A large-scale high-density linkage study of autism identifies multiple genome-wide significant loci. D. E. Arking1, L. A. Weiss2, C. W. Brune3, K. West1, E. H. Cook3, M. J. Daly2, A. Chakravarti1 1) Institute of Genetic Medicine, Johns Hopkins University, Baltimore, MD; 2) Broad Institute of Harvard and MIT, Cambridge MA; 3) Institute for Juvenile Research, University of Illinois at Chicago, Chicago, IL.

   Autism is a childhood neuropsychiatric disorder that, despite high heritability, has largely eluded efforts to identify genetic variants underlying its etiology. Using Affymetrix arrays we genotyped ~4,000 samples comprising 802 affected sib-pairs (ASPs) for ~500,000 SNPs. Samples were obtained from the NIMH Autism Genetics Initiative and AGRE, with each family having at least 1 child diagnosed with autism and a second child classified as autism, broad spectrum, or Not Quite Autism. We performed two genetic studies: a family-based linkage analysis (this abstract) to search for loci that harbor multiple, likely rare, susceptibility alleles, and, an association analysis to examine the role of common variants (Weiss et al., this meeting). For linkage analysis, we removed SNPs with frequency 20%, 1% missing data, 1 Mendelian errors, or in Hardy-Weinberg disequilibrium (P0.001) resulting in ~180,000 high-quality SNPs. Non-parametric ASP linkage was performed using MERLIN with linkage disequilibrium (LD) modeled by clustering SNPs with pair-wise r2 0.1, revealing 4 significant (lod 3.6) peaks at 6q25.2, 15q25.3, 17p12, 17q11.2, and 1 highly significant (lod=5.7) peak at 6q27. To verify that these results were not due to artifacts introduced by clustering or from SNPs in high LD, we removed all SNPs with pair-wise r2 0.2, and observed no significant change in linkage peaks. We also performed analyses using every other SNP, and again, observed no significant change, indicating that these results are unlikely to be influenced by nearby LD or poor genotyping assays. Aside from 15q25.3 and 17q11.2, these loci show little overlap with previous reports, and none of the loci overlap with suggestive regions of association. In summary, we have used high-density SNP genotyping to identify 5 genome-wide significant loci that contribute to autism susceptibility, and likely harbor rare variants exhibiting allelic heterogeneity.