Genetic evidence linking anti-polyomavirus 2 IgG seropositivity to ovarian cancer risk.

1. Introduction
Ovarian cancer remains a leading cause of gynecologic cancer–related mortality worldwide, with marked geographic variation in incidence and outcomes. Despite ongoing advances in surgical management and systemic therapies, the majority of patients are still diagnosed at advanced stages, contributing to unsatisfactory long-term survival rates.[1,2] Epidemiological evidence indicates that ovarian cancer predominantly affects postmenopausal women,[3] with incidence rising sharply after the age of 50 and peaking in the 6th to 7th decades of life.[4] The etiology of ovarian cancer is multifactorial. Although inherited genetic susceptibility – most notably pathogenic variants in breast cancer 1 susceptibility gene and breast cancer 2 susceptibility gene – accounts for ~10% to 15% of cases, most ovarian cancers arise from the interaction of genetic predisposition with reproductive, hormonal, metabolic, and lifestyle-related factors.[5] Established risk factors include advanced age, family history of ovarian or breast cancer, nulliparity or delayed childbearing, hormone-related exposures, and metabolic conditions such as obesity and polycystic ovary syndrome.[6] Clinical manifestations are often nonspecific and may include abdominal discomfort, bloating, urinary symptoms, and constitutional complaints, which partly explains the high proportion of late-stage diagnoses.[7,8]
Current diagnostic strategies rely on a combination of imaging modalities, circulating tumor markers such as cancer antigen 125 and human epididymis protein 4, and histopathological confirmation, while treatment is centered on cytoreductive surgery and platinum-based chemotherapy, with targeted therapies increasingly incorporated into standard care.[9] Risk-reduction strategies, including intensified surveillance and prophylactic interventions, are primarily considered for women with known hereditary susceptibility.[10,11]
In recent years, increasing attention has been directed toward the role of the immune system in ovarian cancer progression and treatment response. Although immunotherapeutic approaches, such as immune checkpoint inhibition, have demonstrated clinical activity in selected patient subgroups, their overall benefit in unselected ovarian cancer populations remains modest.[12,13] These heterogeneous responses highlight the need to better understand immune-related host factors that may influence ovarian cancer susceptibility and immune modulation beyond tumor-infiltrating lymphocyte–centered paradigms.[7,14]
Antibody immune responses represent an integrated reflection of host immunogenetic architecture and cumulative antigen exposure. Unlike acute immune markers, antibody-related phenotypes may capture long-term immune responsiveness shaped by genetic regulation. However, whether specific antibody immune response profiles contribute causally to ovarian cancer risk remains unclear, as conventional observational studies are vulnerable to confounding and reverse causation.
To address this knowledge gap, we applied a Mendelian randomization framework, combined with meta-analytic integration of independent datasets, to investigate the potential causal associations between genetically predicted antibody immune response phenotypes and ovarian cancer risk. This approach provides genetic evidence to clarify the role of antibody-mediated immune processes in ovarian cancer etiology and may inform future immunological and translational research.

2. Methods and materials

2.1. Study design
Summary-level genetic association data for both antibody immune response phenotypes and ovarian cancer were obtained from publicly available genome-wide association studies (GWAS). All datasets were processed using a unified analytical pipeline, including alignment of effect alleles and removal of variants with ambiguous strand orientation, to ensure compatibility across exposure and outcome datasets. A total of 46 genetically predicted antibody immune response phenotypes were evaluated in relation to ovarian cancer risk within a 2-sample Mendelian randomization framework.[15,16]
Ovarian cancer outcome data were derived independently from 2 large-scale consortia, and causal estimates were generated separately for each dataset using inverse-variance weighted (IVW) models as the primary analytical approach. The resulting estimates were subsequently synthesized to obtain overall effect estimates, with statistical adjustment applied to account for multiple testing across immune-related exposures. This integrative design was adopted to enhance statistical power and to reduce the influence of dataset-specific variation.[17]
For antibody immune response phenotypes demonstrating evidence of association with ovarian cancer, additional analyses were conducted to assess the robustness and directionality of the findings. Specifically, reverse-direction Mendelian randomization analyses were performed to examine whether ovarian cancer liability could influence the corresponding antibody-related traits, thereby evaluating the potential for reverse causation.[18,19]
Mendelian randomization relies on 3 core assumptions: relevance – the selected single-nucleotide polymorphisms (SNPs) are associated with anti-polyomavirus 2 immunoglobulin G (IgG) seropositivity; independence – the SNPs are not associated with confounders of the exposure–outcome relationship; and exclusion restriction – the SNPs influence ovarian cancer only through the IgG seropositivity phenotype. The third assumption may be challenged because genetic determinants of antibody responses, particularly variants in the human leukocyte antigen (HLA) region, can reflect broader immune traits (e.g., antigen presentation, inflammatory propensity) that may affect cancer risk via pathways not specific to JCPyV exposure.

