A Mendelian Randomization Analysis Investigates Causal Associations between Inﬂammatory Bowel Diseases and Variable Risk Factors

: The question of whether variable risk factors and various nutrients are causally related to inﬂammatory bowel diseases (IBDs) has remained unanswered so far. Thus, this study investigated whether genetically predicted risk factors and nutrients play a function in the occurrence of inﬂamma-tory bowel diseases, including ulcerative colitis (UC), non-infective colitis (NIC), and Crohn’s disease (CD), using Mendelian randomization (MR) analysis. Utilizing the data of genome-wide association studies (GWASs) with 37 exposure factors, we ran Mendelian randomization analyses based on up to 458,109 participants. Univariable and multivariable MR analyses were conducted to determine causal risk factors for IBD diseases. Genetic predisposition to smoking and appendectomy as well as vegetable and fruit intake, breastfeeding, n -3 PUFAs, n -6 PUFAs, vitamin D, total cholesterol, whole-body fat mass, and physical activity were related to the risk of UC ( p < 0.05). The effect of lifestyle behaviors on UC was attenuated after correcting for appendectomy. Genetically driven smoking, alcohol consumption, appendectomy, tonsillectomy, blood calcium, tea intake, autoimmune diseases, type 2 diabetes, cesarean delivery, vitamin D deﬁciency, and antibiotic exposure increased the risk of CD ( p < 0.05), while vegetable and fruit intake, breastfeeding, physical activity, blood zinc, and n -3 PUFAs decreased the risk of CD ( p < 0.05). Appendectomy, antibiotics, physical activity, blood zinc, n -3 PUFAs, and vegetable fruit intake remained signiﬁcant predictors in multivariable MR ( p < 0.05). Besides smoking, breastfeeding, alcoholic drinks, vegetable and fruit intake, vitamin D, appendectomy, and n -3 PUFAs were associated with NIC ( p < 0.05). Smoking, alcoholic drinks, vegetable and fruit intake, vitamin D, appendectomy, and n -3 PUFAs remained signiﬁcant predictors in multivariable MR ( p < 0.05). Our results provide new and comprehensive evidence demonstrating that there are approving causal effects of various risk factors on IBDs. These ﬁndings also supply some suggestions for the treatment and prevention of these diseases.


Introduction
Chronic, progressive inflammatory bowel disease (IBD) can cause bowel injury, the condition for hospitalization, disability, and a reduction in life quality.Although North America and Europe have the highest rates, developing nations such as Asia have seen rising incidence rates as a result of these regions' increased development [1].Although the exact source of these phenomena is unknown, it can be connected with the intricate interaction between the environment and genetics [2].Due to the statement that the incidence speeds of IBD are higher in developed countries, the latter has been linked to disease development [3].In addition, the urbanization of societies is associated with changes in diet, antibiotic use, hygiene status, microbial exposure, and pollution, which have been implicated as potential environmental risk factors for IBD.Environmental risk factors for individual, familial, community-based, country-based and regionally based origin could all contribute to the pathogenesis of IBD [4][5][6].Lending further support to the critical importance of environmental influences is the recognition of the central role of the gut microbiota in the development and propagation of inflammation in IBD [7].Although host genetics might partly determine gut microbial structure, external environmental exposure from the time of birth to adulthood continue to alter the composition, structure, and function of the gut microbiome, thereby dynamically altering the risk and natural history of disease throughout life [8,9].Discovering how environmental factors influence the onset of IBD and contribute to its pathogenesis could ultimately help to determine how individuals can reduce their risk of disease or have a milder clinical course.The search for pathogenic environmental factors is also important, as many unmet therapeutic needs and suboptimal outcomes in IBD remain.Mechanistic insights obtained from robustly defining environmental influences could also lead to the identification of new therapeutic targets and treatment strategies.The best treatment should be started early in disease progression to avoid problems, because IBD can be a progressive and chronic condition.To conduct clinical remission and mucosal healing, which can improve life quality, the current treatment paradigm for ulcerative colitis (UC), Crohn's disease (CD), and non-infective colitis (NIC) uses biological therapy [10].However, measuring the causal impact of these variable factors on IBD can be challenging due to possible reverse causality and potential residual confounders that can induce spurious associations and mask the effects of real risk factors.Therefore, it is essential to clarify whether these possible risk factors recreate causal functions in the evolution of IBD or serve as transferred risk factor consequences.In addition, genetic and environmental variables work together to cause the disease, but the exact mechanisms are still poorly understood, particularly concerning the non-genetic hazards.Hundreds of variations have already been correlated to IBD by genome-wide association studies (GWASs) [11,12], but our understanding of the different risk variables and genetic relations that influence IBD is still restricted.Numerous factors that may affect the IBD risk have been specified by prior studies, but these investigations have not always come to the same results, in part because of the difficulties in establishing satisfactory statistical authority, avoiding bias, and correcting for confounding aspects [13,14].Additionally, only a small number of research studies have examined the genetic-environmental interactions linked to IBD, including single-nucleotide polymorphisms (SNPs) [15,16].Therefore, both environmental and genetic elements play a function in the complex etiology of IBD.Numerous modifiable factors have been investigated [12], but none are reliably detrimental nor protective.There is currently a lack of solid research to conduct IBD prevention.In addition, there has not been an organized action to compile and evaluate this evidence, despite the vast number of research studies that have looked at environmental factors and IBD, including Mendelian randomization (MR) analyses.MR is an instrumental variable analysis method for examining causal relationships between disease outcomes and risk factors [17].It employs genetic variants that are closely connected with a risk factor as instrumental variables (IVs) and mimics a randomized controlled background in which all other factors excluding exposure are almost distributed over subgroups.Thus, MR analysis can avoid reversing causation and confounding biases that are common in observational studies.

