Shared genetic risk factors for irritable bowel syndrome and mental health disorders found in large genetic studies

We compiled GWAS summary statistics from individuals of European ancestry inferred from genetic data from publicly available sources. First, we obtained the largest-to-date GWAS for irritable bowel syndrome (IBS), termed IBS-UKB/Bellygenes (53,400 cases and 433,201 controls). This dataset was derived from a prior meta-analysis (12) that integrated two independent cohorts: the UK Biobank (UKB; 40,548 cases and 293,220) and the Bellygenes initiative (12,852 cases and 139,981 controls). Specifically, cases were defined using a combination of Rome III symptom criteria, clinician-verified diagnoses, and hospital-linked ICD-10 codes across both cohorts. Controls were defined as digestively healthy individuals, with strict exclusion criteria applied to both cases and controls to remove participants with organic gastrointestinal diseases, such as inflammatory bowel disease or celiac disease, thereby preventing signal contamination. We also incorporated two recently released IBS datasets from the VA Million Veteran Program: IBS-MVP1 (survey-based phenotype; 15,794 cases and 299,874 controls) (17) and IBS-MVP2 (electronic health record diagnosis using PheCode 564.1; 14,867 cases and 427,355 controls) (17). Controls for IBS-MVP1 were defined as participants who did not report an IBS diagnosis in the enrollment survey, while IBS-MVP2 controls were EHR-derived non-cases lacking the IBS PheCode. The larger control pool in IBS-MVP2 reflects the near-universal availability of clinical records, whereas IBS-MVP1 is limited to participants who responded to the specific enrollment questionnaire.

GWAS data for ten additional psychiatric disorders were obtained from the Psychiatric Genomics Consortium (PGC): anorexia nervosa (AN; 16,992 cases and 55,525 controls) (18); attention-deficit/hyperactivity disorder (ADHD; 38,691 cases and 186,843 controls) (19), anxiety disorders (ANX; 7,016 cases and 14,745 controls) (20), autism spectrum disorder (ASD; 18,381 cases and 27,969 controls) (21), bipolar disorder (BD; 41,917 cases and 371,549 controls) (22), major depressive disorder (MDD; 170,756 cases and 500,199 controls) (23), obsessive-compulsive disorder (OCD; 53,660 cases, 2,044,417 controls) (24), panic disorder (PD; 2,147 cases and 7,760 controls) (25), post-traumatic stress disorder (PTSD; 29,556 cases and 166,145 controls) (26), and schizophrenia (SCZ; 74,776 cases and 101,023 controls) (27). Across PGC cohorts, controls were generally defined as cohort-specific non-cases, often screened for the absence of lifetime psychiatric disorders. Stringent control definitions were used in specific traits, such as “super-normal” individuals for ANX and trauma-exposed non-cases for PTSD. Additionally, family-based cohorts (e.g., ASD, OCD, and SCZ) utilized pseudo-controls derived from non-transmitted parental alleles. The GWAS summary statistics for neuroticism (NE), a stable personality trait reflecting negative emotionality, were obtained from the GWAS Catalog (28), comprising data from 523,783 participants. NE was analyzed as a continuous quantitative trait across population cohorts; therefore, no binary case-control definition was applied to this dataset.

Detailed information on all GWAS datasets analyzed in this study is provided in. 1

Inverse-variance-weighted meta-analyses were conducted using the “-meta-analysis” function in PLINK v1.90b7 (29), combining IBS-UKB/Bellygenes separately with IBS-MVP1 and IBS-MVP2. Fixed-effect meta-analytic P-values and odds ratios (ORs) were calculated. SNPs reaching genome-wide significance in the meta-analysis (P < 5 × 10⁻⁸) with suggestive significance (P < 0.01) in the corresponding single-trait GWAS were prioritized.

Each novel significant locus was queried through the GTEx Portal (https://www.gtexportal.org/home/↗). The GTEx project provides eQTL data from 49 human tissues (each with a sample size greater than 70), employing standardized pipelines for gene expression quantification, covariate adjustment, and cis-eQTL mapping. Significant associations were identified using an FDR ≤ 0.05 threshold. For each variant, eQTL associations across multiple tissues were extracted, and results meeting the significance threshold of P < 1 × 10⁻⁵ were retained and compiled.

Single-cell eQTL annotations were obtained using a similar manual lookup strategy via the scQTLbase platform (https://bioinfo.szbl.ac.cn/scQTLbase/Search/↗). Each SNP was individually queried to retrieve cell type-specific eQTL associations. scQTLbase is an integrated human single-cell eQTL database, covering 304 datasets, 57 cell types, and 95 cell states, encompassing approximately 16 million SNPs significantly associated with gene expression.

