Application of Mendelian randomization to assess host gene–gut microbiota correlations in patients with esophageal cancer

Background Increasing evidence suggests that esophageal cancer (ESCA) may be correlated with gut flora. However, their causal connection remains unclear. This study aimed to evaluate potential causal linkages and gene–gut microbiome associations between the gut microbiota and ESCA using Mendelian randomization (MR). Methods We analyzed the data using genome-wide association studies. The exposure factor and outcome variable were the gut microbiota and ESCA, respectively. The MR-Egger method, weighted median, inverse-variance weighted method, heterogeneity test, sensitivity analysis, and multiplicity analysis were used for the MR analysis. And it was validated using an external dataset. Further meta-analysis was performed to validate the robustness of this relationship. Finally, we annotated single nucleotide polymorphisms in the gut microbiota that were causally associated with ESCA to explore possible host gene-gut microbiota correlations in patients with ESCA. Results We identified four species with potential associations with ESCA. Three of these species had a negative causal relationship with ESCA (odds ratio (OR): 0.961; 95% confidence interval (CI): 0.923–0.971; p = 0.047 for Romboutsia; OR: 0.972; 95% CI: 0.921–0.961; p = 0.018 for Lachnospira; OR: 0.948; 95% CI: 0.912–0.970; p = 0.032 for Eubacterium). A positive causal relationship was observed between one bacterial group and ESCA (OR: 1.105; 95% CI: 1.010–1.072; p = 0.018 for Veillonella). External datasets show the same trend. This is further supported by meta-analysis. None of the data showed pleiotropy, and leave-one-out analysis indicated the reliability of these findings. The gut microbiomes of patients with ESCA may correlate with the 19 identified genes. Conclusion Our data indicate a potential causal link between these four gut bacteria and ESCA and identify a correlation between host genes and gut microbiota in ESCA, offering novel therapeutic options.


Background
Esophageal cancer (ESCA) is one of the most common cancers globally, with the sixth highest mortality rate according to global cancer data (Sung et al., 2021).Surgery is an effective treatment for ESCA, but in advanced ESCA, the 5-year survival rate of patients remains less than 25% even after surgery (Oppedijk et al., 2014).Chemotherapy is commonly used as a treatment for ESCA but has unavoidable side effects such as toxicity and drug resistance (He et al., 2021).Additionally, epidemiological data indicates that the incidence of ESCA is increasing annually, gravely endangering human health (Uhlenhopp et al., 2020).Therefore, identifying factors potentially associated with ESCA can provide an essential basis for the early prevention of ESCA.
Increasing evidence has shown that the gut microbiota and ESCA are closely related (Cheung et al., 2022;Muszynski et al., 2022;Ohkusa et al., 2022;Baba et al., 2023;Sugimoto et al., 2023).Further, significant variations have been reported in the composition and abundance of fecal microorganisms between patients with ESCA and healthy controls.Notably, these differences are closely correlated with the severity of the disease, suggesting that the gut microbiota may play a significant role in the development of ESCA (Li et al., 2022;Lin et al., 2022).Moreover, gut microbiota can alter genome-wide methylation levels in ESCA, which may be one of the mechanisms influencing the malignant behavior of ESCA cells (Baba et al., 2023).Exposure to antibiotics leads to changes in the gut microbiota, which in turn increase the risk of developing ESCA; the risk of developing the disease increases with the duration of antibiotic exposure (Thanawala et al., 2023).Therefore, investigating the potential link between the gut microbiota and disease to prevent and treat ESCA is crucial.
Although current research reveals that the gut microbiota and ESCA are related, the results are susceptible to confounding factors.Mendelian randomization (MR) is a genetic technique frequently used to investigate causal links between exposures and outcomes and prevents confounding variables in common observational studies because genetic variants are randomly assigned at conception (Smith and Ebrahim, 2003).
We investigated the potential causative relationship between gut microbiota and ESCA using MR to provide a proper theoretical foundation for understanding the interaction between ESCA and gut microbiota.We further identified genes related to the single nucleotide polymorphisms (SNPs) in the gut microbiota obtained by MR analysis.Our results may help identify novel therapeutic options for ESCA treatment.