R E T R A C T E D
In this study, two-sample Mendelian randomization analysis was performed to investigate the causal effects of 37 genetically predicted potential risk factors and nutrients, including related diseases, drugs, lifestyle, surgeries, lipid and glucose metabolism, blood parameters, and obesity traits, on IBD (including UC, CD, and NIC) in the European population.This study aimed to provide a comprehensive overview of putative variable risk factors for IBD and offer novel insights into the etiology of IBD or colitis disorders.

MR Design
MR was utilized to investigate the relationships between various risk factors and IBD or different types of colitis.A total of 37 primary risk factors were selected and classified into eight categories: exposure to drugs, lifestyle behaviors, surgeries, related diseases, blood parameters, lipid metabolism, glucose metabolism, and obesity traits.SNPs associated with these risk factors were utilized as the instrumental variables.The following three assumptions served as the foundation for the MR study: (1) the SNPs are closely connected with the risk factors; (2) the SNPs are irrelevant to various confounders; (3) the SNPs only influence the outcomes via the risk factors.This study was performed according to the STROBE-MR procedures (Figure 1).
MR is an instrumental variable analysis method for examining causal relationships between disease outcomes and risk factors [17].It employs genetic variants that are closely connected with a risk factor as instrumental variables (IVs) and mimics a randomized controlled background in which all other factors excluding exposure are almost distributed over subgroups.Thus, MR analysis can avoid reversing causation and confounding biases that are common in observational studies.
In this study, two-sample Mendelian randomization analysis was performed to investigate the causal effects of 37 genetically predicted potential risk factors and nutrients, including related diseases, drugs, lifestyle, surgeries, lipid and glucose metabolism, blood parameters, and obesity traits, on IBD (including UC, CD, and NIC) in the European population.This study aimed to provide a comprehensive overview of putative variable risk factors for IBD and offer novel insights into the etiology of IBD or colitis disorders.

MR Design
MR was utilized to investigate the relationships between various risk factors and IBD or different types of colitis.A total of 37 primary risk factors were selected and classified into eight categories: exposure to drugs, lifestyle behaviors, surgeries, related diseases, blood parameters, lipid metabolism, glucose metabolism, and obesity traits.SNPs associated with these risk factors were utilized as the instrumental variables.The following three assumptions served as the foundation for the MR study: (1) the SNPs are closely connected with the risk factors; (2) the SNPs are irrelevant to various confounders; (3) the SNPs only influence the outcomes via the risk factors.This study was performed according to the STROBE-MR procedures (Figure 1).

