Causal Interplay Between Inflammatory Cytokines and Lipid Metabolites in Serous Ovarian Carcinoma: Insights From a Genetic Association Study.

Chong‐ze Yang and Lan‐hui Qin designed the study and drafted the manuscript. Chong‐ze Yang, Lan‐hui Qin, and Wei Lan performed data analysis. Cheng‐can Huang, Xiao‐li Huang, and Ying‐xia Yang prepared the figures. Miao Fan critically revised the manuscript. All authors read and approved the final version of the manuscript.

The authors have nothing to report.

This work was supported by National Natural Science Foundation of China, U24A20256.

In this study, we employed a two‐step MR approach to investigate the potential causal associations between inflammatory cytokines and SOC, and to further assess whether these associations are mediated by blood metabolites. In the first step, a two‐sample MR analysis was conducted to evaluate the causal relationships of inflammatory cytokines and blood metabolites with SOC, thereby identifying candidate exposures associated with SOC risk. In the second step, we examined the potential causal relationships between the selected inflammatory cytokines and metabolites, and subsequently quantified the mediating effects of metabolites on the inflammatory cytokines–SOC pathway (Figure  1 ). This study was reported in accordance with the STROBE‐MR (Strengthening the Reporting of Observational Studies in Epidemiology using Mendelian Randomization) guidelines (Data  S3 ). The flowchart of this study. The primary analytic strategy of this study was two‐sample MR. MR analysis relies on three core assumptions to ensure unbiased causal inference: (1) the IVs are strongly associated with the exposure of interest; (2) the IVs are independent of potential confounders; and (3) the IVs influence the outcome exclusively through the exposure, rather than through alternative pathways. All data used in this study were obtained from publicly available datasets; therefore, no additional ethical approval or informed consent was required. Genome‐wide association study (GWAS) summary data for ovarian cancer were obtained from the FinnGen consortium ( https://www.finngen.fi/en/access_results ). The dataset included 1025 cases of ovarian cancer, of which 852 were SOC, and 167,189 controls. For the present analysis, we specifically utilized the GWAS summary statistics derived from the case–control comparison of SOC (852 cases versus 167,189 controls) to perform MR analyses. The GWAS summary statistics for 91 inflammatory cytokines (GCST90274758–GCST90274848) [ 17 ] and for 1400 blood metabolites and metabolite ratios (GCST90199621–GCST902010209) [ 18 ] were downloaded from the NHGRI‐EBI GWAS Catalog ( https://www.ebi.ac.uk/gwas/ ). Single nucleotide polymorphisms (SNPs) significantly associated with the exposures of interest were selected as IVs, using a threshold of p  < 5 × 10 −5 . Several quality control procedures were applied to ensure the validity of IVs. First, linkage disequilibrium (LD) pruning was performed with a threshold of r 2  < 0.001 and a clumping window size of 10,000 kb. Second, the F‐statistic for each SNP was calculated using the formula F = β 2 /SE 2 . SNPs with an F‐statistic ≥ 10 were considered sufficiently strong instruments, whereas those with F < 10 were regarded as weak instruments and excluded from subsequent analyses. A two‐step MR analysis was conducted to estimate the potential causal relationships of the exposure (inflammatory cytokines) and the mediator (metabolites) with the outcome (SOC). This approach was applied to assess the extent to which metabolites mediated the exposure–outcome relationship. Specifically, the causal effect of the exposure on the mediator (β₁) was first estimated, followed by the effect of the mediator on the outcome (β₂). The mediation effect was then calculated as β₁ × β₂, and the proportion of the mediation effect relative to the total effect was quantified using the delta method (Figure  1 ). MR analyses were performed using five methods: inverse‐variance weighted (IVW), MR‐Egger, weighted median, simple mode, and weighted mode. Among these, IVW was considered the primary analytic approach; therefore, p ‐values derived from the IVW method were regarded as the main results, with the other four methods serving as sensitivity analyses. Results were reported as odds ratios (ORs) with corresponding 95% confidence intervals (CIs), and a p  < 0.05 was considered statistically significant. To ensure the causal direction was from exposure to outcome, reverse MR analyses were additionally conducted. Only associations with significant p ‐values in the forward MR analysis and nonsignificant p ‐values in the reverse MR analysis were considered robust; results failing to meet these criteria were excluded from further interpretation. To further ensure the robustness of causal inferences in the two‐sample MR framework, we conducted a series of sensitivity analyses. Heterogeneity across SNP‐specific causal estimates was assessed using Cochran's Q test, with a p  < 0.05 indicating significant heterogeneity. Horizontal pleiotropy was evaluated using the MR‐Egger regression intercept and the MR‐PRESSO global test; associations were considered to exhibit horizontal pleiotropy if either test yielded a p  < 0.05. Causal relationships with evidence of horizontal pleiotropy were deemed unreliable and excluded from further interpretation. After testing for pleiotropy and heterogeneity, we performed leave‐one‐out analyses for eligible exposures. In this approach, each SNP was sequentially excluded, and the combined effect of the remaining SNPs was recalculated. If the overall confidence interval showed no substantial changes after removing individual SNPs, the result was considered robust. All statistical analyses were performed using R software (version 4.2.1), with the TwoSampleMR package applied for MR analyses.

Through MR analysis, we identified five inflammatory cytokines and 68 blood metabolites with potential causal associations with SOC. Mediation MR analysis further revealed that 1‐palmitoyl‐GPG (16:0) mediated the effect of IL‐8 on SOC (Data  S1 and S2 ). IVW analyses identified five inflammatory cytokines with significant causal associations with SOC, with the number of SNPs used in the MR analyses ranging from 8 to 19. All five inflammatory cytokines were identified as risk factors for SOC. The IVW estimates were as follows: CSF1 (OR = 1.688, 95% CI: 1.174 to 2.426; F‐statistics: 20.88 to 203.67), CXCL1 (OR = 1.396, 95% CI: 1.012 to 1.927; F‐statistics: 20.89 to 547.00), IL‐20 (OR = 1.862, 95% CI: 1.034 to 3.353; F‐statistics: 20.97 to 27.03), IL‐8 (OR = 1.610, 95% CI: 1.082 to 2.394; F‐statistics: 21.10 to 61.76), and VEGF‐A (OR = 1.239, 95% CI: 1.000 to 1.535; F‐statistics: 21.08 to 943.51) (Figures  2 and 3 ). All F‐statistics significantly exceeded the conventional threshold of 10, indicating the absence of weak instrument bias. Furthermore, post hoc power calculations revealed varying levels of statistical power for these identified factors. CSF1 demonstrated adequate statistical power (80.8%, with a minimum detectable OR of 1.68 required to achieve 80% power). IL‐8 and IL‐20 achieved robust power approaching the standard threshold (73.9% and 73.1%, respectively). In contrast, due to their more modest effect sizes and the limited sample size of SOC cases, the analyses for CXCL1 and VEGF‐A were relatively underpowered (30.5% and 29.4%, respectively), suggesting that these specific associations should be interpreted with caution. Forest plot of the causal relationship between circulating inflammatory proteins and SOC. IVW: Inverse variance weighted; nSNP: Number of single nucleotide polymorphisms; OR: Odds ratio. (A) Volcano diagram: Inflammatory cytokines with potential causal relationship with SOC. (B–F) Scatter plots for the causal association between CSF1, CXCL1, IL20, IL8, VEGF‐A, and SOC. MR analyses identified a total of 68 blood metabolites with significant causal associations with SOC, with the number of SNPs used in the analyses ranging from 14 to 46 (F‐statistics: 19.503 to 2297.785). Among these, 39 metabolites were positively associated with SOC risk, while 29 metabolites were negatively associated (Data  S2 ). In this study, MR analyses identified five inflammatory cytokines and 68 blood