Abstract
Background The prevalence of co-occurring heavy alcohol consumption and obesity is increasing in the United States. Despite neurobiological overlap in the regulation of alcohol consumption and eating behavior, alcohol- and body mass index (BMI)-related phenotypes show no or minimal genetic correlation. We hypothesized that the lack of genetic correlation is due to mixed effect directions of variants shared by AUD and BMI.
Methods We applied MiXeR, to investigate shared genetic architecture between AUD and BMI in individuals of European ancestry. We used conjunctional false discovery rate (conjFDR) analysis to detect loci associated with both phenotypes and their directional effect, Functional Mapping and Annotation (FUMA) to identify lead single nucleotide polymorphisms (SNPs), Genotype-Tissue Expression (GTEx) samples to examine gene expression enrichment across tissue types, and BrainXcan to evaluate the shared associations of AUD and BMI with brain image-derived phenotypes.
Results MiXeR analysis indicated polygenic overlap of 80.9% between AUD and BMI, despite a genetic correlation (rg) of −.03. ConjFDR analysis yielded 56 lead SNPs with the same effect direction and 76 with the opposite direction. Of the 132 shared lead SNPs, 53 were novel for both AUD and BMI. GTEx analyses identified significant overexpression in the frontal cortex (BA9), hypothalamus, cortex, anterior cingulate cortex (BA24), hippocampus, and amygdala. Amygdala and caudate nucleus gray matter volumes were significantly associated with both AUD and BMI in BrainXcan analyses.
Conclusions More than half of variants significantly associated with AUD and BMI had opposite directions of effect for the traits, supporting our hypothesis that this is the basis for their lack of genetic correlation. Follow-up analyses identified brain regions implicated in executive functioning, reward, homeostasis, and food intake regulation. Together, these findings clarify the extensive polygenic overlap between AUD and BMI and elucidate several overlapping neurobiological mechanisms.
Introduction
Alcohol use disorder (AUD) and obesity adversely impact millions of individuals and contribute to hundreds of billions of dollars in combined annual economic cost (Apovian, 2016; Sacks et al., 2015). The detrimental health impacts of AUD include increased risk for cancers of the liver, head, and neck; cardiovascular disease; and liver disease including cirrhosis (Kranzler, 2023). Obesity, typically defined as a body mass index (BMI) of greater than 30 kg/m2 (Apovian, 2016), is associated with increased risk for hypertension, type II diabetes, coronary artery disease, liver disease, and various cancers (Hruby et al., 2016). The incidence and prevalence rates of co-occurring heavy alcohol consumption and obesity are increasing in the United States (Raza et al., 2023).
Substance use disorders, including AUD, are thought to share pathophysiological mechanisms with obesity. For example, the neurotransmitter dopamine plays integral roles in both eating and substance use behaviors by impacting motivation, self-regulation, and reinforcement (Volkow et al., 2008, 2017). Evidence of other overlapping neurocircuitry-based mechanisms that contribute to addiction and pathological overeating has also accumulated (Moore et al., 2017), which supports the concept of obesity as an addictive disorder (Lindgren et al., 2018), though this remains controversial (Gearhardt & Hebebrand, 2021). There is also overlap in the interaction of both alcohol and food with appetite-related neuroendocrine pathways such as ghrelin (Deschaine & Leggio, 2022; Farokhnia et al., 2019) and glucagon-like peptide-1 (GLP-1) (Klausen et al., 2022). Interest in this putative overlap has been heightened by recent reports that GLP-1 receptor agonists (GLP-1RAs)—approved for treating type 2 diabetes and obesity—may represent potential new pharmacotherapies for AUD (Leggio et al., 2023). Similarly, medications used to treat AUD reduce weight in individuals with obesity (Elmaleh-Sachs et al., 2023). For example, topiramate, recommended as an off-label treatment for AUD (Perry et al., 2022; Reus et al., 2018), is approved in combination with phentermine as a weight loss medication.
