Introduction

Functional gastrointestinal disorders (FGIDs), now termed disorders of gut–brain interaction1, are prevalent as they affect more than one-third of the population2, and refer to symptoms such as abdominal discomfort, diarrhea, constipation, bloating, fullness, nausea, and vomiting without underlying structural defects3. Irritable bowel syndrome (IBS) and functional dyspepsia (FD) are the two most common types of the 26-adult-classified FGIDs identified using the Rome criteria. However, the pathogenesis of these diseases remains unclear4. The possible pathological mechanisms of FGID, including gut–brain dysfunction, psychological factors, chronic infections, disorders of the intestinal microbiota, systemic immune activation, alterations in intestinal permeability, low-grade mucosal inflammation, abnormalities in the bile salt metabolism, abnormalities in the serotonin metabolism, and genetic factors, have been widely studied currently5. However, most of the present studies are cross-sectional or case–control observational studies, in which potential residual confounding factors and reverse causality issues could affect the results.

Several studies on hormones, blood lipids, and proteins in patients with FGIDs have been conducted because its diagnosis is based on symptom criteria that lack biochemical markers6,7,8,9. Some researchers have found that the concentration of plasma lipids, especially triglycerides, is significantly higher in patients with IBS than that in controls6,10, and that lipid intermediates or end products of cholesterol metabolism are more prominent in IBS than in FD8. Some researchers believe that the triglyceride levels increase due to the pathophysiological changes such as the low-grade mucosal inflammation which caused by IBS, increased intestinal mucosal permeability, and abnormal intestinal motility associated with IBS would reduce the absorption of fats in the gastrointestinal tract10. Further, evidence shows that one of the main characteristics of patients with IBS is poly-unsaturated fatty acid malabsorption11. In contrast, some studies have suggested that hypertriglyceridemia is a risk factor rather than the result of IBS, as the use of lipid-lowering drugs can reduce the clinical symptoms in patients with IBS12,13. These two contradictory views make it difficult to determine a causal relationship between elevated triglyceride levels and IBS in traditional observational studies. Simultaneously, it is not easy to study the results by conducting randomized controlled trials for the high costs, long experiment time and ethical concerns.

Mendelian randomization (MR) is an ideal epidemiological approach that uses genetic instrumental variables (IVs) in non-experimental data to assess the causal effect of exposure on outcomes to enhance a causal inference14. Genetic variants are relatively independent of self-selected behaviors and are established long before disease manifestation, as they are randomly assigned to offspring. This is equivalent to the random process in randomized controlled clinical trials. Therefore, the MR method can be used in cases in which the exposure factors are expensive or difficult to assess, to eliminate residual confounding factors, and prevent reverse causality. Considering the unclear causal relationship between increased plasma lipid levels and FGID, in this study, we conducted a MR study to investigate the association between plasma lipids and the risk of FGID.

Materials and methods

Data source and open genome-wide association studies (GWAS) statistics

Data related to elevated levels of plasma lipids were obtained from the UK Biobank (UKB), a repository of biomedical data and research resources collected the genetic and clinical information on half a million participants in the United Kingdom15. The consortium of FGID, which includes IBS and FD, belongs to the FinnGen study16, a new study that integrates genetic information with digital healthcare data from people in Finland.

Single nucleotide polymorphism (SNP) selection and assumption

The assumptions used for the genetic variation met the following three basic requirements: (1) this variation was strongly associated with exposure (related hypothesis); (2) the variance did not affect the outcome because of the confounding factors (independence hypothesis); and (3) this variant did not directly affect the through indirect exposure (excluding hypothesis) (Fig. 1). For SNP selection, we set the statistical significance level (p < 5 × 10–8) to rigorously confirm genome-wide significant correlations to satisfy our hypothesis (1). In order to attenuate the disequilibrium of the linkage in SNP, we established the threshold (R2 < 0.001) and a specific mutation frequency, the minor allele frequency (MAF > 1%). For hypotheses (2) and (3), we examined each SNP in PhenoScanner, a public database on the association between human genetics and phenotypes, and excluded genome-wide SNPs that were significantly associated with potential confounders. Palindromic variants were removed to verify the harmonization of the variants, as it was challenging to verify the alleles for which they were correctly directed.

Figure 1
figure 1

Flow diagram of the Mendelian randomization (MR) study. (1) The IVs are strongly associated with plasma lipids; (2) The genetic IVs do not affect the outcome through the confounders; (3) The genetic IVs do not affect functional gastrointestinal disease (FGID) directly, only via indirect exposure.

