Genome-Wide Analysis Identifies Global Rare Variation in Autism

Background: Autism spectrum disorders (ASDs) are highly heritable (≈90%), yet the underlying genetic determinants are largely unknown. Objectives: To understand the genetic and phenotypic heterogeneity in ASDs, the Autism Genome Project (AGP) Consortium conducted high-resolution genotyping and applied multiple analytical strategies to examine over 1,500 families for single nucleotide polymorphisms (SNPs) and copy number variation (CNV) affecting risk for ASDs. Methods: We analyzed 1,558 strictly defined ASD families with 1,000,000 SNPs. We conducted genome-wide association (GWA) analyses of SNPs and targeted analysis of SNP-tagged common CNVs to evaluate involvement of common alleles. We also analyzed array data to identify and characterize rare CNVs in autism families in comparison to large control datasets. Results: From four primary association analyses, the P-value for rs4141463 within MACROD2 crossed the GWA significance threshold of 5x10^-8. Two SNP-tagged common copy number variations (CNVs), rs12142922 (tagging exonic CNVR244.1 located at GBP3) and rs7576647 (tagging CNVR742_full), also showed strong association. Compared to controls, ASD cases carried a higher global burden of rare genic CNVs (1.19 fold, P=0.012), especially for loci previously implicated in ASD or intellectual disability (1.69 fold, P=3.4x10^-4). In addition, we identified a plethora of novel, rare, de novo and inherited CNVs implicating novel ASD genes including X-linked PTCHD1 (P=3.1x10^-3), and the autosomal genes SYNGAP1, DLGAP2, and SHANK2. Hallmarks of these CNVs included combinations of de novo and inherited events in a given family, incomplete penetrance, and non-segregation in families. Conclusions: Our results reveal profound etiological heterogeneity in ASDs. These studies further underscore the need for larger datasets to detect common alleles that confer low to modest risk and highlight the important role that rare variation plays in ASD etiology. Since rare CNVs clearly contribute to ASD risk, we anticipate that rare, deleterious sequence variants will be equally important. Large-scale sequencing projects that target large candidate gene sets or whole exome or genome for analysis will test this hypothesis. Ultimately, larger well-characterized ASD cohorts will be required to elucidate the full range of both common alleles of low effect size and rare variants conferring substantial risk, such as those identified in this study. Despite the heterogeneity of ASD etiology, understanding the underlying genetic architecture will reveal gene networks that will ultimately serve as targets for therapeutic intervention.