Both AUD and obesity have substantial genetic contributions, with an estimated heritability of 50% for AUD (Verhulst et al., 2015) and 40-70% for obesity (Loos & Yeo, 2022). Despite shared neurobiological pathways and a high rate of co-occurrence in some populations, the genetic correlation between alcohol- and BMI-related phenotypes is non-significant (Saunders et al., 2022; Zhou et al., 2023). In the two largest studies to date, null genetic correlations were reported between obesity and problematic alcohol use (i.e., a phenotype that combines AUD diagnoses and a quantitative measure of harmful drinking; rg = −0.03) and drinks per week (rg = 0.03) in European-ancestry individuals (Saunders et al., 2022; Zhou et al., 2023).
Although these findings are consistent with a modest amount of shared genetic variation between the two traits, an alternative hypothesis is that the presence of shared variants with both concordant and discordant effects across the two phenotypes obscures evidence of genome-wide correlation. Other analytic methods, such as bivariate causal mixture models, are not influenced by the directionality of the variants’ effects making them more appropriate to evaluate the extent of polygenic overlap between the two conditions (Frei et al., 2019). For example, accounting for concordant and discordant variant effects has revealed substantial shared genetic relationships between psychiatric and medical or cognitive traits despite small or null genetic correlations (Hindley et al., 2022).
In this study, we utilized MiXeR to investigate the overall shared genomic architecture between AUD and BMI. MiXeR is a statistical method that estimates the potential causal variants for each trait and the total degrees of overlap between two traits without regard to the direction of variants’ effects, thereby identifying the extent of unique and shared genetic architecture (Frei et al., 2019). MiXeR has been used to identify genetic overlap across multiple psychiatric disorders (Hindley et al., 2022), psychiatric disorders and irritable bowel syndrome (Tesfaye et al., 2023), and psychiatric disorders and cognitive traits such as educational attainment (Demontis et al., 2023; Hope et al., 2023). To complement MiXeR’s overarching approach, we also used the conjunctional false discovery rate (conjFDR) method to identify specific overlapping loci (Smeland et al., 2020). We hypothesized that the absence of genetic correlation between AUD and BMI is attributable to the presence of shared variants with inconsistent directions of effect and that this would be evidenced by both (1) greater genetic overlap than would be predicted by the observed genetic correlation and (2) shared loci showing a mixture of consistent and inconsistent effect directions. After testing that hypothesis, we conducted follow-up functional annotation and drug repurposing analyses on shared loci identified by conjFDR. Additionally, we examined the associations of AUD and BMI risk with brain image-derived phenotypes (IDPs) to identify potential shared neural underpinnings. In doing so, we aimed to uncover the genomic architecture and physiologic pathways shared by AUD and BMI to advance our understanding of mechanisms contributing to their comorbidity and aid in the development of physiologically-informed interventions.
Methods
Samples
We used summary statistics from two large-scale GWAS for AUD (N = 753,248; Ncase = 113,325) (Zhou et al., 2023) and BMI (N = 681,275) (Yengo et al., 2018). The AUD GWAS summary statistics were derived from a meta-analysis of several cohorts of individuals with AUD or alcohol dependence (AD) diagnoses and controls with no diagnosis. Despite differences in the specific criteria used for diagnosis across cohorts, the genetic correlation across diagnostic categories and cohorts was very high (Million Veteran Program AUD and Psychiatric Genomics Consortium AD rg = .98) (Zhou et al., 2020). We excluded UK Biobank participants from the AUD GWAS to minimize the overlap of participants between the two studies and inflation by cross-trait enrichment (Smeland et al., 2020). The BMI GWAS summary statistics were derived from a meta-analysis of two cohorts: the Genetic Investigation of Anthropometric Traits (GIANT) consortium and the UK Biobank. All participants were of European ancestry (for more details, see original publications and Supplemental Table 1).
