Causal association between Epstein-Barr virus infection and malignant gynecologic and breast tumors: a two-sample Mendelian randomization study.

Background
Breast cancer is a malignant tumor that develops in breast tissue and represents the most frequently diagnosed cancer among women globally [1]. Contributing risk factors for breast cancer include genetic predisposition, advancing age, reproductive and hormonal history, lifestyle choices, certain medical conditions, and exposure to radiation [2]. Cervical cancer arises from the transformation zone of the cervix, most often affecting squamous cells [3]. It ranks as the second most prevalent cancer and is the third leading cause of cancer-related death among women worldwide [1, 3]. Risk factors for cervical cancer include having multiple sexual partners, engaging in sexual activity before age 18, having multiple childbirths, giving birth before age 19, current smoking, being of Black or Hispanic ethnicity, using oral contraceptives for more than five years, a history of sexually transmitted infections, immunosuppression, low socioeconomic status, and a family history of cervical cancer [4]. Endometrial cancer originates from the lining of the uterus, known as the endometrium [5]. Factors that increase the risk of developing endometrial carcinoma include prolonged exposure to estrogen, a greater number of years menstruating, polycystic ovary syndrome, obesity, not having given birth or a history of infertility, a family history of the disease, long-term tamoxifen use, and Lynch syndrome [6]. Ovarian cancer is among the most fatal gynecologic cancers, with a lifetime risk estimated at 1 in 50 to 70 women [1, 7]. Risk factors for ovarian cancer include early onset of menstruation, late menopause, genetic predispositions such as BRCA1 or BRCA2 mutations, and a family history of ovarian cancer [7].
Epstein-Barr virus (EBV) is a ubiquitous herpesvirus with a reported worldwide prevalence of >90% [8]. The primary infection is asymptomatic in some persons but may cause infectious mononucleosis in others [8]. Following the primary infection, EBV establishes a lifelong persistent infection in human B cells [8]. T-cell-mediated immunity is particularly important for the control of latent infection, and immunocompromised patients are at particular risk for developing EBV-related disorders, including posttransplant lymphoproliferative disorder (PTLD), oral hairy leukoplakia, and AIDS-related lymphomas. Though EBV has intrinsic oncogenic potential, persistent infection is benign and well-controlled by the immune system in most healthy people. In other persons, persistent infection is associated with malignancies and other syndromes, such as Hodgkin lymphoma [9], Burkitt lymphoma [10], nasopharyngeal carcinoma [11], and gastric carcinoma [12].
A systematic review revealed that EBV infection was present in 28% of patients with breast cancer and 8% of patients without breast cancer, but that additional evidence was necessary [13]. Molecular evidence also supports the role of EBV in breast cancer pathogenesis [14]. EBV infection might also have causal relationships with cervical cancer [15] and ovarian cancer [16]. Still, the evidence for causality is poor, and additional studies are necessary. Notably, beyond breast and cervical tumors, case reports and small-scale clinical studies have also implicated EBV in gynecological malignancies. For instance, EBV DNA has been demonstrated in cervical lymphoma-like lesion [17], and viral coinfections involving EBV were reported in genital lesions with possible links to cervical carcinogenesis [18]. Moreover, a recent clinical investigation found EBV in advanced stages of ovarian cancer and suggested its possible role in gynecologic tumor progression [19]. Although these studies do not provide definitive causal evidence, they collectively indicate that EBV may participate in tumorigenesis within the female reproductive tract.
Genome-wide association studies (GWASs) have generated data on millions of single-nucleotide polymorphisms (SNPs) and their links to various phenotypes and diseases, greatly advancing our understanding of complex traits and conditions [20]. Mendelian randomization (MR) leverages the characteristics of common genetic variants related to different environmental exposures, enabling researchers to investigate potential causal relationships between exposures and diseases [21]. In two-sample MR, associations between SNPs and exposures, as well as between SNPs and outcomes, are obtained from separate GWASs and then combined to estimate causality. Provided that key assumptions are met, MR helps minimize issues of reverse causation and confounding, which frequently hinder or distort the interpretation of findings from traditional epidemiological studies [22].
Therefore, this MR study aimed to explore the causal association between genetically predicted EBV infection and breast, cervical, endometrial, and ovarian cancers.

