Title
Author
DOI
Article Type
Special Issue
Volume
Issue
Comprehensive analysis of the impact of circulating metabolites on the risk of endometrial cancer: a Mendelian randomization study
1Department of Gynecology, The Fifth Affiliated Hospital of Sun Yat-sen University, 519000 Zhuhai, Guangdong, China
*Corresponding Author(s):yangh353@mail.sysu.edu.cn (Hua Yang)
| History | Submitted: 31 July 2024 | Accepted: 03 September 2024 | Published: 15 June 2025 |
| Copyright: | ©2025 The Author(s). Published by MRE Press. |
Background: Endometrial cancer (EC) is the predominant gynecological malignancy in developed countries and is closely associated with metabolic syndrome. However, the underlying pathogenic mechanisms and the impact of serum circulating metabolites (CMs) on EC risk remain largely unexplored. Methods: To elucidate potential associations between CMs and EC, a two-sample Mendelian randomization (MR) study was conducted. The study utilized summary genome-wide association study (GWAS) data labeled as ebi-a-GCST006464, serving as the outcome dataset, which included data from 12,906 cases and 108,979 controls, all of European descent. Genetic predictors associated with CMs were sourced from three metabolite GWAS datasets compiled by Shin, Kettunen, and Borges. Results: The MR analyses revealed 36 associations between CMs and EC that passed a nominal p-value significance threshold (p-value range: 0.003–0.0492). However, upon multiple testing correction, none of the CMs remained significantly associated with EC. Subgroup analysis found 27 associations between CMs and endometrioid EC passed a nominal p-value significance threshold (p-value range: 5.69 × 10–6–0.0499). Notably, the associations for 4-androsten-3beta,17beta-diol disulfate 2 and Hexadecanedioate survived multiple testing corrections (False Discovery Rate (FDR) = 0.0015 and 0.0422, respectively). Concurrently, 78 associations between CMs and non-endometrioid EC passed a nominal p-value significance threshold (p-value range: 0.0003–0.4997). Furthermore, 23 associations between CMs (all belonging to lipometabolomics) and non-endometrioid EC survived multiple testing corrections (FDR value range: 0.0435–0.0486). Conclusions: This analysis has identified specific CMs potentially associated with EC, especially in non-endometrioid EC. The results offer new evidence of the association between CMs and EC, including its risk factors. This information may guide the development of metabolite-based interventions for EC and its risk factors in forthcoming clinical trials and can also act as candidate targets for further mechanism exploration and drug selection.
Cite this article
Hua Yang. Comprehensive analysis of the impact of circulating metabolites on the risk of endometrial cancer: a Mendelian randomization study. European Journal of Gynaecological Oncology. 2025; 46(6): 69-81. doi: 10.22514/ejgo.2025.081
1. Introduction
Endometrial cancer (EC) ranks among the prevalent gynecological malignancies, with an estimated 420,000 new cases annually, underscoring its significant global burden. Notably, both its morbidity and mortality rates have shown a troubling increase over recent decades [1, 2, 3]. Diagnosis typically hinges on biopsy and/or curettage for histological verification, as the field still lacks non-invasive diagnostic approaches despite advances in biomarker research. EC has two primary subtypes based on clinical, endocrine, and epidemiological observations and four molecular subtypes based on genomic abnormalities. Amongst the primary subtypes, Type I, is estrogen dependent and constitutes 80% of diagnosed cases. It is associated with a comparatively favorable prognosis. Conversely, Type II is estrogen independent and linked to a less optimistic outcome. While it is widely acknowledged that risk factors for EC involve advancing age, obesity, hypertension, metabolic disorders, and sustained high estrogen exposure [4, 5], the precise etiology and pathogenic pathways implicated in EC initiation and progression remain poorly understood, particularly for Type II EC.
