1. Introduction

Digestive system diseases (DSDs) are the most common disorders globally, including esophageal and gastrointestinal diseases (chronic gastritis, colorectal cancer, Crohn’s disease, gastric cancer, gastric ulcer, irritable bowel syndrome and esophageal cancer). Chronic gastritis is persistent inflammatory lesions in the gastric mucosa and it has an influence on above half of the global population in various degrees.[1,2] For colorectal cancer, it is the second most common cause of death from malignant tumor.[3] Risk factors about colorectal cancer consist of family and personal medical history (such as family genetics and history, Crohn’s Disease and so on) and lifestyle (such as patterns of daily dietary and the habit of cigarette smoking).[4] Gastric cancer is the fifth most common cancer with approximately 784,000 deaths all over the world in 2018 and one of the most risk factors of gastric cancer is Helicobacter pylori infection.[5] The study in[6] illustrated that gastric ulcer is a disability in the stomach wall penetrating through the entire mucosa and the muscularis mucosae deeply and it can be also caused by toxic factor of H. pylori infection. Irritable bowel syndrome[7] is one of the most common gut–brain interaction disorders and it has an influence on above one tenth people worldwide with including abdominal pain related to a change in stool form or frequency.[8] For esophageal cancer, it is the seventh most common cancer globally and it leads to around 450,000 patient deaths each year.[9] Risk factors about esophageal cancer are tobacco smoking, alcohol overconsumption, intake of red meat and the consumption of very hot beverages.[10] Of note, some evidences had found out that the gut microbiota might have a causal link with the development of digestive system disorders.[11] Gut microbiota located in the gastrointestinal tract contains thousands of bacterial species and trillions of microorganisms and it play an important role in a variety of diseases.[12] Its inflammation might have a causal association with the initiation, development and progression of digestive cancer.[13,14] Accumulating evidence demonstrated that specific gut bacteria, which are considered as interbacterial communication, are related to the development of gastrointestinal cancers.[15] In,[16] authors have reported that differences in microbial composition have been linked to chronic digestive diseases such as inflammatory bowel disease and colorectal cancer. The study in[17] investigated the significance of a diminution in gut microbial diversity for host metabolism. The research has demonstrated that the gut microbiota is related to the development of colorectal cancer.[18] Although some studies have demonstrated causal relationships between gut microbiota and DSDs, the detailed development of DSDs still remains unclear. Hence, subsequent researches play a crucial role in the exploration of causal links between gut microbiota and DSDs.

Mendelian randomization (MR) is a well-known technique for the assessment of causal relationships between exposure dataset and outcome dataset by using genome-wide association study (GWAS) summary data. In our context, the MR statistical analysis was employed to explore causal associations between gut microbiota as exposure and DSDs as outcome dataset and showcased results based on forest plots.

2. Materials and methods 2.1. Ethical statement

All of dataset employed in our study were large-scale public GWAS summary data. Ethical approval and consent to participate were acquired in all original studies. The flowchart of the process is illustrated in Figure 1. In a nutshell, the human gut microbiota was considered as exposure dataset and DSDs were considered as outcome dataset.

Figure 1.:

The flowchart of our study. GWAS = genome-wide association study, MR = Mendelian randomization, SNP = single nucleotide polymorphism.

2.2. Gut microbiota exposure

Human gut microbiota GWAS summary data were selected from the MiBioGen consortium (https://mibiogen.gcc.rug.nl/) with 16S ribosomal RNA gene sequencing profiles and genotyping data from 18,340 sample sizes. The gut microbiota dataset contains 211 taxa, 131 genera, 35 families, 20 orders, 16 classes and 9 phyla.[19] In our study, we selected a series of parameters to make sure of the accuracy of results. The selection criteria of instrumental variables (IVs) from single nucleotide polymorphisms (SNPs) are as followed[20,21]: the genome-wide statistical significance threshold is less than 1 × 10−6 as candidate IVs; the linkage disequilibrium threshold was set as R2 < 0.01 and distance was set as kb = 10000 kb to avoid linkage disequilibrium; F-statistics[22] of each IV was calculated by F = R2 × (N − 2)/(1 − R2), where R2 represents the genetic variant explanation of the exposure variance and N represents sample sizes. Meanwhile, F > 10 indicated a strong instrument and was retained.[23]