metabolites with potential causal associations with SOC. We subsequently investigated the causal relationships between these five inflammatory cytokines and the 68 metabolites. IVW analysis revealed suggestive evidence for a positive association between IL‐8 and 1‐palmitoyl‐GPG (16:0) (OR = 1.140, 95% CI: 1.001–1.297), based on 16 SNPs used in the MR analysis (Figure  4 ). Forest plot of the causal relationship between IL‐8, 1‐Palmitoyl‐GPG (16:0), and SOC. IVW: Inverse variance weighted; nSNP: Number of single nucleotide polymorphisms; OR: Odds ratio. Mediation MR analysis revealed that 1‐palmitoyl‐GPG (16:0) potential partially mediated the effect of IL‐8 on SOC. Specifically, IL‐8 exhibited a positive causal association with SOC (β = 0.4760, 95% CI: 0.0791–0.8729), and 1‐palmitoyl‐GPG (16:0) was similarly associated with an increased risk of SOC (β = 0.3633, 95% CI: 0.1475–0.5791). Furthermore, a positive association was observed between IL‐8 and 1‐palmitoyl‐GPG (16:0) levels (β = 0.1307, 95% CI: 0.0015–0.2599). The indirect effect mediated by 1‐palmitoyl‐GPG (16:0) was estimated at 0.0475 (95% CI: −0.0073 to 0.1022), accounting for 9.98% of the total effect (Figures  1 and 4 ). Cochran's Q test indicated no significant heterogeneity among the selected SNPs. Neither the MR‐Egger regression nor the MR‐PRESSO global test detected evidence of horizontal pleiotropy. Furthermore, leave‐one‐out analysis demonstrated that the observed causal relationships were not driven by any single SNP.

Ovarian cancer remains a leading cause of tumor‐related mortality among women worldwide. Among its histological subtypes, serous ovarian carcinoma (SOC)—particularly high‐grade serous ovarian carcinoma (HGSOC)—is the most aggressive form and is associated with the poorest prognosis [ 1 , 2 ]. Due to the absence of specific symptoms in the early stages and the lack of effective population‐based screening strategies, the majority of patients are diagnosed at an advanced stage [ 3 ]. Moreover, the high recurrence rate and the development of acquired chemoresistance further contribute to the unfavorable overall survival outcomes. Previous studies have suggested that the occurrence of SOC is associated with a family history of endometriosis, environmental exposures, and adverse lifestyle factors [ 4 ]. However, the precise pathogenesis of SOC remains largely unclear. In the current era of precision and personalized medicine, identifying actionable causal targets and translatable biomarkers from etiological and pathway perspectives holds considerable scientific value and clinical significance. Chronic inflammatory microenvironments are widely recognized as major drivers of disease initiation and progression, with inflammatory cytokines and chemokines playing pivotal regulatory roles within these processes [ 5 ]. Multiple mediators contribute to this pathological state by promoting angiogenesis, inducing epithelial–mesenchymal transition, reprogramming myeloid cells, and recruiting immunosuppressive cell populations, thereby sustaining a “pro‐tumorigenic inflammation–immune imbalance” that facilitates disease development [ 6 , 7 ]. Importantly, inflammation‐related signaling pathways do not operate in isolation or arise from a single causal event; rather, they are tightly coupled and interact dynamically with metabolic networks [ 8 , 9 ]. Parallel to inflammation, metabolic reprogramming is also recognized as a hallmark of cancer [ 10 , 11 ]. Blood metabolites not only reflect the status of energy metabolism and biosynthetic pathways but may also serve as critical bridges linking chronic inflammation to malignant transformation [ 12 ]. Observational studies have provided foundational evidence linking specific metabolomic profiles to ovarian cancer. For instance, using a two‐step metabolomics strategy, Chen et al. identified specific serum metabolites, most notably 27‐nor‐5β‐cholestane‐3,7,12,24,25 pentol glucuronide (CPG), as potential diagnostic biomarkers for epithelial ovarian cancer, highlighting the presence of metabolic disruptions even at early disease stages [ 13 ]. Furthermore, a 23‐year prospective study by Zeleznik et