Correlation between gut microbiota and pancreatitis: a bidirectional Mendelian randomization.

Acute pancreatitis (AP) and chronic pancreatitis (CP) are common gastrointestinal disorders associated with significant morbidity and reduced life expectancy [ 1 ]. AP is frequently encountered in clinical settings as a prevalent acute abdomen condition [ 2 ]. The severity of AP is classified into mild acute pancreatitis (MAP), moderately severe AP, and severe acute pancreatitis (SAP) according to the 2012 revision of the Atlanta Classification [ 3 ]. Mortality rates vary among AP subtypes. Approximately 15–20% of patients develop SAP, which has a mortality rate as high as 25%, whereas MAP carries a mortality rate of just 1% [ 4 , 5 ]. SAP is characterized by high morbidity and mortality due to the development of pancreatic and extrapancreatic necrosis, their subsequent infection, and multisystem organ failure [ 6 ]. It is now widely recognized that infection of necrotic pancreatic tissue is triggered by a failure in intestinal barrier function and the displacement of intestinal microbiota into the visceral vascular bed [ 7 , 8 ]. Thus, gut microbiota may play a critical role in the onset and progression of SAP. The predominant symptom and most common complication of CP is abdominal pain, which significantly impairs quality of life [ 9 ]. Moreover, the most daunting complication of CP – the development of pancreatic cancer – has yet to find effective preventive strategies [ 10 ]. While no therapies currently exist to delay or stop disease progression in CP, ongoing research aims to fully understand its natural history and the mechanisms underlying the disease [ 11 ]. Recently, the potential causal link between the composition of gut microbiota and the risk of pancreatitis has gained widespread interest. Increasing evidence suggests that changes in intestinal flora during AP development may correlate with disease severity [ 12 ]. Gut microbiota has been recognized as a crucial mediator in AP, with its dysbiosis linked to the severity of the condition [ 13 ]. Patients with CP exhibit gut microbiota dysbiosis, characterized by reduced diversity and richness, along with changes in taxa composition [ 14 ]. Through interventions such as probiotic supplementation and fecal microbiota transplantation (FMT), regulating gut microbiota is considered a promising therapeutic approach [ 15 – 17 ]. Mendelian randomization (MR), which addresses the bias from confounding and reverse causality, is a statistical method that infers causal relationships from exposure to outcome [ 18 ]. MR is conceptually similar to a randomized controlled trial, where genetic variables serve as instrumental variables (IVs). These IVs are randomly assigned at birth into ‘case’ or ‘control’ groups and remain constant throughout an individual’s life in accordance with Mendel’s second law. Over the last decade, genome-wide association studies (GWASs) have transformed the field of complex disease genetics by examining millions of genetic variations across the genomes of numerous individuals to identify genotype-phenotype relationships [ 19 ]. Associations between single-nucleotide polymorphisms (SNPs) with exposure and SNPs with outcome, derived from separate GWASs, can be utilized in two-sample MR analysis to generate a singular causal estimate. This methodology has enabled us to implement a two-sample bidirectional MR approach to explore the reciprocal causal relationships between gut microbiota and pancreatitis.

