Correspondence: Stephen H. Montgomery, Department of Genetics, Evolution and Environment, University College London, Gower Street, London WC1E 6BT, UK. Tel.: +44 2076792170; fax: +44 2076797193; e-mail: email@example.com
Microcephaly genes are amongst the most intensively studied genes with candidate roles in brain evolution. Early controversies surrounded the suggestion that they experienced differential selection pressures in different human populations, but several association studies failed to find any link between variation in microcephaly genes and brain size in humans. Recently, however, sex-dependent associations were found between variation in three microcephaly genes and human brain size, suggesting that these genes could contribute to the evolution of sexually dimorphic traits in the brain. Here, we test the hypothesis that microcephaly genes contribute to the evolution of sexual dimorphism in brain mass across anthropoid primates using a comparative approach. The results suggest a link between selection pressures acting on MCPH1 and CENPJ and different scores of sexual dimorphism.
The complex social lives of primates require significant cognitive abilities (Cheney et al., 1986). Comparative studies have demonstrated a positive association between relative neocortex size and measures of social complexity (Dunbar, 1992; Barton, 1996) leading to the ‘social brain hypothesis’ (Dunbar, 1998) which suggests selection for increased social cognition has contributed to cortical expansion. Across primates, males and females are subject to differing selection pressures, related to differences in social ecology, which have had a detectable influence on brain evolution (Lindenfors et al., 2007). In humans, sexual dimorphism in whole brain size and the volume of grey and white matter is established prenatally (Gilmore et al., 2007), and sexual dimorphism in head circumference can also be observed at birth in nonhuman primates (Joffe et al., 2005) suggesting a conserved developmental origin across primates.
Identifying the genetic basis of complex phenotypic evolution and sexual dimorphism are two major goals currently being pursued by evolutionary biologists as they may reveal how nearly identical genomes produce phenotypic variation in sex-specific morphology and behaviour (Vallender et al., 2008; Williams & Carroll, 2009). Genes associated with microcephaly, a human neurodevelopmental disorder characterized by a disruption of cortical neurogenesis, are amongst the most intensively studied candidate genes with potential roles in brain evolution (Thornton & Woods, 2009). So far all primary microcephaly genes investigated show signatures of positive selection across primates but only two of these, ASPM and CDK5RAP2, show significant associations between their rate of molecular evolution and brain size across anthropoids which suggests selection has acted on these loci to bring about changes in primate brain size (Montgomery et al., 2011; Montgomery & Mundy, 2012). The phenotypic relevance of positive selection acting on other microcephaly genes, such as CENPJ and MCPH1, is still unclear. Within human populations, two recent studies have shown associations between brain volume or cortical surface area and variants of microcephaly genes, but in both cases the associations were sex-specific (Wang et al., 2008; Rimol et al., 2010). This suggests microcephaly genes may contribute to the development of sexual dimorphism in brain size (Montgomery & Mundy, 2010; Rimol et al., 2010). Here, we test whether the molecular evolution of four microcephaly genes are associated with levels of sexual dimorphism in anthropoid brain mass.
Materials and methods
Data on targeted partial coding sequence of ASPM, CDK5RAP2, CENPJ and MCPH1 from 21 anthropoids are available from a published study (Montgomery et al., 2011) (Table S1, Fig. S1). The data for each gene include exons with known functional domains and regions previously shown to have high rates of molecular evolution (Zhang, 2003; Evans et al., 2004a,b, 2006; Kouprina et al., 2004) and were collected with the aim of understanding the phenotypic relevance of selection acting on these genes. Data are available for ASPM exons 3 and 18, (totalling 6235 bp – 60% of the coding region), MCPH1 exons 8, 11 and 13 (totalling 1556 bp – 62% of the coding region), CENPJ exons 2 and 7 (totalling 1556 bp – 52% of the coding region), CDK5RAP2: exons 12, 20, 21, 24, 25, 32 and 33 (total 2120 bp – 37% of the coding sequence). These data revealed a signature of pervasive positive selection acting on all four loci across anthropoids (Montgomery et al., 2011). Estimates of the average dN/dS during the evolution of each species as the divergence from their last common ancestor, termed ‘root-to-tip dN/dS’, were used to test for gene-phenotype associations (Montgomery et al., 2011; see below). Here, we use the same root-to-tip dN/dS estimates. It is important to note that the high root-to-tip MCPH1 dN/dS ratio for chimpanzee (Pan) is heavily influenced by a small number of synonymous substitutions on the terminal Pan branch (one synonymous change, compared to six synonymous changes on the Homo branch) and previous studies have found a much lower dN/dS ratio on the terminal Pan lineage (Evans et al., 2004a; Wang & Su, 2004).We therefore repeat the regressions including and excluding this data point as was done in a previous analysis (Montgomery et al., 2011).