2.1.1. Definition of exposure
In the present Mendelian randomization (MR), “anti-polyomavirus 2 IgG seropositivity” is treated as a host serological phenotype indexing genetic liability to mount an IgG response and to be classified as seropositive. This phenotype is most consistent with past exposure and humoral immune responsiveness, rather than with active infection, viral replication, viral load, or the presence of oncogenic viral gene expression in ovarian tissue at the time of carcinogenesis. Therefore, the causal estimand in MR should be interpreted as the effect of genetically proxied liability to this serological response on ovarian cancer risk.
The reporting of this study followed the Strengthening the Reporting of Observational Studies in Epidemiology using Mendelian Randomization guidelines for Mendelian randomization analyses. An overview of the study design and analytical workflow is provided in Figure 1.

2.2. GWAS data sources for antibody immune response phenotypes

2.2.1. Conceptual definition of antibody-related exposures
Anti-polyomavirus 2 IgG seropositivity captures a composite immune phenotype reflecting historical antigen exposure and host humoral immune capacity. Rather than indicating active viral infection or ongoing viral replication, this serological trait is shaped by genetically regulated processes involved in antigen presentation, immune modulation, and adaptive immune responsiveness. Accordingly, genetic instruments for this phenotype are best interpreted as markers of underlying immunogenetic variation influencing antibody responses, rather than as direct proxies for JC polyomavirus (JCPyV)–specific pathogenic effects.

2.2.2. Source and characteristics of GWAS data
Summary-level genetic association data for antibody immune response phenotypes were obtained from a publicly available GWAS repository. The dataset was generated from a cohort of 8735 individuals of European ancestry and included genetic variants associated with inter-individual variability in antibody-mediated immune responses. In addition to genome-wide signals, the analysis captured immune-related loci within the human leukocyte antigen region, which plays a central role in antigen presentation and adaptive immune regulation.
A total of 46 antibody immune response phenotypes were included in the present study, with corresponding GWAS accession numbers ranging from GCST90006884 to GCST90006929.[20] Genetic instruments for anti-polyomavirus 2 IgG seropositivity were derived exclusively from this European-ancestry population. Given the pronounced ancestry specificity of immune-related genetic architecture, particularly within the HLA region, differences in linkage disequilibrium (LD) patterns and allele frequencies should be taken into account when interpreting the generalizability of these findings to non-European populations.

2.2.3. Instrument selection and quality control
Genetic instruments for each antibody immune response phenotype were defined based on established criteria to ensure validity and analytical robustness. Variants showing evidence of association at the genome-wide level (P < 1 × 10−5) were considered, and instrument strength was evaluated using F-statistics (Table S1, Supplemental Digital Content, https://links.lww.com/MD/R460), with weak instruments excluded. Common variants were retained by applying a minor allele frequency threshold of 0.01, and LD pruning was performed to obtain a set of independent instruments using a 10-Mb window and an r2 (squared correlation coefficient) cutoff of 0.001. Following these selection and quality control procedures, 1018 SNPs were included in the Mendelian randomization analyses.