2.3. Digestive system diseases outcome

The summary data of 7 DSDs were derived from GWAS (portal: https://gwas.mrcieu.ac.uk/), including chronic gastritis (N = 361,194, ncase = 1790, ncontrol = 359,404), colorectal cancer (N = 470,002, ncase = 6581, ncontrol = 463,421), Crohn’s disease (N = 40,266, ncase = 12,194, ncontrol = 28,072), gastric cancer (N = 476,116, ncase = 1029, ncontrol = 475,087), gastric ulcer (N = 474,278, ncase = 6293, ncontrol = 467,985), irritable bowel syndrome (N = 486,601, ncase = 53,400, ncontrol = 433,201) and esophageal cancer (N = 372,756, ncase = 740, ncontrol = 372,016). The outcome detailed characteristic of GWAS were illustrated in Table 1.

Table 1 - Detailed characteristics of GWAS related to outcome dataset in the study. Type Population Sample size Number of SNPs ID in GWAS Chronic gastritis European 361,194 10,262,134 ukb-d-K11_CHRONGASTR Colorectal cancer European 470,002 24,182,361 ebi-a-GCST90018808 Crohn’s disease Mixed 40,266 9457,998 ebi-a-GCST004132 Gastric cancer European 476,116 24,188,662 ebi-a-GCST90018849 Gastric ulcer European 474,278 24,178,780 ebi-a-GCST90018851 Irritable bowel syndrome European 486,601 9739,966 ebi-a-GCST90016564 Oesophageal cancer European 372,756 8970,465 ieu-b-4960
2.4. Mendelian randomization analysis

We performed a MR analysis to explore causal relationships between gut microbiota and DSDs. MR statistical analysis should comply with 3 core assumptions to reduce estimate bias.[24] Initially, genetic variants must be associated with gut microbiota exposure dataset. Subsequently, IVs of gut microbiota exposure dataset are supposed to be uncorrelated with confounder, which are linked to both gut microbiota and DSDs. Lastly, IVs can only impact DSDs through gut microbiota to avoid horizontal pleiotropy. If a particular taxon had 1 SNP as an IV, Wald ratio method was leveraged for MR analysis. If taxon had more than 1 SNP as IVs, 4 MR regression methods were utilized, such as inverse variance weighted (IVW),[25] weighted median,[26] weighted mode[27] and MR-Egger.[28] IVW method was considered as the primary analysis (P < .05) and other regression methods were served as complements. Meanwhile, the Benjamini-Hochberg method was used to adjust P value and the corrected p_FDR value was displayed in tables in Section 3. In order to evaluate the sensitive of our MR analysis,[29] Cochrane’Q test[30] was utilized to obtain heterogeneity. For horizontal pleiotropy, MR-Egger and MR-PRESSO global test were employed and there were no evidence of both heterogeneity and horizontal pleiotropy based on all of P > .05. Finally, the leave-one-out method was leveraged to investigate the reliability of harmonizing both exposure and outcome dataset.

In our paper, all MR statistical analyses were implemented using “TwoSampleMR” (v0.5.7)[31] and “MR-PRESSO” (v 1.0)[32] packages in R v 4.3.1. In order to accelerate our statistical analysis, “doParallel” (v 1.0.17) package was employed.

3. Results 3.1. Selection of instrumental variables for gut microbiome

After calculating set parameters, 302 bacterial characteristics with 5 biological levels (class, family, genus, order and phylum) were selected. The F-statistics of IVs ranged from 20,631 to 88,430, thereby all of selected IVs were remarkable >10, which manifested the absence of weak instrument bias.

3.2. Results of Mendelian randomization analysis 3.2.1. Causal relationships between gut microbiome and chronic gastritis

According to results of the IVW method, order Erysipelotrichales [OR = 0.996, 95% CI: 0.992–0.999, P = .015], family Erysipelotrichaceae [OR = 0.996, 95% CI: 0.992–0.999, P = .015] and class Erysipelotrichia [OR = 0.996, 95% CI: 0.992–0.999, P = .015] had a decreased evidence of chronic gastritis. Otherwise, genus Victivallis [OR = 1.002, 95% CI: 1.000–1.004, P = .021], phylum Bacteroidetes [OR = 1.004, 95% CI:1.000–1.008, P = .028], order Bacteroidales [OR = 1.004, 95% CI: 1.000–1.008, P = .029], as well as class Bacteroidia [OR = 1.004, 95% CI: 1.000–1.008, P = .029] had a positive evidence of chronic gastritis in Figure 2.

Figure 2.:

MR results of causal relationships between gut microbiome and chronic gastritis (P < 1 × 10−6). MR = Mendelian randomization.