Full size image

Statistical analysis of primary MR

We chose the inverse-variance weighted (IVW) method, which is considered the most effective analysis method using reliable IVs, as the main analysis technique in our study17. When the pleiotropic effects of IVs were absent, and the sample size was large enough, the IVW estimate was reliable, accurate, and close to the true value18. When heterogeneity was statistically significant (p < 0.05), the multiplicative random effects IVW model was used. Otherwise, a fixed effects model was used. In addition to IVW, other reliable techniques were used to guarantee the accuracy and consistency of the results, including the MR-Egger method19, weighted median18, penalized weighted median estimator20,21, and maximum likelihood method22. Using scatter, funnel plots and forest to visualize the results and demonstrate the validity and stability of MR study.

The heterogeneity test was used to determine the conformance of each SNP using the MR-Egger and inverse-variance weighted techniques to calculate Cochran’s Q statistics. Heterogeneity was considered statistically significant at p < 0.05. A leave-one-out analysis was performed, and the results were compared with those of a global IVW analysis to confirm the reliability and stability of the causal relationship14. The MR-Egger intercept test was used to assess horizontal pleiotropy14. All analyses were performed by the R software (version 4.1.1) and the two-sample MR package.

Results

GWASs of plasma lipids levels

The GWAS of various plasma lipids levels were obtained from the UK Biobank, covering 187,365–403,943 participants, and 125–362 highly associated SNPs were recovered. The FinnGen study, which included 4605 and 4376 participants, produced the GWAS for IBS and FD, respectively (Table 1). Robust genetic IVs for six kinds of plasma lipidi levels were identified by the FinnGen database using open GWAS platforms (https://gwas.mrcieu.ac.uk) to determine the effect of plasma lipids levels-related genetic IVs on IBS and FD. The functions of “extract_outcome_data” and “harmonise_data” was used in conducting the identification process, and 81–315 SNPs were extracted for the subsequent causality analysis (Tables 2, 3).

Table 1 The list of genome-wide summary association studies (GWAS) used in the study.
Full size table
Table 2 MR analysis results of different plasma lipids and irritable bowel syndrome.
Full size table
Table 3 MR analysis results of different plasma lipids and functional dyspepsia.
Full size table

Primary MR analysis

No causal relationship was found between the five plasma lipids (low-density lipoprotein (LDL)-cholesterol, high-density lipoprotein (HDL)-cholesterol, total cholesterol, apolipoprotein A1, and apolipoprotein B) and FGID (including IBS and FD) (Tables 2, 3). Moreover, the IVW method did not reveal a causal relationship between triglycerides and FD. However, it revealed a positive causal relationship between triglycerides and IBS (IVW, OR/95% CI 1.120/[1.010, 1.243], p < 0.05) (Table 2). The risk of IBS was increased by 12.0% for each standard deviation increase in genetically determined plasma lipid levels. Moreover, we have not only used the primary IVW analysis method but also other statistical methods to verify the accuracy of the main results, such as weighted median estimator, the MR-Egger, maximum likelihood estimation, and penalized weighted median estimator methods. Visualizing the effect size of each MR method by using a scatter plot (Figs. 2, 3); visualizing the individual SNP estimates of the outcomes by using a forest plot (Supplementary Figs. 1, 2); and the distribution balance of the single SNP effects is showed by using a funnel plot (Supplementary Figs. 3, 4). These plots suggest that the effects of each SNP and its distribution were in equilibrium.

Figure 2
figure 2

Forest plot showing associations of genetically predicted plasma lipids with risk of irritable bowel syndrome (IBS). CI confidence interval, OR odds ratio. (A) Analysis of apolipoprotein A1 and IBS. (B) Analysis of apolipoprotein B and IBS. (C) Analysis of HDL cholesterol and IBS. (D) Analysis of LDL cholesterol and IBS. (E) Analysis of total cholesterol and IBS. (F) Analysis of triglycerides and IBS.

Full size image
Figure 3
figure 3

Forest plot showing associations of genetically predicted plasma lipids with risk of functional dyspepsia (FD). CI confidence interval, OR odds ratio. (A) Analysis of apolipoprotein A1 and FD. (B) Analysis of apolipoprotein B and FD. (C) Analysis of HDL cholesterol and FD. (D) Analysis of LDL cholesterol and FD. (E) Analysis of total cholesterol and FD. (F) Analysis of triglycerides and FD.