Methods

Study design
This study used publicly available data from GWASs to investigate the causal association between EBV infection and breast and gynecological cancers (Fig. 1). Additional ethical approval was unnecessary since all data were obtained by studies approved by ethical committees and under the guidance of the Declaration of Helsinki. The three key MR assumptions are (1) the relevance assumption (i.e., the genetic variant(s) being used as an instrument for the exposure is associated with the exposure), (2) the independence assumption (i.e., there are no common causes of the genetic variant(s) and the outcome of interest), and (3) the no horizontal pleiotropy assumption (i.e., there is no independent pathway between the genetic variant(s) and the outcome other than through the exposure) [22]. Based on the three MR assumptions [22], this study assumed that the SNPs used as instrumental variables (IVs) for the exposure (EBV infection) are associated with the exposure, that there are no common causes to the SNPs and the outcome (cancers), and that there is no independent pathway between the SNPs and cancers other than through EBV infection. The study was reported according to the “STrengthening the Reporting of OBservational studies in Epidemiology using Mendelian Randomization (STROBE-MR)” checklist [23].

Data source
The GWAS data for breast, endometrial, cervical, and ovarian cancers were from a previous study [24]. The dataset for endometrial cancer was from 2188 cases and 237,839 controls (all of European ancestry) and encompassed 24,135,295 SNPs (ID ebi-a-GCST90018838). The dataset for cervical cancer was from 909 cases and 238,249 controls (all of European ancestry) and encompassed 24,138,337 SNPs (ID ebi-a-GCST90018817). The dataset for ovarian cancer was from 1588 cases and 244,932 controls (all of European ancestry) and encompassed 24,137,758 SNPs (ID ebi-a-GCST90018888). Finally, the dataset for breast cancer was from 17,389 cases and 257,730 controls (all of European ancestry) and encompassed 24,133,589 SNPs (ID ebi-a-GCST90018799).
The GWAS dataset for the exposure (EBV infection) was from a previous study [25]. The dataset contains 8735 individuals (all of European ancestry) and covers 9,170,312 SNPs (ID ebi-a-GCST90006897). In addition, because both the exposure (EBV GWAS) and outcome (cancer GWAS) samples were derived from European populations, complete independence cannot be guaranteed. While the GWAS datasets came from different consortia, some degree of sample overlap cannot be entirely excluded.

Instrumental variable selection
The IVs included in this study had to meet the following three criteria: (1) SNPs with genome-wide significant associations with EBV infection were initially screened at P < 5 × 10⁻⁸. However, because the EBV GWAS had a relatively small sample size (n = 8,735), very few SNPs met this threshold, which limited the feasibility of MR analyses. Therefore, following established practice in underpowered MR settings, the threshold was relaxed to P < 5 × 10⁻⁶. (2) SNPs with a minimum allele frequency (MAF) of >0.01, and (3) SNPs with linkage disequilibrium (LD) and SNPs with LD were removed based on R2 < 0.001 and window size of 10,000 kb. When a selected IV was not present in the summary data of tshe outcome, the SNP with the IV with a high LD (R2 >0.8) was sought as a surrogate SNP [26]. The F-value of each SNP in the IV was calculated to assess IV strength and to rule out possible weak instrumental variable bias between IV and exposure factors. The F-values were calculated as F = R2×(N-2)/(1-R2). F-values >10 were considered to indicate the absence of weak instrumental bias [27].

Mendelian randomization analysis
The primary MR analysis method used in this study to assess the causal association between exposure and outcome risks was the inverse variance weighted (IVW) method [28]. The results were presented as odds ratios (ORs) and 95% confidence intervals (CIs). Significant results with the IVW method were tested for robustness using the MR-Egger, weighted median, and weighted mode methods. All analyses were conducted using the “TwoSampleMR” package in R (The R Project for Statistical Computing, www.r-project.org).

Sensitivity analysis
Heterogeneity among IVs was detected using Cochran’s Q, with P >0.05 indicating low heterogeneity, meaning that the estimates among IVs were randomly distributed and had little impact on the IVW results. The third MD assumption is the absence of horizontal pleiotropy [22], which was detected using the MR-Egger regression method. When the intercept of the MR-Egger regression approached zero or was not statistically significant, it suggested the absence of pleiotropy. The MR pleiotropy residual sum and outlier (MR-PRESSO) method was used to detect potential outliers (i.e., SNPs with P < 0.05) and to re-estimate causal associations after their removal to correct for horizontal pleiotropy. Leave-one-out analysis was employed to assess the robustness and consistency of the results by detecting whether a single SNP was driving the results [29].

Results

Instrument variable selection
In this study, 25 IVs related to EBV infection were screened. The mean F-value was 22.84 (range: 20.84–31.40). All F-values were > 10, indicating no weak instrumental bias. There were no missing SNPs in the outcome summary data. All IVs are presented in Supplementary Table S1.

