Evaluating the association between lipidome and female reproductive diseases through comprehensive Mendelian randomization analyses

Genome-wide association study (GWAS) summary statistics for the lipidome were obtained from the GWAS Catalog (GCST90277238- GCST90277416). The study encompassed 7,174 unrelated Finnish individuals from the GeneRISK cohort. SNPs were evaluated across 179 lipid species, categorized into 13 lipid classes: cholesterol ester (CE), ceramide (Cer), cholesterol (Chol), diacylglycerol (DAG), lysophosphatidylcholine (LPC), lysophosphatidylethanolamine (LPE), phosphatidylcholine (PC), phosphatidylcholine-ether (PCO), phosphatidylethanolamine (PE), phosphatidylethanolamine-ether (PEO), phosphatidylinositol (PI), sphingomyelin (SM), triacylglycerol (TAG), as well as 4 categories: sterols (ST), sphingolipids (SL), glycerolipids (GL) and glycerophospholipids (GP) 6 . The lipid molecules were identified at the subtype levels using the SwissLipids database ( http://www.swisslipids.org ) 22 . Additional details regarding the 179 lipid species can be found in Additional file 1-sTable 1 . GWAS data for six common FRDs, including PCOS, endometriosis, female infertility, uterine fibroid, and uterine cancer, were sourced from the GWAS Catalog. Additional GWAS data specifically for ovarian cancer were obtained from the IEU OpenGWAS project. Participants in these studies were predominantly of European ancestry. To prevent sample overlap between exposures and outcomes, GWAS data for the outcomes were collected from individuals across different batches than those used for exposures 23 . Further details regarding the characteristics of the GWAS studies utilized can be found in Additional file 1-sTable 2 . IVs associated with the lipidomes were selected based on a GWAS threshold of P  < 5 × 10 − 6 and subjected to linkage disequilibrium clumping with an r 2  < 0.001 and clumping distance of 10,000 kb. To mitigate weak-instrument bias in MR analysis, the strength of SNP-exposure associations was assessed using the F statistic, calculated as previous established methods 24 . SNPs with F values below ten were considered weak instruments and excluded from subsequent MR analysis 25 . To minimize the influence of confounding factors, candidate SNPs associated with exposures were cross-referenced with the IEU OpenGWAS project ( https://gwas.mrcieu.ac.uk/ ) to exclude SNPs directly linked to the outcomes. Additionally, SNPs exhibiting a palindromic strand and outliers identified by the MR pleiotropy residual sum and outlier (MR-PRESSO) test were removed prior to further MR analysis 26 . Initially, a univariable TSMR analysis was conducted to explore the relationship between the lipidome and FRDs in this study 27 . To avoid confounding and bias, SNPs associated with outcomes were filtered using a threshold of 5 × 10 -8 28 , followed by harmonization to correct for strand mismatches and ensure consistent alignment of effect sizes. Subsequently, the impact of the lipidome on FRDs was assessed using TSMR through several methods including inverse-variance weighted (IVW), MR-Egger, weighted median (WM), simple mode, and weighted mode approaches. In IVW analysis, each variant was weighted based on its precision to reduce overall variance effectively, which enhances accuracy and efficiency in MR analysis 29 . To ensure robust and reliable results, IVW was primarily used for interpretation, with MR-Egger, WM, simple mode, and weighted mode results providing supplementary insights to support IVW estimates. Results were deemed significant if IVW yielded P values less than 0.05 and MR-Egger, WM, simple mode, and weighted mode estimates were consistent in direction with IVW. TSMR analysis was performed using R (version 3.6.3) and the “Two Sample MR” package (version 0.5.7). False discovery rate (FDR) correction was applied to control for false positives arising from multiple testing, defining statistical significance as FDR-adjusted P values less than 0.05. To address potential biases arising from SNPs associated with multiple lipid types, MVMR analysis was employed to estimate the direct effects of lipidome on FRDs 27 . In this multivariable approach, IVW, MR-Egger, MR-Lasso, and WM methods were utilized to assess the causal relationships between lipid exposures and FRDs 30 . IVW served as the primary method of analysis, supported by MR-Egger, MR-Lasso, and WM as supplementary approaches 31 . Significant results in IVW ( P  < 0.05) further substantiated the potential causal associations identified between exposures and outcomes in MVMR analysis. The analysis was conducted using the “MendelianRandomization” (version 0.8.0) and “TwoSampleMR” (version 0.5.7) packages in R. To mitigate the impact of heterogeneity and pleiotropy in MR analysis, several sensitivity tests were conducted 32 . Heterogeneity was assessed using Cochran’s Q test, with significance set at P  < 0.05. In cases where heterogeneity was detected, multiplicative random-effects IVW was employed for MR analysis. Funnel plots were generated to examine potential directional pleiotropy, and MR-Egger intercept tests were utilized to evaluate horizontal pleiotropy. Additionally, leave-one-out analyses were performed to assess whether the IVW estimate was driven by any single SNP. These sensitivity analyses were crucial in ensuring the robustness and reliability of the MR results by addressing potential sources of bias and confounding in the genetic instruments used 32 . To further investigate the direction of causality between the lipidome and FRDs, reverse MR and Steiger directionality tests were conducted. Reverse MR analysis utilized FRDs-associated SNPs as IVs to determine if FRDs causally influenced the lipidome identified in previous analyses. IVs for FRDs were selected using a threshold of P  < 5 × 10 -6 , with linkage disequilibrium clumping at r 2  < 0.001 and a clumping distance of 10,000 kb. Additionally, MR-Steiger tests were employed to assess the directionality and potential biases in the causal relationships inferred from MR analyses. These complementary analyses provided further insights into the causal pathways between lipidome traits and FRDs, enhancing the robustness of the study findings. To investigate the genetic mechanisms linking lipidome with FRDs, we conducted a 2-step MR analysis to explore potential mediators influencing these associations. In the first step of MR analysis, lipidome traits were used as exposures and several potential mediators including BMI, vitamin D deficiency, alcohol use, smoking, physical activity, obesity, visceral adipose tissue mass, waist-hip ratio, BMI adjusted waist-hip ratio, diabetes, depression, inflammation cytokines, hormone-binding globulin, and testosterone were examined as outcomes 33 – 37 . In the second step, these potential mediators were subsequently analyzed as exposures, with various FRDs as outcomes. Detailed information on each GWAS summary data source for these potential mediators was provided in Additional file 1-sTable 2 . The primary analysis utilized the IVW method, with statistical significance defined as P  < 0.05. By multiplying the beta values obtained from the two-step MR, we quantified the portion of the total effect of the lipidomes on the FRDs that is mediated by the mediators. The statistical power was then evaluated by calculating the confidence intervals for the mediated effect using the delta method 38 . Through mediation MR analysis, we aimed to delineate both direct and indirect causal effects, thereby elucidating the underlying mechanisms involved in the development of FRDs. These comprehensive analyses aimed to provide insights into the genetic pathways linking lipidome profiles with female reproductive health outcomes.

The causal effects of lipidome on FRDs were assessed using TSMR and subsequently validated with FDR adjustment and MVMR methods. The F-statistic values for selected SNPs were all above 10. In the initial TSMR analysis, 56 lipids across four categories (ST, SL, GL, GP) were found to be associated with FRDs, suggesting their potential impact on the female reproductive system. Results from both TSMR and FDR-adjusted analyses generally exhibited consistency, with significant causal associations persisting after FDR adjustment. Subsequent MVMR analysis refined these associations, revealing that six lipid species remained associated with FRDs across four classes (SM, CE, LPE, PC) from three categories (ST, GP, SL). Detailed results from all analyses methods were provided in Additional file 2 . As depicted in Fig.  1 , phosphatidylcholine (O-16:0_18:1) and phosphatidylinositol (16:0_20:4) show positive associations with PCOS, indicating an increased risk for developing PCOS. Conversely, the presence of sphingomyelin (d38:2) is negatively associated with PCOS, suggesting a protective effect against the development of PCOS. Following further analysis using the MVMR method, the effect of genetically predicted sphingomyelin (d38:2) (Estimate = -0.104, 95% CI: -0.199 ~ -0.010, P  = 0.031) on PCOS remained statistically significant. Fig. 1 Forest plot and results of TSMR and MVMR analysis from the effects of lipidomes on PCOS. GP, glycerophospholipids; PCO, phosphatidylcholine-ether; PI, phosphatidylinositol; SL, sphingolipids; SM, sphingomyelin; IVW, inverse-variance weighted; WM, weighted median; OR, odds ratio; CI, confidence interval; Q values, FDR-adjusted P values; MVMR, multivariable Mendelian randomization. Forest plot and results of TSMR and MVMR analysis from the effects of lipidomes on PCOS. GP, glycerophospholipids; PCO, phosphatidylcholine-ether; PI, phosphatidylinositol; SL, sphingolipids; SM, sphingomyelin; IVW, inverse-variance