The metabolome serves as a comprehensive reflection of the human phenotype in health and disease states. This is because metabolic profiles are positioned downstream of the genome, transcriptome and proteome, essentially representing the information that results from the regulation of the upstream control and signaling pathways. A key indicator of cancer progression is metabolic reprogramming [6]. Metabolites associated with cancer are by-products of cellular processes that arise from neoplastic transformation and cellular proliferation, as well as from the body’s immunological (inflammatory) response to malignancy. Furthermore, aberrations in metabolism associated with cancer can result in reprogramming of epigenetic patterns that are caused by enzymatic alterations which then result in the formation or accumulation of so called onco-metabolites. These onco-metabolites can in turn affect the activity of enzymes that influence gene regulation. It is widely acknowledged that EC is strongly linked to obesity and the dysregulation of metabolic pathway activities, such as estrogen and insulin signaling. The disruption of various metabolic pathways has been connected to the development of EC, including exposure to high levels of estrogen [7], hyperinsulinemia [8], as well as chronic inflammation associated with obesity [9]. In recent decades, several metabolomics studies have identified distinct metabolomic signatures that are closely correlated with risk prediction, diagnosis and prognosis of EC [10, 11, 12, 13, 14]. However, there is a persisting need for compelling evidence that EC develops and progresses within the context of profound metabolic dysfunction. This is primarily because most reports originate from case-control studies with restricted sample sizes. Furthermore, cause and effect relationships between metabolite profile alterations in EC and disease development and progression have not been clearly elucidated so far. Nonetheless, a convincing causal metabolomics biomarker panel derived from population-level data could provide novel targets for EC prevention and treatment.
Mendelian randomization (MR) offers a sophisticated approach to transcend the constraints of observational studies, enabling robust causal inferences. This method harnesses germline single nucleotide polymorphisms (SNPs) as instrumental variables (IVs), leveraging these genetic markers to discern causal relationships between exposures and outcomes. In pursuit of elucidating the underlying metabolomic etiology and pathogenic mechanisms associated with EC, a comprehensive analysis was undertaken. This analysis explored the causality between readily available metabolomic GWAS data and EC, employing a two-sample MR analytical framework. This meticulous examination sought to unravel the intricate metabolic factors contributing to EC, offering avenues for further investigation and exploration of targeted intervention strategies.
2. Materials and methods
2.1 GWAS statistics of EC
The GWAS data, designated as ebi-a-GCST006464, was employed in this investigation [15]. This meta-analysis dataset encompassed 9,470,555 SNPs. The cohort under study comprised 9988 cases of EC and 108,979 nationality-matched European ancestry controls. These controls were derived from 17 studies identified by the Endometrial Cancer Association Consortium (ECAC), the Epidemiology of Endometrial Cancer Consortium (E2C2), and the UK Biobank. Owing to the distinct genetic alterations and pathological characteristics that differentiate endometrioid from non-endometrioid EC, the dataset was meticulously subdivided into two principal categories: endometrioid EC, encompassing 8758 cases and non-endometrioid EC, which included 1230 cases. 2918 cases with unclear or controversial pathological subtypes were excluded to ensure reliability of the results to interrogate the hypothesis. All original studies have secured ethical approval and informed consent. Further details, including the recruitment criteria for the population and quality control measures for the genetic data, can be found in the primary literature (https://gwas.mrcieu.ac.uk/datasets/ebi-a-GCST006464/).
2.2 GWAS statistics of CMs
The GWAS data pertaining to CMs was sourced from three recently published studies. The research conducted by Shin et al. [16] represents the most exhaustive analysis of CMs to date, with the full summary statistics made publicly accessible through the Metabolomics GWAS Server. This GWAS analysis incorporated 7824 adult individuals from European cohorts including approximately 2.1 million SNPs. Of the 486 metabolites examined, 309 were identified as distinct metabolites, categorized into eight metabolic groups: lipids, amino acids, carbohydrates, energy, cofactors and vitamins, nucleotides, xenobiotic metabolism components and additional lipid classes. Omitted from subsequent analyses were 177 unidentified metabolites, whose chemical identities remained undetermined.