Mendelian randomization analysis results
The genetic prediction results, based on the IVW method, showed that there were no statistically significant causal associations between EBV infection and the risk of endometrial cancer (OR = 1.55, 95%CI: 0.71–3.37, P = 0.27), cervical cancer (OR = 0.96, 95%CI: 0.42–2.22, P = 0.93), ovarian cancer (OR = 1.55, 95%CI: 0.72–3.33, P = 0.26), and breast cancer (OR = 1.02, 95%CI: 0.79–1.33, P = 0.85) (Table 1; Figs. 2 and 3). The raw MR-PRESSO analyses support the IVW results but detected one outlier SNP (rs2200844) for the endometrial cancer analysis. After removing the outlier, a suggestive causal association was observed between EBV infection and endometrial cancer (OR = 2.43, 95%CI: 1.19–4.95, P = 0.02) (Table 2 and Supplementary Figure S1). Additionally, results from the MR-Egger, weighted median, simple mode, and weighted mode methods were all negative, showing no evidence of causal associations (Table 1).

Sensitivity analyses
Except for endometrial cancer, there were no outliers for the other analyses (Table 2). Cochran’s Q-test detected heterogeneity for endometrial cancer (P = 0.04) (Table 3; Fig. 4), probably related to the outlier detected by the MR-PRESSO analysis. MR-PRESSO analyses were performed for all cancers, and no outliers were detected for cervical, ovarian, or breast cancer. The MR-Egger regression analysis revealed no horizontal pleiotropy (Table 3), satisfying the third MR assumption. The leave-one-out analysis revealed no SNPs driving the results by themselves (Fig. 5).