To investigate causality between gut microbiota influenced by host genetics and AP, as well as CP, we utilized a two-sample bidirectional MR approach with summary statistics from extensive GWASs (Fig. 1 ). Three key assumptions were addressed to minimize bias in our MR analysis: First, IVs must be significantly associated with the gut microbiota. Second, IVs should be independent, ensuring that they are not significantly correlated. Third, there should be no horizontal pleiotropy, meaning IVs affect AP/CP solely through gut microbiota taxa. The flowchart of the study. The whole workflow of MR analysis. AP, acute pancreatitis; CP, chronic pancreatitis; MR, Mendelian randomization. Gut microbiome data from the MiBiogen study represent the largest genome-wide meta-analysis available at the GWAS aggregation level [ 20 ] [MiBioGen Consortium, https://www.mibiogen.org (accessed on 20 February 2023)]. This study consolidated 16S rRNA gene sequencing profiles and whole-genome genotyping data from 18 340 individuals across 24 cohorts, with over 78% (14 363) of participants of European ancestry, as detailed elsewhere [ 20 ]. The consortium provided summary data for nine phyla, 16 classes, 20 orders, 35 families, and 131 genera. Fifteen unidentified bacterial taxa were excluded, leaving nine phyla, 16 classes, 20 orders, 32 families, and 119 genera for MR analysis. Instrument variable summary statistics for AP/CP were obtained from a meta-analysis by the R8 version of the FinnGen consortium, which includes prospective epidemiologic cohorts, disease-specific cohorts, and hospital biobank specimens [ 21 ]. Detailed information is available on the official website ( https://www.finngen.fi/en ). To verify the authenticity and accuracy of conclusions regarding the causal relationship between gut microbiota and the risk of pancreatitis, optimal IVs were chosen following several quality control steps. Initially, SNPs significantly associated with gut microbiota were selected as IVs. Those gut microbiota groups with SNPs exhibiting significance levels smaller than 1 × 10 −5 at the locus level were chosen to yield more comprehensive results [ 22 ]. Second, a fundamental aspect of the MR approach is ensuring no linkage disequilibrium (LD) between the IVs, as high LD can result in biased outcomes. To assess LD, a clumping process with an r 2  < 0.1 and a clumping distance of 500 kbp was implemented. Finally, a crucial step in MR is to confirm that the effects of SNPs on the exposure are consistent with their effects on the outcome, adhering to the principle that the IVs should not include palindromic SNPs. We conducted a bidirectional MR using data from the MiBioGen Consortium and FinnGen study. To evaluate the causal link between exposure factors and outcomes, we employed several methods, including IVW, MR-Egger, weighted median, simple mode, and weighted mode. IVW is a standard method that combines Wald ratio estimates from each IV through a weighted linear regression, analyzing the association between the IVs and the outcome. To ensure consistency, summary statistics were harmonized so each IV was associated with the same effect allele. SNPs with ambiguous strands (A/T and C/G alleles) were excluded, and palindromic SNPs were removed to eliminate their effect on outcome causality between gut microbiota taxa and pancreatitis. Several sensitivity analyses were conducted to evaluate and adjust for the presence of pleiotropy in causal estimates. Horizontal pleiotropy and outliers were assessed using MR-Egger and MR Pleiotropy Residual Sum and Outlier (MR-PRESSO) tests. MR-Egger preliminarily identified the presence of horizontal pleiotropy, with a P -value greater than 0.05 indicating no significant horizontal pleiotropic effect. MR-PRESSO, known for its higher accuracy, is effective in detecting pleiotropy and horizontal outliers [ 23 ]. Cochrane’s Q test was applied to examine heterogeneity among the IVs in the IVW and MR-Egger regressions, whereas leave-one-out sensitivity analysis assessed the outliers and stability of the results. Additionally, we calculated F statistics to address weak instrument bias using the formula [ 24 ]: F = R 2 ( n − 1 − k ) ( 1 − R 2 ) k , where n , k , and R 2 represent the sample size, number of IVs, and variance explained by the IVs, respectively (Supplementary Table 1, Supplemental digital content 1, http://links.lww.com/EJGH/B77 ). An F -value below 10 suggests a weak instrument. The power calculation for MR was also performed to determine the statistical power of the causal effect estimates [ 25 ], using a power threshold of 0.8 to ensure that 4/5 of the false null hypotheses would be rejected [ 26 ]. To investigate reverse causality, we executed an additional reverse MR analysis. A significant finding from the inverse MR analysis suggested reverse causality from pancreatitis (as exposure) to characteristics of the microbiota (as outcome). The procedure for the reverse MR analysis was identical to the MR analysis previously described. All analyses were conducted using R Version 4.1.2 (Foundation for Statistical Computing, Copenhagen, Denmark) and the TwoSampleMR package (version 0.5.6). We adhered to the Strengthening the Reporting of Observational Studies in Epidemiology using Mendelian Randomization (STROBE-MR) guidelines for strengthening the reporting of observational studies in epidemiology utilizing MR for reporting our results [ 27 ].