In the study by Kettunen et al. [17], 123 circulating metabolites were analyzed, encompassing lipoprotein subclasses, involving 24,925 individuals of European descent. The number of included SNPs reached up to 12,133,295. All metabolite concentrations were adjusted for age, sex, time elapsed since the last meal, and the initial ten principal components.
Furthermore, the research by Borges et al. [18] enrolled 249 circulating metabolites identified by nuclear magnetic resonance analyses from 121,000 participants of European descent, as generated by Nightingale Health. These data comprised 168 absolute metabolites and 81 metabolite ratios, primarily focusing on lipids and lipoprotein particle subfractions, and further covering apolipoproteins, amino acids, fatty acids, cholesterol, ketone bodies, esterified cholesterol, free cholesterol, cholines, glycolysis metabolites, phospholipids, lipoprotein particle size and triglycerides.
2.3 Instrumental variables (IVs) estimation
The selection of IVs for MR studies must adhere to three fundamental principles: (1) IVs should have a demonstrable association with the exposure the disease, (2) IVs must not exhibit any association with potential confounders such as diet, exercise, body mass index in individuals not having the disease, and (3) IVs should influence the outcome exclusively through their relationship with the exposure to the disease, without acting through alternative pathways.
In order to fortify the reliability and accuracy of the association inferences connecting CMs and the risk of EC, rigorous quality control procedures were implemented for the optimal extraction of IVs estimation. Initially, SNPs that met a statistical significance threshold of p < 5 × 10−8 and showed a close association with CMs were identified. However, this approach resulted in the inclusion of a limited number of CMs. To enable elucidation of broader relationships, a suggestive p-value threshold of p < 5 × 10−6 was adopted. The minor allele frequency (MAF) criterion for variants of interest was established at 0.01.
Crucial MR analysis protocols were followed to mitigate linkage disequilibrium (LD) effects among IVs estimates, as substantial LD could introduce bias. In the present study, clumping procedures were undertaken (R2 < 0.001 and clumping distance = 10,000 kb) to assess LD among the chosen SNPs. This is an important procedure as clumping helps to identify the most significant SNPs, namely the ones with the lowest p-value. As a consequence, this process decreases the correlation among the remaining SNPs while preserving those with most robust statistical support. Additionally, it was ensured that the SNP effects on exposure corresponded to identical allele effects on the outcome. Measures were taken to avoid palindromic SNPs to prevent uncertainties in strand directionality or allele coding. F statistics were calculated to sidestep biases related to weak IVs, using the formula F = R2 × (N − 1 − K)/(1 − R2) × K. SNPs with F-values below 10 were excluded from subsequent MR analyses [19]. Alleles were harmonized with the human genome reference sequence, and any ambiguous or duplicate SNPs were eliminated. Furthermore, genotype-phenotype associations for each SNP were examined leveraging the PhenoScanner website (https://www.repository.cam.ac.uk/handle/1810/293487). As a result, SNPs related to potential confounding factors of EC were omitted, for examples SNPs related to body fat percentage, total cholesterol level and other variables that might obscure any disease specific variables.
Horizontal pleiotropy was assessed using MR-Pleiotropy Residual Sum and Outlier Test (PRESSO) and MR-Egger regression. For individual SNPs, the MR-PRESSO outlier test provided a p-value indicating the significance of pleiotropy, while the global test evaluated overall horizontal pleiotropy. SNPs were ranked based on their outlier test p-values and subsequently removed one by one. Following each removal, the global test was rerun on the remaining SNPs. This procedure was repeated until the global test p-value exceeded 0.05. The SNPs that remained after the removal of those with pleiotropic effects were utilized in the ensuing MR analyses to determine EC related signals and patterns.
2.4 Mendelian randomization (MR) analysis
The causal relationships between CMs and the risk of EC were inferred using two-sample univariate MR analysis. Three predominant MR methods, employed for analyses involving multiple instrumental variables (IVs), included the inverse-variance weighted (IVW) test, weighted median test and the MR-Egger test. Given its heightened efficacy under specific conditions, the IVW method was primarily utilized for results with multiple IVs, while the other methods provided supplementary insights. For analyses featuring a single IV, the Wald ratio test was applied. This procedure allows for a causal estimate obtained from a single IV by dividing the beta coefficient describing the association between the genetic variant and the exposure to the disease.
