Contribution of autosomal rare and de novo variants to sex differences in autism

Abstract

Autism is four times more prevalent in males than females. To study whether this reflects a difference in genetic predisposition attributed to autosomal rare variants, we evaluated sex differences in effect size of damaging protein-truncating and missense variants on autism predisposition in 47,061 autistic individuals using a liability model with differing thresholds. Given the sex differences in the rates of cognitive impairment among autistic individuals, we also compared effect sizes of rare variants between individuals with and without cognitive impairment or motor delay. Although these variants mediated different likelihoods of autism with versus without cognitive or motor difficulties, their effect sizes on the liability scale did not differ significantly by sex exome wide or in genes sex-differentially expressed in the cortex. De novo mutations were enriched in genes with male-biased expression in the adult cortex, but these genes did not show a significant sex difference on the liability scale, nor did the liability conferred by these genes differ significantly from other genes with similar loss-of-function intolerance and sex-averaged cortical expression. Exome-wide female bias in de novo protein-truncating mutation rates on the observed scale was driven by high-confidence and syndromic autism-predisposition genes. In summary, autosomal rare and damaging coding variants confer similar liability for autism in females and males.

Keywords: ASC; Autism Sequencing Consortium; SPARK; Simons Foundation Powering Autism Research for Knowledge; exome sequencing; rare variant association.

Figure 1

Exome-wide rare variant burden and liability in SPARK and ASC trio-sequenced cohorts (A) The sample size of the trio-sequenced individuals in the Simons Foundation Powering Autism Research for Knowledge (SPARK) study and the Autism Sequencing Consortium (ASC) cohorts. (B) Sex-stratified DNM enrichment and liability. To obtain sex-specific effect sizes on the observed scale, the average DNM rate per trio in autistic males (green dots) or females (blue dots) was divided by the average rate in sex-matched siblings not diagnosed with autism (rate ratio; left). Corresponding effect sizes on the liability scale (Z score; right) were measured as explained in Figure S5A and supplemental methods section 6. For sex differences in enrichment (red dots), the observed DNM rate in autistic females was divided by the rate in autistic males (a rate ratio > 1 thus indicates that females show a higher enrichment). The sex difference in variant liability was estimated by subtracting the Z scores of the male-only analysis from the female-only Z scores (a Z score > 0 indicates that females show a higher effect size on the liability scale). (C) Over-transmission and liability of inherited variants. These were assessed using similar comparisons between parental alleles transmitted to autistic individuals and untransmitted alleles (see subjects and methods). Error bars show 95% confidence intervals. See supplemental results sections 2.1.3 and 2.2.3 for further details on synonymous variant imbalances.

Figure 2

Sex differences in the association between exome-wide burden of damaging de novo and rare variants and co-occurring intellectual disability in ASC and SPARK Trio-sequenced individuals from the Autism Sequencing Consortium (ASC) and Simons Foundation Powering Autism Research for Knowledge (SPARK) study cohorts were divided into those who have co-occurring cognitive impairment (cog. imp.; solid dots) and those who do not have cognitive impairment or have missing information (no cog. imp. or unknown; blank diamonds). The sample size of these subgroups is given in (A). In each group, we examined the risk ratio and average liability attributed to damaging DNMs (versus the same group of siblings) (B) and rare inherited variants (transmitted versus untransmitted) (C). Variant burden and liability estimates were meta-analyzed between ASC and SPARK. Sex differences in DNM rate ratios were estimated by direct comparisons of autistic females and males (ratio > 1 means that females have a higher DNM rate), and sex differences in the effect sizes on the liability scale were estimated by subtracting the Z scores (score > 0 means that females have a higher effect size).

Figure 3

Rare variant burden in autistic individuals with and without cognitive impairment or motor delay in SPARK trio-sequenced cohort (A) The sample size in two sub-cohorts of trio-sequenced individuals from the Simons Foundation Powering Autism Research for Knowledge (SPARK) study divided based on the presence of co-occurring motor developmental delays or cognitive impairment (“motor or cog. imp.”). (B) Sex-stratified observed DNM rates (left) and average liability (right) (see subjects and methods). In addition to the comparisons versus siblings, autistic probands with motor or cognitive difficulties were compared directly to sex-matched autistic individuals without these co-occurring conditions (“with versus without”). (C) Over-transmission of rare inherited variants (left) and their average liability (right) (see subjects and methods). See supplemental results section 4.1.3 for details on the imbalance of synonymous variants. Limiting the analysis in (B) to ultra-rare DNMs in ancestry-matched autistic females and siblings showed well-balanced synonymous DNM burden (p = 0.42; Figure S28).