weighted; WM, weighted median; OR, odds ratio; CI, confidence interval; Q values, FDR-adjusted P values; MVMR, multivariable Mendelian randomization. In the context of endometriosis, TSMR analysis identified 11 lipids, including glycerophospholipids and glycerolipids, that potentially influence the condition (P ivw < 0.05). Detailed findings were presented in Fig.  2 . Specifically, phosphatidylinositol (16:0_20:4) was found to be potentially negatively associated with endometriosis (β = -0.181, 95% CI = 0.698 ~ 0.999, P  = 0.048), indicating a protective effect. Conversely, several triacylglycerol structures showed risk effects on endometriosis. Additionally, certain forms of phosphatidylethanolamine and phosphatidylcholine exhibited inconsistent effects, with different structures within these lipid classes demonstrating varying impacts on endometriosis. Following MVMR correction, the IVW MR estimation revealed a significant positive correlation between phosphatidylethanolamine (18:2_0:0) and endometriosis (Estimate = 0.183, 95% CI = 0.015 to 0.350, P  = 0.033), indicating a potential risk effect. Fig. 2 Forest plot and results of TSMR and MVMR analysis from the effects of lipidomes on endometriosis. GP, glycerophospholipids; PC, phosphatidylcholine; LPE, lysophosphatidylethanolamine; PCO, phosphatidylcholine-ether; GL, glycerolipids; PE, phosphatidylethanolamine; PI, phosphatidylinositol; TAG, triacylglycerol; IVW, inverse-variance weighted; WM, weighted median; OR, odds ratio; CI, confidence interval; Q values, FDR-adjusted P values; MVMR, multivariable Mendelian randomization. Forest plot and results of TSMR and MVMR analysis from the effects of lipidomes on endometriosis. GP, glycerophospholipids; PC, phosphatidylcholine; LPE, lysophosphatidylethanolamine; PCO, phosphatidylcholine-ether; GL, glycerolipids; PE, phosphatidylethanolamine; PI, phosphatidylinositol; TAG, triacylglycerol; IVW, inverse-variance weighted; WM, weighted median; OR, odds ratio; CI, confidence interval; Q values, FDR-adjusted P values; MVMR, multivariable Mendelian randomization. As presented in Fig.  3 , 33 lipids spanning all four lipid categories were associated with uterine fibroid. Ceramide (d40:1), diacylglycerol (18:1_18:2), and several types of triacylglycerol were identified as risk factors for uterine fibroid. Different structures of sterol ester, phosphatidylethanolamine, and phosphatidylcholine exhibited varied effects on uterine fibroid. Higher levels of sterol ester (27:1/18:1) and lower levels of sterol ester (27:1/20:4) were found to be associated with uterine fibroid risk. Following MVMR analysis, sterol ester (27:1/18:1) remained a significant risk factor for uterine fibroid. Phosphatidylcholine (16:1_18:1) showed inconsistent results between TSMR and MVMR analyses, with MVMR indicating an opposite direction of effect. However, the protective effects of phosphatidylcholine (18:0_22:6) on uterine fibroid remained significant after MVMR analysis. These findings underscored the complex associations between specific lipid structures and uterine fibroid, highlighting their potential roles in the pathogenesis of this condition. Fig. 3 Forest plot and results of TSMR and MVMR analysis from the effects of lipidomes on uterine fibroid. ST, sterols; CE, cholesterol ester; SL, sphingolipids; Cer, ceramide; GL, glycerolipids; DAG, diacylglycerol; GP, glycerophospholipids; LPC, lysophosphatidylcholine; LPE, lysophosphatidylethanolamine; PC, phosphatidylcholine; PCO, phosphatidylcholine-ether; PE, phosphatidylethanolamine; PEO, phosphatidylethanolamine-ether; TAG, triacylglycerol; IVW, inverse-variance weighted; OR, odds ratio; CI, confidence interval; Q values, FDR-adjusted P values; MVMR, multivariable Mendelian randomization. Forest plot and results of TSMR and MVMR analysis from the effects of lipidomes on uterine fibroid. ST, sterols; CE, cholesterol ester; SL, sphingolipids; Cer, ceramide; GL, glycerolipids; DAG, diacylglycerol; GP, glycerophospholipids; LPC, lysophosphatidylcholine; LPE, lysophosphatidylethanolamine; PC, phosphatidylcholine; PCO, phosphatidylcholine-ether; PE, phosphatidylethanolamine; PEO, phosphatidylethanolamine-ether; TAG, triacylglycerol; IVW, inverse-variance weighted; OR, odds ratio; CI, confidence interval; Q values, FDR-adjusted P values; MVMR, multivariable Mendelian randomization. As for female infertility, TSMR analysis identified 10 lipids associated with infertility, with sterol ester, phosphatidylethanolamine, and triacylglycerol showing protective effects, and diacylglycerol exhibiting risk effects. Various phosphatidylcholine structures displayed inconsistent effects on infertility, with their impacts varying depending on specific structures. However, MVMR analysis did not reveal any