To assess the reliability of these associations, several sensitivity analyses were undertaken. The leave-one-out analysis applied here for cross validation evaluated whether the association was predominantly influenced by a singular SNP. The causal direction test compared the variance induced by IVs in both exposure and outcome. The identified causality was considered as directionally robust if IVs induced greater variance in exposure than in outcome. Horizontal pleiotropy was examined through Egger regression, where IVs impacting the outcome through pathways other than exposure were deemed pleiotropic and violated MR assumptions. Cochran’s Q statistic and the two-sample MR package were employed for heterogeneity testing. A Q value exceeding the number of IVs minus one or a p-value < 0.05 indicated heterogeneity and invalidated the IVs.
All statistical analyses were executed using R software version 3.5.3, with MR analyses conducted utilizing the “MendelianRandomization”, “TwosampleMR” and “MRPRESSO” packages. The Benjamini-Hochberg corrected p-value controlled the false discovery rate (FDR) during multiple testing, with FDR < 0.05 considered statistically significant [20]. Metabolic pathway analysis was performed via Metaconflict 4.0.
A detailed flowchart of the MR analyses in this study is presented in Fig. 1.
Fig. 1.The detailed flowchart of present mendelian randomization analysis. GWAS: genome-wide association study; CMs: circulating metabolites; EC: endometrial cancer; SNPs: single nucleotide polymorphisms; MR: Mendelian randomization; IVW: inverse-variance weighted; WME: within-method of experimental design; PRESSO: pleiotropy residual sum and outlier test.
3. Results
3.1 Selection of SNPs
SNPs associated with EC were selected by first extracting 1 to 194 SNPs per CM, following a threshold for suggestive significance set at p less than 5 × 10−6 after implementing an exhaustive array of quality control steps. The exact allocation of SNPs within each respective taxon is thoroughly delineated in Supplementary Tables 1,2,3. In the subsequent phase of our analysis, pleiotropic SNPs were eliminated, as pinpointed by results from the MR-PRESSO outlier test. Thus, this phase allowed for the determination of clear absence of horizontal pleiotropy among the IVs. This observation was further corroborated by using the MR-PRESSO global test and MR-Egger regression to affirm the robustness of the IVs in the analysis.
3.2 Association between CMs and EC risk
Within the set of IVs (p < 5 × 10−6) and utilizing GWAS data with ebi-a-GCST006464 (endometrial cancer) as the outcome, there was a significant association noted between elevated levels of arachidonate (20:4n6) and a reduced risk of EC (beta = −0.5722, p = 0.0298 by IVW test). This observation was further corroborated by congruent results derived from the MR-Egger and Weighted median tests, with an absence of horizontal pleiotropy (p = 0.645) or heterogeneity (p = 0.7403) across the SNPs. The leave-one-out approach indicated that no single SNP had a major impact on the association. This clearly suggested a potential relationship between A = arachidonate (20:4n6) and EC. Similarly, elevated levels of margarate (17:0), ascorbate (Vitamin C), aspartylphenylalanine, erythrose, lathosterol, phenylalanylserine, vanillin, total cholesterol in intermediate density lipoprotein (IDL), cholesterol in large very low density lipoprotein (VLDL), concentration of medium VLDL