We utilized mQTL summary statistics from three published studies:1) McRae et al. (30): This study was based on the Brisbane Systems Genetics Study (n = 614) and the Lothian Birth Cohort (n = 1366), using the Illumina HumanMethylation450 array to measure methylation status, covering 485,512 DNA methylation probes. 2) Li et al. (31): This study provided cross-population (European and East Asian ancestry) cis-mQTL summary statistics, generated via inverse-variance weighted meta-analysis, with a significance threshold set at P < 10⁻¹⁰. The study aimed to quantify the extent of shared and heterogeneous genetic regulation across different ancestry groups. 3) Hannon et al. (32): This study included mQTL summary statistics for multiple complex traits and evaluated the extent to which mQTLs explain phenotypic variation. Each novel locus was matched against these datasets to extract relevant results, which were then compiled and organized.

We performed transcriptome-wide association studies (TWAS) using the MashR model (33) pre-trained on GTEx v8 data to leverage cross-tissue genetic regulation. GWAS summary statistics were imputed using the 1000 Genomes EUR reference panel via MetaXcan to maximize SNP compatibility. We analyzed all GTEx v8 tissues, prioritizing gut-brain axis relevance, including 13 brain regions, digestive tissues (colon/small intestine), liver, pancreas, adipose, and blood/immune tissues. Gene-level associations were assessed using S-PrediXcan (34, 35) for tissue-specific effects and S-MultiXcan for cross-tissue integration (condition number threshold=30). Significance was determined by integrating single-tissue p-values, cross-tissue consistency, and multi-tissue meta-analysis results.

To quantify the shared genetic architecture between irritable bowel syndrome (IBS) and psychiatric disorders, genetic correlation (r_g) estimates were computed through linkage disequilibrium score regression (LDSC) (36) using publicly available GWAS summary statistics. The analysis employed precomputed linkage disequilibrium scores derived from HapMap3 SNPs, which were calculated based on European-ancestry individuals from the 1000 Genomes Project Phase 3 reference panel to ensure population-matched LD structure. Standard quality control procedures were implemented, including the exclusion of SNPs with ambiguous strand orientation, low minor allele frequency (MAF < 1%), or poor imputation quality (INFO score < 0.9). The intercept term was constrained to account for potential residual population stratification and cryptic relatedness in the input GWAS data.

Given that the genetic correlations estimated by LDSC aggregate information from all variants across the genome, we performed additional analyses using ρ-HESS (Heritability Estimation from Summary Statistics) to evaluate pairwise local genetic correlations between each pair of traits within 1,703 predefined LD-independent segments (37). ρ-HESS is particularly suitable for estimating local genetic correlations within each of these segments, which have an average length of 1.6 Mb. To determine statistical significance, we applied Bonferroni correction, considering p-values below 0.05/1703 as statistically significant.

MTAG was applied to explore the genetic overlap between irritable bowel syndrome (IBS) and psychiatric traits by leveraging joint GWAS summary statistics (38). This method enhances the detection of trait-relevant loci through the integrated analysis of GWAS summary statistics across multiple traits. Compared to conventional inverse-variance-weighted meta-analysis, MTAG additionally accounts for sample overlap and incomplete genetic correlations between traits. In the initial step of MTAG, variants were filtered by excluding low-frequency single nucleotide polymorphisms (SNPs), duplicate SNPs, and those with strand ambiguity. Pairwise genetic correlations between traits were then estimated using LDSC, and these estimates were used to calibrate the variance-covariance matrix of the random-effects component. Following calibration, MTAG performed a random-effects meta-analysis to generate SNP-level summary statistics. Pleiotropic SNPs that reached genome-wide significance (P < 5×10^-8) in the multi-trait analysis and suggestive significance (P < 0.01) in the single-trait GWAS were prioritized. Due to its complex linkage disequilibrium (LD) structure, the MHC region was treated as a single locus. Results were visualized using Circos (39).