Research methods
In this study, heterogeneity, sensitivity, and multiplicity analyses were conducted in addition to MR analyses using genome-wide association studies (GWAS) information to evaluate the causal relationship between gut microbiota and ESCA.MR studies must satisfy three core assumptions of association, independence, and exclusivity: (i) the selected SNPs should be significantly associated with the exposure (intestinal microbiota); (ii) the SNPs must be independent of potential confounders between the exposure and the outcome; and (iii) there is no direct relationship between the SNPs and the outcome (ESCA), and the causal linkage can only be made through the intestinal flora.The workflow is illustrated in Figure 1.

Data sources
Gut microbiota data from the most recent GWAS meta-analysis, comprising 24 cohorts and 18,340 participants, were used in this investigation (Kurilshikov et al., 2021).ESCA data for the experimental and validation groups were obtained from the UK Biospecimen Repository, 1 the experimental group included 372,756 samples and the validation group included 476,306 samples (Table 1).

Instrument selection
A total of 196 bacterial traits, including 9 phyla, 16 classes, 20 orders, 32 families, and 119 genera, were retained after initially removing 15 bacterial qualities without specific names.SNPs significantly related to the gut microbiota were chosen at the genome-wide level (p < 1.0 × 10 −5 , R 2 < 0.001, and clumping distance = 10,000 kb) to fulfill the first MR hypothesis that SNPs must be strongly associated with the gut microbiota.Second, to ensure that the genetic variants were not associated with potential confounders (smoking, heavy alcohol consumption, hot beverages, pickles), a query was performed in the Phenoscanner database. 2This step was performed to ensure that the SNPs were not associated with known confounders and, ultimately, to obtain SNPs significantly associated with the gut microbiota, serving as instrumental variables.Next, we calculated the proportion of variance (R 2 ).We calculated the F-statistic using the following formula: R = where "MAF" is the minor allele frequency, "N" denotes the exposed GWAS sample size, and "K" is the number of SNPs.F > 10 confirmed the absence of a mild instrumental variable bias.The process was completed by annotating the SNPs using an internet database. 3

Statistical analysis
The weighted median approach, MR-Egger method, and random effects inverse variance weighting (IVW) method were used in the MR analysis.Statistical significance was set at p < 0.05.The IVW approach was primarily employed to analyze these studies (Burgess and Thompson, 2017), and methods such as MR-Egger and weighted median were used to complement the IVW method (Verbanck et al., 2018).Additionally, we applied Cochran's Q method to evaluate heterogeneity among SNPs.We performed several heterogeneity tests, including the MR-Egger intercept test and a sensitivity analysis, to ensure the robustness of our results.The leave-one-out test was used for the sensitivity analysis to determine outliers among the final SNPs.All data were analyzed using the R packages "Two-Sample-MR" and "MR-PRESSO" in R software (version 4.3.0).

Main results of the 196 bacterial traits with the risk of ESCA
The F-statistics for the 196 bacterial characteristics, ranging from 21.63 to 144.84 and with mean values exceeding 10, suggested a robust connection with exposure.We screened for SNPs strongly associated with the gut microbiota (p < 1.0 × 10 −5 ), and the linkage disequilibrium parameter was set (R 2 < 0.001, kb = 10,000).Thereafter, the data of the final ESCA were extracted

Sensitivity analyses
Weighted median

MR-Egger regression Excluded potential pleiotropic SNPs
MR pleiotropy residual sum and out lier test