Selection of Genetic Instruments
GWASs of participants of European ancestry were selected as data sources for genetic agencies associated with the 37 risk factors.Genetic instruments of cigarettes per day, smoking, and alcoholic drinks were extracted from the GSCAN (GWAS and Sequencing Consortium of Alcohol and Nicotine use) consortium [11].GWAS summary statistics for coffee, tea, fruit, and vegetable intake; vitamin D consumption; and breastfeeding were obtained from the MRC-IEU (MRC Integrative Epidemiology Unit) consortium.IVs for

GWAS Summary Statistics for IBD Cohorts
GWAS summarization for UC, CD, and NIC was received from the FinnGen consortium.The R5 release of the FinnGen data was utilized [19]; this data set contains 1213 cases and 164,254 controls for CD, 2155 cases and 186,103 controls for UC, 411 cases, and 103,973 controls for NIC.All selected GWASs from the Biobank obtained ethical approval from FinnGen Steering Committee, and individuals provided informed consent.

Statistical Analysis
The F-statistic was utilized to evaluate genetic instrument strength.F-statistics (F = beta 2 /se 2 ) were calculated for each SNP, and a general F-statistic was calculated for all SNPs for the corresponding exposure.F > 10 was considered to be sufficient strength.All F-statistics were over 10.The random-effect inverse-variance-weighted (IVW) technique was utilized as the main analysis method to estimate the association between genetic liability to modifiable risk factors and the risk of pancreatitis.Given that the analysis is sensitive to outliers and horizontal pleiotropy, three sensitivity analyses, including the weighted median, MR-Egger, and MR-PRESSO methods, were used to examine the consistency of the results.The weighted median model was used; this can produce unbiased estimates under the precondition that at least 50% of the selected IVs are valid [33].MR-Egger regression was used to obtain cogent causal estimates under the influence of pleiotropy [34].The MR-PRESSO method was conducted to specify outliers due to the existing pleiotropy; causal effect estimates were obtained with the IVW approach after removing these outliers [35].In addition, a leave-one-out sensitivity test was applied to examine if the SNPs possessing effective horizontal pleiotropic effects could affect the causal estimates [36].MR-PRESSO and Cochran's Q statistics were used to evaluate pleiotropy and heterogeneity, respectively.The multivariable MR analysis of the genetic associations between the instruments and UC was adjusted for appendectomy, while the multivariable MR analysis of the associations between the instruments and CD was adjusted for alcohol consumption and smoking.All statistical analyses were performed using R 4.2.1, (R Foundation for Statistical Computing, Vienna, Austria), with the R packages "TwoSampleMR" (https://github.com/MRCIEU/TwoSampleMR)and "MRPRESSO".(https://github.com/rondolab/MR-PRESSO(accessed on 24 July 2022)).

R E T R A C T E D
Nutrients 2023, 15, 1202 5 of 17

Baseline Characteristics of the 36 Candidate Risk Factors
Thirty-seven potential risk factors were included in the analyses.The risk factors were classified into eight categories: lifestyle behaviors, surgeries, exposure to drugs, related diseases, blood parameters, lipid metabolism, glucose metabolism, and obesity traits (Table 1).The lifestyle behaviors included smoking, alcohol consumption, coffee consumption, tea consumption, fruit intake, vegetable intake, breastfeeding, cesarean delivery, vitamin D intake, and physical activity.The surgeries included appendectomy and tonsillectomy, and the related diseases include T2D, CKD, and autoimmune diseases (including celiac disease and SLE).The blood parameters included calcium, lactose, CRP, and zinc.Exposure to drugs included antibiotics and isotretinoin.Additionally, three traits related to glucose metabolism, five traits related to lipid metabolism, and five obesity traits were analyzed.SNP numbers ranged from 11 to 523.Across the 37 variable potential risk factors that were explored, the F-statistics of their separate genetic instruments were all greater than the empirical threshold of 10, indicating no possible inadequate instrument bias.

