Review | Published: 17 June 2024
open-url

Mendelian randomisation analysis for intestinal disease: achievement and future

https://doi.org/10.1136/egastro-2023-100058

Request reuse permissionopen-url
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) licenseopen-url

Abstract

Intestinal disease is a group of complex digestive system diseases imposing a significant burden globally. Identifying the risk factors and potential complications of intestinal disease is important for its prevention and treatment. However, traditional observational clinical studies are limited by confounding factors and reverse causation, making causal inference challenging. Mendelian randomisation (MR) method has been developed to effectively mitigate these constraints and assess the causal relationships. This review briefly introduces the MR method, summarises MR research on intestinal disease and delineates the prospective avenues for future research. Conventional risk factors, such as lifestyle behaviours (eg, physical activity, smoking and alcohol consumption), nutrients (eg, selenium), obesity markers (eg, body mass index and waist-to-hip ratio) and inflammatory biomarkers, have been validated in MR studies. Multiomics MR studies are becoming novel hotspots, which provide a theoretical foundation for the exploration of pathogenesis and the investigation of new drug targets. However, most of the recent studies are based on European individuals, and thus it is necessary to replicate the results in other ancestries. Moreover, triangulation integrating MR and other epidemiology methods is suggested as a validated paradigm for causal inference in future MR studies.

Introduction

Intestinal diseases are complex and common digestive system diseases, leading to a heavy burden on the healthcare system globally.1 Investigating potential environmental risk factors would provide valuable insights into the prevention of intestinal diseases while identifying the causal potential therapeutic targets represents an important first step toward the treatment of diseases. Epidemiological studies have investigated these potential associations. An umbrella review incorporating 71 environmental factors revealed the detrimental impact of smoking, soft drink consumption and vitamin D deficiency on inflammatory bowel disease.2 The association of obesity, alcohol consumption and hyperlipidaemia with elevated colorectal cancer risk has been summarised in a meta-analysis.3 However, the unavoidable confounding bias and reverse causation constrained the ability to establish a definitive causal link between risk factors and diseases in observational studies. Therefore, the causal association between the factors identified in previous observational studies and the risk of intestinal diseases remains unverified. There are many clinical trials investigating intestinal diseases, primarily focused on evaluating the safety and effectiveness of various medications and supplements.4 5 Despite the high-quality evidence provided by randomised controlled trials (RCTs), certain inquiries are ill-suited for RCT methodology, particularly those concerning modifiable lifestyles. For instance, ethical considerations preclude the random assignment of participants to sustained smoking or excessive alcohol consumption. Moreover, the execution of large-scale, multidecade RCTs is frequently hindered by prohibitive costs and time constraints, thus constraining their feasibility.

Analogous to an RCT, Mendelian randomisation (MR) is an epidemiological method that uses genetic variants as instrumental variables to explore the potential causal association between exposure and outcome.6 As genetic variants are randomly allocated at conception, MR is less prone to confounding and reverse association in contrast to observational design. To obtain valid and reliable results from MR analysis, instrument variables should meet the following three assumptions (figure 1): (1) instrument variables are strongly associated with the exposure; (2) genetic instruments should not be associated with any confounders; (3) genetic instruments should impact the outcome through exposure, not alternative pathways or direct influence. The current review aims to provide a comprehensive summary of the available evidence regarding potential causal risk factors for intestinal disease. Moreover, this review will also offer valuable insights into the potential future advancements of MR studies in the field of intestinal disease.

Figure 1
Figure 1

Assumptions of Mendelian randomisation analyses.

Causal risk factors for intestinal diseases

Figure 2 presented the MR associations of modifiable lifestyles, nutrients, obesity and lipid metabolism with intestinal disease.

Figure 2
Figure 2

A summary of Mendelian randomisation (MR) studies exploring associations of modifiable lifestyles, nutrients, obesity and lipid metabolism with intestinal disease. Significance and direction of associations were determined by the inverse variance weighted methods in the studies with the largest sample size.