al. demonstrated that elevated pre‐diagnostic circulating levels of specific lipid groups, particularly sphingomyelins (SMs), are prospectively associated with an increased risk of ovarian cancer [ 14 ]. However, these findings are inherently limited by reverse causation and confounding biases, making it difficult to establish causal inference. Consequently, distinguishing between concomitant alterations and true causal relationships, and quantifying their mediating contributions to disease mechanisms, remains a major challenge in aetiological research and therapeutic target discovery. Because of these limitations in traditional observational designs, determining whether inflammatory and metabolic alterations are causes, consequences, or mere bystanders of SOC requires alternative methodological approaches. While randomized controlled trials (RCTs) are the gold standard for establishing causality, they often face ethical and feasibility constraints when addressing etiological questions. Mendelian randomization (MR), which employs genetic variants associated with exposures as instrumental variables (IVs), provides a robust strategy to mitigate confounding and reverse causation, thereby offering a potential solution to these limitations. Based on the principles of Mendel's law of segregation and independent assortment, genetic variants are randomly allocated during gamete formation and are fixed at conception. These genetic assignments precede disease onset and are generally independent of self‐selected environmental or lifestyle factors; MR is highly resilient to reverse causation and unmeasured confounding [ 15 , 16 ]. Furthermore, two‐step mediation MR allows the dissection of the sequential pathway from exposure to mediator to outcome, enabling quantification of the mediator's causal contribution and elevating associations to causal evidence. With the increasing availability of large‐scale, publicly accessible resources such as FinnGen and the NHGRI‐EBI GWAS catalog, the feasibility and reproducibility of these strategies are now greatly enhanced. Building on these considerations, the present study was designed to test two hypotheses: (1) inflammatory cytokines exert a causal effect on the risk of SOC; and (2) specific blood metabolites mediate the relationship between inflammatory cytokines and SOC. To this end, we applied a two‐sample MR framework to investigate the potential causal associations between inflammatory cytokines and SOC, followed by a two‐step mediation MR approach to evaluate whether particular metabolites serve as intermediates in this pathway. Our goal was to further elucidate the mechanistic axis linking inflammation, metabolism, and SOC, thereby providing novel, testable clues for mechanistic understanding, risk stratification, and biomarker development, while laying the foundation for subsequent functional validation and translational research.

In this study, we applied bidirectional MR to screen 91 inflammatory cytokines and identified five with suggestive causal associations with SOC. Furthermore, we employed two‐step mediation MR to examine whether blood metabolites mediated the effects of inflammatory cytokines on SOC. The results demonstrated that CSF1, CXCL1, IL‐20, IL‐8, and VEGF‐A were suggestive risk factors for SOC. Mediation MR analysis further indicated that the lipid metabolite 1‐palmitoyl‐GPG (16:0) partially mediated the effect of IL‐8 on SOC. Multiple sensitivity tests confirmed the robustness of these findings. Collectively, this biomarker‐oriented approach provides novel and reliable insights into the etiology of SOC. To further elucidate the underlying mechanisms, we discuss the implications of these findings in the following sections. Chronic inflammation and immune imbalance constitute a central background for the initiation and progression of SOC [ 19 ]. CXCL1, IL‐8, and IL‐20 belong to the chemokine family, which sustain and amplify inflammatory responses by binding to their respective receptors, CXCR1/2 or IL‐20R. Persistently elevated proinflammatory signaling leads to continuous activation of pathways such as NF‐κB and STAT3 within the ovarian cancer