Novel Risk Loci in Tinnitus and Causal Inference With Neuropsychiatric Disorders Among Adults of European Ancestry

Abstract

Importance: Tinnitus affects at least 16 million US adults, but its pathophysiology is complicated, and treatment options remain limited. A heritable component has been identified in family and twin studies; however, no large-scale genome-wide association studies (GWAS) have been accomplished.

Objective: To identify genetic risk loci associated with tinnitus, determine genetic correlations, and infer possible relationships of tinnitus with hearing loss and neuropsychiatric disorders and traits.

Design, setting, and participants: A GWAS of self-reported tinnitus was performed in the UK Biobank (UKB) cohort using a linear mixed-model method implemented in BOLT-LMM (linear mixed model). Replication of significant findings was sought in the nonoverlapping US Million Veteran Program (MVP) cohort. A total of 172 995 UKB (discovery) and 260 832 MVP (replication) participants of European ancestry with self-report regarding tinnitus and hearing loss underwent genomic analysis. Linkage-disequilibrium score regression and mendelian randomization were performed between tinnitus and hearing loss and neuropsychiatric disorders. Data from the UKB were acquired and analyzed from September 24, 2018, to December 13, 2019. Data acquisition for the MVP cohort was completed July 22, 2019. Data analysis for both cohorts was completed on February 11, 2020.

Main outcomes and measures: Estimates of single nucleotide variation (SNV)-based heritability for tinnitus, identification of genetic risk loci and genes, functional mapping, and replication were performed. Genetic association and inferred causality of tinnitus compared with hearing loss and neuropsychiatric disorders and traits were analyzed.

Results: Of 172 995 UKB participants (53.7% female; mean [SD], 58.0 [8.2] years), 155 395 unrelated participants underwent SNV-based heritability analyses across a range of tinnitus phenotype definitions that explained approximately 6% of the heritability. The GWAS based on the most heritable model in the full UKB cohort identified 6 genome-wide significant loci and 27 genes in gene-based analyses, with replication of 3 of 6 loci and 8 of 27 genes in 260 832 MVP cohort participants (92.8% men; mean [SD] age, 63.8 [13.2] years). Mendelian randomization indicated that major depressive disorder had a permissive effect (β = 0.133; P = .003) and years of education had a protective effect (β = -0.322, P = <.001) on tinnitus, whereas tinnitus and hearing loss inferred a bidirectional association (β = 0.072, P = .001 and β = 1.546, P = <.001, respectively).

Conclusions and relevance: This large GWAS characterizes the genetic architecture of tinnitus, demonstrating modest but significant heritability and a polygenic profile with multiple significant risk loci and genes. Genetic correlation and inferred causation between tinnitus and major depressive disorder, educational level, and hearing impairment were identified, consistent with clinical and neuroimaging evidence. These findings may guide gene-based diagnostic and therapeutic approaches to this pervasive disorder.

Figure 1.. Manhattan Plot of Tinnitus Genome-Wide Association Study Showing the Top Variants in 6 Independent Genome-Wide Significant Loci

Tinnitus is self-reported in UK Biobank participants of European ancestry (n = 172 995). The horizontal red line represents the Bonferroni-adjusted significance threshold of P = 5 × 10⁻⁰⁸. Red dots indicate single-nucleotide variations within the regional locus (see eFigures 2-7 in the Supplement for regional plots).

Figure 2.. Significant Genetic Correlations Between Tinnitus and Other Phenotypes

Data are shown with genome-wide association study summary statistics. Squares indicate significant correlations between tinnitus and other disorders after Bonferroni correction for 248 comparisons (P < 2.0 × 10⁻⁴). Error bars indicate 95% CIs. Genetic correlations were calculated using linkage disequilibrium score regression on publicly available summary statistics comparing nonoverlapping populations with the United Kingdom Biobank cohort.