significant causal relationship between lipidome and infertility. Detailed findings were presented in Fig.  4 . Fig. 4 Forest plot and results of TSMR and MVMR analysis from the effects of lipidomes on female infertility. ST, sterols; CE, cholesterol ester; GL, glycerolipids; DAG, diacylglycerol; GP, glycerophospholipids; LPC, lysophosphatidylcholine; PC, phosphatidylcholine; PCO, phosphatidylcholine-ether; PEO, phosphatidylethanolamine-ether; TAG, triacylglycerol; IVW, inverse-variance weighted; WM, weighted median; OR, odds ratio; CI, confidence interval; Q values, FDR-adjusted P values; MVMR, multivariable Mendelian randomization. Forest plot and results of TSMR and MVMR analysis from the effects of lipidomes on female infertility. ST, sterols; CE, cholesterol ester; GL, glycerolipids; DAG, diacylglycerol; GP, glycerophospholipids; LPC, lysophosphatidylcholine; PC, phosphatidylcholine; PCO, phosphatidylcholine-ether; PEO, phosphatidylethanolamine-ether; TAG, triacylglycerol; IVW, inverse-variance weighted; WM, weighted median; OR, odds ratio; CI, confidence interval; Q values, FDR-adjusted P values; MVMR, multivariable Mendelian randomization. In Fig.  5 , sterol ester (27:1/20:3) was found to be protective, while phosphatidylcholine (16:0_22:6) was identified as a risk factor for uterine endometrial cancer in TSMR analysis. These results remained consistent after FDR adjustment and MVMR analysis. Fig. 5 Forest plot and results of TSMR and MVMR analysis from the effects of lipidomes on uterine endometrial cancer. ST, sterols; CE, cholesterol ester; GP, glycerophospholipids; PC, phosphatidylcholine; PCO, phosphatidylcholine-ether; IVW, inverse-variance weighted; WM, weighted median; OR, odds ratio; CI, confidence interval; Q values, FDR-adjusted P values; MVMR, multivariable Mendelian randomization. Forest plot and results of TSMR and MVMR analysis from the effects of lipidomes on uterine endometrial cancer. ST, sterols; CE, cholesterol ester; GP, glycerophospholipids; PC, phosphatidylcholine; PCO, phosphatidylcholine-ether; IVW, inverse-variance weighted; WM, weighted median; OR, odds ratio; CI, confidence interval; Q values, FDR-adjusted P values; MVMR, multivariable Mendelian randomization. Regarding ovarian cancers, phosphatidylethanolamine (18:1_0:0), phosphatidylcholine (14:0_18:1), and two types of triacylglycerol structures were identified as risk factors. MVMR analysis was employed to estimate their direct effects on ovarian cancer; however, none of these associations reached statistical significance (Fig.  6 ). Fig. 6 Forest plot and results of TSMR and MVMR analysis from the effects of lipidomes on ovarian cancer. GP, glycerophospholipids; LPE, lysophosphatidylethanolamine; PC, phosphatidylcholine; GL, glycerolipids; TAG, triacylglycerol; IVW, inverse-variance weighted; WM, weighted median; OR, odds ratio; CI, confidence interval; Q values, FDR-adjusted P values; MVMR, multivariable Mendelian randomization. Forest plot and results of TSMR and MVMR analysis from the effects of lipidomes on ovarian cancer. GP, glycerophospholipids; LPE, lysophosphatidylethanolamine; PC, phosphatidylcholine; GL, glycerolipids; TAG, triacylglycerol; IVW, inverse-variance weighted; WM, weighted median; OR, odds ratio; CI, confidence interval; Q values, FDR-adjusted P values; MVMR, multivariable Mendelian randomization. All TSMR results underwent consistency analysis using WM, MR-Egger, simple mode, and weighted mode, confirming robustness in establishing causal relationships. Similarly, MVMR results were supported by consistency analysis with MVMR-Egger, MVMR-Lasso, and MVMR-WM, enhancing confidence in the validity of these findings. A series of sensitivity analyses were conducted to assess the robustness of the findings. Heterogeneity was observed in the Cochrane’s Q test analysis for phosphatidylethanolamine (18:2_0:0) (Q = 25.8815, P  = 0.0268), phosphatidylcholine (18:1_18:2) (Q = 23.4823, P  = 0.0362), phosphatidylethanolamine (16:0_18:2) (Q = 21.3769, P  = 0.0451), and phosphatidylethanolamine (18:0_18:2) (Q = 31.1418, P  = 0.0277) on uterine fibroid. In this study, random-effects IVW was employed as the primary estimation method, effectively accounting for acceptable levels of heterogeneity 20 . No heterogeneity was detected in other analyses. Additionally, all MR-Egger intercept tests in this study yielded P-values greater than 0.05, indicating the absence of horizontal pleiotropy and confirming the stability of the TSMR results reported earlier. Due to the limited number of SNPs associated with phosphatidylethanolamine (O-16:1_22:5) (only 3 SNPs), MR-PRESSO tests could not be conducted for this lipid. However, P-values from MR-PRESSO tests for other outcomes were above 0.05, indicating no evidence of pleiotropic bias across the examined FRDs. No pleiotropy or heterogeneity