Identification of Gene-Environment Interactions Associated with Autism

Background: There is increasing interest in understanding genetic and environmental risk factors and their interplay in autism. However genome-wide gene-environment interaction studies have been hindered in the past mainly due to the lack of specific exposure and genome-wide genotyping data from the same individuals. Here we utilize a unique sample source to examine gene-environment interactions in autism: The Study to Explore Early Development (SEED). SEED is one of the only multi-site case-control autism studies with comprehensive phenotyping and genome-wide genetic and prenatal environmental exposure data for thousands of children.  

Objectives: The main purpose of this study is to identify genetic and environmental factors that influence risk for autism. More specifically, we sought to: (1) identify SNPs whose effects on autism risk vary across levels of selected prenatal environmental exposures; and (2) assess copy number variation (CNV) in SEED children to identify CNVs associated with autism.  

Methods: For our GxE analysis, we examined prenatal exposures across four domains including maternal use of tobacco, alcohol, and medication (B2ARs and SSRIs) as well as maternal infection. Prenatal environmental exposure information was derived from maternal self-reported data using a structured interview. Genotypes for 1,348 SEED children (606 cases and 742 controls) were measured using HumanOmni1 and Affymetrix Axiom arrays. After applying data quality control measures, and performing imputation to obtain > 6 million genotypes per person, initial analysis was performed using a new joint likelihood ratio test for marginal genetic main effects and gene-environment interaction. For our CNV analysis, we utilized PennCNV to call copy number variants from the SEED genotyping array data and to perform overall CNV burden analyses. We have also performed association tests to identify autism-specific CNVs.  

Results: Our preliminary GxE analysis, using data from 873 SEED children, revealed a genome-wide significant (P < 5x10^-7) interaction between genotype and smoking for several neighboring SNPs on chromosome 2. In a similar subset of SEED samples, we found a significant (P = 0.006) increase in the overall global burden of large rare CNVs in cases relative to controls and identified autism-associated amplifications and deletions in genes previously implicated in autism. We are currently in the process of completing our final GxE and CNV analyses for the complete set of 1,348 SEED samples and will present our findings at the conference.

Conclusions: We have identified copy number variants and interactions between genomic regions and specific in utero environmental exposures that are associated autism. This suggests coupling genetic and environmental exposure information to determine autism risk is more fruitful than looking for genetic marginal effects alone. SEED has proven to be a particularly useful resource for these types integrative of studies.