Genome-Wide SNP and Environment Interaction Study in Autism Spectrum Disorder

Background: Despite high heritability estimates for autism spectrum disorder (ASD), several recent studies show environmental factors, especially exposures during the prenatal window, play a larger role in ASD risk than previously thought. Genome-wide studies integrating genetic and environmental factors could provide further understanding of the contribution of environmental and genetic factors to ASD risk. Both technical and practical limitations have hindered this type of integrated study in the past. However, recent advances in gene-environment interaction (GxE) methods, as well as availability of environmental exposure, genetic, and phenotypic information on a large representative population of children, now make a genome-wide GxE study (GEWIS) for ASD feasible.

Objectives: Our primary study objective was to identify single nucleotide polymorphisms (SNPs) whose effects on ASD risk vary across levels of selected prenatal environmental exposures, including maternal alcohol, tobacco, and medication use, and maternal infections.

Methods: The Study to Explore Early Development (SEED) is a US multisite epidemiologic study of ASD with comprehensive phenotyping, broad prenatal environmental exposure information, and biospecimens available for DNA measurements. A total of 1,321 children, aged 2-5, enrolled in SEED phase 1 with a classification of ASD or population control (POP) were genotyped on the Illumina Omni1M or Affymetrix Axiom genome-wide genotyping platforms. After applying our rigorous quality control and imputation pipelines, using SHAPEIT, IMPUTE2, and the 1000 genomes reference panel (version3), we had a final analytic dataset of 1,307 samples (582 ASD and 725 POP) and over 16 million SNPs available for GxE analyses. Prenatal environmental exposure data were derived from retrospective maternal self-report using a structured interview. Specific exposures included maternal use of tobacco, alcohol, β-2 adrenergic receptor agonist or antidepressant medications, and any of 36 specific maternal infections. GxE analyses used two approaches: (1) a case-control likelihood ratio test (LRT) that allows for genetic main effects to be modified by exposures, and (2) a case-only LRT that overcomes problems with gene-environment dependence assumptions by utilizing an empirical Bayes-type shrinkage estimator. These two approaches were specifically designed to handle millions of genomic loci making this GEWIS effort on a moderately sized sample plausible.

Results: Using the case-only LRT, we identified a genome-wide significant interaction (P<1x10^-07) for prenatal exposure to alcohol and a locus on chromosome 15 within the TRPM1 gene. This locus also reached a suggestive association threshold level using the case-control LRT method. The interaction effect estimate for both tests suggest a synergist interaction between the genetic variant and prenatal exposure. Several loci on chromosome 6 also reached suggestive significance (P<1x10^-05) for increased ASD risk among individuals with prenatal exposure to maternal infections. These were not located near the major histocompatibility genes but instead are located near a long non-coding RNA shown to have brain-region specific expression in the Genotype-Tissue Expression project.

Conclusions: We identified a genomic locus on chromosome 15 whose effect on ASD risk varies by levels of alcohol exposure during pregnancy, as well as an intergenic region on chromosome 6 showing suggestive association with ASD risk in the context of prenatal exposure to infection.