Sex-specific data on cranial capacities (Isler et al., 2008) were converted to brain mass using the relationship given by Martin (1980): Log10(cranial capacity) = [1.018 × Log10(brain mass)] − 0.025. Body mass was taken from the same source (Table S1). Results obtained using Log10-transformed endocranial volumes are consistent with those presented using brain mass suggesting the conversion does not introduce a bias. To calculate a measure of sexual dimorphism for absolute brain mass female brain mass was regressed against male brain mass using a reduced major axis regression following Forstmeier (2011) in the SMATR package (Warton & Ormerod, 2007). The regression was significant (P < 0.001). The regression equation between male and female brain mass was as follows:
Sexual dimorphism in body mass was calculated in the same manner. The regression equation between male and female body mass was as follows:
Although these measures of dimorphism are not corrected for phylogenetic nonindependence, this is taken to account in the gene-phenotype association tests (see below). We performed additional analyses exploring potential biases in the measures of sexual dimorphism and adopting alternative measures of sexual dimorphism, including ln(male mass/female mass) and multiple regressions between male brain mass, female brain mass and root-to-tip dN/dS. These results are presented in the supplementary information.
Tests for gene-phenotype co-evolution
Gene-phenotype associations were tested for using phylogenetically controlled regressions between Log10-transformed root-to-tip dN/dS and phenotypic traits (Montgomery et al., 2011). These were performed using phylogenetic generalized least squares (PGLS) models implemented in BayesTraits (Pagel, 1999). Several previous analyses which have attempted to test for association between phenotypes and dN/dS ratios have typically used the dN/dS of the terminal branch (e.g. Dorus et al., 2004; Nadeau et al., 2007; Wlasiuk & Nachman, 2010). The root-to-tip dN/dS values are more inclusive of the evolutionary history of a locus, are properties of the species tips, rather than the terminal branch, and are not subject to temporal effects on dN/dS (Wolf et al., 2009). They are therefore more suitable for regressions against phenotypic data from extant species whilst simultaneously controlling for phylogenetic nonindependence (Montgomery et al., 2011). Although the data values for each species are not phylogenetically independent, this is accounted for in the PGLS regression. With PGLS, the phylogeny is converted 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’) (Pagel, 1997). With PGLS regression, the variance-covariance matrix is included in the error term of the regression model, and the resulting estimated regression parameters (i.e. slopes and intercepts) are ‘phylogenetically controlled’ (Pagel, 1997). The analysis is therefore robust to phylogenetic nonindependence.
We envisage two ways in which the evolution of a gene may contribute to sexual dimorphism based on the assumption that cellular processes can act to inhibit or promote brain development in either or both sexes. First, it is possible that evolutionary changes in a protein's structure could affect the way in which it responds to, or interacts with, components of sexually dimorphic gene networks, in a bi-directional way such that selection can act on the locus to increase or decrease brain size in either sex. In this case, for the purpose of the present analysis, the absolute difference between the phenotypes of each sex is more relevant than the direction of deviation, which may be controlled by separate loci. To test for this effect, we therefore used the absolute values of residuals calculated from the male/female regression.
In the second scenario, it is possible that a locus responds in a uni-directional way such that it responds to selection in a sex-specific manner (i.e. selection for increase male phenotype but not female phenotype, or vice versa). In this case, both the direction of deviation and the size of the deviation between sexes are of interest. Under this scenario, an association could be significant either in a positive direction, indicating a unidirectional response to selection acting to increase male brain size, or the negative direction, indicating a unidirectional response to selection acting to increase female brain size. In the PGLS analyses scores of sexual dimorphism values were modelled as the response variable and the Log10(root-to-tip dN/dS) as the predictor variable. Lambda, a parameter which measures the phylogenetic signal, was left as the default value of one as we found maximum likelihood estimates of lambda were not stable most likely due to the relatively small sample size. The significance of both tests was determined using a 2-tailed t-test.
