New Linkage Analysis by the Autism Genome Project (AGP) Reveals Strong Evidence of Linkage to Multiple Loci as Well as Gene-Gene Interactions

Joachim Hallmayer representing the Autism Genetics Cooperative (AGC) and the Autism Genome Project (AGP).

Background: The AGP is an international consortium with the aim to find susceptibility genes for autism spectrum disorders (ASD). As a first step, the AGP conducted the most comprehensive linkage study of ASDs to date; currently a Genome Wide Association Study (GWAS) is in progress using the Illumina Human 1M Bead Chip. The original linkage and CNV study (1181 multiplex families, 10K SNP data), published in Nature Genetics, yielded a peak Z_LR = 3.6 on 11p12-p13 in the whole data set; no other loci crossed the threshold for suggestive linkage.

Objectives: a) To reexamine the linkage data using the model-integrated posterior probability of linkage, or PPL, to account for heterogeneity by dividing the data set into subsets based on IQ status in AS individuals (Low, Normal, or Missing).  b) To explore whether allowing for different male and female penetrances will substantially change the results. c) To test for gene-gene interaction between the engrailed 2 (EN2) gene and other markers in the genome

Methods: We use a quasi-Bayesian framework to reanalyze the AGP data, sequentially updating linkage evidence across data collection sites, separately for LIQ, MIQ, and NIQ families.

Results: Overall 98% of the genome shows PPL < 5%, and 90% of the genome gives evidence against linkage (PPL < 2%). In the NIQ group we find strong evidence of linkage to one locus, with 65% posterior probability of linkage (PPL) on 11p15.4-15.3; as well as PPL=24% over the original 11p12-p13 peak and 22% at 11q14.1.  In the LIQ group we obtain a similarly clean plot, with several striking linkage peaks: 55% on 1q31.3, 25% on 2p25.1, 50% on 11p15.2; 32% on 11p12p-13 (the original peak), 46% on 13q22.1 (a previously implicated AD/specific language impairment locus), and 95% on 16q21. Other than the original peak on 11p12-13, there is no overlap in findings between LIQ and NIQ groups. Small but positive evidence (PPL = 3%-17%) is also seen over CNTNAP2, the 16 microdeletion region, CENTG2, the Prader-Willi/Angelman del/dup region on 15q, and the MET region on 7, all of which have been implicated in AD.  Allowing for different male and female penetrances did not substantively alter the results.

We, then, incorporated genotypes for an SNP in the EN2 gene, that had been previously shown to be associated with autism, and repeated the genome scan allowing for two-locus (2L) epistasis.  We find evidence of interaction with EN2 at the 15q del/dup region (2LPPL=25%) in NIQ; and strong evidence at the 11p14, p15 regions (69%, 46%) and at 13q22 (54%) in LIQ, as well as at several additional loci not detected in the single-locus scan.

Conclusions: We have obtained multiple strong linkage and epistasis signals in the largest ASD linkage study to date.  Follow-up molecular work is underway, as are additional PPL analyses involving sex differences and interactions with other candidate genes.