Objectives: The overall goal of this study is to identify genetic and environmental factors that together influence risk for autism. More specifically, we aim to identify SNPs whose effects on autism risk vary across levels of selected prenatal environmental exposures. We plan to further characterize identified gene-environment interactions (GxE) by determining the relationship between specific features of the exposure, such as which trimester the exposure occurred, and the associated SNP.
Methods: The Study to Explore Early Development (SEED) is a national epidemiologic study of autism with comprehensive phenotypic evaluation, broad prenatal environmental exposure information, and biospecimens available for DNA measurements. Thus, it is unique and particularly well suited for identification of GxE in autism. Using the HumanOmni1 BeadChip we measured genotypes at over 1 million loci. Prenatal environmental exposure information was derived from maternal self reported data using a structured interview. Specific exposures included maternal use of tobacco, alcohol, β-2 adrenergic receptor agonist or antidepressant medications, and maternal infection. For our GxE analysis, we utilize two recently developed approaches specifically designed for GxE to overcome previous limitations with traditional methods when applied to millions of genomic loci making this GEWIS effort on a moderately sized sample plausible. The first method, developed by Kraft et al., is a case-control likelihood ratio test that is sensitive to genetic main effects but unlike traditional methods allows for the possibility that the genetic main effect is modified by an environmental exposure. The second method, developed by Mukherjee et al., is a case-only likelihood ratio test for GxE that overcomes problems with gene-environment dependence assumptions problems by utilizing an Empirical Bayes type shrinkage estimator to allow for uncertainty in the model.
Results: A total of 968 children enrolled in SEED were genotyped, including 418 cases, and 550 general population controls. Several data quality control measures were implemented and potential genotyping errors at both the sample and SNP level were removed, leaving 878 samples (356 cases and 522 controls) and over 800,000 SNPs for GxE analysis. Results from our GEWIS analysis, currently underway, will be presented at the conference.
Conclusions: Here we present the first genome-wide study to examine gene-environment interactions in autism.