Rare X-linked variants carry predominantly male risk in autism, Tourette syndrome, and ADHD

Abstract

Autism spectrum disorder (ASD), Tourette syndrome (TS), and attention-deficit/hyperactivity disorder (ADHD) display strong male sex bias, due to a combination of genetic and biological factors, as well as selective ascertainment. While the hemizygous nature of chromosome X (Chr X) in males has long been postulated as a key point of "male vulnerability", rare genetic variation on this chromosome has not been systematically characterized in large-scale whole exome sequencing studies of "idiopathic" ASD, TS, and ADHD. Here, we take advantage of informative recombinations in simplex ASD families to pinpoint risk-enriched regions on Chr X, within which rare maternally-inherited damaging variants carry substantial risk in males with ASD. We then apply a modified transmission disequilibrium test to 13,052 ASD probands and identify a novel high confidence ASD risk gene at exome-wide significance (MAGEC3). Finally, we observe that rare damaging variants within these risk regions carry similar effect sizes in males with TS or ADHD, further clarifying genetic mechanisms underlying male vulnerability in multiple neurodevelopmental disorders that can be exploited for systematic gene discovery.

Fig. 1. Study schema.

A We identified risk-enriched regions (RERs) on Chromosome X (Chr X) using microarray data from 48 quintets, consisting of two unaffected parents, one male autism spectrum disorder (ASD) proband (box in dark) and two unaffected male siblings, where at least one of the unaffected siblings shares the same Chr X origin as the proband (top panel). We identified 4 peaks (RERs) within Chr X non-pseudoautosomal regions (Chr X non-PAR), encompassing a total of 149 genes (bottom panel). B We then utilized published whole-exome sequencing (WES) data from the Simons Simplex Collection (SSC) and SPARK ASD cohorts for (1) sex-specific burden and transmission analyses (SSC for primary burden analyses, SPARK for validation via assessment of transmission disequilibrium, combined cohort for final estimation of effect sizes based on the extent of overtransmission in probands) and (2) risk gene identification (SSC & SPARK combined). For burden analyses, (1a) we leveraged SSC siblings as controls as they are well-characterized and do not have reported psychiatric or developmental disorders. §: Chr X data is not independent for SSC families with both a male proband and one or more male siblings, and therefore, we trimmed the male probands from such families and kept the male control siblings because they are the more limiting sample set. We similarly trimmed individuals from families with multiple female children, though in this case we removed unaffected female siblings because female probands are more limiting. For transmission analyses (1b) we combined SSC and SPARK samples to investigate whether rare damaging variants are overtransmitted in male probands. In this analysis, as the untransmitted variants in each individual serve as controls, we included all individuals from the SSC cohort. For risk gene identification (2), we integrated all SSC and SPARK male samples and conducted a modified transmission disequilibrium test. C We extended our analyses to Tourette syndrome (TS) and attention-deficit/hyperactivity disorder (ADHD) in order to determine whether RERs carry risk in other male-biased psychiatric disorders. See also Supplementary Tables 1, 2.

Fig. 2. Rare transmitted damaging variants are enriched in risk-enriched regions (RERs).

We defined RERs based on patterns of segregation in a microarray dataset from the Simons Simplex Collection (SSC) (see Fig. 1) and considered any regions outside of the four risk regions to be non-enriched regions (NERs). A We first compared the rate of (maternally) transmitted damaging variants in SSC probands versus SSC siblings, utilizing the rate of synonymous variants to control for potential differences in sequencing metrics and ancestries (see related Fig. 1 and Supplementary Table 1). Rare (minor allele frequency or MAF ≤ 0.1%) transmitted damaging variants are enriched in RERs in male (1014 probands versus 746 siblings) but not female probands (314 probands versus 811 siblings). Rare damaging variants in NERs are not enriched in male or female probands, nor are more common variants (MAF > 0.1%) enriched in RERs or NERs. B We next orthogonally quantified the enrichment of rare damaging variants in RERs in male probands from SSC and SPARK families by comparing the transmission probabilities of rare variants (13,052 male probands versus 2295 male siblings). Separately, likely gene-disrupting (LGD) and missense 3 (Mis3; PolyPhen2 [HDIV] score ≥0.957) variants are overtransmitted. Damaging variants consist of LGD and Mis3 variants. For each bar plot, the gray horizontal line indicates odds ratio (OR) = 1, the height of the bar represents the odds ratio derived from a one-sided Fisher’s exact test, and the black error bars denote the 95% confidence intervals. See also Supplementary Figs. 1–5 and Supplementary Tables 1–3.