microenvironment, thereby shortening the G1 phase, accelerating DNA synthesis and cell division, promoting rapid tumor growth, and increasing genomic instability [ 20 , 21 ]. CXCL1, through CXCR2, recruits immune cells and reprograms them toward immunosuppressive phenotypes, resulting in Arg‐1/iNOS/ROS‐mediated inhibition of CD8 + T cells and the expansion of Tregs [ 22 , 23 ]. Elevated IL‐8 is also closely associated with neutrophil infiltration; high neutrophil levels, partly activated through Jagged2 signaling, impair CD8 + T‐cell cytotoxicity, further exacerbating chronic inflammation and immune evasion [ 24 ]. Clinically, serum CXCL1 levels are higher in patients with advanced ovarian cancer compared with those in early‐stage disease [ 25 ]; patients with elevated IL‐8 are more prone to metastasis and experience poorer overall survival; and animal studies have shown that blockade of IL‐8 signaling can reverse resistance to cisplatin and paclitaxel [ 26 ]. As a proinflammatory member of the IL‐10 family, IL‐20 not only contributes to the chronic inflammatory milieu but also promotes epithelial–mesenchymal transition upon receptor complex binding, thereby enhancing cell migration and invasion [ 27 , 28 ]. Collectively, these findings suggest that inflammatory chemokines drive the disease course and malignant progression of SOC through chronic inflammation and its associated mechanisms. The polarization of tumor‐associated macrophages (TAMs) represents another key mechanism in the progression of SOC. CSF1, a classical macrophage colony‐stimulating factor, binds to CSF1R and activates downstream PI3K–AKT, MAPK/ERK, and JAK/STAT3 signaling pathways, thereby promoting monocyte recruitment and differentiation into immunosuppressive M2‐type TAMs [ 29 ]. M2‐polarized TAMs secrete effector molecules such as VEGF and IL‐10, which act synergistically to promote angiogenesis and stromal remodeling. In addition, they impair antigen presentation and upregulate PD‐L1 expression, further weakening antitumor immune responses and fostering an immunosuppressive microenvironment that facilitates tumor progression and metastasis [ 30 , 31 ]. Notably, CSF1 signaling complements the CXCL1–IL‐8 axis in immune cell recruitment by mobilizing myeloid‐derived suppressor cells (MDSCs). Together with TAMs, this forms a cooperative network that enhances tumor cell invasion, adhesion, and immune evasion [ 32 ]. Angiogenesis and hypoxia‐driven microenvironmental remodeling are hallmarks of tumor progression. VEGF‐A, a classical angiogenic factor, promotes endothelial cell proliferation, migration, and vascular permeability, thereby driving neovascularization in tumors [ 33 ]. VEGF‐A does not act in isolation; recent studies have shown that IL‐8 and IL‐20 can stimulate VEGF expression, exerting strong synergistic effects with VEGF‐A in angiogenesis. On the one hand, VEGF‐A activates the PI3K/Akt/mTOR and MAPK/ERK pathways to enhance endothelial cell migration and proliferation, accelerating vascular network formation. On the other hand, it promotes cytoskeletal reorganization and vascular permeability through the PLCγ/PKC–Src/FAK axis. Together, these mechanisms reinforce neovascular network construction, providing favorable conditions for tumor initiation and progression [ 34 , 35 ]. The mismatch between rapid neovascularization and high metabolic demand creates a hypoxic tumor microenvironment, under which HIF‐1α becomes stabilized and binds to the VEGF‐A promoter, establishing a vicious cycle of hypoxia‐induced angiogenesis [ 36 , 37 ]. Moreover, ovarian cancer cells themselves can express VEGFR‐2, forming an autocrine VEGF‐A/VEGFR‐2 signaling loop that further enhances invasive capacity [ 38 , 39 ]. Clinically, high VEGF expression in serum or tumor tissue is closely associated with aggressiveness and metastatic potential. Patients with elevated VEGF levels typically exhibit shorter progression‐free intervals and overall survival, particularly in those with residual disease after surgery [ 40 ]. These findings indicate that VEGF not only functions as a key driver of angiogenesis and tumor progression but also serves as an important biomarker