particles, triglycerides (TG) in medium VLDL, TG in small high density lipoprotein (HDL), concentration of small VLDL particles, phospholipids in very small VLDL, TG in small VLDL, serum TG, TG in chylomicrons and largest VLDL particles, albumin, and cholesterol in medium low density lipoprotein (LDL) were associated with a diminished EC risk (beta = −1.023, −0.2275, −0.9572, −0.9742, −0.386, −0.2451, −0.9043, −0.0825, −0.1578, −0.118, −0.0997, −0.1427, −0.1198, −0.0915, −0.1178, −0.1231, −0.1048, −0.1571 and −0.1096 respectively). Conversely, an elevated level of glucose was linked to an increased risk of EC (beta = 0.8133, p = 0.0226 by IVW test). This observation was further corroborated by congruent results derived from the MR-Egger and Weighted median tests, with an absence of horizontal polymorphism (p = 0.799) or heterogeneity (p = 0.7527) amongst the SNPs. The leave-one-out approach indicated that no single SNP had a major impact on the association, suggesting a potential relationship between glucose and EC risk. Additionally, elevated levels of androsterone sulfate, carboxy-4-methyl-5-propyl-2-furanpropanoate, isobutyrylcarnitine, ADpSGEGDFXAEGGGVR, 1-arachidonoylglycerophosphoinositol, tetradecanedioate, hexadecanedioate, dimethylarginine, 4-androsten-3beta,17beta-diol disulfate 2, average fatty acid chain length, total cholesterol (TC) in medium HDL, free cholesterol (FC) in medium HDL, total lipids in medium HDL, histidine and phospholipids-to-total lipids ratio in small HDL were linked to an increased risk of EC (beta = 0.2504, 0.2122, 0.4222, 0.6171, 0.5588, 0.2607, 0.4788, 0.9117, 0.8244, 0.1739, 0.1291, 0.3179, 0.1362, 0.2601 and 0.0908 respectively). These findings are graphically represented in Fig. 2 and summarized in Supplementary Table 4. After adjustment for multiple comparisons, the FDR for every causality evaluation was above the threshold of 0.05 (Supplementary Table 4).
Fig. 2.The forest plot summarized the causality of circulating metabolites on risk of endometrial cancer.*: indicates that the compound is a specific form of stereoisomer, specifically the trans configuration. M_LDL_C: Cholesterol in medium LDL; S_HDL_PL_pct: Phospholipids to total lipids ratio in small HDL; IDL.C: Total cholesterol in IDL; L.VLDL.C: Total cholesterol in large VLDL; M.HDL.C: Total cholesterol in medium HDL; M.HDL.FC: Free cholesterol in medium HDL; M.HDL.L: Total lipids in medium HDL; M.VLDL.P: Concentration of medium VLDL particles; M.VLDL.TG: Triglycerides in medium VLDL; S.HDL.TG: Triglycerides in small HDL; S.VLDL.P: Concentration of small VLDL particles; XS.VLDL.PL: Phospholipids in very small VLDL; S.VLDL.TG: Triglycerides in small VLDL; Serum.TG: Serum total triglycerides; XXL.VLDL.TG: Triglycerides in chylomicrons and largest VLDL particles.
Taken together, these analyses identified several CMs that may be associated with the risk of EC. Metabolic pathway analysis found that CMs involved in steroid biosynthesis (p = 0.0063), and histidine metabolism (p = 0.049) are potentially linked to EC risk. Other identified CMs such as neomycin, kanamycin and gentamicin biosynthesis components (p = 0.0063) are considered as xenobiotic metabolites that are unlikely to be directly involved in EC risk. It should be highlighted that since none of the CMs exhibited a significant association with EC upon correction for multiple testing, the possibility of spurious associations cannot entirely be ruled out.