Characterizing polygenic overlap
MiXeR (Frei et al., 2019; Holland et al., 2020) was applied to investigate the overall shared genetic architecture between AUD and BMI. Univariate MiXeR analyses were first conducted to estimate each trait’s polygenicity (i.e., the number of potential causal variants required to explain 90% of single nucleotide polymorphism [SNP] heritability) and discoverability (i.e., the average estimated effect size of causal variants). Next, bivariate models were implemented to identify the number of unique and shared causal variants for each pair of traits. These models also provide estimates of the proportion of causal variants with concordant directions of effect. The Dice coefficient, an indicator of the proportion of polygenic overlap, is also computed. Conditional Q-Q plots were produced to visualize cross-trait enrichment. These plots show the distribution of p-values for a primary phenotype as a function of its association with the secondary phenotype at three p-value strata (p ≤ 0.1, 0.01, and 0.001). As a secondary analysis, we used MiXeR to examine the shared genetic architectures of AUD and BMI with other psychiatric traits (major depressive disorder, attention-deficit/hyperactivity disorder [ADHD], schizophrenia) and height (Demontis et al., 2023; Howard et al., 2019; Trubetskoy et al., 2022; Yengo et al., 2018). This secondary analysis shows how the shared genetic architecture of AUD and BMI compares with that of other psychiatric traits and height. Also, for all traits, LD Score regression (LDSC) v1.0.1 was used to calculate heritability (Bulik-Sullivan et al., 2015), genetic correlations, and standard errors.
We used conjFDR (Andreassen et al., 2013; Smeland et al., 2020) analysis to detect loci significantly associated with both phenotypes, including variants with opposite directions of effect. To achieve this, conditional false discovery rate (condFDR) estimates were first obtained by conditioning the primary phenotype’s (i.e., AUD) test statistics on a secondary phenotype’s (i.e., BMI’s) SNP associations. A condFDR value for the second phenotype conditioned on the first’s SNP associations was calculated by reversing the order of phenotypes from the first condFDR assessment. ConjFDR defines the value for each association as the maximum of the two condFDR values for the given SNP, providing a conservative estimate of the SNP association with both phenotypes. Statistical significance was defined as a conjFDR value < 0.05.
Genomic loci definition and gene-set enrichment
SNPs having a conjFDR<.05, indicating significant SNP effects on both AUD and BMI, were input into Functional Mapping and Annotation (FUMA) v1.5.2 (Watanabe et al., 2017) to identify LD-independent genomic loci. Independent significant SNPs were identified using a LD block distance for merging of < 250 kb, r2 < 0.6, and the European ancestry 1000 Genomes reference panel (Auton et al., 2015). Of the independent SNPs, lead SNPs were identified using r2 < 0.1. Each locus is represented by a single lead SNP with the lowest conjFDR value. The novelty of lead SNP was determined by examining whether variants were genome-wide significant (p < 5×10-8) in the AUD and BMI summary statistics. Lead SNPs were assigned to genes based on presence within the gene or otherwise distance to the nearest gene transcription start site. Annotations for the lead SNPs corresponding to Variant Effect Predictor (VEP), Combined Annotation Dependent Deletion (CADD) scores, and nearest transcription start site were sourced from OpenTargets (https://genetics.opentargets.org/ v22.10) (Ochoa et al., 2023). The presence of lead SNPs within genes was confirmed using dbGaP (https://www.ncbi.nlm.nih.gov/gap/). Genes linked to the lead SNPs were then used to conduct gene expression, tissue enrichment specificity, and gene-set enrichment analyses in FUMA (Watanabe et al., 2017). Analyses were corrected for multiple comparisons using the False Discovery Rate (FDR) correction.