Fig. 3. RERs are correlated with local recombination rates.

A Density curve for risk-enriched regions (RERs) (“RER”, red), and genes with the highest recombination rate (“Top”, blue, generated from HapMap). The overall recombination rates from HapMap project are indicated with a gray smooth line (“HapMap”). The red dots correspond to the top five SNPs identified in a Chr-X-wide association study for loci contributing to the female protective effect. B Top panel RER genes (red) tend to have high rates of recombination compared to all Chr X non-PAR genes (gray). However, the RER genes only partially overlap the distribution of the top 149 genes based on recombination rate (blue). Bottom panel, Venn diagram depicting the overlap between the 149 genes contained within RERs and the 149 genes with the highest recombination rate. These two gene sets significantly overlap (permutation test with 100,000 iterations), but most of the risk for rare transmitted damaging variants resides within RER genes (one-sided Fisher’s exact tests). C We compared the linear regression models for per gene counts of rare variants occurring across all of Chr X with the formula #ssc_pro.Dam ~ #ssc_sib.Dam + log(recombination rate) (x-axis) and #ssc_pro.Dam ~ #ssc_sib.Dam (y-axis). We transformed the recombination rate to a log scale in order to make its distribution more normal. We then performed F-tests to determine whether log(recombination rate) is a significant covariate. There is no significant difference between the two models (F = 0.253, P = 0.62), suggesting that recombination rate is not a significant predictor of the per gene count of rare damaging variants in probands Chr-X-wide. OR odds ratio. See also Supplementary Fig. 6.

Fig. 4. Modified transmission disequilibrium test identifies MAGEC3 as a high-confidence ASD risk gene.

We conducted gene discovery within the risk-enriched regions (RERs) using a modified transmission disequilibrium test for rare damaging variants. As a control, we also conducted gene discovery in the non-enriched regions (NERs). In each case, we created a quantile-quantile plot comparing the distribution of P values to a uniform distribution (red diagonal line). Chr X non-pseudoautosomal region (Chr X non-PAR) and exome-wide significances are indicated with blue and red horizontal dashed lines, respectively. 95% confidence intervals are shown with black dashed lines. The red dot signifies the only exome-wide significant gene. Genes outside the 95% CI but not significant after correction are labeled. See also Supplementary Tables 4–5.

Fig. 5. Rare damaging variants in risk-enriched regions (RERs) are also enriched in males with Tourette Syndrome (TS) or attention-deficit/hyperactivity disorder (ADHD).

We identified rare damaging variants based on proband data alone for affected males in cohorts ascertained based on ASD (n = 995), TS (n = 561), ADHD (n = 329), or epileptic encephalopathies (EE, n = 220). We also identified variants in unaffected male siblings from the SSC ASD cohort (n = 730). A Among these data, rare (minor allele frequency or MAF ≤0.1%) damaging variants are enriched within RERs in males from the ASD, ADHD, and TS cohorts. However, they are not enriched in males with EE, which has minimal sex bias, nor are they enriched in non-enriched regions (NERs) in any disorder. Likewise, more common variants (MAF > 0.1%) are not enriched in RERs or NERs in any disorder. We used unaffected male siblings from the SSC as controls in all comparisons. B Rare damaging variants appear to be more strongly enriched in male probands in the TS and ADHD cohorts, but this could be due to comorbid TS, ADHD, and/or ASD diagnoses in male probands within each cohort. Only individuals affected by one or two disorders are listed. *In the ASD cohort probands were reported with/without Tourette/Tic disorder, whereas in the TS and ADHD cohorts, probands were reported with/without TS. C We therefore conducted a Poisson regression analysis with ASD, TS, and ADHD status as covariates (see (B) for sample sizes), the results of which further suggest that male probands with TS and/or ADHD are the most likely to have rare damaging variants in RERs. We excluded patient groups with fewer than 100 samples (e.g., ASD + TS). For each bar plot in (A), the gray horizontal line indicates odds ratio (OR) = 1, the height of the bar represents the odds ratio derived from a one-sided Fisher’s exact test, and the black error bars denote the 95% confidence intervals. For each bar plot in (B) the gray horizontal line indicates rate ratio (RR) = 1, the height of the bar represents the rate ratio derived from a Poisson regression, and the black error bars denote the 95% confidence intervals. See also Table 1, Supplementary Fig. 2, and Supplementary Table 6.