3.3 The association between CMs and endometrioid EC risk
Within the set of IVs (p < 5 × 10−6) and utilizing GWAS data with ebi−a−GCST006465 (endometrioid EC) as the outcome, a statistically significant correlation was observed between increased concentrations of Myo−inositol and a attenuated risk for endometrioid EC (beta = −0.7444, p = 0.0159 by IVW test). This observation was further corroborated by congruent results derived from the MR−Egger and Weighted median tests, with an absence of horizontal pleiotropy (p = 0.541) or heterogeneity (p = 0.5856) across the SNPs. A leave−one−out analysis further substantiated that this association was not being driven disproportionately by any singular SNP, suggesting a potential relationship between Myo-inositol and endometrioid EC. Similarly, elevated levels of arachidonate (20:4n6), aspartylphenylalanine, margarate (17:0), mannitol, lathosterol, eicosapentaenoate, O-sulfo-L-tyrosine, concentration of medium VLDL, serum TC, saturated-to-total fatty acid ratio, cholesterol-to-total lipids ratio in very large HDL, and cholesterol-to-total lipids ratio in medium LDL indicated an attenuated risk for endometrioid EC (beta = −0.8377, −0.7651, −1.489, −0.3595, −0.4551, −0.9329, −0.922, −0.1167, −0.0851, −0.2253, −0.1071 and −0.1403 respectively). Conversely, an elevated level of dimethylarginine (symmetric dimethylarginine (SDMA) + asymmetric dimethylarginine (ADMA)) was linked to an increased risk of endometrioid EC (beta = 1.408, p = 0.0031 by IVW test). This observation was further corroborated by congruent results derived from the MR-Egger and Weighted median tests, with an absence of horizontal pleiotropy (p = 0.308) or heterogeneity (p = 0.3966) across the SNPs. The leave-one-out approach indicated that no single SNP had a major impact on the association, suggesting a potential relationship between dimethylarginine (SDMA + ADMA) and endometrioid EC. Additionally, elevated levels of 1-eicosatrienoylglycerophosphocholine, threonine, 4-androsten-3beta,17beta-diol disulfate 2, ADpSGEGDFXAEGGGVR, 1-palmitoleoylglycerophosphocholine, uridine, the inflammation associate pepetide HWESASXX, tetradecanedioate, 3-carboxy-4-methyl-5-propyl-2-furanpropanoate, hexadecanedioate, aspartate, estrone 3-sulfate, and description of average fatty acid chain length were linked to an increased risk of endometrioid EC (beta = 0.7948, 0.9348, 1.131, 0.5732, 1.16, 1.58, 0.4631, 0.318, 0.2185, 0.5671, 0.9506, 0.1507 and 0.1719 respectively). These findings are graphically represented in Fig. 3 and summarized in Supplementary Table 5.
Fig. 3.The forest plot summarized the causality of circulating metabolites on risk of endometrioid endometrial cancer.*: indicates that the compound is a specific form of stereoisomer, specifically the trans configuration. SDMA: symmetric dimethylarginine; ADMA: asymmetric dimethylarginine; CMPF: 3-carboxy-4-methyl-5-propyl-2-furanpropanoate; M.VLDL.P: Concentration of medium VLDL particles; Serum.C: Serum total cholesterol; SFA_pct: Ratio of saturated fatty acids to total fatty acids; XL_HDL_C_pct: Cholesterol to total lipids ratio in very large HDL; M_LDL_C_pct: Cholesterol to total lipids ratio in medium LDL.
Upon correcting for multiple testing, the FDRs for 4-androsten-3beta,17beta-diol disulfate 2 and hexadecanedioate were lower than 0.05 (Supplementary Table 5). Metabolic pathway analysis found there CMs involved in the biosynthesis of unsaturated fatty acids displayed a p-value of 0.013, indicating a significant result. Correspondingly, the metabolic routes for valine, leucine, isoleucine, ascorbate, and aldarate demonstrated similar levels of significance with a p-value of 0.04. Taken together, these analyses identified distinct CMs that may be linked to the risk of endometriod EC. Especially, the linkage of 4-androsten-3beta,17beta-diol disulfate 2 and hexadecanedioate with endometrioid EC exhibited strong resilience.