Drug repurposing
We integrated drug-protein interaction/druggability information from the Target Central Resource Database (TCRD) (Kelleher et al., 2023) and OpenTargets (Ochoa et al., 2023). We searched for each gene in each database (https://pharos.nih.gov/ v3.18.0; https://platform.opentargets.org/ v23.12). The TCRD divides target development/druggability into four levels: (1) Tclin targets have approved drugs with known mechanisms of action; (2) Tchem targets have drugs or small molecules that satisfy activity thresholds; (3) Tbio targets have no known drugs or small molecules that satisfy thresholds, but have Gene Ontology (GO) leaf term annotations or Online Mendelian Inheritance in Man (OMIM) phenotypes, or meet at least two of three conditions: a fractional PubMed count >5, >3 National Center for Biotechnology Gene Reference Intro Function annotations, or >50 commercial antibodies; and (4) Tdark targets—proteins that have been manually curated in UniProt but do not meet criteria for the above categories. OpenTargets was utilized to identify approved drugs and drugs in development that target identified druggable genes.
BrainXcan
We used BrainXcan software (Liang et al., 2021) to evaluate the associations of AUD and BMI with 327 brain image-derived phenotypes (IDPs) obtained from structural (T1-weighted) and diffusion magnetic resonance imaging (dMRIs). BrainXcan infers trait-IDP associations using GWAS summary statistics, brain feature prediction weights, and reference LD data. Prediction weights for BrainXcan were derived by training a ridge regression model on brain IDPs in 24,409 individuals from the UK Biobank. Effect sizes and p-values of trait-IDP associations were adjusted using LD block-based permutation, and Bonferroni correction was used to account for multiple testing (T1: 0.05/109 = 4.59 x 10-4; dMRI = 0.05/218 = 2.29 x 10-4). Based on these results, we identified brain IDPs that were associated with both AUD and BMI. We also examined the concordance of effect direction for brain IDPs across the two traits.
Results
Shared genomic architecture (MiXeR)
MiXeR analysis yielded an overall level of polygenic overlap between AUD and BMI of 80.9% (as quantified by the Dice coefficient) despite a minimal genetic correlation (rg = −0.03, SE = .02). Of the 10.5K and 11.2K potential causal variants linked to AUD and BMI, respectively, 8.8K were shared by the two traits (Figure 1). The estimated proportion of shared variants with a concordant direction of effect was 48.4%. Conditional Q-Q plots demonstrated enrichment of SNP associations with AUD that increased with the significance of the associations with BMI, and vice versa for BMI, both reflecting polygenic overlap (Supplemental Figure 1).
Notably, in comparison overlap between BMI and other psychiatric phenotypes, rgs between BMI and other psychiatric phenotypes ranged from −0.11 for schizophrenia to 0.30 for ADHD. There were substantial shared genetic variants between BMI and other traits, including 8.7K of the 10.1K schizophrenia variants (Dice coefficient = 82.0%; concordance = 45.3%), 7.1K of the 7.5K MDD variants (Dice coefficient = 75.9%; concordance = 54.4%), and 5.8K of the 7.9K ADHD variants (Dice coefficient = 60.9%; concordance = 68.4%).
There were moderate genetic correlations (rgs = .35-.41) and shared genetic variants between AUD and other psychiatric traits, including 7.3K of the 10.1K schizophrenia variants (Dice coefficient = 71.1%; concordance = 67.3%), 4.2K of the 7.5K MDD variants (Dice coefficient = 46.3%; concordance = 95.6%), and 4.0K of the 7.9K ADHD variants (Dice coefficient = 43.2%; concordance = 94.3%). MiXeR analysis indicated that of the 3.8K causal variants linked to the control condition height, 0.9K (Dice coefficient = 12.7%; concordance = 36.7%) and 2.4K (Dice coefficient = 32.5%; concordance = 41.1%) were shared with AUD and BMI, respectively (Figure 1).
Shared genetic loci (cond/conj FDR)
At conjFDR<0.05, we identified 132 significant loci associated with both AUD and BMI (Figure 2), 121 of which were unique from the AUD GWAS and 53 of which were unique from the BMI GWAS. Notably, all 53 loci that were novel for BMI were also novel for AUD (Supplemental Table 2). Of the shared loci, 56 lead SNPs (42.4%) had consistent effect directions for AUD and BMI and 76 (57.6%) had opposite effect directions. This supports the hypothesis that the lack of genetic correlation is due to mixed effect directions of variants shared by AUD and BMI.
