Bidirectional genetic overlap between autism spectrum disorder and cognitive traits

Abstract

Autism spectrum disorder (ASD) is a highly heritable condition with a large variation in cognitive function. Here we investigated the shared genetic architecture between cognitive traits (intelligence (INT) and educational attainment (EDU)), and risk loci jointly associated with ASD and the cognitive traits. We analyzed data from genome-wide association studies (GWAS) of INT (n = 269,867), EDU (n = 766,345) and ASD (cases n = 18,381, controls n = 27,969). We used the bivariate causal mixture model (MiXeR) to estimate the total number of shared genetic variants, local analysis of co-variant annotation (LAVA) to estimate local genetic correlations, conditional false discovery rate (cond/conjFDR) to identify specific overlapping loci. The MiXeR analyses showed that 12.7k genetic variants are associated with ASD, of which 12.0k variants are shared with EDU, and 11.1k are shared with INT with both positive and negative relationships within overlapping variants. The majority (59-68%) of estimated shared loci have concordant effect directions, with a positive, albeit modest, genetic correlation between ASD and EDU (r_g = 0.21, p = 2e-13) and INT (r_g = 0.22, p = 4e-12). We discovered 43 loci jointly associated with ASD and cognitive traits (conjFDR<0.05), of which 27 were novel for ASD. Functional analysis revealed significant differential expression of candidate genes in the cerebellum and frontal cortex. To conclude, we quantified the genetic architecture shared between ASD and cognitive traits, demonstrated mixed effect directions, and identified the associated genetic loci and molecular pathways. The findings suggest that common genetic risk factors for ASD can underlie both better and worse cognitive functioning across the ASD spectrum, with different underlying biology.

Fig. 1. MiXeR-modeled genome-wide genetic overlap between autism spectrum disorder (ASD), educational attainment (EDU) and intelligence (INT).

Venn diagrams from MiXer analyses shows the number of shared and trait-specific “causal” genetic variants in thousands for ASD & EDU and ASD & INT. The MiXeR estimated DICE coefficient for ASD & EDU was 0.90 and for ASD & INT it was 0.91. Both analyses had positive AIC values when comparing modeled estimates to minimum possible overlap but negative compared to maximum possible overlap, indicating that the estimates may underestimate genetic overlap. Rg: MiXeR estimated genome-wide genetic correlation.

Fig. 2. Conditional Q–Q plots.

Conditional QQ plots of observed versus expected −log10 p-values in the primary trait (ASD) as a function of significance of genetic association with the secondary traits intelligence (a) and educational attainment (b) at the level of p ≤ 0.1 (red lines), p ≤ 0.01 (yellow lines) and p ≤ 0.001 (purple lines). Blue lines indicate all SNPs. Black dotted line is the expected Q–Q plot under the null hypothesis (no SNPs associated with the trait).

Fig. 3. Manhattan plots showing common genetic variants jointly associated with autism (ASD) and cognitive traits.

The plots show common genetic variants jointly associated with ASD and intelligence (a) and ASD and educational attainment (b) with the –log10 transformed conjFDR values for each SNP on the y-axis and chromosomal positions on the x-axis. The black dotted horizontal line represents the threshold for significant shared associations (conjFDR <0.05, i.e. –log10 (conjFDR >1.3)). Independent lead SNPs are encircled in black.