Validation of external datasets
Sample size:47,6306 The workflow of the present study.from GWAS, and 196 intestinal flora were merged with the final ESCA sequentially.SNPs with palindromic sequences were excluded from the analysis.Finally, we obtained four gut microbiota samples with potential associations with ESCA using the IVW method (Figure 2).Using the IVW method, we found that Romboutsia was negatively associated with ESCA (odds ratio (OR): 0.961; p = 0.047).After screening for F-statistics and excluding chain imbalances, 13 SNPs related to ESCA were included (Supplementary Table S1).In the weighted median approach, the results for the association between Romboutsia and ESCA remained stable (p = 0.02).To evaluate the stability of these findings, we performed the MR-Egger test on the loci of the included SNPs.No possible horizontal pleiotropy was found (p = 0.96), indicating that instrumental variables did not significantly alter the outcomes through mechanisms other than exposure.Cochran's Q test results showed no significant heterogeneity among the SNPs (p = 0.45).
Similarly, Lachnospira was negatively associated with ESCA (OR = 0.972; p = 0.018).After screening, six SNPs associated with ESCA were included (Supplementary Table S2).In the weighted median approach, the genus Lachnospira was weakly associated with ESCA (p = 0.08).The SNPs were then subjected to an MR-Egger test, which revealed no apparent level of multiple effects (p = 0.96).According to Cochran's Q analysis (p = 0.45), no discernible heterogeneity was observed among the selected SNPs.Additionally, using the IVW method, we discovered that the genus Eubacterium was adversely linked with ESCA (OR: 0.948; p = 0.032).Overall, five SNPs associated with ESCA were included (Supplementary Table S3).Again, no pleiotropy (p = 0.60) or heterogeneity (p = 0.79) was observed.In contrast, using the IVW method, we found that Veillonella was positively associated with ESCA (OR: 1.105; p = 0.018).The seven SNPs associated with ESCA identified after screening (Supplementary Table S4) were free of pleiotropy (p = 0.87) and heterogeneity (p = 0.41).
We conducted a "leave-one-out" sensitivity analysis to confirm the impact of each SNP on the overall causation.The findings revealed that none of the SNPs exhibited significant differences when one SNP was excluded (Supplementary Figure S1).Finally, the results were plotted using a pattern map (Figure 3).Using an online database, we identified 19 genes that may be associated with the gut microbiota of patients with ESCA (Table 2).
We brought the above four intestinal flora into the external dataset for further analysis, and by plotting scatter plots for the training group (Figure 4A) and the validation group (Figure 4B), we found that the four intestinal flora in the validation group had the same tendency to have the same effect on ESCA when compared with the experimental group, which further validated our results.Meanwhile, we further counted the heterozygosity and pleiotropy of the training and validation group analyses (Figure 5), which showed that none of them had pleiotropy, indicating the credibility of our results.

Meta analysis
We further meta-analyzed and plotted and mapped the correlation data of the training and validation groups with the four intestinal flora in a forest plot (Figures 6A-D), which showed that the above relationship was robust.For ease of analysis, we summarized the results into a three-line table (Figure 6E), which showed less statistical heterogeneity and statistically significant results, which further justified our conclusions above.