was detected in the results of MVMR analysis. Detailed information was presented in Additional file 3 . The scatter plots illustrating the MR analyses showed no significant heterogeneity, and the funnel plots exhibited symmetrical distributions (Additional file 4 ). Furthermore, leave-one-out analyses indicated that no single SNP markedly influenced the MR estimates. While a few SNPs potentially influenced specific MR results, their odds ratio values consistently remained on the same side of the zero line (Additional file 4 ). The Steiger test confirmed the correct direction of causality between the identified lipids and FRDs ( P  < 0.05). Additionally, our reverse MR analysis, which examined lipidome as the outcome and the aforementioned FRDs as exposures, did not reveal evidence supporting a reverse causal relationship. Detailed information was provided in Additional file 5. No significant heterogeneity or pleiotropy was detected in the sensitivity analysis results of reverse MR, as detailed in Additional files 5 and 6 . To explore potential mediators of the association between lipidome and female reproductive pathological conditions, we assessed the mediation effects of several common risk factors. As depicted in Fig.  7 , five lipid species were found to causally affect the potential mediators; however, none of the mediators were shown to causally influence the evaluated FRDs. Detailed information was provided in Additional file 7 . These results indicated that the causal effect of lipidome on FRDs remained unaffected by the potential risk factors examined, suggesting that lipidome influences the development of FRDs through alternative mechanisms. Fig. 7 Forest plot and results of two-step MR analysis for the mediation effects. ST, sterols; CE, cholesterol ester; GP, glycerophospholipids; PC, phosphatidylcholine; LPE, lysophosphatidylethanolamine; OR, odds ratio; CI, confidence interval. Forest plot and results of two-step MR analysis for the mediation effects. ST, sterols; CE, cholesterol ester; GP, glycerophospholipids; PC, phosphatidylcholine; LPE, lysophosphatidylethanolamine; OR, odds ratio; CI, confidence interval.

In this study, we thoroughly investigated the complex causal relationships between lipidome and FRDs, analyzing 179 lipidome traits as exposures and 6 common FRDs as outcomes. Our results identified significant effects of 56 specific lipids across glycerophospholipids, sphingolipids, sterols, and glycerolipids categories on the risk of FRDs. Notably, 6 specific lipids across four classes (SM, LPE, PC, CE) from three categories (ST, SL, GP) warranted particular attention. We identified significant roles for phosphatidylcholine and phosphatidylinositol within the glycerophospholipids category, highlighting their relevance in PCOS. Previous research has also underscored disruptions in glycerophospholipid metabolism and phosphatidylcholine synthesis in PCOS patients, implicating lipid metabolism in the pathogenesis of this condition 39 , 40 . Our findings further indicated that sphingomyelin (d38:2) may have a protective effect against PCOS, suggesting that decreased sphingomyelin levels could contribute to PCOS development. Similarly, lower sphingomyelin levels observed in PCOS follicles indicate a potential down-regulation in sphingolipid metabolism, which may influence the follicular environment and impact oocyte development in PCOS patients 41 . Sphingomyelin is known to play crucial roles in various biological processes such as cell proliferation, differentiation, apoptosis, migration, membrane trafficking, cell-cell interactions, and signaling pathways, all of which are directly implicated in the manifestation of insulin resistance symptoms in PCOS 42 , 43 . Future studies should intensively explore sphingomyelin’s potential as a biomarker for PCOS. However, it’s noteworthy that some studies have reported no associations between lipidomic profiles and PCOS or have found no notable elevations in sphingomyelin levels in PCOS patients compared to healthy women 8 , 44 . This discrepancy could be attributed to variations in lipid subtype detection. Our study underscored that different subtypes of lipids within the same classes and categories could exert opposing effects on PCOS pathophysiology. In the context of endometriosis, phosphatidylethanolamine, phosphatidylcholine, and phosphatidylinositol within the glycerophospholipids category played significant roles, alongside triacylglycerol from the glycerolipids category. Changes in lipid metabolism have been implicated in fostering a chronic inflammatory state conducive to the increased adhesion, angiogenesis, and tissue remodeling associated with the onset and progression of endometriosis 45 . Phosphatidylcholine has been proposed as a potential biomarker for the semi-invasive diagnosis of