The two most significant loci exerted opposite directions of effect for BMI and AUD. The first was an intronic variant (rs9939973) in the well-known obesity risk gene, FTO (also linked to numerous other traits, including substance use disorders, major depression, and schizophrenia; Chang et al., 2022). The second was an intronic variant (rs13114738) in the highly pleiotropic gene SLC39A8 (Pickrell et al., 2016). The most significant locus (rs10511087) with a concordant direction of effect was CADM2, which encodes cell adhesion molecule 2 and has been linked to array of cognition, pain, substance use, and metabolic phenotypes (Pasman et al., 2022; Sanchez-Roige et al., 2023).
Functional annotations
The Variant Effect Predictor (VEP) analysis of the 132 lead SNPs in the loci shared between AUD and BMI indicated that 74 were intronic, 31 intergenic, 7 downstream, 6 non coding transcript exon variant, 5 missense, 4 upstream, 3 in the 3’UTR, 1 synonymous, and 1 regulatory (Supplemental Table 2). Ten of the 132 variants had CADD scores > 12.37, indicating possible deleteriousness (Kircher et al., 2014).
Analysis of Genotype-tissue Expression (GTEx) (Aguet et al., 2020) samples showed that genes linked to the lead SNPs were significantly overexpressed in the brain (Supplemental Figure 2). Specifically, 47 genes were significantly upregulated in the frontal cortex (BA9), hypothalamus, cortex, anterior cingulate cortex (BA24), hippocampus, and amygdala (Figures 3 and 4).
GO gene-set analysis identified 30 biological processes where shared genes were significantly enriched, with “cell morphogenesis involved in differentiation”, “cell morphogenesis”, “axon development”, and “presynaptic active zone organization” being the top four processes (Supplemental Table 3). Additionally, genes were enriched for seven GO cellular component processes, including “synapse”, “GABAergic synapse”, and “presynapse”. Finally, the genes were enriched for 21 cell type signatures (La Manno et al., 2016), the top seven of which were in the midbrain and included “HGABA”, “HNBGABA”, “HDA1”, “HDA”, “HDA2”, “HNBML5”, and “HSERT”. These cell types are GABAergic, dopaminergic, serotonergic, and neuroblast-related (see Supplemental Table 3 for more information).
Drug repurposing analysis
Of the 131 unique genes associated with the 132 lead SNPs, six (OPRM1, RET, DPYD, ADH1A, PDE4B, PRKCB) were located in the gene or nearest to the transcription start site of genes associated with FDA-approved drugs (i.e., Tclin; Supplemental Table 2), including naltrexone (OPRM1) among others. Eleven other genes were Tchem (i.e., known to bind to small molecules with high potency), followed by 101 Tbio and 13 Tdark (Kelleher et al., 2023). Two of the 11 Tchem targets are targeted by known drugs in OpenTargets. Specifically, FTO is targeted by bisantrene for acute myeloid leukemia (Phase II), and GRM2 is targeted by four investigational drugs for central nervous system disorders including schizophrenia, major depressive disorder, perceptual disorders, bipolar disorder, psychosis, methamphetamine dependence, post-traumatic stress disorder, and seizure disorder.
BrainXcan
AUD was significantly associated with gray matter volume in four subcortical brain regions after applying p-value corrections for LD structure and multiple testing: the bilateral caudate nucleus (positive association), left amygdala (negative association), and right thalamus (positive association) (Supplemental Table 4, Supplemental Figures 4 and 5). After p-value adjustments, BMI had a total of 98 significant associations with brain IDPs. Approximately three-quarters of the BMI associations were with diffusion MRI IDPs (n = 74; 75.51%), with the remainder coming from structural MRIs. The top associations for BMI were with the medial lemniscus, a white matter tract that is part of the somatosensory pathway responsible for transmitting tactile, proprioceptive, and other sensations from the body to higher brain centers (Warren et al., 2017). All significant associations for AUD had the same direction of effect for BMI (Figure 5), 76 of the BMI associations had the same direction of effect as AUD, and 22 had effects in the opposite direction for the two traits (Supplemental Table 4, Supplemental Figures 6 and 7). Gray matter volume in three brain regions—the bilateral caudate and the left amygdala— was significantly associated with both AUD and BMI.
