Objectives: The goal of this study was to identify sex-specific genomic variants that underlie susceptibility to ASD using a family-based genome-wide association study (GWAS) approach.
Methods: We performed a GWAS analysis on 783 ASD families from the Autism Genetic Resource Exchange (AGRE), genotyped with the Affymetrix SNP Array v. 5.0. A series of stringent quality control (QC) procedures was performed on SNPs and individuals; Mendelian errors were reset to missing in the analyses. We used the multi-dimensional scaling (MDS) procedure for inferring population structure and restricted our final analysis to 2,423 individuals of European ancestry from 667 ASD families. Sixty percent of the multiplex European families were classified as male only (MO) while 40% contained at least one affected female (FC). A total of 337,319 autosomal SNPs, including 110 in the pseudoautosomal region (PAR) were analyzed. Association analysis was performed with the Pedigree-Based Association Test (PBAT) program, assuming an additive model. Applying a conservative Bonferroni correction for multiple testing resulted in a genomewide significance level of 1.48E-07.
Results: In the MO family set analysis we identified two novel loci for autism that reached statistical significance at the genomewide level (rs2535443, p = 3.8E-08, and rs311150, p = 1.2E-07); a third SNP, rs311149, in the same region showed evidence for suggestive association (p = 3.6E-07). These three markers, located in the introns of the XG gene in the PAR boundary of the short arm (PAR1) of Xp22.3, are in high linkage disequilibrium (LD) with each other (pairwise r2 = 0.96-1.0). In addition, markers rs1328250, rs9521354, rs9521355, rs9521356, and rs1328244, located in a 600 kb intergenic region on chromosome 13q33.3 between the MYO16 and IRS2 genes, showed suggestive association to ASD (p = 5.3E-07, 9.2E-07, 1.5E-06, 9.2E-07, 2.6E-06, respectively) in MO families. MYO16 encodes an unconventional myosin protein predominantly expressed during brain development, and has shown protein-protein interactions with neurexin 1-α, a presynaptic membrane cell adhesion molecule that is associated with autism. These SNPs had moderate to complete LD with each other (r2 = 0.63-1.0) and may serve as a proxy that captures the association of functional variants which regulate the expression or action of MYO16. These markers were not associated with ASD in the FC families despite the fact that the power analysis indicated there was adequate power to detect a risk effect in the FC family set, similar to that in the MO family set. Therefore the markers associated in the MO families may be sex-influenced loci.
Conclusions: Our results suggest that the pseudoautosomal boundary of PAR1 and the intergenic region on 13q33.3 may harbor novel male-specific genetic variants for ASDs.