Causal association between EBV and gastritis, gastric cancer, and gastric ulcer: a two-sample Mendelian randomization study.

Introduction
Epstein–Barr virus (EBV), a ubiquitous member of the human herpesvirus family, has long been recognized for its role in various lymphoid and epithelial malignancies1. Specifically, EBV’s oncogenic potential is attributed to its ability to immortalize B cells and dysregulate cellular pathways, leading to abnormal proliferation and transformation1. In the context of gastrointestinal health, mounting evidence implicates EBV in the pathogenesis of stomach diseases, beyond its traditional associations with nasopharyngeal carcinoma and Burkitt’s lymphoma.
Gastritis, characterized by inflammation of the gastric mucosa, is a common condition with diverse etiologies, including Helicobacter pylori infection and autoimmune mechanisms2. Of note, recent epidemiological studies have hinted at an intriguing link between EBV seropositivity and certain forms of gastritis, suggesting a potential role for the virus in the inflammatory cascade3.
Gastric ulcer, another significant gastric disorder, arises from an imbalance between aggressive factors (such as acid and pepsin secretion) and mucosal defense mechanisms4. Although the primary cause is often attributed to H. pylori infection, the potential contribution of EBV in certain cases of gastric ulceration, particularly in immunocompromised individuals, remains an area of active investigation5.
Stomach cancer, also known as gastric cancer (GC), is a malignant carcinoma that can affect any part of the stomach. GC commonly affects older people, and the lifetime risk of developing stomach cancer is higher in men than in women6. GC is relatively rare when compared to other types of cancer. Also, early-stage detection of GC is challenging because it often produces no symptoms7.
So far, many risk factors related to GC have been identified8. Several studies suggested that people with a history of ulcers have a 3 times higher risk of developing GC than those without a history of ulcers9–11. Also, 80% of GC seems to be related to Helicobacter pylori (H. pylori) gastritis12,13. This type of lesion is commonly atrophic and exhibits intestinal metaplasia14. Besides H. pylori, Epstein–Barr virus (EBV), also known as human herpesvirus 4, accounts for 10% of all gastric cancers15,16. When the local or systemic immune function is compromised, EBV-positive cells proliferate, allowing the virus to infect gastric epithelial cells17. The expression of EBV genes causes epithelial cells to acquire proliferative properties and resist apoptosis17. In particular, the molecular biology of EBV-associated GC is characterized by frequent and extensive methylation of the promoter regions of tumor cell genes18. Nevertheless, little is known about the causality, frequency, and genetic diversity of EBV in GC or its risk factors.
MR is an epidemiological method that can analyze the potential causal association between exposure and outcome19. MR can minimize conventional confounding and reverse causation because genetic variation is randomly distributed and is independent of the environment, meiosis, disease onset, and progression20. Thus far, MR has been increasingly applied to examine causal inference during GC, reflecting the increasing availability of large datasets such as the UK Biobank and multiple GWASs for potential risk factors21,22. For example, Yang et al.23 assessed a potential causative link between inflammatory biomarkers and GC risk via a two-sample MR approach. They found that interleukin 6 receptor potentially mitigates the pathogenesis of gastric cancer (GC), while fatty acid-binding protein 4 may contribute to the same. However, the relationship between EBV infection and GC using the MR method has not yet been investigated.
This study aimed to evaluate the causality and evolutionary mechanism of EBV infection and GC, as well as ulcer and gastritis (Fig. 1).

Results
Relevant IVs were selected after assessing the causal association between EBV infection (as exposure factors) and GC, ulcer, or gastritis (outcomes). A total of 93 IVs related to EBV were identified and screened (mean F statistic of 35.5, with a minimum value of 20.8 and a maximum value of 342.3) (see Table 1). Detailed information regarding IV screening and F values is shown in Table S3.

The main findings of the study are that the genetic prediction results demonstrated a statistically significant association between EBV EBNA-1 antibody levels and GC [OR (95% CI): 1.231 (1.0385–1.4592), p = 0.01, see Fig. 2; Table 1] and gastritis [OR (95% CI): 0.9021 (0.8154–0.998), p = 0.04, see Fig. 3; Table 1]. The robustness of these results was supported by the MR Egger, WM, and Weighted mode methods (all P < 0.05, Table 1). No associations between EBV EBNA-1 antibody levels and gastric ulcer were found [OR (95% CI): 0.994 (0.9171–1.0773), p = 0.88, see Table 1]. Genetic prediction results suggested no association between other factors and GC, gastritis, or gastric ulcer (all P > 0.05, Figure S1-S7, Table 1).

