Polygenic Adaptation Underlies Evolution of Brain Structures and Behavioral Traits

Abstract

Many of the characteristics of psychiatric disorders, including their early age of onset, moderate to high prevalence (i.e., 1% for schizophrenia, >15% for major depression), reduced fecundity, and high heritability have led many to speculate on how risk alleles have persisted throughout evolutionary history. While several potential mechanisms for maintaining high allele frequency of risk variants have been hypothesized (e.g., weak positive selection, balancing selection), few have been empirically tested. Given the recent analytic advancements in detecting polygenic adaptation, we systematically evaluated evidence for selection across a total of twenty-five complex traits including ten neuropsychiatric disorders, three personality traits, total intracranial volume, seven subcortical brain structure volume traits, and four complex traits with no known neuropsychiatric associations. We tested each set of trait-associated variants for evidence of classical hard sweeps (i.e., extreme integrated haplotype scores, iHS), partial sweeps (i.e., extreme population differentiation, Fst), rapid evolution since divergence from Neanderthal (i.e., Neanderthal depletion score, NDS), ancient polygenic selection (i.e., Qx scores), and very recent polygenic selection within the past 2,000 years (i.e., trait singleton density scores, tSDS). Variants associated with schizophrenia (Qx = 208.36, p < 0.001), extraversion (Qx = 88.04, p = 0.001), subjective well-being (Qx = 113.90, p < 0.001), hippocampus volume (Qx = 108.98, p < 0.001), and putamen volume (Qx = 115.76, p < 0.001) were enriched for signatures of weak polygenic adaptation after accounting for genetic drift. However, a limitation of this approach is that the null model of neutral drift is computed with respect to ancestral and derived allelic annotations, therefore it is not possible to determine whether polygenic adaptation has favored risk/trait-increasing alleles or protective/trait-decreasing alleles. Thus, we also employed a directional test of selection (tSDS) which overcomes this limitation. Remarkably, tSDS is powered to detect very recent selection that has been active only in the past ~2,000 years. Here, we find that signatures of recent polygenic adaptation are enriched among SCZ-protective alleles (Z=-3.06, p=0.002) total intracranial volume-increasing alleles (Z=3.09, p = 0.002), but putamen (Z=-4.93, p < 0.001), amygdala (Z=-3.49, p<0.001) and pallidum (Z=-3.39, p<0.001) volume-decreasing alleles. These results provide the first genome-wide genetic evidence in support of the mosaic theory of brain evolution in humans, suggesting that individual brain regions can evolve and vary in size independent of one another. No trait studied displayed significant consistent evidence of classical hard sweeps or partial sweeps. To characterize the biological processes and functional elements potentially targeted by polygenic adaptation, we conducted expression Quantitative Trait Locus (eQTL) and gene set enrichment analyses. SNPs associated with schizophrenia were enriched for eQTLs in brain (p = 0.007) while immune eQTLs were overrepresented among putamen (p = 0.001) and intracranial volume (p = 0.001) associated variants. Our results suggest that alleles associated with behavioral and brain structure volume traits have experienced polygenic adaptation acting through neurodevelopmental and immune-mediated pathways.