We applied the condFDR approach to identify loci associated with IBS by conditioning the genetic associations of IBS on each psychiatric trait. The condFDR method extends the standard FDR framework and operates within an empirical Bayesian framework, leveraging the combined power of two GWAS to enhance the discovery rate of genetic variants (40). This approach utilizes associations of single nucleotide polymorphisms (SNPs) with both the primary phenotype (IBS) and conditional phenotypes (e.g., schizophrenia) to prioritize variants that are more likely to represent true associations, even if their p-values do not reach the conventional genome-wide significance threshold (40, 41). The condFDR procedure reorders the SNP p-values from the primary phenotype (IBS) based on the strength of association in the conditional phenotypes (five psychiatric traits). As a result, IBS variants that show association with the five psychiatric traits are assigned lower condFDR estimates (40–42). Similarly, the conjFDR approach—implemented by reversing the roles of the primary and conditional phenotypes in an inverse condFDR analysis—enables the detection of SNPs associated with both the primary and conditional phenotypes. The conjFDR statistic is defined as the maximum of the two conditional FDR values. In our analysis, a threshold of 5% was applied for both conditional FDR and conjoint FDR.

We performed two independent inverse-variance weighted meta-analyses by integrating the IBS-UKB/Bellygenes dataset with each of the two MVP cohorts separately, accounting for the partial sample overlap between IBS-MVP1 (survey-based phenotype) and IBS-MVP2 (diagnosis-based phenotype). The resulting datasets—IBS-UKB/Bellygenes-MVP1 and IBS-UKB/Bellygenes-MVP2—identified nine and ten genome-wide significant loci (P < 5 × 10^-8), respectively. After excluding previously reported loci and the extended MHC region (chr6: 26–34 Mb), we detected six novel risk loci in IBS-UKB/Bellygenes-MVP1 and five in IBS-UKB/Bellygenes-MVP2 that were not genome-wide significant in the original IBS-UKB/Bellygenes dataset (Figure 1). All of these loci had previously shown suggestive evidence of genetic overlap between IBS and related traits, but did not reach genome-wide significance in the original IBS-UKB/Bellygenes GWAS due to limited statistical power. (Supplementary Table 2) (14, 40). Notably, rs1580180 (from IBS-UKB/Bellygenes-MVP1) and rs4738680 (from IBS-UKB/Bellygenes-MVP2) mapped to the same LD block (R² = 0.96), indicating convergence between the two phenotyping approaches despite differences in how IBS was defined in MVP1 (survey-based) versus MVP2 (diagnosis-based). In addition, rs20551 was robustly replicated across both meta-analyses, further supporting their potential roles in IBS susceptibility.

To explore the functional relevance of the newly identified loci from the meta-analysis, we conducted tissue- and cell-type-specific QTL annotation. Using GTEx v8 data, we identified multiple SNPs—including rs2014252 (2p23.3), rs2728934 (3p25.3), rs20551 (22q13.2), and rs1580180/rs4738680 (8q12.1)—as expression quantitative trait loci (eQTLs) in colonic tissue (Supplementary Table 3). Several genes near these variants, such as ATRAID, MEI1/XPNPEP3, UBXN2B, and SRGAP3, also demonstrated coordinated expression in both colon and brain, supporting the gut-brain regulatory axis in IBS pathophysiology (13, 14, 43). To further characterize cell-type specificity, we integrated single-cell eQTL datasets from peripheral blood mononuclear cells (44), human iPSC-derived midbrain neurons (45), and CD4⁺ T cells (46), revealing serotonergic and immune cell co-expression of ATRAID and XPNPEP3, with MEI1, UBXN2B, and SRGAP3 showing restricted expression in glial or progenitor subtypes (Supplementary Table 4). These genes align with prior findings implicating impaired serotonergic and dopaminergic signaling in IBS (47–50). Furthermore, methylation QTL (mQTL) profiling identified 81 novel associations, with rs2014252 and rs20551 showing broad regulatory effects on 16 and 55 CpGs, respectively, suggesting potential pleiotropic action via both transcriptional and epigenetic mechanisms (Supplementary Table 5).