Discussion
The esophagus in the gastrointestinal tract is colonized by various microorganisms and is not sterile (Laserna-Mendieta et al., 2021).Healthy individuals have a relatively stable pH (approximately 7) that provides a stable environment for microbial survival (Hasan et al., 2021).The gut microbiota can directly or indirectly affect human health and disease and is considered a new "organ" (Baquero and Nombela, 2012).Gut microbes significantly impact cell formation, differentiation, metabolism, and growth.Dysbiotic gut microbiota may contribute to the body's carcinogenic process (Zhou et al., 2021;Gou et al., 2023;Guevara-Ramirez et al., 2023).Notably, the gut microbiota may be involved in the development of ESCA (Lv et al., 2019;Zhou et al., 2021).For example, human papillomavirus (HPV) and alterations in intestinal bacteria may cause ESCA (Meng et al., 2018).Moreover, in regions with a high incidence of ESCA, a high prevalence of HPV has often been reported (Yano et al., 2021) individuals and observed that patients with ESCA have higher levels of Firmicutes and Actinobacteria and lower levels of Bacteroidetes (Deng et al., 2021).This suggests that the gut microbiota and development of ESCA are closely related.In addition, changes in the gut microbiota can increase the levels of pro-inflammatory cytokines and immune cells, thereby inducing tumorigenesis (Proano-Vasco et al., 2021).Notably, gut microbiota can induce the overexpression of nitric oxide synthase, potentially leading to ESCA (Gillespie et al., 2021).The gut microbiota can also interact with the host by secreting bioactive substances, such as vitamins (Malesza et al., 2021), that can be beneficial or harmful for the organism (Riwes and Reddy, 2018).These explanations help to clarify how the gut microbiota and ESCA are related.
In this study, MR analysis was used to assess the causal relationship between gut microbiota and ESCA for providing a theoretical foundation for the interactions between ESCA and gut microbiota, given the lack of conclusive evidence to support the potential relationship between gut microbiota and ESCA.The results of this study showed that four intestinal microbiota were potentially associated with ESCA.Specifically, the genera Romboutsia, Lachnospira, and Eubacterium were negatively associated with ESCA, whereas Veillonella had a positive causal relationship with ESCA.
The genus Romboutsia is a group of gram-positive bacteria first proposed in 2014 by Ricaboni et al. (2016) from the right half of the human colon using colonoscopy.Most studies concluded that Romboutsia belongs to the natural gut microbial community and plays an essential role in host health.For example, Lactobacillus acidophilus ameliorates colitis by increasing the abundance of Romboutsia (Han et al., 2023).Several studies have found that Romboutsia is significantly less abundant in patients with inflammatory bowel disease (e.g., Crohn's disease, ulcerative colitis) (Qiu et al., 2020;Wang et al., 2023), and is associated with hepatocellular liver cancer and postherpetic neuralgia (Feng et al., 2023;Jiao et al., 2023).Consequently, Romboutsia may influence immune regulation and intestinal health (Liu et al., 2023).At the same time Romboutsia is able to increase the antioxidant capacity of the body (Zhang et al., 2023), attenuates intestinal inflammatory damage and inhibits endoplasmic reticulum stress, which may be the mechanism by which it affects human health (Li et al., 2023).Moreover, Romboutsia is closely associated with esophageal epithelial atrophy (Pan et al., 2021), an essential stage in the development of ESCA.The potential relationship between Romboutsia and ESCA identified in this study is consistent with the above findings, which suggests that our analysis is logical.
Lachnospira is integral to the gut microbiota and colonizes the intestinal lumen from birth (Vacca et al., 2020).It is a group of potentially beneficial bacteria involved in the metabolism of various carbohydrates.Fermentation produces acetic and butyric acids, which provide energy to the host (Devillard et al., 2007;Wong et al., 2014).Lachnospira has been implicated in various diseases, including obesity (Natividad et al., 2018), liver disease (De Minicis et al., 2014), and chronic kidney disease (Yasuno et al., 2023), and can also lead to depression via the gut-brain axis (Bonaz et al., 2018).The Lachnospira spp.have anti-inflammatory and antioxidant effects and can protect the intestinal mucosal barrier by inhibiting inflammatory responses and scavenging free radicals, which may be a mechanism for preventing esophageal diseases (Mukherjee et al., 2020).A cohort study showed that Lachnospira, which has anti-inflammatory properties, was