Discussion
Analyses revealed several key findings regarding the genetic relationship between AUD and BMI. Both MiXeR and conjFDR analyses showed that more than half the variants shared by AUD and BMI exerted opposite directions of effect on the traits, supporting our hypothesis that this underlies their lack of genetic correlation. Leveraging conjFDR, we identified 132 shared genomic loci, including 53 that were novel from the original GWAS for AUD and BMI. Expression analysis of genes linked to both phenotypes identified heightened expression in brain regions implicated in executive functioning, reward, homeostasis, and food intake regulation. BrainXcan analyses of brain IDPs from the UK Biobank reinforced these findings, identifying significant shared associations with caudate nucleus and amygdala. Overall, these findings detail the extensive polygenic overlap between AUD and BMI, elucidate several overlapping neurophysiological mechanisms, and suggest possible targets for intervention.
The low genetic correlation found in the present and prior studies (Saunders et al., 2022; Zhou et al., 2023) is explained by the mixed directionality of genetic effects of the phenotypes. MiXeR analysis indicated high polygenic overlap (81%) between AUD and BMI, with 48% concordance in the variants’ effect directions. Similarly, using conjFDR, 42.4% (n = 56) of the 132 loci significantly associated with both AUD and BMI had the same direction of effect. Comparison analyses examining the overlap of AUD and BMI with other psychiatric phenotypes (ADHD, MDD, schizophrenia) yielded a range of genetic correlations and effect direction concordance, all showing substantial genetic overlap, consistent with prior work (Bahrami et al., 2020; Torgersen et al., 2022; Zhou et al., 2023). AUD and BMI exhibited among the highest proportions of estimated shared causal variants of all other phenotype pairings, despite having the lowest absolute genetic correlation of any pairing, which underscores the importance of accounting for variants’ effect direction.
The FTO and SLC39A8 gene loci exhibited the most significant discordant effects, and the CADM2 gene locus exhibited the most significant concordant effects. Specifically, the effect allele of rs9939973, intronic within the FTO gene, was associated with a reduced likelihood of AUD and increased BMI. While BMI and alcohol-related GWAS have identified intronic FTO variants (Kranzler et al., 2019; Loos & Yeo, 2014; Zhou et al., 2023), this conjunctional analysis highlights this locus as the region that is most significantly associated with both phenotypes. Obesity research suggests that the FTO gene region alters the function of nearby genes (IRX3 and IRX5), which impact the involvement of fat cells in thermogenesis (Laber et al., 2021; Zhang et al., 2023). The psychiatric literature suggests that this region also affects neuronal activity, namely dopamine receptor type 2 and 3 function (Hess et al., 2013; Zhang et al., 2019). The SLC39A8 gene, which encodes the metal ion transporter protein ZIP8 (Nebert & Liu, 2019), was associated with both AUD and BMI, but with opposite directions of effect. This locus is highly pleiotropic and has been linked to an array of psychiatric and neurological diseases and Crohn’s disease, possibly due its role in maintaining manganese homeostasis (Pickrell et al., 2016; Sazonovs et al., 2022; Sunuwar et al., 2020). Conversely, the most significant variant with concordant effects on AUD and BMI was rs10511087, intronic within the CADM2 gene. CADM2 is expressed in some brain regions implicated in the present study, namely the frontal cortex (BA9) (Figure 3), and has been associated with cognition, pain, impulsivity, substance use, other risky behaviors, obesity, and other metabolic traits (Arends et al., 2021; Koller et al., 2024; Morris et al., 2019; Pasman et al., 2022; Sanchez-Roige et al., 2023).