endometriosis. Studies have reported decreased concentrations of specific phosphatidylcholines (18:1/22:6, 20:1/14:1, 20:3/20:4) in early-stage endometriosis patients compared to controls 46 . Conversely, other studies have suggested elevated levels of phosphatidylcholines in endometriosis, which may suppress apoptosis and influence lipid-associated signaling pathways 47 . Discrepancies between these findings may stem from differences in disease stages and variations in phosphatidylcholine subtypes, as highlighted in our study where specific phosphatidylcholine (18:0_20:2) levels were increased while others were decreased in inducing endometriosis. Similarly, phosphatidylinositol (PI 16:0/18:2) was significantly elevated in the follicular fluid of endometriosis patients compared to controls, whereas phosphatidylinositol (16:0_20:4) levels were reduced in endometriosis patients 48 . Additionally, elevated phosphatidylethanolamine (18:2_0:0) was implicated in the causation of endometriosis in our study. This is consistent with findings suggesting alterations in the phosphatidylcholine/phosphatidylethanolamine ratio in serum may contribute to the disease, potentially through upregulation of phosphatidylethanolamine N-methyltransferase gene expression, providing new insights into endometriosis etiology 49 . The association between lipidome and uterine fibroids encompasses several lipid categories, including glycerophospholipids, sterols, glycerolipids, and sphingolipids. Uterine fibroids, benign tumors of the myometrium in women, are characterized by abnormal extracellular matrix deposition and neoplasia of uterine smooth muscle cells. Understanding the lipidomic influence on uterine fibroids reveals insights into various biological processes such as inflammation, immune response, and cell proliferation. Research has indicated that inhibition of sterol O-acyltransferases can lead to excessive accumulation of free fatty acids, triggering endoplasmic reticulum stress and promoting extracellular matrix deposition, thereby contributing to uterine fibroid development 50 . Furthermore, studies by Vignini et al. have demonstrated that women with uterine fibroids exhibit significantly increased visceral fat deposits, which can alter cholesterol fractions and provoke chronic inflammation. This process is crucial for cellular differentiation and proliferation, both essential for uterine fibroid pathogenesis 51 . Sphingolipids, highlighted in our study and others, also play a critical role in uterine fibroids. Sphingosine kinase, up-regulated notably after transforming growth factor-beta (TGF-β) activation, correlates with changes in sphingolipid levels, particularly ceramide 52 . Ceramide is pivotal in TGF-β-induced extracellular matrix deposition by stimulating collagen and other matrix proteins while inhibiting matrix degradation 52 . This highlighted the importance of sphingolipid metabolism in influencing the pathophysiological processes involved in uterine fibroid development. Overall, these findings underscored the multifaceted roles of lipidomic profiles in influencing biological pathways crucial to uterine fibroid development, providing potential avenues for targeted therapeutic interventions in the future. Moreover, we found that the same lipid species, or lipid species from the same classes or categories, exerted similar associations between endometriosis and uterine fibroids, which aligns with previous studies 53 . Genomic studies have demonstrated a molecular basis for the co-occurrence of endometriosis and uterine fibroid. Dissecting this overlap has identified shared genes and pathways, providing insights into the biological mechanisms for the development of endometriosis and uterine fibroids 53 . On the other hand, our findings demonstrated that the lipid associations observed for PCOS and endometriosis were opposite, consistent with studies suggesting that PCOS and endometriosis represented two opposite extremes 54 . In our study, glycerophospholipids and sterols emerged as significant factors associated with uterine endometrial cancer. Glycerophospholipids, including phospholipids, are integral components of lipid metabolism, playing critical roles in maintaining metabolic balance. One notable enzyme, lysophosphatidylcholine acyltransferase (LPCAT), regulates the intracellular levels of various lipids such as phosphatidylethanolamine, phosphatidylcholine, and triglycerides. Research has highlighted LPCAT1’s pivotal role in endometrial cancer development: its inhibition suppresses cancer cell growth, whereas its overexpression enhances stemness and metastasis in endometrial cancer cells 55 . Elevated levels of phosphatidylcholine, particularly an increased ratio of phosphatidylcholine to sphingomyelin, can enhance membrane fluidity, thereby promoting invasiveness