Modifiable lifestyles

Alcohol consumption has been found to have a direct detrimental effect on the integrity of the intestinal wall barrier and disrupt the composition of the gut microbiota, thereby eliciting an inflammatory response within the gastrointestinal tract.7 Evidence from MR studies has revealed that genetically predicted alcohol consumption, frequency of alcoholic drinks consumed and alcohol use disorder are risk factors for colorectal cancer in both European ancestry and East Asian ancestries.8–12 An MR investigation further identified two alcohol-related CpG sites (mapped the COLCA1/COLCA2 gene) that were associated with a higher risk of colorectal cancer.10 MR studies demonstrated that alcohol consumption was associated with an increased risk of Crohn’s disease,13 14 whereas no associations were observed between alcohol consumption and ulcerative colitis (UC).11 13–16 Inconsistent with some previous observational studies, a recent MR analysis that employed three large consortia datasets uncovered that genetically predicted alcohol consumption was suggestively associated with decreased irritable bowel syndrome risks.11 This discrepancy may be explained by residual or unmeasured confounding in observational studies.

In accordance with observational investigations, cigarette smoking was identified as a significant risk factor for colorectal cancer,17 18 diverticular disease,11 19 irritable bowel syndrome11 and Crohn’s disease11 13 in MR studies. Evidence was interesting that genetically predicted smoking initiation was associated with increased UC risks which was inconsistent with traditional observational studies.2 It is worth noting that associations of genetically predicted cigarettes per day and lifetime smoking index with some intestinal diseases were found null.13 14 20 21 These null associations may be attributed to the insufficient statistical power caused by a small phenotypical variance explained by using genetic instruments (approximately 0.52% for lifetime smoking index and 2.3% for smoking initiation) or a small sample size of the outcome of genome-wide association study (GWAS). Furthermore, MR studies have demonstrated the association between smoking cessation and a reduced colorectal cancer risk,18 22 which suggests that disseminating public health messages regarding smoking cessation could potentially enhance the overall health outcomes for individuals.

An MR study reported that poor sleep characterised by insomnia and short sleep duration was associated with an increased risk of irritable bowel syndrome.23 Although the statistical significance was not observed after conducting multiple comparisons, genetically predicted short sleep duration (<7 hours) was associated with elevated colorectal cancer risk in an MR analysis encompassing a comprehensive examination of site-specific cancers.24 Previous MR investigations also explored the associations of physical activity and sedentary behaviour with some intestinal diseases. Arriving at the same conclusion, physical activity was associated with a decreased risk of inflammatory bowel disease and colorectal cancer,13 21 25 26 whereas leisure screen time was associated with increased irritable bowel syndrome risk.27

MR studies have evaluated the causal effects of single-food intake on the risk of multiple intestinal diseases. Consistent with observational studies, milk intake has been linked to a reduced risk of inflammatory bowel disease.13 28 MR studies also revealed that milk consumption was related to a lower colorectal cancer risk.29 30 With regard to the consumption of beverages, the findings from MR studies indicated that coffee consumption is inversely associated with the risk of colorectal cancer in East Asian individuals,9 while such association was not observed in individuals of European ancestry.12

The most critical step in MR analysis is to identify genetic instruments that proxy the exposure of interest. However, it is important to note that not all single nucleotide polymorphisms (SNPs) used as instrumental variables have a proven or plausible biological effect on the target exposure. The SNP–exposure associations identified in GWASs are often difficult to be interpreted mechanistically, increasing the chance of pleiotropy. Horizontal pleiotropy violates the core MR assumption by allowing alternative causal pathways between genetic variants and outcomes not merely through exposure. In contrast, vertical pleiotropy does not invalidate the MR estimate. For instance, physical activity levels may reduce colorectal cancer risk by lowering body mass index.26 This represents vertical pleiotropy whereby the causal pathway from genetic variants to the outcome passes through intermediates, which does not violate the third assumption. Distinguishing horizontal from vertical pleiotropy requires understanding the biological relationship between genetic variants, exposure, outcome and other potential pleiotropy traits. Caution is warranted when interpreting the causal nature of MR findings, particularly for complex lifestyle exposures where pleiotropy is a greater concern. Improving functional annotation of genetic variants can clarify mechanisms and strengthen future MR utility.