Full size image

Sensitivity analysis

Using the Cochran’s Q test to measure heterogeneity. The effects of HDL-cholesterol on IBS were significantly and statistically heterogeneous among the genetic IVs (IVW, p = 0.01) (Table 4). The IVW multiplicative random-effects model was applied to the associations to calculate causal effects. Furthermore, no significant statistical heterogeneity was found among the genetic IVs for the effects of the remaining five plasma lipids on IBS and the effects of all six plasma lipids on FD (IVW, p ≥ 0.05). Accordingly, the fixed-effects IVW model was performed in the primary MR analysis.

Table 4 The heterogeneity test of plasma lipids genetic variants in IBS and FD Genome-wide summary association study (GWAS) datasets.
Full size table

Leave-one-out analysis was used to assess the effects of a single SNP on the final MR results. After the sequential omission of the single SNP, the other causal effects of various plasma lipids on FGID found in the leave-one-out analysis were consistent with those found in the primary MR studies, showed that no single SNP significantly affected the final result (Supplementary Figs. 5, 6). This suggests that these MR studies are credible and reliable.

A horizontal pleiotropic effect test was used to determine whether plasma lipid-related genetic IVs could lead to FD via other potential pathways. There is no significant horizontal pleiotropy in our MR analyses (all p values > 0.05). This indicates that our MR studies were probably not affected by potential confounding pathways, and the results were robust, credible, and reliable (Table 5).

Table 5 The pleiotropic test of plasma lipids genetic variants in IBS and FD genome-wide summary association study (GWAS) datasets.
Full size table

Discussion

Till date, several metabolomic analyses have identified that plasma lipid levels, including triglyceride, fatty acid, cholesterol, and several other lipid metabolite levels, differ between patients with FGID and controls, and this variability also occurs in different types of FGID8,11,23,24, however, it is difficult to explain the exact causal relationship between the diseases and the change in plasma lipid levels because of the limitations of the traditional observational study methods. In contrast, MR is used to study the causal relationship between exposure and outcome using genetic variation that is randomly distributed and not influenced by confounding factors such as IVs25. This enables us to assess the relationship between increased plasma lipid levels and FGID. This two-sample MR study showed that increased triglyceride levels increase the risk of IBS rather than that of FD. However, there is not any causal relationship between other plasma lipids (LDL-cholesterol, HDL-cholesterol, total cholesterol, apolipoprotein A1, and apolipoprotein B) and FGID.

Plasma lipids include free fatty acids, triglycerides, and cholesteryl esters. Triglycerides are the major form of intracellular and plasma fatty acids storage and transport, which stores most of the energy for regulating physiological functions26. However, abnormal levels of triglyceride lead to several kinds of diseases, including coronary heart disease, acute pancreatitis, metabolic syndrome, and microvascular complications in patients with diabetes such as diabetic nephropathy and other diseases10,27,28,29.

After eliminating potential confounders such as sex, age, and the first 10 genetic principle components, we found that 312 genetic IVs acquired from the UK biobank were tightly associated with triglycerides. The maximum likelihood estimation, penalized weighted median estimator, weighted median estimator, and MR-Egger, also proofed that increased triglyceride levels resulting from genetic factors may increase the risk of IBS. The pleiotropic analysis did not show significant pleiotropic effect between triglycerides genetic variants and IBS (Table 2), and the leave-one-out analysis detected there is not a statistically significant single SNP associated with the results (Fig. 3). Based on the results, we found that plasma lipids, particularly triglycerides, affected the pathogenesis of IBS. Moreover, triglycerides showed an exact causal relationship with IBS, and no other causal relationships were detected in the remaining groups, including LDL-, HDL-cholesterol, total cholesterol, apolipoprotein A, and apolipoprotein B. The correctness and reliability of the relationship were proven by sensitivity analyses, such as the pleiotropic test. Supplementary MR methods also validated these results. Among the 12-MR analysis groups, the HDL-cholesterol groups showed significant statistical heterogeneity. We cannot determine the exact source of the heterogeneity because of the limited access to the original data; however, we speculate that it may be caused by factors such as depression, anxiety, and education. A multiplicative random-effects IVW model was performed to alleviate this effect. Further, this investigation was carried out to explore the causal effects of plasma lipids on FGID. However, traditional epidemiological studies have a difficult problem as the value of plasma lipid levels cannot be accurately detected and the intake value cannot be monitored. This could be solved by using MR. In a traditional epidemiological study, plasma lipid levels can be affected by diet, exercise, body conditions, and other factors that change the relationship with IBS. The influence of diet or other habits on plasma lipid levels could not be judged without an original article on plasma lipid GWAS statistics. But, our study explored variations in the FGID risk based on plasma lipid levels determined by genetic variants. We designed a MR analysis to avoid traditional confounders like diet and other habits by introducing IVs and monitoring their interference using sensitivity analysis. Therefore, the influence of confounding factors did not affect the results.