2.3. GWAS data sources for ovarian cancer outcomes
Ovarian cancer outcome data were obtained from 2 independent GWAS to minimize potential bias related to sample overlap and to enable the validation of causal estimates across datasets. The primary outcome dataset was derived from the FinnGen R12 release, which included 2339 ovarian cancer cases and 222,078 controls.[21] FinnGen integrates nationwide health registry information with genetic data, providing a population-based resource with comprehensive phenotypic coverage.
An additional ovarian cancer dataset was obtained from the Ovarian Cancer Association Consortium through the OpenGWAS platform (GWAS ID: ieu-a-1120), comprising 25,509 cases and 40,941 controls, for a total of 66,450 participants.[22] Both outcome datasets were restricted to individuals of European ancestry to ensure compatibility with the antibody immune response GWAS used for exposure and to reduce potential bias arising from population stratification and ancestry-specific LD patterns. While this ancestry-matched design enhances internal validity, it may limit the generalizability of the findings to non-European populations.

2.4. Instrumental variable selection and quality control
Genetic variants used as instrumental variables were selected to balance instrument relevance and analytical robustness. Variants associated with antibody immune response phenotypes at a relaxed genome-wide threshold (P < 1 × 10−5) were initially considered to ensure sufficient instrument availability, with each antibody trait represented by at least 3 independent single-nucleotide polymorphisms.
Instrument strength was subsequently assessed using F-statistics calculated for individual variants, and SNPs with insufficient strength were excluded. Common variants were retained by applying a minor allele frequency threshold of 0.01 to reduce instability associated with rare alleles.[23]
Following harmonization of exposure and outcome datasets, LD pruning was performed to derive a set of independent instruments using a 10-Mb window and an r2 cutoff of 0.001.[24] This filtering strategy yielded a final set of variants suitable for Mendelian randomization analyses.

3. Statistical analysis

3.1. Mendelian randomization of antibody immune phenotypes in relation to ovarian cancer
All analyses were performed using the R statistical environment (version 4.5.1). Genetic variants were harmonized across antibody immune response exposure datasets and ovarian cancer outcome datasets to ensure consistent allele orientation and effect alignment. Only variants present in both datasets were retained, and palindromic variants were processed using conservative matching rules, with variants failing harmonization or internal quality control excluded prior to analysis.
Robustness of the Mendelian randomization estimates was evaluated using multiple complementary approaches. Heterogeneity across instrumental variables was assessed, and pleiotropy-robust estimators were applied to account for potential violations of instrumental variable assumptions. In addition, outlier detection and correction were conducted using Mendelian Randomization Pleiotropy RESidual Sum and Outlier (MR-PRESSO) with 3000 permutations, providing global and outlier-corrected assessments of horizontal pleiotropy. These sensitivity analyses were particularly relevant given the enrichment of immune-related loci, including the HLA region, among antibody-associated genetic instruments.
Causal estimates were generated using IVW models, with model specification informed by the presence or absence of heterogeneity. Alternative Mendelian randomization methods, including weighted median and regression-based approaches, were implemented to further assess consistency across analytical assumptions. Effect estimates were reported as odds ratios with corresponding confidence intervals (CIs).
Estimates derived independently from the 2 ovarian cancer outcome datasets were subsequently synthesized to obtain overall effect estimates. Adjustment for multiple testing across antibody immune response phenotypes was applied to control the false-positive rate and ensure conservative inference.

3.2. Reverse Mendelian randomization of ovarian cancer on antibody responses
Reverse-direction Mendelian randomization analyses were performed by modeling ovarian cancer liability as the exposure and the antibody immune response phenotype, showing evidence of association as the outcome. This analysis was undertaken to examine whether the observed association was compatible with a causal effect operating in the opposite direction.
Instrument selection, data harmonization, and statistical modeling followed the same analytical framework used in the primary Mendelian randomization analyses, including consistent thresholds for variant selection and LD pruning. This ensured methodological symmetry between forward and reverse analyses and facilitated direct comparison of effect estimates.

3.3. Robustness assessment of causal estimates
Potential horizontal pleiotropy was evaluated using a combination of complementary sensitivity analyses. Evidence of directional pleiotropy was assessed using the Mendelian randomization Egger regression (MR-Egger) intercept, while heterogeneity across instrumental variables was examined using Cochran Q statistic. In addition, MR-PRESSO was applied to identify potential outlier variants and to generate outlier-corrected estimates (NbDistribution = 3000; SignifThreshold = 0.05). These analyses were considered particularly relevant given the enrichment of immune-related loci, including variants within the HLA region, among antibody-associated instruments.
Heterogeneity across genetic instruments was formally assessed prior to causal estimation. When evidence of heterogeneity was detected (Q_p < .05), causal effects were estimated using an IVW random-effects model; otherwise, a fixed-effects IVW approach was applied. This strategy allowed effect estimates to appropriately account for variability across instruments while maintaining consistency in model selection (Tables S2 and S3, Supplemental Digital Content, https://links.lww.com/MD/R460).