Nutrients

Align with evidence from observational studies, MR analyses have strengthened the protective role of circulating vitamin D levels in Crohn’s disease and UC.13 31 Besides, more evidence regarding the oxidative-related vitamins was provided by the MR investigation that genetically predicted circulating vitamin A and E levels were inversely associated with UC risks,31 circulating vitamin K levels were associated with decreased Crohn’s disease risk31 and circulating vitamin C levels were associated with decreased risk of colorectal cancer.32 33 However, not all vitamins have the same protective effects. Some MR analyses have reported that higher vitamin B12 levels were associated with an increased risk of Crohn’s disease21 31 and colorectal cancer,12 32 34 35 which deserved further investigation. Limited by few GWASs, findings regarding circulating antioxidants were not as much as the vitamins. It has been reported that genetically-predicted lycopene was inversely associated with the risk of inflammatory bowel disease including both Crohn’s disease and UC.31

Aberrant fatty acid metabolism in intestinal diseases, especially inflammatory bowel disease and colorectal cancer, was also captured in MR investigations. Genetically predicted omega-3 fatty acids and omega-6 fatty acids levels were inversely associated with Crohn’s disease and UC risks; however, this association was towards null when different datasets were used.13 21 36–39 Notably, the omega-3 fatty acids and omega-6 fatty acids consist of diverse fatty acids with distinct biochemical mechanisms. An MR study that investigated various types of omega-3 fatty acids on inflammatory bowel disease found a causal link between eicosapentaenoic acid and a lower risk of Crohn’s disease.39 Multiple circulating fatty acids were found to be associated with colorectal cancer (figure 2).40 41 However, the observed associations were predominantly influenced by SNPs in the FADS1/2 genes which encode a key enzyme in fatty acid metabolism. This constraint hindered the ability to separate the correlation between specific fatty acids and bowel disease, necessitating further sensitivity analyses to assess pleiotropy. Consequently, it is imperative to take into account the potential pleiotropic effects introduced by these genetic variants and infer the MR results with caution.

The harmful effects of magnesium and calcium, and the protective effects of phosphorus, selenium and zinc on the inflammatory bowel disease risks have been found in MR studies.13 31 Genetically predicted higher iron levels were associated with an increased risk of colorectal cancer.12 32 35 Also, these studies included SNPs located in the HFE gene were related to the predominant type of hereditary haemochromatosis and thus have substantial pleiotropic effects.42 Therefore, a further investigation that excluded genetic instruments from HFE needs to be conducted to confirm the effect of iron levels rather than hereditary haemochromatosis on colorectal cancer. Selenium was considered to be protective for cancer, as it is antioxidative and anti-inflammatory, and findings from MR studies confirmed the protective effect of selenium against colorectal cancer.12 35

Obesity and lipid metabolism

The relationship between various obesity indicators and the risk of various intestinal diseases has been thoroughly studied in MR studies. Similar to observational studies, genetically predicted body mass index, body fat percentage, waist circumference and waist-to-hip ratio were related to a higher risk of colorectal cancer in MR studies.12 43–48 Interestingly, in sex-specific MR analyses, higher body mass index was associated with an increased risk of colorectal cancer among males, while waist-to-hip ratio was associated with an increased risk of colorectal cancer in females.49 Consistent with observational studies, MR studies indicated the effects of body mass index on increased risk of inflammatory bowel disease.21 49 Notably, visceral adiposity, especially the mesenteric adipose tissue, was considered to play a crucial role in the onset of Crohn’s disease, which also was confirmed in MR investigations.49 50 MR studies also confirmed the detrimental role of obesity (indicated by body measurements) in diverticular disease,19 48 49 duodenal ulcers49 51 52 and coeliac disease.49 Positive associations between childhood body mass index and diverticular disease risk, early body size and colorectal cancer risks were observed in life-trajectory MR studies.49 53 One possible explanation is that the effect is largely mediated via adult body mass index.

