The MR is established based on three core assumptions, as shown in Figure 1A : (i) relevance: IVs such as SNPs are strongly associated with exposures (p < 5 × 10^–8); (ii) independence: instrumental SNPs should be independent from the potential confounders that might influence the outcome and exposures; and (iii) exclusion hypothesis: SNPs are only associated with outcome through the exposures without other alternative ways, that is, if SNPs can directly affect the outcomes without through the exposure, the results of MR have horizontal pleiotropic effects and violated the third assumption (18). The most important step in MR analysis is the selection of the appropriate SNPs to be used as instruments ( Supplementary Figure 1 ), which must be strongly linked to exposures with low linkage disequilibrium (LD) (r² < 0.001, window size = 10000 kb). To meet the second assumption and reduce interference from confounders in Phenoscanner (http://www.phenoscanner.medschl.cam.ac.uk/), each instrumental SNP was estimated for possible connections with confounders, including BMI, white matter microstructure, sleep duration, insomnia, household income, and smoking. SNPs associated with these confounders were subsequently excluded from MR analysis. For harmonization, palindromic SNPs were excluded because the directions of the positive and negative chains could not be determined for the same alleles on both strands.

We used univariable two-sample MR study to investigate the causal associations between psychiatric disorders and IBS. Then we further investigated potential pathways from psychiatric disorders to IBS by using univariable and multivariable MR studies: First, the causal effects of the potential mediators proposed in Part 2.1.2 on IBS were investigated through a univariable two-sample MR study (β_B). Second, the causal associations between psychiatric disorders and these mediators that showed significant causal effects on IBS were also acquired (β_A). Lastly, MR mediation analysis was utilized to investigate whether the mediators associated with both psychiatric disorders (β_A; P < 0.05) and IBS (β_B; P < 0.05) mediated any causal effects of psychiatric disorders on IBS (45). For the mediation analysis, multivariable MR was performed as previously described (46). Briefly, IVs for both psychiatric disorders and mediators were included in the same MR model to investigate their direct effects on IBS after mutual adjustment. Any attenuation of the direct MR effects (β_C’; P<0.05) of psychiatric disorders, after adjusting for mediators, would support mediation through mediators ( Figures 1B, C ) compared with those observed in the two-sample MR measuring the total effects (β_C) of psychiatric disorders on IBS (47, 48). Additionally, if the β_A showed nonsignificant P values or the GWAS summary data for psychiatric disorders or mediators were not fully available, limiting multivariable MR analysis, the Sobel test was used to infer mediation effects (49).

All statistical analyses were performed using R 4.1.0 with the R packages “TwoSampleMR”, “MR-PRESSO”, and “MRlap”. In the MR estimation, we chose the inverse-variance weighted (IVW) method as the primary analysis, which can obtain an unbiased result only when all IVs are valid with no directional pleiotropy effects. MR analysis was also conducted using other methods to check the robustness of the results, such as the weighted median method and MR-Egger method, which can estimate the causal effect in cases where some IVs are invalid (31). The weighted median method estimates allow up to 50% of the information in the analysis to come from invalid IVs; thus, it provides consistent effect estimates even in the presence of IVs with heterogeneous effects (50). Considering pleiotropic effects, the MR-Egger method provides a valid causal estimate if the SNP-exposure associations are not correlated with the direct effects of IVs on the outcome (51). If the P values of MR estimation were less than 0.05, it indicates significant causal associations. Additionally, the MR-pleiotropy method was used to verify the horizontal pleiotropy, and the MR-Pleiotropy Residual Sum and Outlier method (MR-PRESSO) was used to detect outliers that might introduce bias into the results. Subsequently, the outliers were removed from further analyses. In addition, the presence of heterogeneity among different IVs was assessed using Cochran’s (IVW) and Rücker’s (MR-Egger) Q tests.

Given that our GWAS summary statistics are mostly drawn from European populations; the possibility of sample overlap cannot be ignored. To mitigate potential bias arising from sample overlap, we applied the MRlap function to correct the IVW results (52). If the corrected effects are closely consistent with the observed effects and do not exhibit significant differences, then we can be reasonably confident in the IVW-MR estimates. Conversely, if substantial differences occur, the corrected effect should be given priority because it is independent of sample overlap. GWAS summary data on exposures and outcomes were needed in MRlap analysis, and if it was not available, we assessed the bias and Type 1 error rates caused by sample overlap in MR (53).

