Causal associations between plasma proteins and premature ovarian failure through multi-omics Mendelian randomization analyses

Introduction

Premature ovarian failure (POF) is a complex and heterogeneous disorder characterized by the premature cessation of ovarian function before the age of 40 years. It affects approximately 1–3.5% of women of reproductive age and is associated with significant reproductive and health-related complications, including infertility, hormonal imbalances, and psychosocial consequences (1). The condition exhibits considerable heterogeneity due to a range of potential etiologies, such as genetic abnormalities, autoimmune disorders, and environmental influences, all of which contribute to the complexity of diagnosis and clinical management (2,3).

Despite advances in reproductive medicine, current diagnostic and therapeutic strategies for POF remain suboptimal. Hormone replacement therapy (HRT) is widely regarded as the primary intervention for alleviating symptoms related to estrogen deficiency; however, it does not rectify the underlying causes of ovarian dysfunction (4,5). For women seeking fertility, assisted reproductive technologies (ART) such as oocyte donation are commonly pursued. Nevertheless, these interventions do not offer a comprehensive resolution to the multifaceted challenges associated with POF (2). Therefore, there is an urgent need for innovative research approaches aimed at identifying reliable biomarkers that can enable early diagnosis and the development of targeted therapeutic strategies addressing the root pathophysiological mechanisms of the condition.

Utilizing multi-omics approaches, recent investigations have uncovered a panel of promising plasma proteins for POF (6). Liu et al. conducted comparative proteomic and metabolomic analyses of serum samples from 14 patients with POF and 16 healthy controls, revealing that the combination of afamin (AFM) and 2-oxoarginine exhibits high diagnostic accuracy as a biomarker panel for the clinical identification of POF (7). Notably, proteins involved in immune regulation and metabolic processes have been linked to POF, suggesting that dysregulation in these biological pathways may play a key role in disease pathogenesis (8,9). Despite these promising findings, much of the existing research has relied on observational and correlational data, which limits the ability to establish clear causal relationships between plasma proteins and the risk of developing POF.

In this study, we selected instrumental variables (IVs) associated with expression quantitative trait loci (eQTLs) and protein quantitative trait loci (pQTLs), enabling direct causal inferences regarding the relationships between gene expression, protein levels, and the risk of POF. Given that Mendelian randomization (MR) alone may not be sufficient to reliably identify causal proteins within complex biological pathways, we supplemented our analysis with complementary approaches, including colocalization analysis, summary-data-based MR (SMR), heterogeneity in dependent instruments (HEIDI) testing, and a series of sensitivity analyses. Additionally, we performed mediation analysis to investigate the potential mechanisms through which plasma proteins may influence POF via intermediary pathways, particularly those involving immune function and metabolic processes. Finally, a phenome-wide MR analysis (MR-PheWAS) was performed to systematically evaluate their associations with diverse clinical traits across multiple disease domains. We present this article in accordance with the STROBE-MR reporting checklist (available at https://gpm.amegroups.com/article/view/10.21037/gpm-2025-1-68/rc).

Methods

Study design

The overall study design is illustrated in Figure 1. First, we applied MR to examine the potential causal associations between cis-acting pQTLs (cis-pQTLs) and the risk of POF. Following this, we validated these causal relationships using SMR and colocalization analysis. To further strengthen the reliability of our findings, we conducted external validation through MR analyses based on corresponding cis-acting eQTLs (cis-eQTLs). To investigate whether the identified candidate proteins contribute to POF pathogenesis, we integrated data on 731 immune cell traits and 1,400 metabolites, and established a two-step mediation model to systematically assess the mediating roles of immune and metabolic pathways in the associations between proteins and POF. Finally, to comprehensively evaluate the potential clinical implications of these candidate therapeutic targets, we conducted an MR-PheWAS to identify possible off-target effects or unintended consequences associated with these proteins. This study utilized previously published research data that had already obtained approval from the ethics review committee in the original studies, thus requiring no additional ethical approval for the current analysis. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Figure 1 Overview of the study design. cis-eQTL, cis-acting expression quantitative trait loci; cis-pQTL, cis-acting protein quantitative trait loci; FDR, false discovery rate; HEIDI, heterogeneity in dependent instrument; MR-pheWAS, phenome-wide mendelian randomization analysis; POF, premature ovarian failure; PP.H4, posterior probability for H4; SMR, summary-data-based Mendelian randomization; SNP, single nucleotide polymorphism.

