Detecting tandem repeats associated with alcohol use and abuse

Alcohol use disorder (AUD) is characterized by problematic impulsivity, risk-taking, and impaired ability to control alcohol consumption. Individuals with AUD tend to have a poor quality of life with low educational attainment, poor sleep, and adverse social, occupational and health outcomes. Current treatment of AUD is ineffective, leading to adverse effects or relapse, partly due to diagnostic complexity and a paucity of knowledge of diagnostic and prognostic biomarkers. Tandem repetitive elements (TREs) in coding regions are an untapped source of genomic information that may inform actionable genetic influences on AUD beyond those detected by traditional genome-wide association studies. Genotyping of autosomal TREs from whole-exome sequences was performed with GangSTR v2.5.0. Genotypes for European (EUR) ancestry (N=85,173 participants were converted to a locus-level burden by summing the allele lengths at each locus. We used generalized linear models to associate TRE locus burden with items from the alcohol use disorder identification test (AUDIT) considering age, sex, sex*age, age2, and with the first 10 within-ancestry principal components as covariates. A genome-wide significance threshold (p < 5 × 10-8) was applied to account for multiple testing. We also investigated cross-population replication of EUR-discovered TREs associated with AUD. AlphaFold was used to predict the protein structural changes conferred by expanded and contracted TREs that localize to amino acid stretches. We report variable protein folding confidence relative to canonical amino acid sequences. We found 10 TREs that passed genome-wide multiple testing correction (p < 5 × 10-8). The most significant associations include TBX6-[TGA]N (Intron; β= -0.0142, P= 2.07 × 10-10) and SH2B1-[GCC]N (5’UTR; β= -0.00442, P= 6.58 × 10-10) with alcohol intake frequency, BRN1/POU3F3-[CCG]N (exon; β= -0.151, P= 6.73 × 10-8) with frequency of failure to fulfil normal expectations due to drinking alcohol in the last year; and SRF-[GAG]N (exon; β= -0.123, P= 6.34 × 10-11) and DPF3-[GGA]N (intron; β= -0.115, P= 2.12 × 10-10) with frequency of needing a morning drink after a heavy drinking session in the last year. Gene-based associations from the GWAS Atlas PheWAS were enriched for psychiatric pleiotropic effects (BRN1/POU3F3: 2.65-fold, P=5.63 × 10-12; TBX6: 2.13-fold, P=4.78 × 10-10; SH2B1: 1.66-fold, P=1.60 × 10-5; SRF: 2.12-fold, P=1.89 × 10-7; DPF3: 1.88-fold, P=3.05 × 10-7). The most enriched trait domain was social interaction and BRN1/POU3F3 (5.51-fold, P=1.31 × 10-4). TRE association with AUDIT items has the potential to inform alcohol use, misuse, and use disorder etiologies by linking genetic variation and biological function. The most promising locus identified by our study is BRN1/POU3F3, an intronless gene encoding a transcription factor essential to neurodevelopment. The stretch of prolines detected in this study is essential for the formation of a helix-turn-helix domain that facilitates DNA binding. These findings add to the robust genetic epidemiology work surrounding substance use disorder traits and highlight several strong and biologically relevant targets for further investigations.