Causal Effects of Various Factors on NIC
Next, the causal relationships between risk factors and NIC were examined (Figure 4).Notably, genetic liability to alcohol consumption (OR = 1.545, p = 0.001) and fruit intake (OR = 0.497, p = 0.001) was strongly associated with increased odds of NIC.Genetic liability to smoking and appendectomy, as well as decreased vegetable intake, n-3 PUFA intake, and vitamin D deficiency, was suggestively related to raised risk of NIC.

R E T R A
C T E D take, and vitamin D deficiency, was suggestively related to raised risk of NIC.

Discussion
Genetic susceptibility factors play crucial functions in IBD development [37].The development and progression of IBD are multidimensional, with interactions between environmental and genetic elements [38].Therefore, a critical evaluation of the environmental and genetic factors related to IBD is offered in this study, which encompasses MR analyses of observational research.A total of 37 variables, such as lifestyle behavior, dietary intake, blood parameters, obesity traits, related diseases, drug exposure, and glucose and lipid metabolism were examined.Among them, we found 10 risk variables and 10 defensive factors with epidemiological evidence of moderate-to-high strength.The methodological standards among the MR analyses greatly differed.Several variables were connected with CD, UC, or NIC.The found specificity might represent various pathogeneses and traits of these diseases [39].Subgroup analyses for smoking [40][41][42][43][44][45], breastfeeding [46], cesarean delivery [23,47], and high vitamin D levels have found significant differential relationships

R E T R A C T E D
among groups [48].These variations could be explained by unique genetic predispositions and environmental exposures happening in particular geographical regions.The investigation by publication year revealed more conventional calculations for fiber intake and breastfeeding in analyses disseminated after the year 2000 [12,49,50].
Cigarette smoking and alcohol use are two well-recognized lifestyle risk factors for colitis.Smoking promotes the progression from US to CD and accelerates the development of IBD [43,51].MR analysis confirmed that smoking was related to a higher risk of IBD.The impacts of smoking on NIC were partially attenuated after adjusting for alcohol consumption, suggesting that this association is not robust enough in NIC.Alcohol consumption contributes to the progression or initiation of IBD and amplifies the relationship between the risk factors of genetics and CD in a dose-dependent manner [52].Our results verified the causal associations between CD or NIC and alcohol consumption.Notably, the risk of CD due to genetically predicted alcohol consumption was higher than that of NIC.However, there was no proof of a positive connection between UC and alcohol consumption.Furthermore, alcohol accounts for 40-70% of CD etiologies and only 20% of NIC etiologies; thus, this causal relationship may be statistically attenuated by other NIC risk factors.The associations between coffee or tea consumption, and IBD are controversial.Some studies have reported that tea and coffee reduce the risk of colitis, while another prospective cohort analysis discovered no relationship between tea consumption and the risk of colitis [53,54].Our MR study found no associations between genetically predicted coffee consumption and colitis risk.Moreover, appendectomy, autoimmune diseases, and CKD were associated with increased risk of IBD events [55][56][57][58][59][60].Autoimmune diseases, including celiac disease, and SLE have been reported to be associated with IBD in previous studies.The current MR analysis supports a suggestive association between autoimmune diseases and CD; however, this association did not persist when corrected for smoking and alcohol consumption [61][62][63][64].No effect was seen for causal connections between IBD and celiac disease, SLE, nor CKD in this study.Additionally, appendectomy was associated with IBD.However, the risk was reduced if more than five years had expired between colitis diagnosis and appendectomy, proposing that the connection may be biased by unneeded appendectomies operated in individuals with developing colitis [56,57].Tonsillectomy was linked with CD but not NIC or UC.However, patients undergoing tonsillectomy have often been exposed to previous antibiotics [65].
Furthermore, the current results indicate that physiologically normally higher blood calcium levels raised CD risk, while there were no connections between IBD and genetic predisposition to higher lactose and CRP.
Retrospective case-control studies were used to determine the pre-illness diet in order to infer the significance of nutrition in CD and UC development.According to pediatric case-control research study by Amre et al. [66], there is a negative correlation between vegetable or fruit intake and the development of CD.The impacts of dietary micronutrients and macronutrients on the risk of disease have also been more robustly estimated by previous research in North America and Europe.Women in the most elevated percentile of dietary fiber consumption had significantly decreased risk of incident CD than those in the lowest quintile in a considerable cohort of 170,776 females followed for 28 years [67].In contrast, dietary fiber from bran, cereal, and whole grains was not linked to modified risk of disease.The highest percentage of dietary fiber intake from fruits was related to decreased CD risk [67].Consuming fruit lowered the risk of developing CD and UC [68][69][70][71].In addition, eating vegetables reduced the risk of UC.These results have some biological plausibility.Fibers can impact the function of the intestine, which is compromised in inflammatory bowel disease (IBD), and plant ingredients can influence the microorganism translocation across the intestine mucosa [72].CD was also linked to lactose level but did not impact UC, which is consistent with a previous study [73].Dietary n-3 PUFA consumption has been also shown, in two prospective cohort studies, to be inversely related to UC risk, but dietary n-6 PUFA consumption is positively connected with incident UC risk [74,75].n-3 PUFA consumption reduced the clinical colitis severity in an investigation