In the current observational studies, the link between lipid and colorectal cancer is still controversial. MR studies have been conducted to investigate the causal effects of several lipids on colorectal cancer. Findings from MR studies conducted on individuals of European and East Asian ancestry indicated a significant association between total cholesterol levels and an elevated colorectal cancer risk.12 54 55 Results from MR analysis on finer subdivisions of the anatomical site showed that genetically predicted total cholesterol levels were associated with colon cancer rather than rectal cancer.56 High-density lipoprotein was related to a lower risk of inflammatory bowel diseases, while the association was only significant in Crohn’s disease but not in UC.57 Total cholesterol, triglycerides and low-density lipoprotein seem to be protective factors for UC in MR studies.13 58

Inflammatory protein biomarkers

Genetically predicted interleukin (IL)-16,59 IL-1859 and CXCL1059 were related to an increased risk of inflammatory bowel disease significantly, whereas IL-12p7059 and CCL2359 were inversely associated with the inflammatory bowel disease.59 A bidirectional MR investigation revealed that CXCL9, CCL11 and CASP8 were associated with an increased risk of UC; furthermore, it was observed that genetic predisposition to UC did not exert any influence on these three inflammation protein levels.60 Genetically predicted C reactive protein level was inversely associated with the increased risk of inflammatory bowel disease in an MR study, while the association was not significant after excluding SNPs with heterogeneity.61 No causal effect of C reactive protein on colorectal cancer was observed in MR studies.12 62 63 The findings from these studies suggest that the C reactive protein functions as an indicator of inflammation rather than a causative factor in intestinal diseases.

Circulating metabolites and proteins

Large-scale GWASs focusing on circulating metabolites and proteins have significantly advanced our understanding of disease mechanisms using MR methodology. However, current research predominantly concentrates on their associations with inflammatory bowel disease and colorectal cancer. Expanding these investigations to encompass a broader spectrum of intestinal diseases remains an area of growing interest and potential impact.

Examining genetic associations with 1300 serum metabolites in patients with inflammatory bowel disease and healthy controls, Di’Narzo et al confirmed the protective role of lipids with omega-6 polyunsaturated fatty acids (eg, 1-(1-enyl-palmitoyl)−2-arachidonoyl-GPC and 1-arachidonoyl-GPC) and the harmful role of linoleic acid-containing lipids (eg, 1-oleoyl-2-linoleoyl-GPE and 1-palmitoyl-2-linoleoyl-GPE) in Crohn’s disease risk through an MR analysis.64 An MR analysis focused on the association of obesity-related metabolisms with colorectal cancer and identified that pyruvate was inversely associated with colorectal cancer risk.43 Yun et al systemically investigated the relationship between 486 blood metabolites and colorectal cancer, consistently identifying the protective role of pyruvate.65 Additionally, they found 1,6-anhydroglucose and gamma-glutamyl threonine were associated with the increased risk of colorectal cancer, while nonadecanoate, 1-linoleoylglycerophosphoethanolamine and 2-hydroxystearate were associated with the decreased risk.65

Currently, many targeted drugs are developed based on proteomics. High-throughput proteomics holds great promise for revealing the underlying molecular mechanisms of intestinal diseases and identifying potential drug targets.66 A proteome-wide MR and colocalisation study investigated the potential association between 4907 circulating proteins and inflammatory bowel disease risk, and identified the MST1, HGFAC, STAT3, ITPKA and CXCL5 as potential therapeutic targets.67 Additionally, the crucial effect of ST2 and CSF-1 in inflammatory bowel disease was explored in other MR studies.68 These findings greatly advance the landscape of the new drug development for inflammatory bowel disease. In fact, clinical trials also validated the results from proteome-wide MR. For example, small-molecule inhibitors of Janus kinase signalling, which acts upstream of the transcription factor STAT3, have demonstrated efficacy in moderate to severe UC.69 A recent proteome-wide MR study integrated seven proteomic studies, and systematically revealed the association between 4853 circulating proteins and colorectal cancer and identified the harmful role of GREM1 and CHRDL2 and the protective role of 11 proteins, including CLSTN3, POLR2F, ADPGK, CSF2RA, CSAG1, STXBP6, CD86, CXADR, FUT3, MMP2 and TIMP2.70 Protein–protein network analysis found that GREM1 and CHRDL2 were involved in the differentiation of osteoclast, and MMP2 and TIMP2 were involved in the tumorigenesis.70 Additionally, the druggability of these proteins was further evaluated, and POLR2F, CSF2RA, CD86 and MMP2 were identified as potential therapeutic targets for colorectal cancer.70