Data source

Exposure data

The pQTL dataset was obtained from the UK Biobank Pharma Proteomics Project (UKB-PPP, http://ukb-ppp.gwas.eu), a joint initiative involving the UK Biobank and multiple biopharmaceutical partners. This large-scale study systematically analyzed plasma protein profiles across 54,219 UK Biobank participants, identifying 14,287 statistically significant genetic associations through genome-wide pQTL mapping of 2,923 distinct proteins (10). The eQTL dataset, comprising genomic regulatory variants for 16,989 genes, was generated through large-scale analysis of 31,684 peripheral blood samples from healthy individuals of European ancestry by the eQTLGen Consortium (https://eqtlgen.org/) (11).

The immunophenotyping dataset, derived from genomic analyses of 3,757 Sardinian individuals, encompasses 731 quantitative traits cataloged in the Genome-Wide Association Studies (GWAS) Catalog (accession range GCST0001391 to GCST0002121). These metrics systematically quantify immune cell characteristics through four measurement modalities: 118 absolute cell counts, 389 median fluorescence intensities, 32 morphological parameters, and 192 relative counts (12). The metabolomic dataset originated from a large-scale GWAS comprising 8,299 participants of European ancestry within the Canadian Longitudinal Study on Aging (CLSA) cohort. Conducted by Chen et al. [2023], this research systematically evaluated 1,400 quantitative metabolic traits—encompassing 1,091 individual metabolites and 309 derived ratios. The complete GWAS summary statistics are publicly accessible via the GWAS Catalog under accession codes GCST90199621 to GCST90204063 (13).

Outcome data

The data on POF were obtained from the FinnGen database R12 release (access No. finngen_R12_E4_OVARFAIL), comprising 655 cases and 267,780 controls. POF was rigorously defined as the premature decline or cessation of ovarian function attributable to intrinsic ovarian pathology, with diagnostic confirmation based on International Classification of Diseases (ICD)-10 code E28.3 and endpoint validation by the FinnGen expert panel (https://risteys.finngen.fi/endpoints/E4_OVARFAIL).

Statistical analysis

MR analysis

Our two-sample MR analysis systematically assessed the causal relationships between cis-pQTLs and POF using rigorously selected genetic instruments. To ensure robust IVs selection for MR analysis, we applied three key criteria: (I) relevance assumption: direct association with the exposure; (II) independence assumption: independence from confounders of the exposure-outcome association; (III) exclusion restriction assumption: influence on the outcome exclusively through the exposure (14). The screening criteria for IVs include: (I) single nucleotide polymorphisms (SNPs) located within ±1 megabase of their corresponding protein-coding genes with genome-wide significant associations (P<5×10⁻⁸); (II) minimal linkage disequilibrium (r²<0.001, kb <10,000) to ensure independence among variants; (III) F-statistics ≥10 to guarantee instrument strength and mitigate weak instrument bias (15).

For multiple SNPs, we employed a comprehensive analytical approach incorporating five complementary MR methods: inverse-variance weighted (IVW) as the primary method, supplemented by MR-Egger, simple mode, weighted median, and weighted mode approaches to enhance robustness. Individual SNP effects were quantified using Wald ratio estimates (16,17). All statistical analyses were implemented in R version 4.3.1, leveraging the specialized “TwoSampleMR” and “ieugwasr” packages. Moreover, the causal direction between cis-pQTLs and POF was further validated through Steiger filtering analysis to ensure the robustness of the inferred relationship (18).

SMR analysis and HEIDI test

The SMR method extends conventional MR principles by systematically integrating GWAS data with eQTL datasets to infer potential causal relationships between protein levels and disease phenotypes. In our investigation of POF, we leveraged cis-pQTLs as exposures, with POF serving as outcome variable, utilizing SMR software version 1.3.1 for analyses. Our analytical framework restricted the genomic window to ±1,000 kilobases surrounding each target gene’s chromosomal coordinates, exclusively incorporating only top SNPs (P<5×10⁻⁸) as IVs to ensure robust genetic instruments. To mitigate potential false-positive associations arising from linkage disequilibrium among correlated variants, we implemented the HEIDI test, where a HEIDI P value below 0.05 signifies substantial heterogeneity and consequently invalidates the assumption of a causal relationship (19).

Colocalization analysis

To rigorously evaluate whether the plasma proteins identified through SMR analysis share common genetic causal variants with POF, we conducted a Bayesian colocalization analysis using the R package “coloc”. This analytical approach enables robust differentiation between genuine colocalization and spurious associations arising from linkage disequilibrium (20). The analysis assessed five mutually exclusive hypotheses (21): H0, no causal variant for either trait; H1, a causal variant influencing only the exposure (protein); H2, a causal variant influencing only the outcome (POF); H3, distinct causal variants affecting each trait independently; and H4, a shared causal variant underlying both traits. A posterior probability for H4 (PP.H4) greater than 0.8 was used as the threshold to indicate strong evidence of colocalization, and only associations meeting this criterion were carried forward for further analysis.

