Association Between Brain Structure and Alcohol Use Behaviors in Adults: A Mendelian Randomization and Multiomics Study

Abstract

Importance: Past studies have identified associations between brain macrostructure and alcohol use behaviors. However, identifying directional associations between these phenotypes is difficult due to the limitations of observational studies.

Objective: To use mendelian randomization (MR) to identify directional associations between brain structure and alcohol use and elucidate the transcriptomic and cellular underpinnings of identified associations.

Design, setting, and participants: The main source data comprised summary statistics from population-based and case-control genome-wide association studies (GWAS) of neuroimaging, behavioral, and clinical phenotypes (N = 763 874). Using these data, bidirectional and multivariable MR was performed analyzing associations between brain macrostructure and alcohol use. Downstream transcriptome-wide association studies (TWAS) and cell-type enrichment analyses investigated the biology underlying identified associations. The study approach was data driven and did not test any a priori hypotheses. Data were analyzed August 2021 to May 2022.

Main outcomes and measures: Brain structure phenotypes (global cortical thickness [GCT] and global cortical surface area [GCSA] in 33 709 individuals and left-right subcortical volumes in 19 629 individuals) and alcohol use behaviors (alcoholic drinks per week [DPW] in 537 349 individuals, binge drinking frequency in 143 685 individuals, and alcohol use disorder in 8845 individuals vs 20 657 control individuals [total of 29 502]).

Results: The main bidirectional MR analyses were performed in samples totaling 763 874 individuals, among whom more than 94% were of European ancestry, 52% to 54% were female, and the mean cohort ages were 40 to 63 years. Negative associations were identified between genetically predicted GCT and binge drinking (β, -2.52; 95% CI, -4.13 to -0.91) and DPW (β, -0.88; 95% CI, -1.37 to -0.40) at a false discovery rate (FDR) of 0.05. These associations remained significant in multivariable MR models that accounted for neuropsychiatric phenotypes, substance use, trauma, and neurodegeneration. TWAS of GCT and alcohol use behaviors identified 5 genes at the 17q21.31 locus oppositely associated with GCT and binge drinking or DPW (FDR = 0.05). Cell-type enrichment analyses implicated glutamatergic cortical neurons in alcohol use behaviors.

Conclusions and relevance: The findings in this study show that the associations between GCT and alcohol use may reflect a predispositional influence of GCT and that 17q21.31 genes and glutamatergic cortical neurons may play a role in this association. While replication studies are needed, these findings should enhance the understanding of associations between brain structure and alcohol use.

Figure 1.. Study Overview

GWAS indicates genome-wide association study; SNV, single-nucleotide variant.

Figure 2.. Bidirectional Mendelian Randomization of Brain Structures and Alcohol Use Behaviors

Results from the bidirectional 2-sample mendelian randomization analyses of brain structures and alcohol use behaviors. All Z scores (β [SE]) were derived using inverse-variance weighted estimator (IVW) methodology. A, Z scores of the 19 brain structures included in the primary MR analyses on alcohol use behaviors. B, Z scores of alcohol use behaviors on the 19 brain structures. Asterisk indicates significance at a false discovery rate of 0.05. IVW results were prioritized in our interpretation of the results, while other methods served to assess the robustness of IVW estimates. See eTables 18-21 in Supplement 2 for full results.

Figure 3.. Cell-Type Expression-Specific Integration for Complex Traits–Cell-Type Enrichment Analyses

A total of 120 cortical cell types were analyzed from the Allen Brain Map Human Multiple Cortical Areas SMART-sequence data set. Cell types are organized by broad class: excitatory (56 cell types), inhibitory (54 cell types), and nonneuronal (10 cell types). Thirty distinct cell types were nominally significant (P < .05); 27 were excitatory, and 3 were inhibitory. See eTable 43 in Supplement 1 for full results and eMethods in Supplement 1 for a full explanation of cell type nomenclature. AUD indicates alcohol use disorder; LDSC, linkage disequilibrium score regression.