MR studies on circulating proteins commonly use cis-SNPs within coding gene regions as instrumental variables, which is more plausible.71 First, cis-SNPs are located near their corresponding genes and thus directly affect the gene products, strengthening the relevance of the SNP–exposure association. Limiting instruments to the cis-region around each protein-coding gene can help minimise horizontal pleiotropy that violates MR assumptions, because variants in remote areas have higher probability of influencing multiple traits through interconnected biological pathways. Additionally, using cis-variants as instrumental variables may not entirely eliminate horizontal pleiotropy, where a gene can regulate multiple proteins that might not operate within the same biological pathway. When large-scale studies enable multiple instruments for one protein (even after clumping), selecting those variants from the same gene region can increase the chance of assumption violations due to similar pleiotropy patterns.72 This concern is exacerbated when only one instrumental variable is available for a phenotype, limiting the ability to perform sensitivity analyses.

Drug target exploration

Most pharmacological agents exert their therapeutic or adverse effects by interacting with specific protein targets; thus, elucidating these drug targets is foundational for pharmaceutical development. Leveraging genetic variants influencing the expression or function of intestinal disease drug target proteins as instrumental variables, MR can investigate target perturbation consequences to evaluate the efficacy and safety of emerging and current interventions.73 Table 1 summarised MR studies on the exploration of intestinal disease drug targets.

Table 1
|
Summary of Mendelian randomisation studies on exploration of intestinal disease drug targets

Drug target investigations using MR in intestinal diseases can be broadly categorised into two approaches. Proteome-wide analyses related numerous circulating proteins to disease outcomes, as we introduced above, typically validating established while uncovering novel pharmacological targets that could inform drug development efforts. Repurposing studies focused on interrogating whether existing drugs developed for other indications might harbour additional therapeutic utility in intestinal diseases and evaluated the safety in intestinal, exemplified by evaluating antihyperlipidaemic and antihypertensive agents. Recently, several repurposing studies have evaluated lipid-lowering medications for inflammatory bowel disease treatment. Specifically, genetic variants linked to key lipid drug targets like PCSK9,57 74 HMGCR,74 CETP57 and NPC1L175 served as instruments within an MR framework to probe effects on inflammatory bowel disease. PCSK957 74 and NPC1L175 inhibition is associated with increased inflammatory bowel disease risk, HMGCR inhibition may confer Crohn’s disease risk, while CETP inhibition lowers the risk of Crohn’s disease.57 74 In addition, the association between genetically proxied long-term ACE inhibition and the risk of colorectal cancer has been explored.76 Though findings remain preliminary, the analyses suggested that potential safety concerns across drug classes required further investigation in additional observational studies.

Intestinal microbiota

The intestinal microbiota and its metabolites play an important role in the health of humans, influencing the immune system, nutrient absorption and metabolism. The effect of microbiota and its metabolites on inflammatory bowel disease,77–79 colorectal cancer80–83 and coeliac disease84 has been indicated in MR studies. For example, genetically predicted higher levels of Verrucomicrobiaceae, Akkermansia, Enterobacteriaceae, Coprococcus, Oxalobacter and Ruminococcaceae genera were associated with increased risks of inflammatory bowel disease.77–79 Genetically predicted genus Bifidobacterium was related to an increased risk of coeliac disease, while phylum Lentisphaerae and genus Coprobacter were associated with a decreased risk.84 However, the small sample size of GWAS of intestinal microbiota in current MR studies limits the selection of instrument variables.83 The population samples of intestinal flora that Li et al83 used are relatively small which is attributed to the potential insufficient statistical efficacy. Larger GWAS datasets are needed in future studies.