Next, sensitivity analyses were performed to validate the MR assumptions. MR-Egger regression indicated that horizontal pleiotropy did not influence the analysis results (Table 2). In addition, in the presence of potential heterogeneity or uncertainty in IVs validity, the causal association between EBNA-1 and GC or gastritis remained significant (Table 2). Cochran’s test revealed the presence of heterogeneity when assessing anti-EBV IgG seropositivity and peptic ulcer, as well as EBV virus EBNA-1 antibody levels and EBV VCA p18 antibody levels and gastritis (Table 2). On the other hand, although the MR-PRESSO analysis (Table 3) detected no outliers for EBV EBNA-1 antibody levels and GC, it detected two outliers for EBV EBNA-1 antibody levels and gastritis (after exclusion of the corresponding SNPs, the lack of association persisted between these variables). Interestingly, when rs117705227 and rs6927022 were excluded, there were no associations between EBV EBNA-1 antibody levels and gastritis [OR (95% CI): 1.0747 (0.8531–1.3538), p = 0.75, see Figure S6, Table 1]. The MR-PRESSO analysis suggested the presence of one outlier when assessing anti-EBV IgG seropositivity and peptic ulcer. Upon removal of the corresponding SNP, no associations remained.

Discussion
This study used a two-sample Mendelian randomization (MR) analysis to investigate the causal relationship between EBV and GC and conditions that may lead to GC, such as gastritis and gastric ulcer. The present MR analysis provides evidence supporting a causal role for genetically predicted EBNA-1 antibody levels in the etiology of GC. In addition, the findings suggest a potential association between EBNA-1 antibody levels and gastritis, although this association was attenuated after accounting for potential outlier SNPs and requires further validation.
Epstein–Barr nuclear antigen 1 (EBNA1) was the first EBV protein detected. EBNA1 is expressed in both lytic and latent modes of infection24. Observational studies have reported that EBNA1 is the only EBV protein expressed in all EBV-positive tumors and latency types in proliferating cells and is sometimes the only EBV protein expressed25. For example, EBNA1 is consistently expressed in all EBV-associated gastric carcinomas26. Sivachandran et al.27 concluded that BNA1 protein increases the survival of GC cells, at least in part, due to its ability to induce the loss of promyelocytic leukemia nuclear bodies, which in turn results in impairment of p53 function and apoptosis. Nevertheless, some of these studies failed to control for confounding factors, and none established a definite causal relationship between EBV-EBNA1 and GC. The EBV genome has been thoroughly investigated to determine if its variability is due to ethnic or geographic correlates or a specific disease28. EBV types were classified based on variations in a few genes, such as EBNA1, EBNA2, and EBNA3, corresponding to two EBV strains28. Variations of the EBV EBNA1 gene have been found in GC and nasopharyngeal carcinomas from Northern China29. Three major patterns of the EBNA1 variations, V-val, P-thrV, and V-leuV, were observed in GC patients, and V-val was the most common subtype in all three groups, followed by P-thrV and V-leuV29. Moreover, EBNA-1 sequence variation in Danish and Chinese EBV-associated tumors: evidence for geographical polymorphism but not for tumor-specific subtype restriction30. To the best of our knowledge, this is the first study that reported a causal effect of genetic variations in the EBNA1 gene on GC, thus shedding new light on the intricate complexity of this condition.
Some studies have suggested that certain forms of chronic gastritis, such as H. pylori gastritis and atrophic gastritis of the autoimmune type, are risk factors for GC. Gastritis may lead to stomach ulcers and bleeding. Also, the risk of developing GC increases if a patient has extensive thinning of the stomach lining and changes in the lining cells31,32. This study also found a statistically significant association between EBV EBNA-1 antibody levels and gastritis. These data were further suggested by MR Egger, WM, and the Weighted mode method. MR-Egger regression further indicated that horizontal pleiotropy does not influence the analysis results. In addition, in the presence of potential heterogeneity or uncertainty in IVs validity, the causal association between EBNA-1 and gastritis remained significant. Nevertheless, MR-PRESSO detected two outliers for EBV EBNA-1 antibody levels and gastritis. After excluding the corresponding SNPs, the lack of association persisted between these variables. Still, this data may indicate a potential correlation between EBNA-1 and gastritis, so larger or more diverse datasets are needed to confirm this data. The EBNA-1-gastritis association becoming non-significant after outlier removal could be due to limited power since the F-values of all IVs were > 10, indicating no weak instrumental bias. At the same time, it could also be because the association was driven by a single outlier, and no actual association is present. Nevertheless, these outlier SNPs may have pleiotropic effects, meaning that they might influence the risk of gastritis through pathways independent of EBNA-1 antibody levels. Removing them aimed to obtain a more robust causal estimate, but it may also reduce statistical power. Therefore, the evidence for a causal relationship between EBNA-1 antibody levels and gastritis is currently suggestive but inconclusive, and larger-scale or different population datasets are needed for validation.
This study found no association between other factors and GC, gastritis, or gastric ulcer. Nevertheless, Cochran’s test revealed the presence of heterogeneity when assessing anti-EBV IgG seropositivity and peptic ulcer, as well as EBV virus EBNA-1 antibody levels and EBV VCA p18 antibody levels and gastritis. EBV seropositivity was defined as seropositivity to at least one of the three antibodies: EBNA-1 IgG, VCA IgG, or VCA IgM33. A previous study found lower anti-EBNA antibody titers and higher anti-VCA-IgG antibody titers in patients with systemic chronic active EBV disease than those in healthy controls34. Moreover, Wang et al.35 assessed the role of serum EBV-VCA IgG in assessing GC risk and prognosis, finding that it has the potential to predict the risk of GC and its precursor as well as the prognosis of histologically classified GC. Although this study found no genetic association between anti-EBV IgG and stomach conditions, including gastritis, gastric ulcer, and GC, our data may imply that larger or more diverse datasets are needed to confirm this data.
The strengths of the present study include using a rigorous MR design, multiple sensitivity analyses, and large-scale GWAS data. Nevertheless, this study also has a few limitations. For example, the potential for residual confounding (e.g., H. pylori infection) and the reliance on predominantly European ancestry populations limit generalizability. Nevertheless, MR analyzes the causal associations at the genetic level, minimizing the influence of confounders. Although not all potential confounders can be completely ruled out, genetic variants associated with H. pylori infection status (if independent of the genetic variants for EBV) would theoretically not confound the association between EBV and gastric diseases. The MR-Egger regression analysis did not show significant directional pleiotropy, which, to some extent, alleviates the concerns about systematic bias caused by pleiotropy, although it cannot be completely ruled out. Regarding temporality, most GWAS datasets used for MR analyses do not include a temporal component. The individuals are classified as having or not having the exposure or outcome, without data on the timing between the exposure and the outcome. MR studies typically assess the lifelong impact of genetic predisposition to exposure (e.g., EBV antibody levels) on the risk of outcomes (gastric diseases). It cannot distinguish the specific timing of infection (childhood vs. adulthood), whether the virus is in a latent or active phase, or the impact of changes in viral load/antibody levels over time on disease development. Although the temporal dynamics of the relationship between exposure and outcome are important, the lack of temporality is a limitation common to all MR studies. Future research could benefit from more diverse ethnic samples, functional validation of implicated genetic variants, and exploration of potential mediating pathways. Subsequent studies and clinical trials are warranted to elucidate the relationship between EBV infection and gastric alterations.
Using the MR approach, this study provides genetic evidence that EBNA-1 antibody levels may play a causal role in the development of GC, which goes beyond traditional observational associations. The results suggest that strategies targeting EBV (especially EBNA-1-related pathways), such as vaccine development and targeted therapies, may be valuable in GC prevention or treatment and warrant further exploration. Although the association between EBV and gastritis is not definitive, it highlights the need for further research on the role of EBV in gastric mucosal inflammation. The clinical implications could be that EBV infection should be included as an additional factor in the decision to screen for gastric cancer. Patients with a history of EBV infection and gastric signs and symptoms could be prioritized for diagnostic procedures. Although interesting, the present results are primarily of research significance and require further validation before they can be directly applied to clinical decision-making. Functional studies are needed to elucidate the specific mechanisms by which the relevant genetic variants and EBV contribute to the development of gastric diseases.

