A higher mutational burden in females supports a "female protective model" in neurodevelopmental disorders

Abstract

Increased male prevalence has been repeatedly reported in several neurodevelopmental disorders (NDs), leading to the concept of a "female protective model." We investigated the molecular basis of this sex-based difference in liability and demonstrated an excess of deleterious autosomal copy-number variants (CNVs) in females compared to males (odds ratio [OR] = 1.46, p = 8 × 10(-10)) in a cohort of 15,585 probands ascertained for NDs. In an independent autism spectrum disorder (ASD) cohort of 762 families, we found a 3-fold increase in deleterious autosomal CNVs (p = 7 × 10(-4)) and an excess of private deleterious single-nucleotide variants (SNVs) in female compared to male probands (OR = 1.34, p = 0.03). We also showed that the deleteriousness of autosomal SNVs was significantly higher in female probands (p = 0.0006). A similar bias was observed in parents of probands ascertained for NDs. Deleterious CNVs (>400 kb) were maternally inherited more often (up to 64%, p = 10(-15)) than small CNVs < 400 kb (OR = 1.45, p = 0.0003). In the ASD cohort, increased maternal transmission was also observed for deleterious CNVs and SNVs. Although ASD females showed higher mutational burden and lower cognition, the excess mutational burden remained, even after adjustment for those cognitive differences. These results strongly suggest that females have an increased etiological burden unlinked to rare deleterious variants on the X chromosome. Carefully phenotyped and genotyped cohorts will be required for identifying the symptoms, which show gender-specific liability to mutational burden.

Figure 1

Excess of Autosomal CNVs in Females Ascertained for NDs Odds ratios (ORs) and associated p values represent the enrichment of CNVs in females compared to males ascertained for NDs. The CNVs are stratified on the basis of criteria previously associated with deleteriousness: frequency (<1% and <1/1,000), size (400 kb and 1 Mb as cutoffs), and de novo variants. An additional and previously published filter was applied on the basis of the presence of an ND gene (see Material and Methods). ORs and p values were calculated with a two-tailed Fisher’s exact test (ns, not significant). Data on gender were available for 476 de novo CNVs.

Figure 2

The Deleteriousness of Autosomal SNVs in Males and Females Ascertained for ASD The deleteriousness of SNVs was significantly higher in females than in males ascertained for ASD. Truncating variants (gain or loss of stop mutations and frameshift mutations) mainly drove this increased burden in females, and the gender bias was most apparent for variants involving ND genes. Missense variants showed no or only a marginal excess of deleteriousness in female probands. Raw C-scores of nonsynonymous variants were compared between males and females. To perform the analysis on the most deleterious variants, we stratified the sample on the basis of the top 1% and 5% of the C-score distribution. The p values were computed by means of a Wilcoxon rank-sum test. Significant p values demonstrate higher C-scores in females than in males. Significant p values with similar medians indicate that the differences lay at the tail of the distribution, as demonstrated by the quantile-quantile (Q-Q) plots. Abbreviations are as follows: M, male; F, female; med, median; p val, p value; and ns, not significant. False-discovery-rate correction was not applied because only one hypothesis was tested in several nested subsets of the same sample. (A) Q-Q plot comparing the distribution of C-scores for all SNVs in males and females. Red dots with tick marks indicate the top 1%, 5%, 10%, 25%, and 50% for males. The excess of high C-scores (deleterious) in females was only visible in the top 1% and 5% of the C-score distribution. Low C-scores showed equal distribution across gender. (B) Q-Q plot comparing the distribution of C-scores for all SNVs intersecting ND genes in males and females. Red dots with tick marks indicate the same percentiles as in (A).

Figure 3

Excess of Maternally Inherited Deleterious Autosomal CNVs (A) Data on inheritance (maternal, paternal, or de novo) were available for 1,826 and 1,735 CNVs from Signature Genomics and the ISCA, respectively. These CNVs were selected by cytogeneticists, and inheritance was tested on the basis of the likelihood of their association with the proband’s neurodevelopmental phenotype. Maternal ratio in percentage, associated 95% confidence interval, and p values represent the enrichment of maternally versus paternally inherited CNVs. The p values were computed with a binomial test, and the null hypothesis was a balanced 50/50 inheritance. The CNVs were stratified on the basis of size (400 kb and 1 Mb as cutoffs). An additional and previously published filter was applied on the basis of the presence of an ND gene (see Material and Methods). Compared to small CNVs, large CNVs showed increased maternal inheritance. (B) Data on inheritance were available for all CNVs identified in 762 SSC probands ascertained for ASD. The ratio of maternally inherited CNVs is represented with the 95% confidence interval and associated p value. The CNVs were stratified on the basis of size and the disruption of an ND gene. Large CNVs disrupting ND genes were preferentially maternally inherited. ns, not significant.