IV screening was rigorously performed under the previously mentioned conditions (Supplementary Table 2, Supplemental digital content 1, http://links.lww.com/EJGH/B77 ). After clumping and harmonization, the number of IVs linking gut microbiota with pancreatitis varied from 3 to 20. For example, there were 2173 IVs linked with AP, and the gut microbiota taxa associated with these IVs are categorized into nine bacterial phyla (106 SNPs), 16 classes (189 SNPs), 20 orders (231 SNPs), 32 families (378 SNPs), and 119 genera (1269 SNPs). For CP, the taxa with the largest number of SNPs associated included class Actinobacteria (20 SNPs), order Bifidobacteriales (20 SNPs), family Bifidobacteriaceae (20 SNPs), and genus RuminococcaceaeUCG002 (20 SNPs), whereas the taxon with the fewest was LachnospiraceaeND3007group (three SNPs). No feature at any level contained only a single SNP. We utilized all five MR methods (IVW, weighted median, MR-Egger, simple mode, and weighted mode) to analyze the causal relationship between each bacterial taxon and pancreatitis pair (Supplementary Table 3, Supplemental digital content 1, http://links.lww.com/EJGH/B77 ). We concluded there was a significant causal relationship between gut microbiota and pancreatitis if the following three criteria were met: (1) The P -value in IVW was significant ( P  < 0.05); (2) there were consistent estimation directions across IVW, weighted median, MR-Egger, simple mode, and weighted mode methods; and (3) the MR-Egger intercept test and MR-PRESSO global test showed no significant results ( P  > 0.05) [ 28 ]. The IVW test indicated that the relative abundance of genus Flavonifractor [odds ratio (OR): 0.76, 95% confidence interval (CI): 0.58–0.99, P  = 4.10 × 10 −2 ], genus Erysipelatoclostridium (OR: 0.87, 95% CI: 0.75–1.00, P  = 4.80 × 10 −2 ), genus Methanobrevibacter (OR: 0.83, 95% CI: 0.70–0.99, P  = 3.40 × 10 −2 ), genus Prevotella9 (OR: 0.87, 95% CI: 0.76–1.00, P  = 4.60 × 10 −2 ), and phylum Firmicutes (OR: 0.82, 95% CI: 0.68–0.99, P  = 4.10 × 10 −2 ) were negatively correlated with the risk of AP. Conversely, the genetically examined relative abundance of genus Eubacteriumeligensgroup (OR: 1.38, 95% CI: 1.01–1.87, P  = 4.10 × 10 −2 ), genus Eubacteriumfissicatenagroup (OR: 1.20, 95% CI: 1.06–1.36, P  = 4.00 × 10 −3 ), genus Coprococcus3 (OR: 1.28, 95% CI: 1.01–1.61, P  = 4.10 × 10 −2 ), and genus Haemophilus (OR: 1.20, 95% CI: 1.00–1.43, P  = 4.60 × 10 −2 ) were positively associated with the risk of AP (Fig. 2 ). Scatter plots of the MR analyses for the association of nine gut bacterial taxa and the risk of AP. Scatter plots of genus Eubacteriumeligensgroup (a), genus Eubacteriumfissicatenagroup (b), genus Coprococcus3 (c), genus Erysipelatoclostridium (d), genus Flavonifractor (e), genus Haemophilus (f), genus Methanobrevibacter (g), genus Prevotella9 (h), and phylum Firmicutes on AP (i). SNP effects were plotted as lines. The slope of the line corresponded with the estimation of causality. ACUTPANC, acute pancreatitis; AP, acute pancreatitis; MR, Mendelian randomization; SNP, single-nucleotide polymorphism. The IVW estimates suggested that the genetically predicted relative abundance of genus LachnospiraceaeFCS020group (OR: 0.79, 95% CI: 0.62–1.00, P  = 4.60 × 10 −2 ), genus Prevotella9 (OR: 0.80, 95% CI: 0.67–0.96, P  = 1.90 × 10 −2 ), and family Clostridiaceae1 (OR: 0.74, 95% CI: 0.56–0.97, P  = 3.30 × 10 −2 ) were negatively associated with the risk of CP, whereas family Victivallaceae (OR: 1.17, 95% CI: 1.02–1.34, P  = 2.60 × 10 −2 ) was positively associated with the risk of CP (Fig. 3 ). Scatter plots of the MR analyses for the association of four gut bacterial taxa and the risk of CP. Scatter plots of family Clostridiaceae1 (a), family Victivallaceae (b), genus LachnospiraceaeFCS020group (c), and genus Prevotella9 on CP (d). SNP effects were plotted as lines. The slope of the line corresponded with the estimation of causality. CHRONCPANC, chronic pancreatitis; CP, chronic pancreatitis; MR, Mendelian randomization; SNP, single-nucleotide polymorphism. Employing MR-Egger and MR-PRESSO global tests ( P  > 0.05), we detected no pleiotropic effects among the selected SNPs (Supplementary Table 4, Supplemental digital content 1, http://links.lww.com/EJGH/B77 ). There was also no evidence of heterogeneity between the chosen IVs and pancreatitis outcomes (Supplementary Table 5, Supplemental digital content 1, http://links.lww.com/EJGH/B77 ). Forest plots of the causal effects using individual SNPs indicated no extremely significant associations with pancreatitis (Supplementary Figure 1, Supplemental digital content 2, http://links.lww.com/EJGH/B78 ). The leave-one-out sensitivity analysis showed no single SNP driving the causal association signal (Supplementary Figure 2, Supplemental digital content 3, http://links.lww.com/EJGH/B79 ). Setting the genome-wide significance level at P  < 5 × 10 −6 , as in previous publications [ 29 ]. We conducted reverse MR using the IVW method to explore causal effects between pancreatitis and gut microbiota. This analysis confirmed significant causality from gut microbiota to pancreatitis (Supplementary Table 6, Supplemental digital content 1, http://links.lww.com/EJGH/B77 ). However, no evidence of reverse causality from pancreatitis back to gut microbiota was supported by the IVW method in the bidirectional MR analysis.

To our knowledge, this is the first comprehensive MR study to reveal the potential causal relationship between gut microbiota taxa and pancreatitis, utilizing extensive summary-level GWAS data. Previous investigations into the correlation between gut bacterial taxa and pancreatitis primarily consisted of animal model experiments and cross-sectional studies [ 13 , 30 , 31 ]. Consequently, the causal relationship between gut microbiota and pancreatitis has remained ambiguous. Meanwhile, our findings offer new perspectives for future pancreatitis prevention and therapeutic strategies: specifically, the targeted regulation of ecological disorders in genetically specified taxa to prevent and manage pancreatitis. In our study, nine and four gut microbiota features were identified as causally associated with AP and CP, respectively. Conversely, AP and CP were each found to be causally associated with two and five gut bacterial taxa, respectively. Pancreatitis, an inflammatory disease of the pancreas, is a major cause of severe morbidity, mortality, and hospitalization [ 32 ]. In pancreatic-related diseases, alterations in the gut microbiota are linked with both acute and CP [ 31 ]. Furthermore, there is accumulating evidence that pancreatic exocrine function influences gut immunity, reinforcing the pancreas–intestinal axis [ 33 ]. Compared with other host factors, the composition of the gut microbiome is significantly impacted by pancreatic exocrine activity [ 34 , 35 ]. Acute necrotizing pancreatitis, a more severe form of the disease, can arise from an increase in gut microorganisms that impair epithelial barrier function [ 36 ]. Researchers have used in-situ hybridization of pancreatic sections from models of AP to confirm that the severity of pancreatitis is exacerbated by a high bacterial load and barrier collapse [ 30 ]. Previous studies have also shown that the NLRP3 inflammasome is active not only in the intestines of mice with AP but also in their pancreas, and that the extent of AP severity correlates with intestinal NLRP3 activation [ 37 , 38 ]. Eight protective causal associations were identified, including phylum Firmicutes, Erysipelatoclostridium, Flavonifractor, Methanobrevibacter, and Prevotella9 at the genus level with AP, along with family Clostridiaceae1, LachnospiraceaeFCS020group, and Prevotella9 at the genus level with CP. Over 90% of gut microbiota consists of four bacterial phyla: Bacteroidetes, Firmicutes, Proteobacteria, and Actinobacteria [ 39 ]. Previous studies have noted that Firmicutes were decreased in the feces of the AP group compared with the sham operation group (control group) in mouse models [ 40 ]. Similarly, AP samples contained more Bacteroidetes and Proteobacteria but fewer Firmicutes and Actinobacteria than samples from healthy volunteers [ 41 ]. These observations suggest that specific components of the gut microbiome may contribute to bacterial translocation, secondary infection, and exacerbation of AP [ 7 , 41 , 42 ]. However, the role