Methods
Data regarding GC were sourced from a FINN cohort (see Table S1). Data regarding EBV were obtained from a previous study36, which used a UK biobank cohort of up to 10,000 serological measurements of infectious diseases and genome-wide genotyping, and Open Forum Infectious Diseases; more information can be found in Table S2. Our analyses did not require approval from the ethics committee.
The analysis adhered to the following three assumptions of MR studies (Fig. 1)37: it was associated with the exposure; (2) it was not associated with the outcomes due to confounding pathways; (3) it did not affect the outcome. The following procedure was applied: (1) single-nucleotide polymorphisms (SNPs) significantly associated with EBV were selected based on P < 5 * 10− 6; (2) SNPs with a minimum minor allele frequency (MAF) > 0.01 were screened; (3) R2 < 0.001 and window size = 10,000 kb were used for linkage disequilibrium (LD) between SNPs; (4) when the selected IVs were not present in the summary data of the outcome, we searched for SNPs with high LD (R2 > 0.8) with the IVs as proxy SNPs to replace the existing ones; (5) the F-value for each SNP in the IV was calculated to assess IV strength, excluding potential weak instrument bias between the IV and exposure factors, using the following formula: F = R2
* (N-2) / (1-R2), where R2 represented the proportion of exposure variance explained by the SNP in the IV. The requirement for the F-value was >10.
Inverse variance weighting (IVW) was primarily used to estimate the causal relationship between the exposure and outcome by calculating the odds ratio (OR) and 95% confidence interval (CI)38. WM, MR-Egger, and the weighted model were further used to validate the results. All analyses were performed using R 4.1.2 software, incorporating packages such as TwoSampleMR and MR-PRESSO. Heterogeneity was tested using Cochran’s Q, and pleiotropy was examined through MR-Egger regression of intercept values. Scatter plots and sensitivity analysis plots were used for visualization39. As there were four outcome factors in this study, the false discovery rate (FDR) correction method was applied to correct for multiple testing, with PFDR < 0.05 representing statistical significance.

Electronic supplementary material
Below is the link to the electronic supplementary material.