After removing the SNPs in LD or palindromes, the outliers identified in the MR-PRESSO method, as well as the SNPs associated with confounders between psychiatric disorders and IBS (i.e., broad depression: white matter abnormalities (54, 55); MDD: sleep disorders (56–58); PTSD: household income (59, 60); schizophrenia: insomnia, smoking, and white matter abnormalities (61–64); bipolar disorder: BMI (65, 66); ADHD: insomnia (67)), the remaining SNPs utilized as IVs are shown in Supplementary Tables S1-S9 . Both the IVW MR method and weighted median method showed statistically significant associations between genetically predicted broad depression (IVW OR: 3.92, P: 0.007; weighted median OR: 3.11, P:0.023), MDD (IVW OR: 2.53, P: 3.20×10^-10; weighted median OR: 1.65, P: 0.006), anxiety disorder (IVW OR: 1.12, P: 0.006; weighted median OR: 1.09, P: 0.032), PTSD (IVW OR: 1.92, P: 5.02×10^-5; weighted median OR: 1.74, P: 0.001), schizophrenia (IVW OR: 1.08, P: 0.016; weighted median OR: 1.10, P: 0.023) and risk of IBS, although the MR-Egger method showed no statistical significance. The MR pleiotropy method identified non-significant horizontal pleiotropy in these associations, while the Q tests showed significant heterogeneity except for anxiety disorders ( Figures 2A–E , 3 ). Moreover, non-significant associations were found between genetically predicted bipolar disorder, bipolar I disorder, ASD, ADHD, AN, and the risk of IBS using all three methods (P > 0.05), and neither heterogeneity nor horizontal pleiotropy was identified in these associations, suggesting that these results were consistent ( Figures 2F–I , 3 , 4 ).

We investigated 24 potential mediators of the causal effects of psychiatric disorders on IBS after a literature review. As shown in Supplementary Table S10 , the gut microbiota genera, Erysipelatoclostridium (IVW OR: 0.895, P =0.028) and Streptococcus (IVW OR: 0.958, P: 0.002) had protective effects on IBS. However, the causal effects of the Actinobacteria phylum and its genus Bifidobacterium on IBS showed inconsistent results, i.e., the estimates of the IVW method and weighted median method showed the protective effects of the Actinobacteria phylum (IVW OR:0.944, P: 1.826E-5) and Bifidobacterium genus (IVW OR: 0.982, 0.045) on IBS, while the MR-Egger method showed anti-protective effects (Actinobacteria phylum, MR-Egger OR: 1.348, P: 0.046; Bifidobacterium genus, MR-Egger OR: 1.179, P: 0.002). No causal associations were found between IBS and the genera Allisonella, Enterorhabdus, Ruminococcus1, Ruminococcustorques, and Tyzzerella3.

Regarding blood metabolites, inflammatory biomarkers, neurotransmitters, and acetate showed protective effects on IBS (IVW OR: 0.719, P: 0.048); however, β-hydroxybutyrate showed anti-protective effects (weighted median, OR: 1.294, P: 0.004). Simultaneously, inconsistent results were also observed in the association between IL-6 and IBS, in which the IVW (OR: 1.137, P: 6.923E-16) and weighted median method (OR: 1.069, P: 0.009) showed that increasing levels of IL-6 had a causal effect on IBS, whereas, the MR-Egger method showed that IL-6 had a protective effect on IBS (OR:0.883, P =0.023). Additionally, there was no evidence supporting the causal associations among genetically predicted lipopolysaccharide (LPS), C-reactive protein (CRP), tryptophan, histamine, lactate, pyruvate, catalase, or brain-derived neurotrophic factor (BDNF) and IBS in any of the three methods of our MR study.

Insomnia was a significant risk factor for IBS (IVW OR: 2.051; P: 1.548E-5), as shown by all three MR methods. However, there was no significant causal relationship between sleep duration and IBS (IVW OR: 0.916; P: 0.740). Furthermore, we found no significant association between genetically predicted BMI and IBS using the three MR methods (IVW OR: 1.056, P:0.994) ( Supplementary Tables S11-S34 ).

