SA31: GENETIC VARIANTS AS MODIFIERS OF THE ASSOCIATION OF BODY MASS INDEX WITH BIPOLAR DISORDER

Abstract

Background Bipolar Disorder (BD) is associated with higher Body Mass Index (BMI). In a prior Genome-Wide Association Study (GWAS) of BD, we found that rs12772424 in TCF7L2, a transcription factor in the Wnt signaling pathway, modified the association between BMI and BD (Winham et al., 2014), and subsequently replicated this finding in an independent sample (Cuellar-Barboza et al., 2016). Preliminary data further suggested that the TCF7L2-BMI interaction association with BD may be sex-specific, with stronger effects observed in females. Additional studies are needed to validate these findings and determine the role of other clinical factors and genetic modifiers in the BMI-BD association. Here, using data from the Psychiatric Genomics Consortium (PGC), we seek to identify additional genetic variations that may be contributing to the association of BD with elevated BMI. Methods Data from nine PGC sites were included in this study (3165 cases, 2645 controls); five sites provided BMI data for cases and controls, whereas four sites had BMI data only for cases. In all analyses, BMI was log-transformed. Separately for each site, case-only GWAS analyses were conducted to investigate SNP⁎BMI interaction effects by examining SNP effects with BMI as the outcome variable using linear regression, for SNPs that have not shown evidence of genome-wide significant BMI association in prior studies. Additionally, for sites with controls, case-control analyses were conducted using logistic regression, adjusted for BMI. Two tests were performed: 1 df test of SNP⁎BMI interaction, and 2 df test of SNP effect and the SNP⁎BMI interaction effect. Results of the site-specific analyses were then meta-analyzed for SNPs observed across all sites. Sex-stratified analyses were also conducted. Results No evidence of study heterogeneity was observed for the top SNPs. In case-only and case-control analyses of SNP-BMI interaction effects, no results were genome-wide significant (overall or sex-stratified). The top result in the case-control analysis for SNP-BMI interaction was rs736893 in GLIS3 (OR=1.03, P=3.6E-5), with a stronger result in males in sex-stratified analyses (OR=1.07, P=5.7E-6). The top result in the case-only analysis of SNP-BMI interaction was rs4285452 in GSDMC (beta=0.02, P=3.3E-6), which showed similar results in males and females. In sex-stratified case-only analyses, the strongest result was in females for rs7994174 in DCLK1 (beta=0.05, p=4.5E-6), where no effect was observed in males. Discussion Although no genome-wide significant findings were identified in analyses of SNP⁎BMI effects on BD risk, overall or stratified by sex, interesting candidate variants were observed in genes including GLIS3 and DCLK1. GLIS3 encodes a zinc finger protein that has been associated with neonatal diabetes and congenital hypothyroidism. DCLK1 encodes doublecortin like kinase 1, which is involved in neuronal migration, transport, neuronal apotosis, and neurogenesis; it is upregulated by BDNF, previously implicated in both BD and BMI. The retrospective data analyzed here can only identify associations, and does not allow us to determine directionality of these associations, so further investigation is needed to determine if SNP effects modify the effect of BMI on BD or vice versa. Future directions include analysis of imputed data, gene set analyses, analyses of secondary phenotypes related to BMI (e.g. binge eating disorder and weight gain), and analyses to investigate possible medication effects.