R E T R A C T E D
of generated colitis in mice [76].These results are consistent with other studies that showed the preventive effect of n-3 PUFAs on the risk of IBD [77][78][79].In addition, the current results suggested that high vitamin D levels were related to diminished risk of UC, CD, and NIC, and these associations persisted in the CD cohort after adjusting for alcohol consumption [80,81].
Furthermore, high zinc intake was found to be negatively correlated with women's chance of improving CD in some cohort investigations [77,82].With 16 mg daily zinc consumption, or double the suggested everyday dose, the risk was detected to be reduced.Low blood zinc was linked to raised risk of surgeries, hospitalizations, and illness-related complications in colitis patients [83].Additionally, improvement in results was linked to the normalization of zinc levels.Zinc supplementation was linked to a decline in intestinal permeability, as specified by the mannitol:lactulose ratio, in small interventional research [84].In addition, apolipoprotein A-I was previously reported to be related to the severity of UC [85].However, these results provided no effects of any associations between IBD risk and genetically predicted apolipoprotein A-I.Notably, genetically predicted total cholesterol was suggestively connected with a lower odds ratio of UC, whereas these associations were not significant after adjustment for appendectomy.There were no relationships between total cholesterol and IBD, in agreement with a previous investigation [86].A previous meta-analysis documented a positive link between CD risk and T2D [87,88].However, there were no associations between UC nor NIC and T2D in the findings of our study.Despite this, we observed a suggestive relationship between CD risk and T2D; however, this relationship was not significant after adjusting for alcohol consumption.
Causal associations between IBD and fasting glucose, fasting insulin, or HbA1c were not observed in this study.Furthermore, suggestive associations between obesity traits and UC were observed, which is consistent with previous research [89,90].However, these associations did not persist after adjusting for appendectomy, suggesting that elevated risk of appendectomy due to higher BMI or whole-body fat mass may explain this relationship.
It is worth mentioning that some protective factors for colitis were also identified in the present MR study.Relationships between colitis, and both physical activity and breastfeeding have not been reported in previous studies.Genetically predicted breastfeeding was also associated with lower risk of UC, CD, and NIC.Lack of breastfeeding has been linked to immune-mediated illnesses and clostridium difficile colonization [91].The protective effect might be mediated by increased innate mucosal immunity development as a result of microbiome interaction [92,93].Physical activity may modulate colitis risk by affecting multiple pathways, including individuals' health behaviors, living environments, and lifestyles [94].Our results also verified the causal associations between CD and cesarean delivery.Notably, the risk of CD due to genetically predicted cesarean delivery was higher than that of UC and NIC, which is consistent with previous research [23,47].However, there was no proof of a positive connection between UC or NIC and cesarean delivery.
Furthermore, antibiotics raised the likelihood of eventual CD but not UC or NIC.In initial experiments, a dose-response association was found; therefore, IBD risk was elevated by all antibiotic classes [33].Antibiotics can change the taxonomic richness and diversity of the human gut microbiota while impairing its metabolic condition, altering the composition of the microbiome [26,95].Antibiotics have been observed to exacerbate the dysbiosis prevalent in CD patients, and a more changed microbiota has been detected in CD than in NIC or UC [34,96].Isotretinoin was not linked to UC, CD, or NIC, according to the results.
There are several strengths of the attending investigation related to the data source and research design.First, the MR design promoted the computation of the causal links between heritable complex traits, which avoids the biases inherent in conventional observational epidemiological studies.We applied multiple sensitivity computations to verify the plausibility of the instrumental variable assumptions and interpreted the outcomes after viewing horizontal pleiotropy and outliers.Second, this study systematically analyzed the most extensive number of variable causal factors for IBD to date.In this regard, no MR