External validation

To further validate the robustness of our SMR and colocalization findings, we performed external validation by integrating corresponding cis-eQTL data with POF outcome datasets. For this validation step, we employed cis-eQTLs from the eQTLGen consortium (P<5×10⁻⁸) as IVs and applied them to GWAS data from the FinnGen cohorts for two-sample MR analysis, adhering to the same criteria used for pQTLs. This transcriptional-level validation provides additional evidence supporting our protein-level causal inferences, thereby enhancing the reliability and robustness of our findings. Additionally, the directionality of the link between cis-eQTL and POF was verified using Steiger filtering.

Mediation analysis

To examine whether the proteins identified through SMR and colocalization analyses exert indirect effects on the development of POF via immune or metabolic pathways, we constructed a two-step MR mediation model, incorporating 731 immune cell traits and 1,400 plasma metabolites as potential mediators. In the first stage, a two-sample MR approach was applied to estimate the genetic causal effects of cis-pQTLs on each mediator (β₁). In the second stage, MR analyses were conducted to evaluate the causal influence of these mediators on POF (β₂). The mediation effect was quantified as the product of β₁ and β₂ (β₁ × β₂). The total effect was obtained from prior two-sample MR results assessing the direct effect of cis-pQTLs on POF, while the direct effect was calculated as the difference between the total effect and the mediation effect (direct effect = total effect − mediation effect). The mediation proportion was determined using the formula: mediation proportion = (mediation effect/total effect) ×100%. The 95% confidence intervals (CIs) for both the mediation effect and the mediation proportion were calculated using the delta method (22). A significant mediation effect was concluded when both the first-stage (exposure to mediator) and second-stage (mediator to outcome) MR results reached statistical significance (P<0.05), and the mediation effect also showed a P value less than 0.05.

MR-PheWAS

To systematically evaluate the systemic effects and potential adverse reactions of candidate therapeutic targets for POF, we performed an MR-PheWAS analysis using the FinnGen R12 database (23). This comprehensive dataset encompasses 2,469 clinically defined phenotypes derived from FinnGen cohorts, spanning multiple physiological systems including immunity, oncology, metabolism, and endocrinology (24). Building upon our prior SMR and colocalization findings, we selected significant protein candidates and employed their corresponding cis-pQTLs as IVs. Causal relationships were assessed using complementary analytical approaches: the Wald ratio method for single-SNP analyses and IVW regression for multi-SNP models. To ensure methodological rigor, we implemented a multi-tiered quality control framework: (I) instrument strength validation (F-statistic >10); (II) directionality confirmation via Steiger filtering; (III) pleiotropy assessment using MR-Egger intercept tests. All statistical outputs underwent Benjamini-Hochberg false discovery rate (FDR) correction, with significant associations defined at FDR <0.05.

Results

Putative plasma proteins on POF

Based on stringent selection criteria, we identified 5,622 SNPs as IVs for 1,925 plasma proteins, with F-statistics ranging from 29.73 to 15,013.28, indicating no weak instrument bias (table available at: https://cdn.amegroups.cn/static/public/gpm-2025-1-68-1.xlsx). Using either the Wald ratio method or IVW approach, we detected 96 plasma proteins showing causal associations with POF (table available at: https://cdn.amegroups.cn/static/public/gpm-2025-1-68-2.xlsx). The MR results revealed that 46 plasma proteins, including C-C motif chemokine 23 (CCL23) [IVW: odds ratio (OR) =0.704, 95% CI: 0.551–0.900, P=0.005] and tumor necrosis factor receptor superfamily member 14 (TNFRSF14) (Wald ratio: OR =0.224, 95% CI: 0.088–0.570, P=0.002), exhibited protective effects on POF. The remaining 50 plasma proteins were identified as risk factors for the disease (Figure 2).

Figure 2 Volcano plot of the 96 identified proteins with POF. POF, premature ovarian failure.

To further validate the causal relationships of plasma proteins identified through preliminary MR analysis, we conducted additional SMR and HEIDI tests using cis-pQTLs as exposure variables and disease phenotypes as outcome variables. The results confirmed 58 plasma proteins with significant causal associations, including CCL23 (OR =0.648, 95% CI: 0.494–0.850, P_smr =0.002, P_HEIDI =0.13) and TNFRSF14 (OR =0.224, 95% CI: 0.086–0.580, P_smr =0.002, P_HEIDI =0.44). However, three proteins—HMOX1 (OR =0.177, 95% CI: 0.032–0.975, P_smr =0.047, P_HEIDI =0.008), AHSP (OR =14.382, 95% CI: 1.306–158.38, P_smr=0.03, P_HEIDI =0.02), and SUGP1 (OR =1.92, 95% CI: 1.10–3.36, P_smr =0.02, P_HEIDI =0.02)—exhibited marked effects but failed the HEIDI test, suggesting potential pleiotropy or linkage disequilibrium confounding (Figure 3, table available at: https://cdn.amegroups.cn/static/public/gpm-2025-1-68-3.xlsx).