Drug repurposing analysis yielded targets of FDA-approved drugs and others that are in development or have not been examined. In particular, PDE4B, which is targeted by FDA-approved medications for chronic obstructive pulmonary disease and psoriasis/psoriatic arthritis, has shown evidence of reducing alcohol consumption (Grodin et al., 2021; Grigsby et al., 2023) and is under investigation as a treatment target for AUD (NCT05414240), weight loss, and other metabolic conditions (Crocetti et al., 2022; Ferguson et al., 2022; Wu & Rajagopalan, 2016). Additionally, OPRM1 is targeted by naltrexone, which is FDA-approved for treating AUD and opioid use disorder, and as a combination drug, naltrexone-bupropion, for the treatment of obesity (Grunvald et al., 2022; Reus et al., 2018). Importantly, both the PDE4B and OPRM1 loci exhibited concordant effects on AUD and BMI. However, studies have not examined the effectiveness of these medications in treating both conditions simultaneously and this study underscores the relevance of examining treatment options focused on genes linked to both phenotypes.
Several brain regions and neural cell types were identified across the various downstream analyses. Genes linked to the lead SNPs were significantly up-regulated in the prefrontal cortex, hypothalamus, hippocampus, and amygdala. Additionally, amygdala and caudate nucleus gray matter volumes were significantly associated with AUD and BMI risk in BrainXcan analyses. Furthermore, GO gene-set analysis yielded enrichment for signatures in the midbrain, including GABAeric, dopaminergic, and serotonergic cell types. The brain regions and cell types identified here consistently been implicated in obesity (Gómez-Apo, 2021; Meng et al., 2020), binge eating disorder (Donnelly et al., 2018; Giel et al., 2022), and AUD (Fritz et al., 2022; Koob & Volkow, 2016). The prefrontal cortex and amygdala are dysregulated after drug use, resulting in withdrawal, craving, impulsivity, and negative affect, which drive continued use (Koob & Volkow, 2010, 2016). Similarly, the frontal cortex, amygdala, caudate, and hypothalamus exhibit hyperreactivity to food-associated cue exposure in obese and overweight individuals (Meng et al., 2020). Evidence also suggests that the caudate nucleus plays a key role in mediating external stimuli and internal preferences to guide behavior (Doi et al., 2020). The hypothalamus is integral to stress responses and homeostatic regulation of caloric intake to meet real and perceived nutrition needs (Gupta et al., 2020; Stover et al., 2023). Dysfunction in these areas is progressive with eating and drug use, resulting in altered reward processing and a shift in ‘liking’ vs. ‘wanting’ the hyper-fixated substance (Bodell et al., 2023; Morales, 2022; Morales & Berridge, 2020). Overall, these findings underscore the utility of leveraging human genomic and transcriptomic data analysis and align with insights from preclinical and human neuroimaging studies on the neurophysiological mechanisms driving AUD and obesity.
There are notable limitations to this study. BMI may not optimally measure obesity, because there is no distinction between weight from fat, bone, or muscle mass. However, BMI is easily and inexpensively measured, and is a longstanding, well-studied surrogate measure of obesity (Fryar et al., 2015). Nonetheless, future studies should expand these analyses to examine complementary phenotypes, such as body composition, waist-to-hip ratio (Pulit et al., 2019), and binge-eating disorder (Munn-Chernoff et al., 2021), and to clinical subpopulations (e.g., through GWAS of BMI in individuals with and without AUD) (Polimanti et al., 2017).
A second limitation is that the data used in this study are limited to individuals of European ancestry to ensure compatible genetic architectures. As sample sizes increase, a high priority must be placed on including individuals of non-European ancestry. The potential consequences of excluding diverse ancestral groups include the inequitable distribution of the benefits of genetic research and exacerbation of existing health disparities (Lofton et al., 2023; Ogden et al., 2020; Zapolski et al., 2014). Thus, future studies should replicate these findings using more diverse ancestral samples as biobanks continue to grow. This shift towards more diverse samples will improve result generalizability and understanding of cross-population genetic variation.
