12. SHARED GENETIC ARCHITECTURE AND EVOLUTION OF HUMAN LANGUAGE AND MUSICAL RHYTHM TRAITS

Rhythm-related skills such as beat synchronization are phenotypically correlated with language traits, with implications for shared neural and genetic architectures. In particular, individuals with rhythm impairment may be predisposed towards disorders such as dyslexia and developmental language disorder. Shared risk is thought to be rooted biologically, given that the evolution of rhythm synchrony has been theoretically linked to multiple aspects of human communication. Genomic approaches bring a new dimension to explore theories on the co-evolution of language and rhythm. We hypothesize that identifying shared genetic architecture of rhythm and language and probing the evolutionary past of the common genetic factor, can reveal neural and biological underpinnings of rhythm and language traits. We used summary statistics from two large-scale genome-wide-association studies (GWAS) in samples of European ancestry from our collaborations with 23 and Me, inc.: 1) GWAS of a musical rhythm phenotype (beat synchronization) in N=606,825 individuals, and 2) GWAS of dyslexia in N=1,138,870. In line with phenotypic evidence, we found significant genetic correlations between beat synchronization and a set of language-related traits including dyslexia (rg(SE)=-0.28(0.02), PFDR=2.05 × 10-31) and language resting-state functional connectivity (rg(SE)=0.28(0.05), PFDR=8.72 × 10-32). We then performed a multivariate GWAS (mvGWAS) using Genomic Structural Equation Modelling in a bivariate common factor framework to capture the shared genetic factor associated with rhythm impairment and dyslexia, which we call D-RI. Our mvGWAS identified 18 genome-wide significant loci, with the strongest signal coming from the SNP rs28576629 (P=3.79 × 10-14) on chromosome 3, an intronic SNP located on PPP2R3A (implicated in calcium ion binding and protein phosphatase regulator activity). Next, we used S-PrediXcan to integrate gene expression information with D-RI summary statistics, and to identify genes whose predicted expression levels are associated with D-RI. We used the whole-blood and 13 brain joint-tissue imputation phenotype weights, yielding 342 unique genes significantly associated (PFDR < 0.05) with D-RI, including AC072039.2 expression in brain nucleus (Z-score=-7.74, PFDR=1.17 × 10-9). LDSC partitioned heritability analysis yielded significant SNP-heritability enrichments in various cell type-specific regulatory regions, including neuronal promoters (Enrichment(SE)=8.14(1.55), PFDR=3.38 × 10-5) and enhancers (Enrichment(SE)=4.43(0.35), PFDR=7.96 × 10-18). We investigated the associations between D-RI and white-matter tracts of evolutionary significance using LAVA, and identified a significant local genetic correlation between D-RI and the left hemisphere superior longitudinal fasciculus I (rg=1, PFDR=0.02), a white-matter tract implicated in auditory-motor connectivity. Overall, our results reveal the complexity of shared genetic and neural foundations of rhythm and language, and groundwork towards co-evolution debates.