Polygene – By – Prenatal Environment Interaction in Autism Spectrum Disorder Using Copy Number Variant Burden

Background:   Autism Spectrum Disorder (ASD) is a complex disease with both genetic and environmental risk factors.  Increased copy number variation (CNV) burden has previously been associated with (ASD).  The prenatal period is also of increasing importance towards understanding the environmental risk factors contributing to ASD etiology.  However, evidence for specific environmental risk factors has been inconclusive.  A joint approach allowing for the possibility that environmental effect may be modified by polygenic risk, as measured by CNV burden, may increase the ability to detect environmental effects. 

Objectives:   The purpose of this work is to evaluate the association between four prenatal self-reported environmental exposures – smoking, alcohol, beta-2 adrenergic receptor (B2AR) agonists, and selective serotonin reuptake inhibitors (SSRI) – and ASD, while allowing for potential effect modification by CNV burden in the Study to Explore Early Development (SEED). 

Methods:   SEED is a multi-site case-control study of children aged 3-5 years with ASD and a control group drawn from the general population. All children were born between September 2003 and August 2006.  Prenatal exposure to environmental risk factors was ascertained through a detailed maternal interview that occurred 3-5 years after pregnancy.  A Hidden Markov Model approach (PennCNV) was used to call CNVs from Illumina genotype array data.  Measures of CNV burden were assessed on both a genome-wide scale and restricted to ASD candidate regions determined by the SFARI gene online database.  Associations with environmental exposures were first assessed by likelihood ratio test (LRT) comparing the likelihood of a full model with environmental exposure plus covariate terms to a null model constraining environmental exposure term to zero (H₀: β_E = 0). Interactions were then tested by a 2-degree-of-freedom LRT comparison of a full model with CNV burden, environmental exposure, and interaction terms to a null model where environment and interaction terms are jointly constrained to zero (H₀: β_CNVxE = β_E = 0). All models contained sex and the first five ancestral principal components derived from principal component analysis on measured single nucleotide polymorphisms (SNPs) as covariates.

Results:   Allowing for effect modification by copy number burden revealed a significant association between prenatal SSRI use and ASD in the presence of large (>400kb) CNVs (LRT p = 0.033), where the CNV-free SSRI association was not statistically significant.  Additionally, a negative association between prenatal alcohol exposure and ASD was revealed among those with high CNV kb burden in SFARI candidate genes (LRT p = 0.011).  Prenatal smoking exposure was associated with ASD risk with and without consideration of CNV burden, although adjustment for potential SES confounders was not yet considered.

Conclusions:   To our knowledge, this is the first effort to evaluate genome-wide CNV burden by environment interaction in an ASD case and population-derived control sample. Allowing for the presence of effect modification by CNV burden permitted us to detect associations of ASD with self-reported prenatal alcohol and SSRI exposure where the marginal testing failed to do so.  Accounting for genetic risk may potentially elucidate mechanisms of environmental risk for ASD.