and metastasis of endometrial cancer cells 56 . Sterols, on the other hand, were found to potentially inhibit the development of uterine endometrial cancer in our study. Sterol metabolism plays a dual role in tumorigenesis, acting both as a tumor suppressor and a tumor promoter across different stages of cancer progression. Alterations in sterol metabolism can profoundly impact critical cellular processes such as growth, proliferation, and differentiation 57 . Sterol regulatory element binding protein 1 (SREBP1), a nuclear transcription factor, regulates lipogenic processes by activating genes involved in lipid biosynthesis. Downregulation of SREBP1 has been linked to suppressed cell proliferation and demonstrated associations with endometrial cancer 58 . In our study, we identified several key lipids, including phosphatidylcholine, phosphatidylethanolamine, sterol ester, diacylglycerol, and triacylglycerol, that showed close associations with infertility and ovarian cancer. These lipids play pivotal roles in cellular processes that influence reproductive health and cancer development. For infertility, our findings suggest that alterations in phosphatidylcholine, phosphatidylethanolamine, and triacylglycerol levels may impact reproductive outcomes. Specifically, the endometrial lipid profile, characterized by increased concentrations of phosphatidylcholine, phosphatidylethanolamine, and diacylglycerol, has been linked to conditions such as premature progesterone rise at the end of the follicular phase. This hormonal disturbance can impair endometrial receptivity and early embryo implantation, potentially leading to infertility 59 . In the context of ovarian cancer, phosphatidylethanolamine has been implicated in regulating human phosphatidylethanolamine-binding protein 4 (hPEBP4), an anti-apoptotic molecule found to be highly expressed in ovarian cancer cells. Targeting hPEBP4 can potentiate apoptosis signaling in ovarian cancer cells, suggesting a potential therapeutic strategy 60 . Additionally, phosphatidylcholine has been explored for its role in drug delivery systems. Nanoliposomes loaded with gene silencing therapies encapsulated in phosphatidylcholine have shown enhanced antitumor efficacy against ovarian cancers, highlighting phosphatidylcholine’s potential as a targeted drug delivery vehicle 61 . However, despite these associations observed in our initial analysis, our MVMR analysis did not find significant causal relationships between lipid profiles and infertility or ovarian cancer. This discrepancy could be attributed to the heterogeneous nature of lipid subtypes and the complex etiology underlying infertility and ovarian cancer. Studies have reported distinct lipid disturbances between benign ovarian tumors and epithelial ovarian cancers, with variations in glycerophospholipid subtypes exhibiting different tendencies 62 . For instance, lysophosphatidylcholine and lysophosphatidylethanolamine levels were elevated in ovarian tumor patients, while phosphatidylcholine and ether phosphatidylcholine levels were significantly reduced 62 . To the best of our knowledge, this study represents a pioneering effort in thoroughly investigating the influence of lipidomes on FRDs. In this study, we followed the strengthening the reporting of observational studies in epidemiology using Mendelian randomization (STROBE-MR) guidelines as previously described 63 , 64 . Employing a range of advanced MR methods such as TSMR, MVMR and mediation MR, we revealed a complex interplay between lipid metabolism and various FRDs. These comprehensive approaches not only enhanced our understanding of the molecular mechanisms underlying these disorders but also paved the way for the development of personalized medicine approaches, where lipidomic profiles could serve as indicators for early diagnosis, prognosis, and therapeutic interventions tailored to individual patients. However, there are several limitations to our study that warrant considerations. Firstly, the participants in our study were predominantly of European ancestry. Therefore, the generalizability of our findings to other ethnic or racial groups is uncertain. Secondly, in order to include a greater number of instruments in our MR analysis, we relaxed the significance threshold, which could increase the risk of weak instrument bias and horizontal pleiotropy. Although we conducted extensive sensitivity analyses to address horizontal pleiotropy, residual biases may still affect the accuracy of our causal estimates. Thirdly, despite conducting mediation MR to explore potential mediation pathways between lipidome and FRDs, we did not identify specific mediators influencing the causal relationships in this study. The precise signal pathways through which these lipids impact FRDs remain unclear and require further investigation.