Using TWAS analyses (IBS-UKB/Bellygenes-MVP1/2) centered on newly identified loci, we detected convergent signals across gut, brain, and immune-related tissues, consistent with gut-brain axis involvement in IBS. The most robust and replicated association was SUPT7L in cultured fibroblasts (P ≈ 2.4-2.97 × 10^-7). Notably, several genes showed cross-tissue patterns linking colon and brain. AGMO exhibited its strongest association in colon sigmoid (IBS-UKB/Bellygenes-MVP1, P = 3.94 × 10^-3), while TP53TG1 showed a tissue “switch,” peaking in transverse colon (IBS-UKB/Bellygenes-MVP1) and cerebellum (IBS-UKB/Bellygenes-MVP2) (P = 1.25 × 10–² and 5.90 × 10^-3, respectively). SIRT3 reached significance with primary signals in brain regions (e.g., anterior cingulate; P = 3.38 × 10–³ and 6.45 × 10^-3), and HLA-DQA1 demonstrated strong associations with marked tissue heterogeneity across brain regions (P ≈ 1-2 × 10^-3; z spanning negative to near-null). Additional brain-enriched signals included ZNF143 (nucleus accumbens), DLGAP1, KCNV2, PTBP1, and PHRF1, while lncRNA/transcript signals at the AARSD1/RP11-326C3 locus appeared across colon, blood, salivary gland, and cerebellar tissues. Immune modulators (PDCD1LG2, IRF7, AZU1) also showed suggestive associations, aligning with single-cell eQTL evidence in CD4⁺ T cells and neuronal/glial populations (serotonergic-like and dopaminergic neurons, oligodendrocytes). Collectively, TWAS highlights metabolic/mitochondrial (SIRT3) (51), lipid/membrane (AGMO) (52), immune-inflammatory (HLA-DQA1, PDCD1LG2, IRF7), and neuronal-glial regulatory nodes (DLGAP1) (53), supporting a multi-tissue regulatory architecture along the gut-brain axis in IBS (Supplementary Table 6).

SNP-based heritability estimates (h²) for irritable bowel syndrome (IBS) were generally low to modest across cohorts. The highest estimate was observed in the UK Biobank (UKB; h² = 0.0247 ± 0.0015), while both MVP1 (h² = 0.0110 ± 0.0019) and MVP2 (h² = 0.0089 ± 0.0012) exhibited lower values. Meta-analytic models modestly improved precision, with heritability estimates of 0.0189 ± 0.0010 and 0.0163 ± 0.0009 for IBS-UKB/Bellygenes-MVP1 and IBS-UKB/Bellygenes-MVP2, respectively. These heritability levels are notably lower than those observed in core psychiatric conditions such as schizophrenia (SCZ, h² = 0.379), Parkinson’s disease (PD, h² = 0.308), or autism spectrum disorder (ASD, h² = 0.204), and are more comparable to traits like obsessive-compulsive disorder (OCD, h² = 0.010), bipolar disorder type II (BD-II, h² = 0.013), and post-traumatic stress disorder (PTSD, h² =0.019) (). 1

To examine shared genetic architecture between IBS and psychiatric conditions, we performed linkage disequilibrium score regression (LDSC) across all IBS GWAS datasets, using the IBS-UKB-MVP1 meta-analysis as the representative result set (the independent IBS-UKB-MVP2 meta-analysis produced highly similar findings). Strong and highly consistent positive genetic correlations were detected between IBS and multiple brain-related phenotypes, particularly NE (rg = 0.576, P = 4.89 × 10^–¹²⁵) and MDD (rg = 0.566, P = 5.72 × 10^–¹¹⁸), as well as OCD (rg = 0.501, P = 2.67 × 10^–³⁸), PTSD (rg = 0.556, P = 3.02 × 10^–¹⁴), and ADHD (rg = 0.258, P = 3.74 × 10^–¹⁴). More moderate associations were observed for SCZ (rg = 0.206, P = 1.43 × 10^–¹³), ASD (rg = 0.246, P = 6.89 × 10^-9), BD (rg = 0.160, P = 1.03 × 10^-7), BD-II (rg = 0.264, P = 1.61 × 10^-6), and AN (rg = 0.169, P = 2.61 × 10^-6). Additional significant positive correlations were also observed for panic disorder (PD) (rg = 0.280, P = 1.64 × 10^-5), BD-I (rg = 0.116, P = 1.00 × 10^-4), and ANX (rg = 0.563, P = 2.00 × 10^-4), although these estimates were less precise. Compared with the meta-analyses involving UKB/Bellygenes and MVP, the single-cohort MVP analyses showed larger standard errors but maintained consistent directions of effect (Figure 2A; Supplementary Table 8). To visualize the broader architecture of shared liability, we constructed a correlation network based on pairwise genetic correlations (Supplementary Table 8; Figure 2B). The resulting network revealed that core affective and compulsive disorders—namely MDD, ANX, OCD, and NE—formed a tightly connected module. This cluster further linked to neurodevelopmental and psychotic traits such as SCZ, PTSD, and ADHD. Strikingly, multiple IBS datasets were positioned in close proximity to this central psychiatric cluster, underscoring a shared gut-brain axis genetic architecture.

To localize genomic regions driving these shared effects, we next performed local genetic correlation analysis. After correcting for multiple testing, significant regional correlations were observed in the extended major histocompatibility complex (MHC) region, particularly between IBS-UKB/Bellygenes-MVP1 and MDD (Figure 3A), and similarly between IBS-UKB/Bellygenes-MVP2 and MDD (Figure 3B). These results highlight potential immunogenetic mechanisms contributing to IBS-psychiatric comorbidity.