3.2.2. Causal relationships between gut microbiome and colorectal cancer

Using the IVW method, genus RuminococcaceaeUCG005 [OR = 0.718, 95% CI: 0.535–0.963, P = .027] and genus Ruminococcusgauvreauiigroup [OR = 0.745, 95% CI: 0.571–0.972, P = .030] were negative related with the risk of colorectal cancer. On the contrary, family Victivallaceae [OR = 1.128, 95% CI: 1.013–1.257, P = .028] were positive related with the risk of colorectal cancer in Figure 3.

Figure 3.:

MR results of causal relationships between gut microbiome and colorectal cancer (P < 1 × 10−6). MR = Mendelian randomization.

3.2.3. Causal relationships between gut microbiome and Crohn’s disease

Based on the IVW method in Figure 4, family Oxalobacteraceae [OR = 1.246, 95% CI: 1.066–1.457, P = .0057], phylum Cyanobacteria [OR = 1.400, 95% CI: 1.078–1.819, P = .012], order NB1n [OR = 1.268, 95% CI: 1.026–1.566, P = .028], genus Oxalobacter [OR = 1.171, 95% CI: 1.012–1.355, P = .034] had a great risk of Crohn’s disease. Genus Erysipelatoclostridium [OR = 0.740, 95%CI:0.579–0.946, P = .016], genus Tyzzerella3 [OR = 0.781, 95% CI: 0.630–0.969, P = .025], family Ruminococcaceae [OR = 0.639, 95% CI: 0.424–0.962, P = .032], genus FamilyXIIIAD3011group [OR = 0.711, 95% CI: 0.520–0.972, P = .033] and genus Paraprevotella [OR = 0.743, 95% CI: 0.564–0.978, P = .033] acted as preventive results of Crohn’s disease.

Figure 4.:

MR results of causal relationships between gut microbiome and Crohn’s disease (P < 1 × 10−6). MR = Mendelian randomization.

3.2.4. Causal relationships between gut microbiome and gastric cancer

When identifying gut microbiome with gastric cancer, we found that genus Ruminococ caceaeUCG002 [OR = 0.574, 95% CI: 0.385–0.857, P = .0065] and family Ruminococcaceae [OR = 0.400, 95% CI: 0.189–0.848, P = .017] reduced the risk of gastric cancer. On the other hand, genus LachnospiraceaeUCG010 [OR = 1.539, 95% CI: 1.149–2.062, P = .0038], genus Ruminococcusgnavusgroup [OR = 1.397, 95% CI: 1.113–1.752, P = .0039] as well as family FamilyXI [OR = 1.258, 95% CI: 1.050–1.506, P = .013] increased the risk of gastric cancer in Figure 5.

Figure 5.:

MR results of causal relationships between gut microbiome and gastric cancer (P < 1 × 10−6). MR = Mendelian randomization.

3.2.5. Causal relationships between gut microbiome and gastric ulcer

The results provided in Figure 6 from the IVW method showed that genus Holdemania [OR = 0.793, 95% CI: 0.666–0.945, P = .0097] had a decreased incidence of gastric ulcer. Alternatively, order Erysipelotrichales [OR = 1.530, 95% CI: 1.042–2.248, P = .030], family Erysipelotrichaceae [OR = 1.530, 95% CI: 1.042–2.248, P = .030] and class Erysipelotrichia [OR = 1.530, 95% CI: 1.042–2.248, P = .030] had an increased incidence of gastric ulcer.

Figure 6.:

MR results of causal relationships between gut microbiome and gastric ulcer (P < 1 × 10−6). MR = Mendelian randomization.

3.2.6. Causal relationships between gut microbiome and irritable bowel syndrome

When determining gut microbiome with irritable bowel syndrome in Figure 7, we demonstrated that phylum Proteobacteria [OR = 0.808, 95% CI: 0.680–0.959, P = .015] and genus Butyricimonas [OR = 0.887, 95% CI: 0.793–0.991, P = .035] were linked to a low risk of irritable bowel syndrome. Class Melainabacteria [OR = 1.173, 95% CI: 1.062–1.294, P = .0016] and order Gastranaerophilales [OR = 1.173, 95% CI: 1.062–1.294, P = .0016] were linked to a high risk of irritable bowel syndrome.

Figure 7.:

MR results of causal relationships between gut microbiome and irritable bowel syndrome (P < 1 × 10−6). MR = Mendelian randomization.