Based on the present study, the potential mechanisms suggesting that triglycerides are associated with an increased risk of IBS remain unknown. Therefore, we propose several potential possibilities based on the mechanisms contributing to IBS such as disturbances in the intestinal microbiota, immune activation, low-grade mucosal inflammation, and altered intestinal permeability5, in addition to notable influences from obesity and diet30. First, a previous study showed that high levels of triglyceride lead to triglyceride-rich lipoproteins enriched with apoC-III, which affects the signaling pathways that activate NFKβ and increase inflammatory molecules31, which suggesting that it could increase intestinal mucosal inflammation. Second, researchers suggest that the environment for microbial survival in the gut is altered by hyperlipidemia, leading to intestinal flora dysbiosis that aggravates a lipid metabolism disorder, which is an important risk factor for IBS32,33. Third, zonulin is considered a serum biomarker that reflects intestinal permeability, which is a manifestation of IBS34,35. However, the zonulin expression level in the IBS group was not different from that in the control group after adjusting for confounders. Zonulin levels are related to glucose levels, dyslipidemia, and insulin resistance36. Therefore, the increase in intestinal permeability could be caused by dyslipidemia, which increases zonulin levels and induces IBS. Fourth, diet has long been linked with IBS as a double-edged sword, it is a cause of diseases but also a treatment for symptom relief37. Some study suggested that gastrointestinal symptoms were frequently reported after intake of high-fat foods38,39,40, which can increase the level of plasma triglyceride41,42. According to these results, we hypothesized that there may be an increased triglyceride component to the exacerbation of IBS symptoms caused by high-fat foods. Last but not least, a systematic review suggested that obesity is also a relative risk factors of IBS43 and we all known that IBS patients weigh more than healthy controls, and some researches also proved that the obesity patients always have elevated plasma triglyceride levels44. Combined with our results it is easy to find that the mechanisms of obesity aggravating IBS may include the high-level of triglyride. In summary, the possible mechanisms of triglyceride-induced IBS are still unclear; here, we only present speculations and hypotheses based on existing studies and our results, which suggest that increased triglycerides increase the risk of IBS rather than that of FD. Our MR study had several strengths. First, we analyzed the causal relationship between plasma lipid concentrations and FGID, and this could provide a clear direction for pathological studies. Second, this study was established based on the three main assumptions of the IVs, which conform to the checklist for performing MR investigations and make our conclusions reasonable and reliable45. Third, this study benefited from large-scale GWAS of European ancestry, which helped in preventing bias in population stratification. Finally, we used five different MR analysis methods to ensure the consistency of the causal effects.

However, this study had some limitations. First, the outcomes cannot be generalized for all species, and the relationship may change in individuals of other ancestries. The results of the exposure (different plasma lipids) and outcome (IBS and FD) statistics for the genetic IVs were all obtained from the GWAS of European ancestry. Second, the diagnostic methods for IBS and FD could differ between hospitals and health-care systems as they follow a diagnosis of exclusion; similarly, different information acquisition and data processing methods may bias the results. Third, we did not include all dimensions of different plasma lipids in our study, as mentioned previously, because of the lack of an explored GWAS. Finally, although we removed the confounding effects of linkage disequilibrium and pleiotropy, biological mechanisms and genetic co-inheritance, such as gene expression, gene–gene interactions, and gene–environment interactions, could still affect the accuracy of our results.

In conclusion, our study provides genetic evidence that triglycerides are an important risk factor for IBS rather than FD, and this might help in the prevention of IBS if a larger, randomized, controlled cohort study is conducted. Moreover, LDL-cholesterol, HDL-cholesterol, total cholesterol, apolipoprotein A1, and apolipoprotein B levels were not associated with the risk of IBS or FD. However, our study was based on plasma lipid levels determined by genetic variants, which could only account for a part of the IBS risk variation. Therefore, a large-sample, randomized, controlled cohort study is required to clarify the association between plasma lipid levels and IBS. More types of plasma lipids should be explored to determine their causal effects on FGID through further mining and improvement of databases.

Conclusion

Our MR study showed that triglycerides had a positive causal effect on IBS rather than on FD. LDL-cholesterol, HDL-cholesterol, total cholesterol, apolipoprotein A1, and apolipoprotein B levels were not associated with the risk of IBS or FD.