Figure 3 Forest plot of the summary-data-based Mendelian randomization analyses for 58 identified proteins associated with premature ovarian failure risk. CI, confidence interval; HEIDI, heterogeneity in dependent instruments; OR, odds ratio; SNP, single nucleotide polymorphism.

Colocalization assessment between cis-pQTL and POF

Colocalization analysis was conducted on the protein loci that passed both SMR and HEIDI tests, providing further evidence to support the causal involvement of CCL23 (PP.H4 =72.1%, PP.H3 =10.9%) and TNFRSF14 (PP.H4 =54.9%, PP.H3 =21.7%) in POF. These findings indicate that these loci are likely to have genuine pathogenic effects on POF susceptibility (Figure 4). In contrast, the remaining genes did not exhibit evidence of colocalization with POF (table available at: https://cdn.amegroups.cn/static/public/gpm-2025-1-68-4.xlsx), underscoring the necessity of utilizing multiple analytical strategies to thoroughly evaluate gene-phenotype associations and differentiate true causal relationships from potential confounding factors.

Figure 4 Regional plot of colocalization evidence of significant proteins and POF. (A) CCL23; (B) TNFRSF14. CCL23, C-C motif chemokine 23; POF, premature ovarian failure; TNFRSF14, tumor necrosis factor receptor superfamily member 14.

External validation

External validation was performed using cis-eQTL data for CCL23 and TNFRSF14 obtained from the eQTLGen database in conjunction with POF outcome data. The MR results indicated that both CCL23 (IVW: OR =0.543, 95% CI: 0.347–0.849, P=0.007) and TNFRSF14 (Wald ratio: OR =0.673, 95% CI: 0.503–0.900, P=0.008) act as protective factors against POF (Figure 5). No significant heterogeneity or horizontal pleiotropy was detected based on Cochran’s Q test and MR-Egger intercept test, respectively (table available at: https://cdn.amegroups.cn/static/public/gpm-2025-1-68-5.xlsx). These results collectively support the robustness and validity of our prior conclusions concerning the protective roles of these plasma proteins in relation to POF.

Figure 5 Forest plot for the MR results of CCL23 and TNFRSF14 on POF. CCL23, C-C motif chemokine 23; CI, confidence interval; MR, Mendelian randomization; OR, odds ratio; POF, premature ovarian failure; SNP, single nucleotide polymorphism; TNFRSF14, tumor necrosis factor receptor superfamily member 14.

Mediation analysis

To investigate the indirect effects of CCL23 and TNFRSF14 on POF outcomes through immune cells and plasma metabolites, we conducted a two-step MR analysis that integrated 731 immune cell signatures and 1,400 plasma metabolites. The results revealed that 29 immunophenotypic traits and 65 metabolites serve as causal factors in the development of POF (Figure 6A,6B). Subsequently, MR analysis revealed that CCL23 causally modulates CD16-CD56 on natural killer, CCR2 on plasmacytoid dendritic cell, and betaine levels, while TNFRSF14 drives alterations in ceramide (d18:1/24:1) levels and sphingomyelin (d18:1/25:0, d19:0/24:1, d20:1/23:0, d19:1/24:0) levels (table available at: https://cdn.amegroups.cn/static/public/gpm-2025-1-68-6.xlsx).

Figure 6 Circular heatmap illustrating the impact of immune cells and plasma metabolites on POF. (A) Immune cells; (B) plasma metabolites. IVW, inverse-variance weighted; MR, Mendelian randomization; POF, premature ovarian failure.

Mediation analysis further indicated that in the causal pathway between CCL23 and POF, CCR2 on plasmacytoid dendritic cells exhibits a mediating effect of −0.030 (95% CI: −0.072 to 0.012, P=0.14), with a mediation proportion of 8.57% (95% CI: −3.31% to 20.46%). In the causal pathway between TNFRSF14 and POF, ceramide (d18:1/24:1) demonstrates a mediating effect of −0.088 (95% CI: −0.194 to 0.018, P=0.09), with a mediation proportion of 5.86% (95% CI: −1.21% to 12.93%); Sphingomyelin (d18:1/25:0, d19:0/24:1, d20:1/23:0, d19:1/24:0) shows a mediating effect of −0.048 (95% CI: −0.115 to 0.019, P=0.14), with a mediation proportion of 3.19% (95% CI: −1.31% to 7.71%) (Table 1).