In conclusion, our study investigated the association between FRDs and specific lipidomes categorized as sterols, sphingolipids, glycerolipids, and glycerophospholipids. While the evidence for a causal association with glycerolipids was inconclusive in our study, sterols, sphingolipids, and glycerophospholipids emerged as significant contributors to the progression of FRDs. By highlighting these lipidomic associations, we provided insights that could pave the way for personalized approaches in the prevention, diagnosis, and treatment of FRDs. Further research is warranted to validate these findings across diverse populations and to elucidate the underlying biological mechanisms through which these lipid molecules influence the development and progression of FRDs. Such endeavors will deepen our understanding of lipid metabolism in reproductive health and may lead to novel strategies aimed at improving clinical outcomes for individuals affected by FRDs.

Female reproductive diseases (FRDs) encompass a wide range of conditions affecting the ovaries, fallopian tubes, uterus, and cervix in women. Key reproductive disorders include endometriosis, uterine fibroids, gynecologic cancers, polycystic ovary syndrome (PCOS), and related conditions 1 . These diseases not only result in significant morbidity but also lead to chronic pain, infertility, and psychological distress, greatly impacting the quality of life for millions of women worldwide and posing substantial public health challenges 2 , 3 . However, the exact etiologies of these diseases remain unclear, and their clinical manifestations are highly heterogeneous. Therefore, there is an urgent need to elucidate the underlying mechanisms and develop more personalized and effective treatment strategies tailored to individual needs 4 , 5 . The lipidome represents the complete set of lipids present within a biological system, playing critical roles in cellular structure, energy storage, signaling, and metabolism. It encompasses diverse molecules such as phospholipids, sphingolipids, sterols, and glycolipids 6 . Recently, there has been increasing interest in the association between lipids and FRDs within biomedical research. Emerging evidence suggests that dysregulation of lipid metabolism may contribute to the pathogenesis and progression of conditions like PCOS, endometriosis, infertility, and gynecologic cancers 7 – 10 . Mishra et al. reported changes in serum triglycerides (TC), cholesterol, low-density lipoprotein (LDL), and high-density lipoprotein (HDL) levels in women with PCOS, indicating a potential link between dyslipidemia and PCOS-related complications 11 . Furthermore, studies have identified abnormal lipid levels, including HDL, LDL, and TC, as risk factors for female infertility 12 , 13 . Zhang et al. demonstrated a positive correlation between elevated serum TC and LDL levels and increased ovarian cancer risk 14 , while Li et al. highlighted dyslipidemia’s role in endometrial carcinogenesis via inflammation and estrogen signaling pathways 15 . Additionally, lipid-lowering drugs used in treating these diseases may impact the female reproductive endocrine system 16 . However, previous studies, primarily observational and not randomized controlled trials (RCTs), may be subject to biases and confounding factors. Therefore, further investigation is needed to establish causal relationships between lipid levels and FRDs. Moreover, existing research predominantly focuses on general lipid classes such as cholesterol, LDL, HDL, and TC, with other lipid classes warranting further exploration. Mendelian randomization (MR) has emerged as a powerful tool in epidemiological research, utilizing genetic variants as instrumental variables (IVs) to infer causality between exposures and outcomes 17 . This approach leverages principles of Mendelian inheritance, exploiting genetic variants robustly associated with modifiable exposures to mitigate common biases such as confounding and reverse causation observed in observational studies 18 . By using genetic variants as proxies for exposures of interest, MR provides a quasi-randomized experimental design analogous to a RCT, but without the ethical or practical constraints 19 . Advanced MR methodologies include two-sample MR (TSMR), which estimates instrument-exposure and instrument-outcome associations in separate samples 20 ; multivariable MR (MVMR), which explores complex relationships involving multiple exposures that may interact with each other; and mediation MR, specifically addressing the mediation effects of intermediate variables in the causal pathway between exposures and outcome 21 . Therefore, the objective of this study is to explore the causal relationship between lipidomes and FRDs using advanced MR methodologies. We employed TSMR to assess the association between lipidome traits and FRDs. Subsequently, we utilized MVMR to corroborate these findings and mediation MR to investigate potential underlying mechanisms.

Below is the link to the electronic supplementary material. Supplementary Material 1. Supplementary Material 1. Supplementary Material 2. Supplementary Material 2. Supplementary Material 3. Supplementary Material 3. Supplementary Material 4. Supplementary Material 4. Supplementary Material 5. Supplementary Material 5. Supplementary Material 6. Supplementary Material 6. Supplementary Material 7. Supplementary Material 7.