4. Results

4.1. Mendelian randomization of antibody immune phenotypes in relation to ovarian cancer
All 46 antibody immune response phenotypes were evaluated for their potential causal associations with ovarian cancer across 2 independent genome-wide association datasets. After meta-analysis of IVW estimates and adjustment for multiple comparisons, only anti-polyomavirus 2 IgG seropositivity showed evidence of a statistically significant association with ovarian cancer risk.
Specifically, the MR analysis results for anti-polyomavirus2 IgG seropositivity with ovarian cancer in the Finngen R12 database showed that the IVW estimate suggested a modest positive association between genetically proxied anti-polyomavirus 2 IgG seropositivity and ovarian cancer risk (odds ratio [OR] = 1.127; 95% CI: 1.026–1.239, P = .013). Directions were consistent across IVW, MR-Egger, and weighted median, supporting robustness under pleiotropy-robust estimators (Table S4, Supplemental Digital Content, https://links.lww.com/MD/R460). Additionally, an MR summary plot was generated for this result (Fig. 2). And the leave-one-out analysis suggested that the overall association was not driven by any single instrument (Fig. 2A). SNP-specific Wald ratio estimates and the pooled IVW and MR-Egger estimates are shown in the forest plot (Fig. 2B). The funnel plot showed no strong asymmetry, suggesting limited evidence of directional pleiotropy (Fig. 2C). The scatter plot visualizes the SNP–exposure and SNP–outcome associations, with regression slopes corresponding to different MR estimators (IVW, MR-Egger, weighted median, weighted mode, and simple mode; Fig. 2D).
Similarly, the MR analysis results for anti-polyomavirus 2 IgG seropositivity with ovarian cancer in the OpenGWAS database showed that the IVW estimate suggested a modest positive association between genetically proxied anti-polyomavirus 2 IgG seropositivity and ovarian cancer risk (OR = 1.053; 95% CI: 1.015–1.092, P = .0053). Directions were consistent across IVW, MR-Egger, and weighted median, supporting robustness under pleiotropy-robust estimators (Table S4, Supplemental Digital Content, https://links.lww.com/MD/R460). An MR summary plot was also generated for this result (Fig. 3), and leave-one-out analyses indicated that no individual SNP disproportionately influenced the MR estimate (Fig. 3A). SNP-specific Wald ratios together with the pooled IVW and MR-Egger estimates are presented in Figure 3B. The funnel plot did not indicate pronounced asymmetry (Fig. 3C). The scatter plot shows the instrument–exposure versus instrument–outcome effects and the fitted lines from multiple MR methods (Fig. 3D).
To obtain an overall estimate, results from the 2 datasets were combined using meta-analysis. After correction for multiple testing, genetically predicted anti-polyomavirus 2 IgG seropositivity remained associated with ovarian cancer risk (OR = 1.062, 95% CI: 1.027–1.099, adjusted P = .022) (Tables S5 and S6, Supplemental Digital Content, https://links.lww.com/MD/R460; Fig. 4).
The consistency of effect estimates across 2 independent European-ancestry datasets supports the robustness of this association within the studied population, although extrapolation to non-European populations requires further investigation.

4.2. Reverse Mendelian randomization of ovarian cancer on antibody responses
Reverse-direction Mendelian randomization analyses did not provide evidence for a causal effect of ovarian cancer liability on anti-polyomavirus 2 IgG seropositivity. Using ovarian cancer as the exposure, no significant association was observed in either the FinnGen R12 dataset (OR = 1.023, 95% CI: 0.941–1.113, P = .593) or the OpenGWAS dataset (OR = 0.917, 95% CI: 0.818–1.027, P = .135). These findings were consistent across both outcome sources and indicate a lack of support for reverse causality (Table S7, Supplemental Digital Content, https://links.lww.com/MD/R460).