for prognostic stratification. Blood metabolites may mediate the effects of inflammatory cytokines on SOC. Our two‐step mediation MR analysis suggested that the lipid metabolite 1‐palmitoyl‐GPG (16:0) potentially mediated the effect of IL‐8 on SOC. Given that this mediation effect reached nominal significance but was not robust to multiple testing correction, it should be interpreted as a hypothesis‐generating finding. This finding indicates that the pro‐tumorigenic effect of IL‐8 may not solely depend on chronic inflammation and immune evasion but may also be linked to lipid metabolic pathways. However, current evidence specifically addressing this metabolite remains limited. Abnormal abundance of 1‐palmitoyl‐GPG (16:0) reflects lipid metabolic reprogramming in ovarian cancer cells, with phospholipase A2 (PLA2) activity serving as a major pathway connecting 1‐palmitoyl‐GPG (16:0) to IL‐8 signaling [ 41 , 42 ]. PLA2 generates various lysophospholipids, including 1‐palmitoyl‐GPG (16:0), which potentially amplify and sustain IL‐8 signaling in SOC, forming a metabolism–inflammation cascade [ 43 , 44 ]. Similarly, a recent MR study reported that 1‐palmitoyl‐GPG (16:0) acts as a regulatory mediator of immune cell function in the pathogenesis of diabetes [ 45 ]. Collectively, our findings provide evidence that inflammatory cytokines may exert causal roles in the initiation and progression of SOC, with blood metabolites mediating part of these effects. This discovery offers novel molecular insights into the etiology of SOC, supporting the identification and development of more targeted biomarkers, and laying the foundation for new therapeutic strategies directed at inflammation, immunity, metabolism, and angiogenesis, thereby advancing personalized and precision medicine for SOC. Nevertheless, several limitations should be acknowledged. First, the datasets employed in this study were primarily derived from European populations; thus, the generalizability of our results to other ancestries requires further validation. Second, the number of SNPs available for MR analyses remained relatively limited, and the datasets did not comprehensively cover all inflammatory cytokines or metabolites, which may have led to underestimation or omission of relevant mediation effects. Third, to ensure an adequate number of SNPs for these exposures, a relatively lenient significance threshold was utilized for IVs selection. While necessary, this relaxed threshold inherently carries a potential risk of introducing horizontal pleiotropy or weak instrument bias. Post hoc power calculations for the identified cytokines revealed a statistical power range of 29.4% to 80.8%, indicating that the analysis for factors with modest effect sizes may be underpowered and susceptible to false‐negative findings. Future studies should aim to replicate these findings in larger, multi‐ancestry cohorts and leverage integrative multi‐omics approaches, including single‐cell sequencing, to dissect the dynamic remodeling of the SOC tumor microenvironment. Such efforts will be critical for exploring the translational potential of these mechanisms in early diagnosis, risk stratification, and therapeutic decision‐making.

The authors have nothing to report.

The authors declare no conflicts of interest.

Data S1: Detailed datasets and Mendelian randomization (MR) analysis results concerning the causal association between circulating inflammatory cytokines and SOC. This file includes the primary MR results, assessments of heterogeneity and horizontal pleiotropy, MR‐PRESSO analysis, and the specific single nucleotide polymorphisms (SNPs) utilized in the analysis. Data S2: Detailed datasets and Mendelian randomization (MR) analysis results concerning the causal association between metabolites and SOC. This file includes the primary MR results, assessments of heterogeneity and horizontal pleiotropy, MR‐PRESSO analysis, and the specific single nucleotide polymorphisms (SNPs) utilized in the analysis. Data S3: The STROBE‐MR checklist. This document outlines the reporting standards and guidelines followed in this study to ensure transparency and reproducibility.