MR-PheWAS

To assess the therapeutic safety profiles of CCL23 and TNFRSF14 as candidate targets for POF, we performed an MR-PheWAS analysis using FinnGen cohort data (table available at: https://cdn.amegroups.cn/static/public/gpm-2025-1-68-7.xlsx). Our findings revealed distinct phenotypic association patterns for each protein: CCL23 demonstrated protective effects, showing significant negative associations with disorders of the thyroid gland and fracture of forearm, suggesting potential benefits for endocrine and musculoskeletal health.

TNFRSF14 exhibited widespread pleiotropy across three key domains: (I) immune regulation: strong positive associations with autoimmune pathologies including seropositive rheumatoid arthritis (strict/wide, β=0.75/0.47), multiple sclerosis (β=1.07), and second line medication for Crohn’s disease (β=0.37), indicating broad involvement in inflammatory pathways. (II) Oncogenic potential: elevated risks for lymphohematopoietic malignancies, particularly follicular lymphoma (β=1.41), non-Hodgkin lymphoma (β=0.77), and primary lymphoid and hematopoietic malignancies (β=0.51), highlighting potential oncogenicity. (III) Metabolic modulation: negative correlations with body mass index (β=−0.12) and weight (β=−0.09), suggesting negative metabolic regulatory functions (Figure 7).

Figure 7 Phenome-wide MR analysis results showing associations between CCL23 (blue dots) and TNFRSF14 (red dots) and multiple disease phenotypes. CCL23, C-C motif chemokine 23; MR, Mendelian randomization; MS, multiple sclerosis; TNFRSF14, tumor necrosis factor receptor superfamily member 14.

Discussion

To the best of our knowledge, this study represents the most comprehensive MR investigation to date examining the causal relationship between plasma proteins and POF, significantly strengthening causal inference by minimizing confounding biases. By integrating multi-omics data and employing multiple analytical methods, our analysis identifies several plasma proteins, such as CCL23 and TNFRSF14, as potential causal factors in POF pathogenesis. Although a recent study also applied MR to explore the role of plasma proteins in POF (25), our research distinguishes itself by utilizing the most recent Finnish database (FinnGen R12). Furthermore, mediation analyses indicate that these proteins may partially influence POF risk through immune and inflammation-related pathways. Although not statistically significant, this pattern warrants further investigation in larger studies to determine if a robust mediating effect exists.MR-PheWAS was conducted to assess the potential pleiotropic effects of the identified proteins. These findings underscore their potential as diagnostic biomarkers and therapeutic targets.

POF, often considered the end stage of premature ovarian insufficiency (POI), is characterized by the cessation of normal ovarian function before the age of 40 years. Clinically, it is defined by elevated levels of follicle-stimulating hormone (FSH), typically exceeding 40 IU/L on two separate measurements spaced at least one month apart, along with symptoms such as amenorrhea or irregular menstrual cycles (26). A central biomarker for evaluating ovarian reserve is anti-Müllerian hormone (AMH), produced by granulosa cells of preantral and small antral follicles. AMH levels closely reflect the number of primordial follicles and, unlike other reproductive hormones that vary cyclically, remain relatively stable throughout the menstrual cycle. This stability makes AMH a dependable and practical indicator for assessing ovarian reserve at any time (27,28). Together, elevated FSH, low AMH, and clinical manifestations form an integrative diagnostic framework to identify women at risk of premature ovarian dysfunction.

The etiology of POI exhibits considerable heterogeneity, involving a range of contributing factors such as genetic predisposition, autoimmune mechanisms, infectious agents, iatrogenic causes, and environmental exposures (5,29). Notably, in some patients, different etiological factors may interact or coexist. For instance, triple X syndrome in women with POI can occur alongside autoimmune thyroid disorders (30). Recent proteomic studies have identified several proteins that are differentially expressed in the serum of POF patients, including ceruloplasmin, complement C3, fibrinogen α and β chains, and sex hormone-binding globulin (SHBG). These proteins are implicated in various biological processes such as oxidative stress, immune-mediated inflammation, and hormone transport, suggesting that systemic physiological dysregulation contributes to the pathogenesis of POF (6). Building upon these findings, our study significantly expands the current understanding of POF by identifying 96 plasma proteins that exhibit statistically significant causal associations with the condition. This represents a substantial increase in the number of potential molecular contributors to POF and provides a more comprehensive view of the biological networks involved. Further colocalization analysis provided additional support for these causal relationships, revealing shared causal variants between POF and specific proteins, notably CCL23 and TNFRSF14.