Causal association between intestinal diseases and their comorbidity

The exploration of the risk or protective factors related to intestinal diseases contributes to the prevention or treatment of these diseases. It is notable that intestinal diseases are frequently related to the occurrence of multiple diseases rather than occurring in isolation. Concepts of the gut–brain axis and other frameworks highlight the bidirectional effects between intestinal and other diseases. MR has been performed to reveal the relationships among various intestinal diseases and explore the intricate association between intestinal diseases and a custom of health conditions (figure 3). MR studies provided further evidence on the link between intestinal and mental health,85 86 renal disorders,87 88 hepatopancreatobiliary diseases,89 90 autoimmune diseases91 and type 2 diabetes.19 92 These findings emphasised the importance of establishing systemic prevention and treatment strategies in the area of intestinal and associated health diseases.

Figure 3
Figure 3

A summary of Mendelian randomisation associations between intestinal diseases and other systemic diseases. CD, Crohn’s disease; CeD, coeliac disease; CRC, colorectal cancer; DD, diverticular disease; IBD, inflammatory bowel disease; IBS, irritable bowel syndrome; UC, ulcerative colitis.

Future perspectives of MR study in intestinal diseases

Integrating multiomics data through MR represents a promising future direction to uncover more complex mechanistic pathways. This can clarify how genetic regulation of one omic marker (eg, DNA methylation) associates with activity changes in related functional layers (eg, gene expression, protein abundance) to ultimately affect intestinal disease risk. For example, recently developed multiomics MR methods combine different molecular quantitative trait loci (QTLs) to probe relationships from genotype through intermediating -omics layers to disease outcome.93 By integrating three aspect associations, this study ultimately unveiled the association between two mitochondrial genes including PARK7 and ACADM with inflammatory bowel disease.93 The current landscape of multiomics resources includes various QTL, such as methylation QTL, expression QTL, splicing QTL and protein QTL. These diverse QTLs offer a rich avenue for expanding multiomics MR investigations, potentially unveiling intricate molecular mechanisms underlying intestinal diseases.

Leveraging research designs with different strengths and sources of bias allowed the triangulation of findings. Integrating MR studies using extensive GWAS data with prospective observational research can complement each other, and thus reinforce the conclusions. One study has emerged as a prominent manifestation in this regard. Sadik et al, using four complementary methodological approaches including cohort study, linkage disequilibrium score regression, polygenic risk score analysis and two-sample MR, identify potential causal links between inflammatory bowel disease in parents and autism in children.94

The quality of MR investigation is also noteworthy. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE)-MR guidelines outline essential components required for drafting an MR paper, including study design, data sources, selection and definition of variables, statistical analysis methods and interpretation of results.95 When composing an MR paper, researchers should refer to the STROBE-MR guidelines to ensure the quality of their manuscript, which contributes to enhancing the quality of MR studies and ensuring their credibility and reproducibility.

The GWAS sample size could limit the power to detect and test for causality, which may partly explain the null finding from previous MR studies. Thus, larger sample sizes are needed to achieve the power required to detect variants especially of smaller effect for exposure, and larger sample sizes will be also needed for intestinal disease particularly for less heritable disorders. Besides, an MR analysis could only explore a linear relationship between exposure and outcome using summary-level datasets, and therefore null association could also be found when the true relationship is non-linear (eg, U shape). It is imperative to conduct MR analysis using individual-level data obtained from large-scale GWAS in order to investigate the potential non-linear associations.

Most of the MR studies’ populations were European, which may not be generalisable to other ancestries. MR associations explored in populations with different ancestries are encouraged.

Conclusion

In this review, we integrated the MR studies on intestinal diseases and summarised the causal risk factors and complications. Lifestyle behaviours, diet characteristics, nutrients, obesity, inflammatory biomarkers, gut microbiota, metabolites and proteins, and some specific diseases are causally associated with intestinal diseases in MR studies. Multiomics data have been applied in recent MR studies, providing new insights into the pathogenesis of intestinal diseases. Triangulation was suggested to be used as a validated method for strengthening the causal inference in future MR studies.