5. Discussion
In this bidirectional Mendelian randomization study integrating evidence from 2 independent ovarian cancer datasets, we identified a modest but statistically robust association between genetically predicted anti-polyomavirus 2 IgG seropositivity and ovarian cancer risk, with no evidence supporting reverse causality. Among the 46 antibody immune response phenotypes evaluated, this was the only phenotype that remained significant after meta-analysis and multiple testing correction.
Importantly, the MR exposure in this study represents genetic liability to anti-polyomavirus 2 IgG seropositivity, which reflects prior viral exposure and host humoral immune responsiveness rather than active JCPyV infection, viral replication, or oncogenic activity within ovarian tissue. Consequently, the observed association should not be interpreted as evidence that active JCPyV infection directly causes ovarian cancer. Instead, the findings are best understood as implicating immune-related pathways captured by antibody-response phenotypes.
Anti–JCPyV IgG antibodies are highly prevalent in adult populations and serve as markers of long-term immune memory following infection. Given the widespread seropositivity of JCPyV, the pooled MR estimate indicates a small effect size (OR = 1.06), corresponding to a modest increase in relative risk. While statistically significant, this magnitude suggests limited clinical relevance at the individual level and does not support serological screening or risk stratification in the general population.[25–27]
A key methodological consideration is the exclusion restriction assumption. Genetic instruments for antibody-response phenotypes are frequently enriched in immune regulatory regions, particularly within the HLA locus, which may influence cancer susceptibility through generalized immune functions such as antigen presentation, immune surveillance, or inflammatory regulation. Although pleiotropy-robust methods and extensive sensitivity analyses were applied, residual pleiotropy related to broad immunogenetic mechanisms cannot be fully excluded. Accordingly, the association observed here may reflect JCPyV-related immune responsiveness and/or more general immune traits rather than virus-specific oncogenic effects.[28,29]
A central caveat is that IgG seropositivity is a marker of prior exposure and host humoral immune responsiveness, not a direct measure of active JCPyV infection, viral replication, or viral oncogenic activity within ovarian tissue. Consequently, our MR results should not be interpreted as demonstrating that active JCPyV infection per se causally drives ovarian carcinogenesis. Instead, the MR estimate can be understood as the effect of genetic liability to being seropositive/ mounting an IgG response on ovarian cancer risk.
This distinction matters for the exclusion restriction assumption. Genetic determinants of antibody responses, particularly variants in immune-related regions such as HLA, may capture broader immunogenetic traits (e.g., antigen presentation efficiency, immune surveillance capacity, inflammatory propensity) that could influence cancer susceptibility through pathways independent of JCPyV-specific biology. Although we implemented pleiotropy-robust estimators (weighted median, MR-Egger) and outlier/heterogeneity diagnostics (MR-PRESSO, Cochran Q, leave-one-out), these approaches cannot fully exclude residual pleiotropy when instruments index general immune function. Therefore, we interpret the findings as supporting a modest association consistent with the involvement of JCPyV-related immune pathways and/or broader immune mechanisms, and we emphasize the need for mechanistic studies and prospective serology/virology studies (including tissue-based viral markers) to clarify specificity.
Although MR-Egger and MR-PRESSO were applied, instruments for antibody-response phenotypes are often enriched in immune regulatory loci (notably the HLA region), which may influence ovarian cancer susceptibility through broader pathways such as antigen presentation, immune surveillance, and inflammatory propensity. Such mechanisms could operate independently of JCPyV-specific biology, challenging the exclusion restriction assumption. Pleiotropy-robust estimators can mitigate some violations but cannot fully rule out residual pleiotropy when instruments proxy generalized immune function. Therefore, our findings are best interpreted as indicating a modest association consistent with immune-mediated mechanisms (JCPyV-related and/or generalized immunogenetic), rather than definitive evidence that active JCPyV oncogenic activity causally drives ovarian carcinogenesis.
JCPyV has been proposed to have oncogenic potential in experimental and tissue-based studies, including mechanisms involving large T-antigen–mediated perturbation of cell-cycle control and inflammatory changes.[29–31] However, evidence across tumor types remains heterogeneous, and our MR results – based on a serological antibody-response phenotype – do not directly test viral presence, viral gene expression, or integration in ovarian tissue.[32] Thus, these mechanistic pathways should be viewed as a biological context rather than causal mechanisms established by the current analysis. Future studies integrating serology with tissue-based virological markers (e.g., JCPyV DNA/RNA or T-antigen in ovarian tumors) are needed to clarify specificity.[33,34]
Experimental and tissue-based studies have proposed oncogenic roles for JCPyV through mechanisms involving large T-antigen activity and host inflammatory responses. However, the present MR analysis – based on a serological antibody-response phenotype – does not directly assess viral presence, gene expression, or genomic integration in ovarian tissue. These mechanistic hypotheses, therefore, provide biological context rather than causal explanations established by the current findings.[35]
The present findings are hypothesis-generating and point to several directions for future research. First, studies that integrate serology with tissue-based virological assessments (e.g., JCPyV DNA/RNA detection and T-antigen expression in ovarian tumors) are needed to clarify whether the observed association reflects virus-specific biology or broader immunogenetic mechanisms.[36] Second, complementary epidemiologic and experimental work should examine how host immune responsiveness to JCPyV relates to inflammatory pathways and the ovarian tumor immune microenvironment. Importantly, because our MR exposure captures a serological antibody-response phenotype rather than active infection, the current evidence does not support anti-polyomavirus 2 IgG testing as a screening tool or immediate therapeutic/vaccine development; any translational implications would require rigorous validation of predictive performance and biological specificity in prospective cohorts.[18,37]
Restricting analyses to a single ancestry reduces confounding from population stratification, but it does not address external validity. Future work should replicate these MR analyses in multi-ancestry datasets and perform ancestry-stratified instrument evaluation (including HLA-focused sensitivity analyses and colocalization) to assess robustness and trans-ethnic transferability.
Several limitations warrant consideration. First, both exposure and outcome datasets were restricted to individuals of European ancestry, which enhances internal validity but limits generalizability to other populations. Given substantial ancestry-related differences in immune genetic architecture, JCPyV seroprevalence, and LD patterns – particularly within the HLA region – future multi-ancestry analyses are needed to evaluate transferability. Second, the antibody-response phenotype does not distinguish between timing, intensity, or anatomical relevance of viral exposure, underscoring the need for studies integrating serology with tissue-based virological markers.
In summary, this study provides genetic evidence supporting a modest association between anti-polyomavirus 2 IgG seropositivity and ovarian cancer risk, consistent with the involvement of immune-mediated mechanisms. These findings are hypothesis-generating and highlight the need for integrated epidemiologic, immunologic, and tissue-based investigations to clarify biological specificity and clinical relevance.