CCL23, also known as myeloid progenitor inhibitory factor-1 (MPIF-1), is a chemokine that binds to the C-C chemokine receptor type 1 (CCR1), playing a pivotal role in immune cell trafficking, inflammatory responses, and angiogenesis (31,32). It has been extensively studied in the contexts of cancer, neuroinflammation, and pregnancy physiology, with emerging evidence supporting its potential as a biomarker in ovarian cancer and Alzheimer’s disease (33-35). However, its role in normal ovarian function or in the context of POF remains largely unexplored. Accumulating research indicates that immune dysregulation significantly impacts folliculogenesis and contributes to the development of POF (36). Abnormalities in monocyte differentiation and dendritic cell antigen presentation have been observed in individuals with POF, leading some researchers to classify POF as an endocrine autoimmune disorder characterized by immune-mediated ovarian dysfunction (37,38). In our study, CCL23 was identified as a protective factor against POF, which aligns with the findings from a recent MR analysis on POF (25). Mediation analysis further revealed that CCR2 on plasmacytoid dendritic cells mediates the protective effect of CCL23, with a mediation effect size of −0.030 and a mediation proportion of 8.57%. These findings suggest that CCL23 may alleviate ovarian dysfunction, at least in part, through immune-regulatory mechanisms involving CCR2 on plasmacytoid dendritic cells.

TNFRSF14, also referred to as the herpesvirus entry mediator (HVEM), is a member of the tumor necrosis factor (TNF) receptor superfamily and has been extensively investigated in fields such as tumor microenvironment, autoimmune diseases, and inflammation (39-41). A growing body of evidence highlights its functional duality, acting both as a receptor for canonical TNF ligands like LIGHT and lymphotoxin-α, and as a ligand for immunoglobulin superfamily members such as BTLA and CD160, thereby modulating immune activation and inhibition (42,43). Researchers found that TNFRSF14 participated in ovariectomy-induced adipose tissue inflammation via upregulation of CD11c, resulting in metabolic perturbation (44). In this study, TNFRSF14 was identified as a protective factor against POF. Mediation analysis revealed that Ceramide (d18:1/24:1) levels exhibited a mediating effect with a value of −0.088, accounting for a mediation proportion of 5.86%; sphingomyelin (d18:1/25:0, d19:0/24:1, d20:1/23:0, d19:1/24:0) levels exhibited a mediating effect with a value of −0.048, accounting for a mediation proportion of 3.19%. Furthermore, MR-PheWAS analysis demonstrated that TNFRSF14 exhibits broad pleiotropic effects across immune regulation, oncogenesis, and metabolic modulation. These findings highlight its potential as a multifaceted therapeutic target, offering avenues for intervention not only in POF but also in mitigating associated comorbidities.

The limitations of this study warrant careful consideration. Firstly, external validation was performed using eQTL data rather than independent protein-level measurements. Secondly, the absence of wet-lab experiments restricts the validation of our findings, as the results are derived solely from computational analyses. Additionally, the relatively small sample size may limit the generalizability of our findings across diverse populations. Furthermore, the variability inherent in the datasets utilized could introduce batch effects, potentially confounding the observed associations. The lack of clinical validation analyses further emphasizes the need for caution in interpreting the results, as real-world applicability remains to be established. Therefore, while our findings are promising, they necessitate further exploration through rigorous experimental and clinical studies.

Conclusions

In conclusion, this research elucidates the causal relationships between specific plasma proteins and POF, underscoring their potential as biomarkers and therapeutic targets. The identification of proteins such as CCL23 and TNFRSF14 provides valuable insights into the pathophysiology of POF, paving the way for future investigations aimed at enhancing diagnostic and therapeutic strategies. By addressing the multifactorial nature of POF through targeted interventions, this study contributes to the broader understanding of reproductive health and offers hope for improved patient outcomes in managing this challenging condition.

Acknowledgments

We are grateful for the valuable contributions of all participants and investigators involved in the GWAS project.

Footnote

Reporting Checklist: The authors have completed the STROBE-MR reporting checklist. Available at https://gpm.amegroups.com/article/view/10.21037/gpm-2025-1-68/rc

Peer Review File: Available at https://gpm.amegroups.com/article/view/10.21037/gpm-2025-1-68/prf

Funding: None.

