Refining the scope of genetic influences on alcohol misuse through environmental stratification and gene-environment interaction

Abstract

Background: Gene-environment interaction (G × E) is likely an important influence shaping individual differences in alcohol misuse (AM), yet it has not been extensively studied in molecular genetic research. In this study, we use a series of genome-wide gene-environment interaction (GWEIS) and in silico annotation methods with the aim of improving gene identification and biological understanding of AM.

Methods: We carried out GWEIS for four AM phenotypes in the large UK Biobank sample (N = 360,314), with trauma exposure and socioeconomic status (SES) as moderators of the genetic effects. Exploratory analyses compared stratified genome-wide association (GWAS) and GWEIS modeling approaches. We applied functional annotation, gene- and gene-set enrichment, and polygenic score analyses to interpret the GWEIS results.

Results: GWEIS models showed few genetic variants with significant interaction effects across gene-environment pairs. Enrichment analyses identified moderation by SES of the genes NOXA1, DLGAP1, and UBE2L3 on drinking quantity and the gene IFIT1B on drinking frequency. Except for DLGAP1, these genes have not previously been linked to AM. The most robust results (GWEIS interaction p = 4.59e-09) were seen for SES moderating the effects of variants linked to immune-related genes on a pattern of drinking with versus without meals.

Conclusions: Our results highlight several genes and a potential mechanism of immune system functioning behind the moderating effect of SES on the genetic influences on AM. Although GWEIS seems to be a preferred approach over stratified GWAS, modeling G × E effects at the molecular level remains a challenge even in large samples. Understanding these effects will require substantial effort and more in-depth phenotypic measurement.

Keywords: SES; alcohol; genome‐wide environment interaction; stratified GWAS; trauma.

Figure 1.

Miami and Manhattan plots of the genomic associations for three simulated disease-causing loci for the a) exposed (top; n=54,617) and non-exposed (bottom; n=45,383) subgroups of a stratified GWAS, and for the b) SNP main effects (top) and SNP × environment interaction (bottom) effects, and c) joint test of SNP main and interaction effects in a genome-wide gene-environment interaction study (GWEIS; n=100,000). The red line represents the genome-wide significance threshold (p < 5e-8). Highlighted SNPs, in red, were simulated to have large (rs2044766), moderate (rs1521011), and no (rs7908745) main effects, respectively, but equal SNP × environment interaction effects.

Figure 2.

Manhattan plots of the genomic associations for interaction effects from a GWEIS of (a) trauma exposure with BroadAUD, (b) socioeconomic status (SES) with QuantAlc, (c) SES with FreqAlc, (d) trauma exposure with DrinkMeals, and (e) SES with DrinkMeals. The red line represents the genome-wide significance threshold (p < 5e-8).