Here, we assumed that psychiatric disorders with significant causal associations with IBS were exposures (broad depression, MDD, anxiety disorder, PTSD, and schizophrenia), and potential mediators with significant causal associations with IBS were outcomes to explore their causal associations (β_A). We found that broad depression (IVW OR: 0.598, P: 0.031) and PTSD (IVW OR: 0.867, P: 0.021) causally led to lower levels of acetate and an increased risk of insomnia (broad depression, IVW OR: 1.356, P: 0.001; PTSD, IVW OR: 1.279, P: 4.085E-10). In addition to broad depression and PTSD, an increased risk of insomnia was found in MDD (IVW OR: 1.234, P: 1.397E-07); however, all showed significant heterogeneity. Furthermore, the MR-Egger method showed a significant causal effect of genetically predicted schizophrenia on increased blood β-hydroxybutyrate levels (IVW OR: 1.164, P:0.007), with significant horizontal pleiotropy (P = 0.015). Unfortunately, since the GWAS summary data for IL-6 were not available, we were unable to study the causal effects of psychiatric disorders on IL-6 levels, and we did not find any potential mediators affected by anxiety disorders, which requires further exploration ( Supplementary Table S35 ).

Based on the results of step 2 and step 3 ( Figure 4 ), we conjectured that 1) the reduction in acetate might mediate the effects of broad depression and PTSD on IBS; 2) insomnia might mediate broad depression, MDD, and PTSD-induced IBS; and 3) the increase in β-hydroxybutyrate levels in the blood might mediate schizophrenia-induced IBS (P values of β_A, β_B, and β_C were all lower than 0.05). Thus, we performed a multivariable MR to investigate β_C’. We found a significant attenuation in the effect of genetically predicted broad depression on IBS after adjusting for acetate or insomnia (β_C’, P < 0.05), which mediated 12.60% and 16.00% causal effects of broad depression on IBS, respectively. Unfortunately, since the GWAS summary data for MDD, PTSD, and schizophrenia were not fully available, we were unable to perform multivariate MR analysis to identify their mediators. Therefore, Sobel tests were performed instead, in which we verified that insomnia mediated 16.20% and 27.14% of the causal effect of genetically predicted MDD and PTSD on IBS, respectively, and the increase in β-hydroxybutyrate levels in the blood mediated 50.76% of the causal effect of genetically predicted schizophrenia on IBS ( Table 2 ).

We found that the samples in GWAS on broad depression, MDD, anxiety disorder, PTSD, AN, blood metabolites, CRP, sleep duration, insomnia, and BMI might be overlapped with the samples from GWAS on IBS, because they all contained the participants from UK Biobank. The GWAS summary data on broad depression, blood metabolites (acetate, lactate, pyruvate, and β-hydroxybutyrate), sleep durations, insomnia, and BMI were available, thus the MRlap analysis was performed, which showed that the causal effects of these indicators on IBS, as determined by the MRlap correction, align with those obtained through the primary MR analyses ( Supplementary Table S36 ). The GWAS summary data on MDD, anxiety disorder, PTSD, AN, histamine, and CRP were not available in our included studies, however, low bias and Type 1 error rates were found ( Supplementary Table S37 ). These outcomes reaffirmed the robustness of the IVW method.

To the best of our knowledge, this is the first study to illustrate the causal associations and potential mediators between psychiatric disorders and IBS from a genetic perspective. In this two-sample MR study, we found that broad depression, MDD, lifetime anxiety disorder, schizophrenia, and PTSD were risk factors for IBS.

Depression and anxiety are common psychiatric symptoms of IBS. In a Chinese population, diarrhea-predominant IBS (IBS-D) and depressive disorder showed two shared genetic variants (SYT8 rs3741231 G allele and SSPO rs12536873 TT genotype) associated with neurogenesis and neurotransmission, providing a genetic basis for the high comorbidity of IBS-D and depressive disorder (68). A recent meta-analysis showed that patients with IBS have an eight-fold greater risk of anxiety and a seven-fold greater risk of depression than controls in the Indian population (69). Two prospective studies have found that higher levels of anxiety and depression at baseline were significant predictors of IBS at the 1-year and 12-year follow-ups; conversely, IBS patients without anxiety and depression at baseline reported significantly higher levels of anxiety and depression at the 1-year follow-up, suggesting that independent gut-brain and brain-gut pathways operate in the pathophysiology of IBS (15, 70). However, it is likely that underlying confounders, such as changes in medication, life stressors, environmental exposure, socioeconomic burden, lifestyle, or other conditions unrelated to gastroenterology, will affect these data over a long or even a short period of time, leading to biased results (70). MR studies can methodologically exclude interference from these confounders. The results of our study agree with those of these longitudinal studies, further confirming the above evidence.