Conflicts of Interest: Both authors have completed the ICMJE uniform disclosure form (available at https://gpm.amegroups.com/article/view/10.21037/gpm-2025-1-68/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Li J, Liao Q, Guo Y, et al. Mechanism of crosstalk between DNA methylation and histone acetylation and related advances in diagnosis and treatment of premature ovarian failure. Epigenetics 2025;20:2528563. [Crossref] [PubMed]
2. Rebar RW. Premature ovarian failure. Obstet Gynecol 2009;113:1355-63. [Crossref] [PubMed]
3. Komorowska B. Autoimmune premature ovarian failure. Prz Menopauzalny 2016;15:210-4. [Crossref] [PubMed]
4. Chen ZJ, Qin YY. L SJ. Etiology of primary ovarian insufficiency and premature ovarian failure. Zhonghua Fu Chan Ke Za Zhi 2008;43:897-9.
5. Jankowska K. Premature ovarian failure. Prz Menopauzalny 2017;16:51-6. [Crossref] [PubMed]
6. Lee DH, Pei CZ, Song JY, et al. Identification of serum biomarkers for premature ovarian failure. Biochim Biophys Acta Proteins Proteom 2019;1867:219-26. [Crossref] [PubMed]
7. Liu Z, Zhou Q, He L, et al. Identification of energy metabolism anomalies and serum biomarkers in the progression of premature ovarian failure via extracellular vesicles’ proteomic and metabolomic profiles. Reprod Biol Endocrinol 2024;22:104. [Crossref] [PubMed]
8. Pu X, Zhang L, Zhang P, et al. Human UC-MSC-derived exosomes facilitate ovarian renovation in rats with chemotherapy-induced premature ovarian insufficiency. Front Endocrinol (Lausanne) 2023;14:1205901. [Crossref] [PubMed]
9. Chen J, Zhou Q, Zhang Y, et al. Discovery of novel serum metabolic biomarkers in patients with polycystic ovarian syndrome and premature ovarian failure. Bioengineered 2021;12:8778-92. [Crossref] [PubMed]
10. Sun BB, Chiou J, Traylor M, et al. Plasma proteomic associations with genetics and health in the UK Biobank. Nature 2023;622:329-38. [Crossref] [PubMed]
11. Võsa U, Claringbould A, Westra HJ, et al. Large-scale cis- and trans-eQTL analyses identify thousands of genetic loci and polygenic scores that regulate blood gene expression. Nat Genet 2021;53:1300-10. [Crossref] [PubMed]
12. Orrù V, Steri M, Sidore C, et al. Complex genetic signatures in immune cells underlie autoimmunity and inform therapy. Nat Genet 2020;52:1036-45. [Crossref] [PubMed]
13. Chen Y, Lu T, Pettersson-Kymmer U, et al. Genomic atlas of the plasma metabolome prioritizes metabolites implicated in human diseases. Nat Genet 2023;55:44-53. [Crossref] [PubMed]
14. Richmond RC, Davey Smith G. Mendelian Randomization: Concepts and Scope. Cold Spring Harb Perspect Med 2022;12:a040501. [Crossref] [PubMed]
15. Pritchard JK, Przeworski M. Linkage disequilibrium in humans: models and data. Am J Hum Genet 2001;69:1-14. [Crossref] [PubMed]
16. Burgess S, Thompson SG. Interpreting findings from Mendelian randomization using the MR-Egger method. Eur J Epidemiol 2017;32:377-89. [Crossref] [PubMed]
17. Burgess S, Davey Smith G, Davies NM, et al. Guidelines for performing Mendelian randomization investigations: update for summer 2023. Wellcome Open Res 2019;4:186. [Crossref] [PubMed]
18. Hemani G, Tilling K, Davey Smith G. Orienting the causal relationship between imprecisely measured traits using GWAS summary data. PLoS Genet 2017;13:e1007081. [Crossref] [PubMed]
19. Zhu Z, Zhang F, Hu H, et al. Integration of summary data from GWAS and eQTL studies predicts complex trait gene targets. Nat Genet 2016;48:481-7. [Crossref] [PubMed]
20. Giambartolomei C, Vukcevic D, Schadt EE, et al. Bayesian test for colocalisation between pairs of genetic association studies using summary statistics. PLoS Genet 2014;10:e1004383. [Crossref] [PubMed]
21. Foley CN, Staley JR, Breen PG, et al. A fast and efficient colocalization algorithm for identifying shared genetic risk factors across multiple traits. Nat Commun 2021;12:764. [Crossref] [PubMed]
22. Carter AR, Sanderson E, Hammerton G, et al. Mendelian randomisation for mediation analysis: current methods and challenges for implementation. Eur J Epidemiol 2021;36:465-78. [Crossref] [PubMed]
23. Gagliano Taliun SA, VandeHaar P, Boughton AP, et al. Exploring and visualizing large-scale genetic associations by using PheWeb. Nat Genet 2020;52:550-2. [Crossref] [PubMed]