of ecological imbalance in the gut microbiota in the pathogenesis of pancreatitis remains unclear. In our results, the intestinal flora genus Erysipelatoclostridium belongs to the phylum Firmicutes. Erysipelatoclostridium plays adverse roles in host lipid metabolism [ 43 ]. The pathogenesis of hypertriglyceridemia-induced AP, which is closely related to high free fatty acids, may be influenced by this bacterium [ 44 ]. Despite most studies indicating the harmful effects of Erysipelatoclostridium on health, identifying prebiotics and probiotics that target this bacteria could be beneficial. The role of Erysipelatoclostridium in pancreatitis warrants further investigation. Flavonifractor and Prevotella9 , belonging to Actinobacteria and Bacteroidetes , respectively, have unique roles. Members of the Flavonifractor genus are involved in degrading flavonoids, compounds known for their inverse correlation with various human diseases, including obesity [ 45 ]. There is minimal literature on the detection and reporting of this microbiota in AP. Recent studies suggest that Methanobrevibacter may be beneficial by reducing the production of reactive oxygen species (ROS), trimethylamine-N-oxide, and intestinal permeability while activating the Nrf2 pathway [ 46 ]. Inflammatory bowel disease (IBD) can affect multiple organ systems, and pancreatic manifestations are not uncommon [ 47 ]. Compared with the general population, patients with IBD are thrice as likely to experience AP [ 47 , 48 ]. In regulating autophagy, gut microbiota may influence ROS production and participate in the progression of IBD [ 49 ]. Interestingly, Methanobrevibacter smithii bacterial load is markedly higher among healthy subjects compared with IBD patients [ 50 ]. Targeting gut archaea modulation may be a strategic approach for managing the increased risk of pancreatitis in patients. In the MR of CP, family Clostridiaceae1 and genus LachnospiraceaeFCS020group are from the class Clostridia. Class Clostridia produces short-chain fatty acids, primarily butyrate, which is widely recognized for its ability to reduce oxidative damage, inhibit inflammation, and increases intestinal epithelial cell integrity [ 51 ]. Previous studies have shown that Clostridium butyricum strains suppress acute experimental pancreatitis by maintaining gut homeostasis [ 52 ], supporting the theory behind our findings. Interestingly, genus Prevotella9 had a protective causal effect on both AP and CP. Although Prevotella spp. are commonly associated with humans, their relationship to health and disease, their role in the microbiome, and their contribution to host–microbiome crosstalk remain somewhat ambiguous [ 53 ]. Lower levels of beneficial symbiont genera such as Prevotella were found in AP patients, aligning with prior research [ 13 , 54 ]. As a commensal microbe in the human intestine, Prevotella has attracted increasing attention for its ability to degrade a wide range of foods, including plant polysaccharides, its role in improving glucose metabolism induced by dietary fiber, and its essential function in vitamin B1 biosynthesis [ 55 – 57 ]. Building on previous findings, targeted regulation of bacterial richness may represent a novel treatment strategy for pancreatitis [ 58 , 59 ]. Given the close relationship between the gut microbiota and pancreatic lesions, interventions aimed at modulating the composition of the gut microbiome, such as using probiotics [ 60 ], prebiotics [ 61 ], synbiotics [ 61 ], and postbiotics [ 62 ] products, as well as performing FMT [ 63 ], hold promise for improving pancreatic health. In a trial conducted in 2002, researchers investigated the efficacy of using lactobacilli as probiotics in the treatment of AP through a randomized double-blind controlled trial. The results of the study indicated that patients treated with probiotics had a significant reduction in the incidence of infectious complications (such as infected pancreatic necrosis and pancreatic abscess) compared with the control group not receiving this treatment, suggesting that probiotic therapy may have a positive impact on the prognosis of patients with AP [ 16 ]. Prebiotics provide nondigestible food components