. 2008), suggesting that the expansion of the Cebus brain had a greater lateral than radial dimension in terms of neuron number.

Conclusions

We have presented evidence that shows NIN, a gene encoding a centrosomal protein involved in the maintenance of asymmetric division of neural progenitors, has evolved under positive selection across anthropoids. The pattern of molecular evolution is consistent with the hypothesis that NIN may have a role in the evolution of brain size of anthropoid primates, and may have played a role in the evolution of the number of neurons in each ontogenetic column. However, the dataset is small and further data on neuron number and surface area from a wider range of species are necessary to confirm this hypothesis. Understanding the phenotypic importance of positive selection on NIN through comparative developmental or functional studies could further clarify the contribution of NIN to brain evolution.

Finally, we emphasize the importance of continued collection of different types of phenotypic data that are of primary interest but also necessary for comparative studies on the genetic basis of brain evolution. Long-standing datasets of adult brain volumes and structure (Stephan et al. 1981) combined with data on neonatal brain size (Capellini et al. 2011) and new data on the cellular architecture of neural tissue (Azevedo et al. 2009; Gabi et al. 2010; Herculano-Houzel et al. 2007, 2008) provide a means of deriving and testing hypotheses relating to gene–phenotype links using a comparative approach in taxa, such as primates, where functional tests are challenging.

Acknowledgments

We are grateful to Andrew Kitchener and Drew Bain (National Museums Scotland), Mike Bruford (Zoological Society London) and Leona Chemnick (Center for Reproduction of Endangered Species, San Diego Zoo) for providing tissue samples, and Leslie Knapp and Chris Ponting for comments on aspects of this work. We thank the BBSRC, the Leverhulme Trust and Murray Edwards College for financial support.

Ancillary