The association between PTSD and IBS remains controversial. A meta-analysis (10) comprising eight studies showed that PTSD was a significant risk factor for IBS. This finding is consistent with another study, which revealed a correlation between early-life abuse and trauma and a higher risk for the development of IBS in adulthood (8). However, some studies have disputed this association, as they found that a history of severe sexual/physical abuse was associated with higher pain thresholds for rectal distension in women with IBS (71). Additionally, a single-arm study showed that the prevalence of PTSD was not higher among patients with IBS than in the general population (72). Interestingly, our study confirmed that genetically predicted PTSD is a risk factor for IBS, opening new directions for future mechanistic studies.

Schizophrenia is a complex and debilitating brain disorder characterized by behavioral abnormalities, including cognitive dysfunction, psychosis, delusions, apathy, and withdrawal (73). Some risk factors for schizophrenia, such as inflammation, food intolerance, and Toxoplasma gondii exposure, partly involve the biological pathways of the gut, indicating that the gut-brain interaction may play a pivotal role in the pathophysiology of schizophrenia (73). To date, only a few studies have explored the association between schizophrenia and IBS, and meta-analyses have been conducted to verify this relationship. A cohort study by Lee et al. (74) showed no statistically significant difference in the incidence rate of schizophrenia between the IBS cohort and the cohort without IBS, even after 5 years of follow-up. Another small-scale retrospective study in 1997 suggested that the schizophrenia group exhibited a higher risk of IBS than the control group (11). After over 25 years, this result was reconfirmed by our MR study.

Abnormal levels of cytokines, including TNF-α, IL-8, and IL-10, are significantly associated with IBS and bipolar disorder symptoms, indicating that IBS and bipolar disorder share similar pathophysiological mechanisms (75–79). Thus, the comorbidity of both disorders was hypothesized. However, our MR study suggests no causality between bipolar disorder (including bipolar I disorder) and IBS incidence, which is consistent with the results of a meta-analysis of 11 studies that provided data on IBS (80). Conversely, IBS was found to increase the incidence of subsequent bipolar disorder in a nationwide cohort study; however, this finding needs to be further confirmed in future studies (74).

Prior to the mediation analysis, we explored the causal associations between genetically predicted psychiatric disorders and potential mediators that are significantly associated with IBS. The findings indicated that broad depression and PTSD causally led to lower levels of acetate, as well as an increased risk of insomnia. Additionally, MDD was found to causally increase the risk of insomnia, while schizophrenia causally increased blood β-hydroxybutyrate levels. Therefore, acetate, insomnia, and β-hydroxybutyrate might play crucial roles in psychiatric disorders-induced IBS.

Acetate, the most abundant short chain fatty acid (SCFA) produced by the gut microbiota and a precursor used by many gut commensals to produce propionate and butyrate, affects the metabolic pathway through the G protein-coupled receptor and free fatty acid receptor 2 in colonic cells to improve bowel function (81, 82). A reduction in acetate, propionate, and butyrate levels was found in patients with IBS and linked to specific IBS symptoms such as colonic hyperalgesia and hypersensitivity (83). Additionally, a study showed that acetate supplementation produces antidepressant-like effects by increasing histone acetylation and improving synaptic plasticity in the hippocampus (84). Interestingly, in our MR mediation analysis, we found that the reduction in blood acetate mediated 12.6% of the causal effects of broad depression on IBS, which further supports the key beneficial role of acetate in the brain-gut axis. Unfortunately, in our MR study, although PTSD was causally associated with lower levels of acetate, which was a risk factor for IBS, we found no mediating effects of acetate on PTSD-induced IBS in the mediation analysis.

Another blood metabolite, β-hydroxybutyrate, a ketone body that serves as an energy source during starvation or exercise, shows several beneficial effects in the treatment of seizures, hypertension, NLRP3-mediated inflammation, and neurodegenerative diseases (85). The correlation between β-hydroxybutyrate and IBS has not been documented; however, in the present MR study, β-hydroxybutyrate showed anti-protective effects on IBS. Huang et al. (86) found that patients with schizophrenia had significantly higher serum levels of β-hydroxybutyrate than healthy controls, which was significantly correlated with fasting glucose and triglycerides, thus speculating that serum levels of β-hydroxybutyrate may act as a potential indicator of energy utilization impairment in schizophrenia. Consistently, our study showed that genetically predicted schizophrenia was a risk factor for increasing the β-hydroxybutyrate level in the blood, which further mediated 50.76% of total effects of schizophrenia on IBS.