R E T R A C T E D
studies have analyzed the causal consequences of genetic liability on potential risk factors for IBD or colitis.Third, the GWAS data utilized in this analysis were primarily taken from participants of European ancestry, which can decrease the bias of population stratification.Aside from autoimmune diseases, this study avoided sample overlap between most exposure types and outcomes, thereby controlling the type 1 error to be as low as possible.Nonetheless, there are some limitations of the current study that also need to be considered.First, as in all MR studies, it is difficult to confirm a lack of bias for horizontal pleiotropy.Thus, the MR-PRESSO global analysis and MR-Egger regression were operated to detect widespread horizontal pleiotropy [29,30].Importantly, the results of this study remained robust after the discarding of outlier variants identified using the MR-PRESSO outlier test.Second, the sample size for NIC was rather small, which could limit the power of statistics to detect true causal relationships.

Conclusions
In this study, an MR investigation comprehensively reveals the causal relationship between IBD and a variety of lifestyle factors, associated disorders, drug exposure, surgeries, blood markers, lipid metabolism, glucose metabolism, nutrients, and obesity.Additionally, this study lists the precise genera of variable risk factors implicated in UC, CD, NIC, or IBD pathogenesis.Our discovery may potentially provide fresh perspectives on the design of targeted IBD, UC, NIC, and CD prevention and therapy strategies.

Figure 1 .
Figure 1.The entire flow chart of the study, the red X represents potential pleiotropic or direct causal effects between instrumental variables that would violate Mendelian randomization assumptions.

Figure 1 .
Figure 1.The entire flow chart of the study, the red X represents potential pleiotropic or direct causal effects between instrumental variables that would violate Mendelian randomization assumptions.

Figure 2 .
Figure 2. Forest plot of odds ratio and 95% confidence intervals (lines).Overview impact calculations of MR indicating relationships between UC and risk factors; significant impacts are exhibited in bold.

Figure 2 .
Figure 2. Forest plot of odds ratio and 95% confidence intervals (lines).Overview impact calculations of MR indicating relationships between UC and risk factors; significant impacts are exhibited in bold.

Figure 3 .
Figure 3. Forest plot of odds ratio and 95% confidence intervals (lines).Overview impact calculations of MR indicating relationships between CD and risk factors; significant impacts are exhibited in bold.

Figure 3 .
Figure 3. Forest plot of odds ratio and 95% confidence intervals (lines).Overview impact calculations of MR indicating relationships between CD and risk factors; significant impacts are exhibited in bold.

Figure 4 .
Figure 4. Forest plot of odds ratio and 95% confidence intervals (lines).Overview impact calculations of MR indicating relationships between NIC and risk factors; significant impact are exhibited in bold.

Figure 4 .
Figure 4. Forest plot of odds ratio and 95% confidence intervals (lines).Overview impact calculations of MR indicating relationships between NIC and risk factors; significant impact are exhibited in bold.

Figure 5 .
Figure 5. Forest plot of odds ratio and 95% confidence intervals (lines).Summary effect estimates of multivariable MR reporting associations between (A) UC, (B) CD, and (C) NIC, and risk factors; significant estimates are shown in bold.

Figure 5 .
Figure 5. Forest plot of odds ratio and 95% confidence intervals (lines).Summary effect estimates of multivariable MR reporting associations between (A) UC, (B) CD, and (C) NIC, and risk factors; significant estimates are shown in bold.

Table 1 .
Characteristics of the GWAS summary data.