Esophageal Cancer
Patterns of the relationship between four gut microbiota and ESCA.
significantly reduced in ESCA patients compared to the normal group (Cheung et al., 2022).This is consistent with our findings in the current study.
Eubacterium has been found to be associated with age-related macular degeneration (AMD) (Mao et al., 2023), female infertility (Xi et al., 2023), multiple sclerosis (MS) (Vacaras et al., 2023) and other diseases.Although a previous Mendelian analysis done by Yang et al. showed that Eubacterium reduced the risk of Barrett's esophagus (Yang et al., 2022), the literature exploring the relationship between Eubacterium spp.and ESCA remains scarce to date.This is the innovative finding of our present study, and it also suggests an interesting research direction for us.Now, our experiments on Eubacterium for the prevention of ESCA are in progress, and the results will be published in a follow-up study.
The genus Veillonella includes gram-negative, anaerobic, non-motile, and non-spore-forming coccus bacteria (Djais et al., 2019).Veillonella is strongly associated with the development of several diseases, and has been found to promote the proliferation of lung adenocarcinoma (Zeng et al., 2023), whilst Veillonella activates macrophages to promote inflammatory responses via the LPS-TLR4 pathway (Zhan et al., 2022), for example, Veillonella correlated with the severity of radiation esophagitis (Lin et al., 2022), and inflammation is one of the factors leading to esophageal cancer, suggesting that Veillonella may indirectly contribute to esophageal carcinogenesis through inflammation.In our study, only the genus Veillonella positively correlated with ESCA.An extensive body of literature describes the pathogenic role of Veillonella in the esophagus.For example, a study using whole genome sequencing (wGS) and RNA sequencing (rNAseq) of tumors from 61 patients with ESCA found a high abundance of Veillonella in ESCA (Nomburg et al., 2022).It has also been shown that Veillonella levels gradually increase during the development of esophageal reflux (GR)-Barrett's esophagus (BE)-esophageal adenocarcinoma (EA) (Di Pilato et al., 2016;Hao et al., 2022).Gram-negative bacteria are mainly associated with esophageal abnormalities, e.g., Veillonella is mostly associated with BE (Lv et al., 2019;Park and Lee, 2020), which aligns with our findings.These findings have important implications for research in ESCA in the future and may inspire new prevention and treatment strategies for this disease.Many previous studies have demonstrated a potential relationship between the gut microbiota and ESCA (Baba et al., 2023;Co et al., 2023;Sugimoto et al., 2023).However, most of these studies were observational and susceptible to confounding   Gut flora plays a dual role in cancer development, and the use of gut flora in conjunction with traditional antitumor treatment strategies, as well as the use of probiotics, FMT, and dietary control, can improve the efficacy of anticancer treatments, while reducing the incidence of side effects and improving prognosis (Sun et al., 2023).However, this study had certain limitations.First, most of the participants were European.Additional research is required to determine whether these conclusions apply to other ethnic groups.Second, the flora in this study was limited to the genus level and was not further subdivided.

Conclusion
We analyzed the potential relationship between 196 common intestinal microbiota and ESCA.We found that the genera Romboutsia, Lachnospira, Eubacterium, and Veillonella may be causally associated FIGURE 4 Scatterplot of correlation between training and validation groups.(A) Scatterplot of the causal relationship between the four intestinal flora and ESCA in the training group.(B) Scatterplot of the causal relationship between the four intestinal flora and ESCA in the validation group.
factors.In contrast, the present study used a genetic epidemiological approach to minimize the influence of confounding factors and provide a compelling insight into the relationship between gut microbiota and ESCA.And the use of external datasets to validate trends and Meta-analysis of the results ensures maximum stability of the results.
FIGURE 6Meta-analysis of the causal association between host gene-gut microbiota and ESCA.(A-D) Meta-analysis of the causal relationship between four intestinal flora and ESCA.(E) Meta-analysis results.

TABLE 1
Details of the genome-wide association studies and datasets used in our analyses.
. Deng et al. studied 23 patients with ESCA and 23 healthy Forest plot of the associations between genetically determined 4 bacterial traits with the ESCA.

TABLE 2
SNP annotation of intestinal flora IVs.
Heterogeneity and pleiotropy results of training and validation group analyses.