Emerging evidence suggests that insomnia is positively correlated with many psychiatric disorders, such as ADHD, bipolar disorder, PTSD, MDD, OCD, and schizophrenia; bidirectional causal associations between insomnia and psychiatric disorders have also been observed (87), which might be explained by shared common abnormalities in hypothalamic–pituitary–adrenal (HPA) axis activation, serotonin system dysfunction, and overexpression of immune system peptides (88). A Korean population-based cohort study (56) found that subjects with insomnia also showed a higher prevalence of IBS than those without insomnia, which further highlights the pivotal role of insomnia in brain-gut axis disturbances. In the present study, we found significant causal effects of genetically predicted broad depression, MDD, and PTSD on insomnia, which is consistent with the results of a previous study (87). Insomnia was also a risk factor for IBS, mediating proportion of 16.00%, 16.20%, and 27.4% of the total effects of broad depression, MDD, and PTSD on IBS, respectively.

Overall, our study revealed that acetate, β-hydroxybutyrate, and insomnia were important mediators in psychiatric disorders-induced IBS, supporting their key roles in brain-gut axis. Discovering the key mediators in the causal association between two diseases can not only help to reveal the mechanisms, but also identify markers of disease evolution or exacerbation, so as to facilitate early intervention. For instance, our study found blood β-hydroxybutyrate level mediated schizophrenia-induced IBS, suggesting β-hydroxybutyrate as a potential biomarker. If the blood β-hydroxybutyrate level is elevated in patients with schizophrenia, IBS is likely to be induced, which will worsen abdominal symptoms and further reduce the patients’ quality of life. Therefore, early intervention to improve blood β-hydroxybutyrate level and mental health is important. Of course, these outcomes need further validations by more studies.

Overall, our study has several strengths. First, in observational settings, MR studies can simulate randomized controlled trials, which are costly, laborious, and time-consuming. In contrast, MR studies can technically reduce cost and effort and efficiently avoid confounding bias for SNPs that are randomly assigned at conception (89). Second, compared to other observational studies, MR studies can avoid the reverse causal effect and are less likely to be affected by confounders between exposure and outcome. Third, we selected significant genome-wide SNPs for diseases in the GWAS, which provided large and repeatedly checked samples. After a rigorous process of removing outliers, SNPs in LP or palindromes, and SNPs associated with underlying confounders, our study was more reliable. Fourth, we focused on nine psychiatric disorders with a high or low probability of comorbidity with IBS and further verified the causal association between them at the genetic level, filling the gap of insufficient research on the relationship between IBS and certain psychiatric disorders, such as bipolar disorder, schizophrenia, ASD, ADHD, and AN. Fifth, the utilization of mediation analysis and the multivariable MR method to identify potential mediators between psychiatric disorders and IBS is conducive to identifying the mechanisms of gut-brain axis abnormalities, which may improve health policies for managing these patients. Lastly, although sample overlap between GWAS summary data on exposures and outcomes in the analysis was inevitable, the results of MRlap analysis as well as low bias and Type 1 error rates in MR indicated the bias induced by sample overlap should be minimal in these causal estimates.

Although the present MR study was rigorously performed and had several strengths, some limitations remain. First, to minimize spurious causal relationships induced by individuals with diverse genetic backgrounds, the IVs in our study were selected from sources of European ancestry and genomic control research; thus, our results may not be generalizable to other races. Second, the heterogeneity in some results was challenging to eliminate; thus, we used three methods in addition to the IVW analysis to test the robustness of the results. Third, owing to the lack of a GWAS exploring the subtypes of IBS, we were unable to study the causal associations of these mental disorders with different types of IBS. Forth, the prevalence rate of IBS was higher in women than in men (5.2% vs. 2.9% in Rome-IV) and increasing evidence suggested the pathogenesis of IBS might differed by genders (90). However, since there were no GWAS summary data on IBS by genders, exploring the causal association of psychiatric disorders with IBS in different genders cannot be achieved through MR study. Furthermore, many gut microbiota, blood metabolites, and neurotransmitters are essential in mediating brain-gut axis communication. However, due to limited GWAS summary data, we could only explore the mediation role of some of them in psychiatric disorder-induced IBS; thus, larger, higher-quality, and more detailed GWASs are needed.