6. Conclusion
This Mendelian randomization study identifies a modest association between genetically predicted anti-polyomavirus 2 IgG seropositivity and ovarian cancer risk. As the exposure reflects an antibody-response phenotype rather than active infection, the findings should be interpreted with caution, and further studies integrating tissue-based virological markers across diverse populations are warranted.

Acknowledgments
We thank the investigators and participants of the genome-wide association studies whose data were used in this research, as well as the curators of the publicly available databases that made these analyses possible.

Author contributions
Conceptualization: Weichu Zhang, Zhuorong Gu, Guanghui Song, Shaohan Zou, Shihao Hong.
Data curation: Weichu Zhang, Zhuorong Gu, Shihao Hong.
Formal analysis: Weichu Zhang, Zhuorong Gu, Guanghui Song, Shaohan Zou, Shihao Hong.
Funding acquisition: Zhuorong Gu.
Investigation: Weichu Zhang, Zhuorong Gu.
Methodology: Weichu Zhang, Zhuorong Gu, Guanghui Song, Shaohan Zou, Shihao Hong.
Project administration: Weichu Zhang, Zhuorong Gu, Guanghui Song, Shaohan Zou, Shihao Hong.
Software: Weichu Zhang.
Supervision: Shihao Hong.
Validation: Shihao Hong.
Visualization: Weichu Zhang, Guanghui Song.
Writing – original draft: Weichu Zhang, Zhuorong Gu, Guanghui Song, Shaohan Zou.
Writing – review & editing: Shihao Hong.

Supplementary Material