3.2.7. Causal relationships between gut microbiome and esophageal cancer

Figure 8 illustrated that genus Turicibacter [OR = 0.998, 95% CI: 0.996–1.000, P = .028] reduced the risk of esophageal cancer. Conversely, phylum Verrucomicrobia [OR = 1.003, 95% CI: 1.000–1.005, P = .023], genus Oxalobacter [OR = 1.001, 95% CI: 1.000–1.002, P = .036] and family Oxalobacteraceae [OR = 1.001, 95% CI: 1.000–1.002, P = .043] were positive associated with the risk of esophageal cancer.

Figure 8.:

MR results of causal relationships between gut microbiome and esophageal cancer (P < 1 × 10−6). MR = Mendelian randomization.

3.3. Sensitivity analysis

The robustness of our MR results between gut microbiome and DSDs were evaluated by the sensitive analysis. Significant heterogeneity and horizontal pleiotropy were both identified. Meanwhile, correct causal directions and Steiger P value[29] were also identified. As shown in Table 2, we did not find significant heterogeneity and horizontal pleiotropy due to all P > .05. For Benjamini-Hochberg corrected method, it is worth noting that none of MR results in our study adapt to a significant level owing to false negatives.[33] In addition, the leave-one-out test demonstrated that excluded any SNP had no impact on our MR estimations.

Table 2 - MR estimates for causal relationships between gut microbiome and DSDs (P −6). Outcome Gut microbiota p_FDR Correct causal direction Steiger pval P heterogeneity P pleiotropy Chronic gastritis Order Erysipelotrichales 0.972 TRUE 4.12E–17 0.858 0.919 Family Erysipelotrichaceae 0.972 TRUE 4.12E–17 0.858 0.919 Class Erysipelotrichia 0.972 TRUE 4.12E–17 0.858 0.919 Genus Victivallis 0.972 TRUE 4.96E–12 0.807 NA Phylum Bacteroidetes 0.972 TRUE 3.25E–18 0.395 0.478 Order Bacteroidales 0.972 TRUE 1.13E–17 0.397 0.478 Class Bacteroidia 0.972 TRUE 1.13E–17 0.397 0.478 Colorectal cancer Genus RuminococcaceaeUCG005 0.775 TRUE 8.14E–19 0.376 NA Family Victivallaceae 0.775 TRUE 1.47E–23 0.332 0.305 Genus Ruminococcusgauvreauiigroup 0.775 TRUE 1.12E–16 0.365 0.448 Crohn’s disease Family Oxalobacteraceae 0.283 TRUE 1.43E–18 8.66E-01 0.725 Phylum Cyanobacteria 0.283 TRUE 4.81E–09 9.68E-01 NA Genus Erysipelatoclostridium 0.283 TRUE 9.22E–13 7.48E-01 0.829 Genus Tyzzerella3 0.283 TRUE 5.26E–11 8.68E-01 NA Order NB1n 0.283 TRUE 1.67E–13 2.95E-01 0.469 Family Ruminococcaceae 0.283 TRUE 3.77E–07 3.66E-01 NA Genus FamilyXIIIAD3011group 0.283 TRUE 9.96E–14 5.45E-01 0.479 Genus Oxalobacter 0.283 TRUE 2.66E–20 5.53E-01 0.319 Genus Paraprevotella 0.283 TRUE 9.53E–09 5.33E-01 NA Gastric cancer Genus LachnospiraceaeUCG010 0.150 TRUE 1.06E–10 0.525 NA Genus Ruminococcusgnavusgroup 0.150 TRUE 3.10E–12 0.450 NA Genus RuminococcaceaeUCG002 0.168 TRUE 4.85E–21 0.347 0.478 Family FamilyXI 0.245 TRUE 8.11E–12 0.405 NA Family Ruminococcaceae 0.260 TRUE 6.96E–12 0.546 NA Gastric ulcer Genus Holdemania 0.581 TRUE 5.94E–17 0.897 0.729 Order Erysipelotrichales 0.581 TRUE 1.54E–15 0.186 0.454 Family Erysipelotrichaceae 0.581 TRUE 1.54E–15 0.186 0.454 Class Erysipelotrichia 0.581 TRUE 1.54E–15 0.186 0.454 Irritable bowel syndrome Class Melainabacteri 0.060 TRUE 2.37E–12 0.859 NA Order Gastranaerophilales 0.060 TRUE 2.49E–12 0.844 NA Phylum Proteobacteria 0.386