Discussion
This two-sample MR explored the causal association between genetically predicted EBV infection and breast, cervical, endometrial, and ovarian cancers. The results, after removing an outlier, suggest the possibility of a causal association between EBV infection and the risk of endometrial cancer. However, this association was highly sensitive to a single outlier SNP (rs2200844), indicating that the observed effect may lack robustness and should be interpreted with caution.
EBV is considered an oncovirus because the genes encoded by its genome can lead to oncogenesis by activating oncogenes in the host cells or interacting with proteins that activate oncogenic pathways in the host cell. EBV infection can promote cell immortalization, proliferation, and division and inhibit apoptosis [30]. EBV infection activates multiple signaling pathways (e.g., NF-κB and PI3K/Akt) by expressing latency proteins (e.g., LMP1 and LMP2A) [31, 32], which play important roles in cell proliferation and apoptosis. These changes may cause abnormal cell proliferation, eventually leading to cancer. EBV also inhibits the host’s antiviral immune response through immune escape mechanisms, such as the expression of BCRF1, leading to chronic infection [33]. Long-term exposure of normal cells to EBV may lead to a decrease in immune surveillance, which increases the risk of cancer [34]. Chronic inflammation caused by chronic EBV infection can lead to cell damage and genetic mutations [35], which are known risk factors for cancer. EBV infection can cause epigenetic alterations in host cells, such as DNA methylation and histone modifications. These alterations may lead to tumor suppressor gene silencing or oncogene activation in endometrial cells, which can promote carcinogenesis [36]. EBV-encoded miRNAs may regulate gene expression in host cells, and these miRNAs can target oncogenes or tumor suppressor genes in cells, affecting cell proliferation and apoptosis [37]. Finally, EBV infection can alter the microenvironment of tissues, promoting the growth and invasion of tumor cells. EBV-related inflammatory responses and immune escape mechanisms can alter the cellular and molecular composition of tissue and promote cancer progression [38].
EBV infection is a well-known risk factor for several cancers, including Hodgkin lymphoma [9], Burkitt lymphoma [10], nasopharyngeal carcinoma [11], and gastric carcinoma [12]. On the other hand, although several pieces of evidence are available regarding EBV infection as a possible risk factor for breast [13, 14], cervical [15], and ovarian [16] cancers, the general level of evidence remains low. The primary analyses in the present study revealed no causal associations between EBV infection and breast, cervical, endometrial, or ovarian cancer, but a possible association was observed between EBV infection and endometrial cancer after removing an outlier. It is important to note that MR-PRESSO analyses were conducted for all cancers, and only endometrial cancer showed an outlier, while breast, cervical, and ovarian cancers did not. This indicates that the signal is specific but fragile.
Nagashima et al. [39] reported differentially methylated genes related to EBV infection in endometrial cancer cells in obese women. A study of EBV, papillomavirus, and cytomegalovirus infections in relation to endometrial and ovarian cancers revealed that endometrial cancer was related to cytomegalovirus infection but not to EBV or papillomavirus infection [40]. In addition, case-based and epidemiological evidence provides further insight into the role of EBV in gynecologic tumors. For example, a case report demonstrated EBV DNA in lymphoma-like cervical lesions during silent infection [17], highlighting that EBV may occasionally manifest in unusual genital tract pathologies. Another study of genital lesions found that CMV or HSV infections were linked to cervical cancer, but EBV infection was also detected and might act synergistically in HPV-DNA–negative cases [18]. More recently, a clinical study assessing viral coinfections in ovarian and endometrial cancers observed that while HCMV was more strongly associated with endometrial cancer, EBV infection was correlated with advanced-stage ovarian cancer, suggesting a role in tumor progression [19]. These findings suggest that while EBV is not consistently detected across all gynecologic cancers, it may contribute to tumorigenesis in specific contexts or subgroups.
Nevertheless, why EBV appears to be associated with endometrial cancer and not with breast, cervical, or ovarian cancers remains to be investigated. Several hypotheses can be formulated for future studies. (1) Endometrial tissue may have a unique susceptibility to EBV infection. Different tissue types differ in gene expression, immune response, and microenvironment, which may result in different cancers responding differently to EBV infection. (2) Endometrial cancer is closely related to hormones such as estrogen and progesterone [41]. EBV infection may interact with these hormones to further affect the carcinogenesis process of endometrial cells, a mechanism that may not be evident in other cancers. (3) Endometrial cells may be more susceptible to long-term exposure to EBV infection than other cell types, which may be due to the anatomical location or immune environment of the endometrium. Prolonged exposure to EBV may increase the risk of cancer, while other tissues may not have the conditions for such long-term exposure.
Importantly, the absence of causal associations between EBV infection and breast, cervical, and ovarian cancers should not be interpreted as definitive evidence of no involvement. Negative MR results may reflect limited statistical power, insufficient variance explained by genetic instruments, or heterogeneity across cancer subtypes. Moreover, MR studies assess causality only from a genetic perspective and cannot capture the influence of environmental exposures such as lifestyle or pollutants [42]. Hence, null findings do not rule out a role for EBV, but rather suggest that EBV is unlikely to be a strong universal driver of these cancers, while subtype-specific associations (e.g., triple-negative breast cancer) remain possible [43].
Furthermore, several other factors are involved in oncogenesis, including systemic inflammation and various comorbidities [44]. Other oncoviruses are also involved in cancer development, and they were not examined here [45]. Cancer is a complex multifactorial disorder involving multiple biological pathways [46, 47]. The immune system plays an important role in cancer development [48], and whether EBV-infected B cells can influence the tumor microenvironment remains unknown. Further complicating the analyses is that cancer is a very heterogeneous disease, including etiologies, histological subtypes, cancer marker expression, and prognosis [49]. Thus, future MR studies stratified by molecular or histological subtypes are warranted to clarify potential heterogeneous effects. From a clinical perspective, even if the observed association between EBV infection and ovarian cancer (OR = 0.05) is real, its magnitude is smaller than that of established risk factors such as obesity [50, 51]. However, EBV infection is nearly ubiquitous (>90% lifetime prevalence), which means that even modest effect sizes could translate into substantial public health implications. Moreover, EBV infection represents a potentially modifiable or preventable exposure, unlike many traditional risk factors. These aspects underscore the importance of continued research on EBV as a possible target for prevention and intervention.
This study has strengths. It used a two-sample MR design and GWAS data from hundreds or thousands of patients. It also evaluated the causal association between OSA and four different cancers frequently found in women. On the other hand, the study also had limitations. First, we used relatively relaxed selection criteria (P < 5 × 10⁻⁶) to obtain sufficient SNPs as instruments due to the limited sample size of the EBV GWAS. To address this, we additionally conducted a sensitivity MR analysis using the stricter threshold of P < 5 × 10⁻⁸, and these results are provided in the Supplementary Materials (Table S2). Second, this study did not evaluate cancer subtypes, and potential subtype-specific associations (e.g., TNBC, Type II endometrial cancer) may have been diluted when all cases were analyzed together. Third, both the exposure and outcome GWAS were based on European populations. Although they came from different consortia, some degree of sample overlap cannot be entirely excluded, which may introduce bias. Finally, the observed association between EBV infection and endometrial cancer was dependent on the removal of a single outlier SNP (rs2200844). This result should be considered exploratory and interpreted with caution.

Conclusions
In conclusion, the IVW analysis results showed no causal associations between EBV infection and breast, cervical, endometrial, or ovarian cancer, but after removing an outlier, the MR-PRESSO analysis suggested the possibility of a causal association between EBV infection and the risk of endometrial cancer. Still, this possible association must be taken cautiously and warrants additional study.

Supplementary Information
Below is the link to the electronic supplementary material.