Conclusions
In summary, our study found that the absence of genetic correlation between AUD and BMI is attributable to the presence of shared variants with opposite directions of effect (i.e., a variant protective for obesity increases risk for AUD and vice versa). Follow-up analyses specified overlapping genomic loci and identified brain regions implicated in executive functioning, reward, homeostasis, and food intake regulation. Together, these findings increase our understanding of the shared polygenic architecture of AUD and BMI and lend further support to the notion that eating behavior and AUD share overlapping neurobiological mechanisms.
Data Availability
BMI summary statistics were downloaded from: https://portals.broadinstitute.org/collaboration/giant/index.php/GIANT_consortium_data_files AUD summary statistics (excluding UKB) were provided by the MVP SUD consortium.
Footnotes
↵* Equal contributions
Funding Source: This work was supported by funding from the National Institute on Alcohol Abuse and Alcoholism (R01 AA030041 and R01 AA030056), the Department of Defense (HU0001-22-2-0066), and the Veterans Integrated Service Network 4 Mental Illness Research, Education and Clinical Center of the Crescenz Veterans Affairs Medical Center. LL is a federal employee and is supported by the National Institute on Drug Abuse and the National Institute on Alcohol Abuse and Alcoholism Intramural Research Programs.
Disclaimer: The opinions and assertions herein are those of the authors and do not necessarily reflect the official views of the Henry M. Jackson Foundation for the Advancement of Military Medicine, Inc. Moreover, the opinions and assertions herein do not necessarily reflect the official views of the Department of Defense, Uniformed Services University, the National Institute on Alcohol Abuse and Alcoholism, the US Government, and do not imply endorsement by the Federal Government.
Disclosures: HRK is a member of advisory boards for Dicerna Pharmaceuticals, Sophrosyne Pharmaceuticals, Enthion Pharmaceuticals, and Clearmind Medicine; a consultant to Sobrera Pharmaceuticals; the recipient of research funding and medication supplies for an investigator-initiated study from Alkermes; and a member of the American Society of Clinical Psychopharmacology’s Alcohol Clinical Trials Initiative, which was supported in the last three years by Alkermes, Dicerna, Ethypharm, Lundbeck, Mitsubishi, Otsuka, and Pear Therapeutics. JG and HRK hold U.S. patent 10,900,082 titled: “Genotype-guided dosing of opioid agonists,” issued 26 January 2021. JG is paid for editorial work for the journal “Complex Psychiatry.”
References
- 1.↵
- 2.↵
- 3.↵
- 4.↵
- 5.↵
- 6.↵
- 7.↵
- 8.↵
- 9.↵
- 10.↵
- 11.↵
- 12.↵
- 13.↵
- 14.↵
- 15.↵
- 16.↵
- 17.↵
- 18.↵
- 19.↵
- 20.↵
- 21.↵
- 22.↵
- 23.↵
- 24.↵
- 25.↵
- 26.↵
- 27.↵
- 28.↵
- 29.↵
- 30.↵
- 31.↵
- 32.↵
- 33.↵
- 34.↵
- 35.↵
- 36.↵
- 37.↵
- 38.↵
- 39.↵
- 40.↵
- 41.↵
- 42.↵
- 43.↵
- 44.↵
- 45.↵
- 46.↵
- 47.↵
- 48.↵
- 49.↵
- 50.↵
- 51.↵
- 52.↵
- 53.↵
- 54.↵
- 55.↵
- 56.↵
- 57.↵
- 58.↵
- 59.↵
- 60.↵
- 61.↵
- 62.↵
- 63.↵
- 64.↵
- 65.↵
- 66.↵
- 67.↵
- 68.↵
- 69.↵
- 70.↵
- 71.↵
- 72.↵
- 73.↵
- 74.↵
- 75.↵
- 76.↵
- 77.↵
- 78.↵
- 79.↵
- 80.↵
- 81.↵
- 82.↵
- 83.↵
- 84.↵
- 85.↵
- 86.↵