24. Kurki MI, Karjalainen J, Palta P, et al. FinnGen provides genetic insights from a well-phenotyped isolated population. Nature 2023;613:508-18. [Crossref] [PubMed]
25. Wang W, Li C, Chen L, et al. Identification of Therapeutic Targets for Premature Ovarian Failure Through Mendelian Randomization and Colocalization Analysis Using Human Plasma Proteomics. J Mol Neurosci 2025;75:111. [Crossref] [PubMed]
26. Pastore LM, Christianson MS, Stelling J, et al. Reproductive ovarian testing and the alphabet soup of diagnoses: DOR, POI, POF, POR, and FOR. J Assist Reprod Genet 2018;35:17-23. [Crossref] [PubMed]
27. Nelson SM, Davis SR, Kalantaridou S, et al. Anti-Müllerian hormone for the diagnosis and prediction of menopause: a systematic review. Hum Reprod Update 2023;29:327-46. [Crossref] [PubMed]
28. Jiao X, Meng T, Zhai Y, et al. Ovarian Reserve Markers in Premature Ovarian Insufficiency: Within Different Clinical Stages and Different Etiologies. Front Endocrinol (Lausanne) 2021;12:601752. [Crossref] [PubMed]
29. De Vos M, Devroey P, Fauser BC. Primary ovarian insufficiency. Lancet 2010;376:911-21. [Crossref] [PubMed]
30. Goswami R, Goswami D, Kabra M, et al. Prevalence of the triple X syndrome in phenotypically normal women with premature ovarian failure and its association with autoimmune thyroid disorders. Fertil Steril 2003;80:1052-4. [Crossref] [PubMed]
31. Kim J, Kim YS, Ko J. CK beta 8/CCL23 induces cell migration via the Gi/Go protein/PLC/PKC delta/NF-kappa B and is involved in inflammatory responses. Life Sci 2010;86:300-8. [Crossref] [PubMed]
32. Hwang J, Son KN, Kim CW, et al. Human CC chemokine CCL23, a ligand for CCR1, induces endothelial cell migration and promotes angiogenesis. Cytokine 2005;30:254-63. [Crossref] [PubMed]
33. Krishnan V, Tallapragada S, Schaar B, et al. Omental macrophages secrete chemokine ligands that promote ovarian cancer colonization of the omentum via CCR1. Commun Biol 2020;3:524. [Crossref] [PubMed]
34. Faura J, Bustamante A, Penalba A, et al. CCL23: A Chemokine Associated with Progression from Mild Cognitive Impairment to Alzheimer’s Disease. J Alzheimers Dis 2020;73:1585-95. [Crossref] [PubMed]
35. Jeong W, Bae H, Lim W, et al. Differential expression and functional roles of chemokine (C-C motif) ligand 23 and its receptor chemokine (C-C motif) receptor type 1 in the uterine endometrium during early pregnancy in pigs. Dev Comp Immunol 2017;76:316-25. [Crossref] [PubMed]
36. Huang N, Liu D, Lian Y, et al. Immunological Microenvironment Alterations in Follicles of Patients With Autoimmune Thyroiditis. Front Immunol 2021;12:770852. [Crossref] [PubMed]
37. Hoek A, van Kasteren Y, de Haan-Meulman M, et al. Dysfunction of monocytes and dendritic cells in patients with premature ovarian failure. Am J Reprod Immunol 1993;30:207-17. [Crossref] [PubMed]
38. Hoek A, Schoemaker J, Drexhage HA. Premature ovarian failure and ovarian autoimmunity. Endocr Rev 1997;18:107-34. [Crossref] [PubMed]
39. Han Y, Zou C, Liu T, et al. Inhibiting interferon-γ induced cancer intrinsic TNFRSF14 elevation restrains the malignant progression of glioblastoma. J Exp Clin Cancer Res 2024;43:212. [Crossref] [PubMed]
40. Shui JW, Steinberg MW, Kronenberg M. Regulation of inflammation, autoimmunity, and infection immunity by HVEM-BTLA signaling. J Leukoc Biol 2011;89:517-23. [Crossref] [PubMed]
41. Kim HM, Jeong CS, Choi HS, et al. LIGHT/TNFSF14 enhances adipose tissue inflammatory responses through its interaction with HVEM. FEBS Lett 2011;585:579-84. [Crossref] [PubMed]
42. Steinberg MW, Cheung TC, Ware CF. The signaling networks of the herpesvirus entry mediator (TNFRSF14) in immune regulation. Immunol Rev 2011;244:169-87. [Crossref] [PubMed]
43. Ware CF, Sedý JR. TNF Superfamily Networks: bidirectional and interference pathways of the herpesvirus entry mediator (TNFSF14). Curr Opin Immunol 2011;23:627-31. [Crossref] [PubMed]
44. Choi EK, Kim WK, Sul OJ, et al. TNFRSF14 deficiency protects against ovariectomy-induced adipose tissue inflammation. J Endocrinol 2014;220:25-33. [Crossref] [PubMed]