for beneficial bacteria in the gut, including various types of dietary fiber and polysaccharides. Before inducing SAP, treating mice with chitosan oligosaccharides for four consecutive weeks significantly reduced the extent of pancreatic damage, possibly by reducing oxidative stress and modulating the gut microbiota [ 42 ]. FMT alleviates dysbiosis of the gut microbiota induced by AP and mitigates the severity of AP, including mitochondrial dysfunction, oxidative damage, and inflammation, making FMT a potential strategy for the treatment of AP [ 17 ]. However, the specific impact and mechanism of FMT in the treatment of pancreatitis require further validation and clarification through more animal experiments and clinical studies. We approached pancreatitis as an exposure and gut microbiota as an outcome to evaluate all reverse causality. We found that AP was causally linked with genus Paraprevotella and genus Parasutterella . In cases of CP, it retains causal relationships with Eubacteriumrectalegroup , Ruminococcusgauvreauiigroup , Ruminococcustorquesgroup , Enterorhabdus , and Lachnospira at the genus level. No bidirectional causal relationship between pancreatitis and gut microbiota was observed. Furthermore, reverse MR analysis and sensitivity analysis indicated no evidence of pleiotropy or heterogeneity, affirming the statistical robustness of our results. There are several limitations to our study. First, although we utilized aggregated statistics from GWASs on AP, CP, and gut microbiota, we identified a limited number of SNPs achieving genome-wide significance, which might lead to weak IVs. To mitigate this issue, the P -value threshold for SNPs in the gut microbiota was extended to ( P  < 1 × 10 −5 ). Furthermore, ensuring an F -value greater than 10 helped eliminate potential weak instrument bias, improving the robustness of our statistical results. Second, our use of the MiBioGen Consortium and the FinnGen consortium to infer causality was challenged by the absence of prospective studies tracking microbiome changes pre- and postdisease onset, complicating the assessment of the impact of microbiome level changes on disease evaluation. Additionally, the loci identified are still limited in number compared with those associated with pancreatitis, and our microbiota analysis lacked aggregated GWAS statistics at the species level. Consequently, we were unable to identify taxa associated with pancreatitis causality at the species level, leading to restricted causal inference. Finally, due to the lack of detailed explanation of inclusion and exclusion criteria for patients with acute and CP in the database, such as dietary habits, history of antibiotic use, or chronic disease status of control group members, our positive results had certain limitations in the discussion. Overall, our comprehensive analysis analyzed the potential causality between gut microbiota and pancreatitis. This bidirectional MR study identified predisposing or protective effects of nine features on AP and four features on CP, with Prevotella9 emerging as a common protective gut bacterial taxon for both AP and CP. These findings highlight promising gut biomarkers and novel therapeutic targets for pancreatitis.

The authors are grateful to the FinnGen consortium and MiBioGen Consortium for providing GWAS summary statistics data for our analysis. This work was supported by the Natural Science Foundation of Liaoning Province (2022-YGJC-11 and 2021JH2/10300084). B.N., L.J., and W.Z. designed the study. B.N. and T.W. performed the analysis and drafted the manuscript. C.L. and H.Z. participated in revising the manuscript. L.J., C.W., and W.Z. reviewed the manuscript for its intellectual content and revise the entire work. All authors read and approved the final manuscript. All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors, and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher. Summary data used for this study can be accessed through the following links: microbiota, https://mibiogen.gcc.rug.nl/ ; acute pancreatitis, https://r8.finngen.fi/pheno/K11_ACUTPANC ; and chronic pancreatitis, https://r8.finngen.fi/pheno/K11_CHRONPANC . There are no conflicts of interest.