Building upon the enhanced genetic correlation between IBS and psychiatric disorders achieved through meta-analysis of IBS-UKB/Bellygenes and IBS-MVP, we sought to further identify their shared pleiotropic loci. Given the well-established strong brain–gut axis association in IBS (Figure 1), we first applied multi-trait analysis of GWAS (MTAG) to pair IBS with five psychiatric traits showing robust genetic correlations with IBS via LDSC (MDD, NE, OCD, PTSD, and ADHD). After excluding the MHC region, we identified 124 independent loci, 83 of which were novel to the original IBS-UKB/Bellygenes (Figure 4)—though many had been reported in subsequent joint analyses (14, 15). As expected, most novel loci were detected in MTAG analyses pairing IBS with MDD and neuroticism. Notably, several established loci reside near genes implicated in brain development and synaptic function, such as CADM2 (3p12.1) and NCAM1 (11q23.2). However, the MTAG analysis revealed additional novel loci linked to neuronal development, including RERE (1p36.23), SLC4A10 (2q24.2), SETD5 (3p25.3), CACNA2D1 (7q21.11), CDK5RAP2 (9q33.2), PPP1R3A (7q31.1), and LRFN5 (14q21.1). Furthermore, several novel pleiotropic loci were consistently identified across multiple MTAG analyses of IBS and various psychiatric disorders, including CCNH, SLC4A10, DCC, and MRO (Supplementary Tables 9-13). Importantly, we also detected several truly novel loci—unreported in either IBS-UKB/Bellygenes or prior joint analyses—such as three in the analysis of IBS-UKB/Bellygenes-MVP with MDD: SLC35D1 (1p31.3), RNU7-66P (6q12), and SORCS1 (10q25.1) (Supplementary Table 10). Among these, SORCS1 has been associated with Alzheimer’s disease (54) and regulates receptor levels involved in adhesion and neurotransmission (55). Meanwhile, the MTAG analysis with NE uncovered nine additional novel loci: PTCH2 (1p34.1), RNF150 (4q31.21), LINC02143 (5q34), COA1 (7p13), TLE1 (9q21.32), OR1B1 (9q33.2), ASCL3 (11p15.4), BATF (14q24.3), and EFTUD2 (17q21.31) (Supplementary Table 11). Notably, COA1 and OR1B1 have been linked to colorectal cancer risk or progression (56, 57), TLE1 is associated with Crohn’s disease, and both TLE1 and BATF are involved in T-cell-mediated immunity (58, 59). EFTUD2, crucial for neuroprotection, has been implicated in mental disorders when dysregulated (60, 61). These findings not only reinforce the brain-gut axis hypothesis in IBS but also unveil potential genetic bridges connecting IBS to cancer risk and immune mechanisms, further corroborating insights from prior TWAS analyses.

To further validate and refine the genetic overlap between IBS and psychiatric traits suggested by MTAG-identified pleiotropic loci, we applied condFDR analysis. This approach identified 57 independent loci, 52 of which were novel compared to the original IBS-UKB/Bellygenes—though most had been reported in subsequent joint IBS studies. The condFDR results strongly confirmed core pleiotropic hubs identified by MTAG, with key loci such as CCNH (5q14.3), LRFN5 (14q21.1), and NCAM1 (rs7106434) showing highly significant conjoint false discovery rates (conjFDR) (Supplementary Tables 14-17). Notably, the NCAM1 locus was further supported by proxy SNPs (rs10891490, rs2186710), reinforcing its role as a key genetic nexus in the gut-brain axis (43). Additional novel loci consistently identified by both methods—including RERE (1p36.23), SLC4A10 (2q24.2), and SETD5 (3p25.3)—highlight their involvement in neuronal development and synaptic function. The condFDR analysis also revealed novel IBS-associated loci such as SLC35D1 (1p31.3), LSAMP (3q13.31), COA1 (7p13), and TLE1 (9q21.32). Among these, SLC35D1, COA1, and TLE1 corroborate previous MTAG findings. Notably, LSAMP—previously linked with ulcerative colitis in those of African genetic ancestry (62) and linked to psychiatric traits including bipolar disorder and suicidal behavior (63, 64)—may suggest a potential role in gut-brain axis mechanisms. In contrast to these robust findings, while AN exhibited a significant positive genetic correlation with IBS in LDSC, subsequent condFDR analysis identified only one shared locus (lead SNP: rs750350), a signal already reported in prior IBS GWAS.