Results and Discussion
A significant positive association was found between raw scores of sexual dimorphism in absolute brain size and the evolution of CENPJ (Table 1, Fig. 1). For MCPH1, including the Pan data point, we find a nonsignificant trend with absolute scores of sexual dimorphism (Table 1). However, when Pan, which has an inflated dN/dS, is removed this positive association with absolute scores of sexual dimorphism becomes significant (Table 1, Fig. 2). When strict Bonferroni correction for multiple tests is applied (two tests/gene) both associations remain significant. We next sought to test the robustness of the results using a Jack-Knife approach. The association between MCPH1 and raw scores of sexual dimorphism is revealed to be robust and is not dependent on any single data point (Table S2). The Jack-Knife analysis of the CENPJ association also suggests this association is relatively robust (Table S2). Although Ateles appears to be an outlier prior to phylogenetic correction, with a particularly high, female-biased sexual dimorphism score (−0.076; Table S1) its removal does not affect the significance of either locus (Table S2). In addition, we explored the possibility that the association may be driven by sexual dimorphism in body mass. For CENPJ, the result is not explained by an association with raw scores of sexual dimorphism in body mass (t19 = 0.664, P = 0.514, R2 = 0.022) and for MCPH1 the result is not explained by an association with absolute sexual dimorphism in body mass (t19 = 0.151, P = 0.881, R2 = 0.001). Neither MCPH1 nor CENPJ shows associations with either absolute or relative brain mass (Montgomery et al., 2011). Tests for an association with either measure of sexual dimorphism brain size were nonsignificant for ASPM and CDK5RAP2 (Table 1) and are not influenced by any data points with a major effect.
Table 1. Phylogenetically controlled regression analysis between Log(root-to-tip dN/dS) and sexual dimorphism in brain mass
Absolute male/female residuals
Raw male/female residuals
MCPH1 (no Pan)
These results suggest a possible link between the molecular evolution of MCPH1, CENPJ and the evolution of sexual dimorphism in anthropoid brain mass. Interestingly, the association is found with different measures of sexual dimorphism. The rate of evolution of MCPH1 is associated with absolute scores of dimorphism in brain size, suggesting a role in increased brain size, relative to the opposite sex, in both males and females. In contrast, CENPJ's association with the raw residuals indicates selection on this locus is specifically associated with increases in male brain size relative to female brain size; this could indicate some interaction with sex-specific developmental signals. This latter result appears particularly robust to alternative methods of measuring sexual dimorphism (supplementary information).
An important question is whether these relationships are causative, and if so how could MCPH1 and CENPJ control the development of sexual dimorphism? Sexual dimorphism in brain mass, grey matter and white matter first appears during early prenatal development (Gilmore et al., 2007) potentially before the second trimester (Joffe et al., 2005) and hence overlaps with the onset of neurogenesis (Rakic, 1995). Evidence linking sexual dimorphism to sex-specific apoptosis (Morris et al., 2004; Forger, 2006) suggests sexual dimorphism in whole brain size potentially results from differences in survival of neural progenitor cells. Apoptosis amongst neural progenitors is thought to have a major influence on the development of brain size (Haydar et al., 1999) and hence sex-dependent differences in cell survival at this early stage could result in sexual dimorphism in adult brain size.
MCPH1 is expressed in the foetal brain during neurogenesis and there is strong evidence linking it to the DNA damage response pathway (Wood et al., 2007, 2008). Wood et al. (2008) suggest MCPH1 acts early on in the DNA damage response cascade, potentially promoting the amplification of the response by binding to H2AX, which marks DNA damage sites. In this respect, it is interesting that the region of the MCPH1 gene that encodes the domain that interacts with H2AX, the C-terminal BRCT domain (Wood et al., 2007), contains a SNP (V761A) which is associated with variation in male but not female cranial capacity (Wang & Su, 2004). Hence, there is good evidence that MCPH1 interacts with the right cellular processes during neurogenesis to effect sex-specific development potentially by mediating sex-specific apoptosis of neural progenitors.
There are fewer clues as to how CENPJ may influence brain development in a sex-specific manner. CENPJ plays a crucial role in centriole and centrosome development, which may affect spindle orientation and therefore cell fate during mitosis, but the precise role it has in neurogenesis is not well defined (Thornton & Woods, 2009). However, the finding that several genes involved in centrosome and spindle pole structure and function are differentially expressed in males and females in developing chick brains, suggests genes involved in the centrosome may have a role in the control of sexual differentiation and brain development (Lee et al., 2009).
Until a clearer causative link can be established between these loci and sex differences in brain development, their role in sexual dimorphism must remain a hypothesis. It is important to note that SNPs of both ASPM and CDK5RAP2, which show no association with sexual dimorphism in brain size in this study, are associated with aspects of human brain size in a sex-specific manner, whilst there is currently no evidence for sex-specific associations for CENPJ in humans (Rimol et al., 2010). Hence, if microcephaly genes do indeed contribute to sexually dimorphic traits in the brain the mechanism for this effect is likely to be complex. However, the results presented here in Wang et al. (2008) and in Rimol et al. (2010) suggest understanding how microcephaly genes might have a sex-specific effect on brain development is an interesting endeavour.
We thank Marie Pointer and Judith Mank for helpful comments on a draft manuscript. We are grateful to Eric Vallender, Gavin Thomas and a third anonymous reviewer for constructive comments and advice. We thank the BBSRC, Leverhulme Trust and Murray Edwards College for funding.