3.4 The association between CMs and non-endometrioid EC risk
Within the set of IVs (p < 5 × 10−6) and utilizing GWAS data with ebi-a-GCST006466 (non-endometrioid EC) as the outcome, a marked correlation was observed between increased carnitine levels and a diminished risk of non-endometrioid EC (beta = −1.843, p = 0.0202 as determined by the IVW test). This finding was further corroborated by congruent results derived from the MR-Egger and Weighted median tests, with an absence of horizontal pleiotropy (p = 0.47) or heterogeneity (p = 0.8612) across the SNPs as well as with an absence of horizontal pleiotropy (p = 0.645) or heterogeneity (p = 0.7403) across the SNPs. The leave-one-out approach indicated that no single SNP had a major impact on the association, suggesting a potential relationship between carnitine and non-endometrioid EC. Similarly, elevated levels of aspartylphenylalanine, trans-4-hydroxyproline, 3-dehydrocarnitine, cysteine-glutathionedisulfide, p-cresol sulfate, free cholesterol (FC)-to-esterified cholesterol ratio, total cholesterol (TC) in intermediate density lipoprotein (IDL), FC in IDL, total lipids in IDL, concentration of IDL particles, phospholipids in IDL, TC in large low density lipoprotein (LDL), cholesterol esters in large very low density lipoprotein (VLDL), FC in large LDL, Total lipids in large LDL, concentration of large LDL particles, phospholipids in large LDL, TC in LDL, TC in LDL, Cholesterol esters in medium LDL, concentration of medium LDL particles, phospholipids in medium LDL, phospholipids in medium LDL, total lipids in small LDL, concentration of small LDL particles, TC in small VLDL, TG in small VLDL, serum TC, phospholipids in very small VLDL, apolipoprotein B-to-apolipoprotein A1 ratio, cholesterol-to-total lipids ratio in IDL, cholesteryl esters in IDL, cholesteryl esters-to-total lipids ratio in IDL, cholesterol-to-total lipids ratio in large LDL, phospholipids in large LDL, LDL cholesterol, FC in LDL, total lipids in LDL, phospholipids in LDL, cholesterol in medium LDL, cholesterol-to-total lipids ratio in medium LDL, FC in medium LDL, total lipids in medium LDL, phospholipids-to-total lipids ratio in medium LDL, Cholesteryl esters-to-total lipids ratio in medium VLDL, FC in medium VLDL, cholesterol-to-total lipids ratio in small HDL, cholesteryl esters-to-total lipids ratio in small HDL, cholesterol in small LDL, cholesterol-to-total lipids ratio in small LDL, cholesteryl esters in small LDL, FC in small LDL, phospholipids in small LDL, cholesterol-to-total lipids ratio in small VLDL, cholesterol-to-total lipids ratio in small VLDL, phospholipids-to-total lipids ratio in small VLDL, Cholesteryl esters-to-total lipids ratio in very large HDL, cholesterol-to-total lipids ratio in very large VLDL, and FC-to-total lipids ratio in very small VLDL were linked to a diminished risk of non-endometrioid EC (beta = −2.066, −2.518, −2.011, −1.206, −0.9114, −0.2565, −0.325, −0.2417, −0.2972, −0.3077, −0.2121, −0.2828, −0.2975, −0.2512, −0.2988, −0.2789, −0.2733, −0.3099, −0.3008, −0.2928, −0.284, −0.3061, −0.3509, −0.3016, −0.2883, −0.3457, −0.222, −0.2815, −0.2267, −0.2289, −0.2845, −0.2559, −0.2351, −0.2516, −0.291, −0.2623, −0.2938, −0.2685, −0.3072, −0.322, −0.329, −0.283, −0.3114, −0.3003, −0.2721, −0.2421, −0.2551, −0.2375, −0.2491, −0.504, −0.3257, −0.339, −0.3223, −0.257, −0.312, −0.3099, −0.3445, −0.269 and −0.248 respectively). Conversely, an elevated level of pantothenate was linked to an increased risk of non-endometrioid EC (beta = 1.926, p = 0.0284 by IVW test). This observation was further corroborated by congruent results derived from the MR-Egger and Weighted median tests, with an absence of horizontal pleiotropy (p = 0.645) or heterogeneity (p = 0.7403) across the SNPs.
The leave-one-out approach indicated that no single SNP had a major impact on the association, suggesting a potential relationship between pantothenate and non-endometrioid EC. Additionally, elevated levels of methyl-2-oxobutyrate, 1-heptadecanoylglycerophosphocholine, 2-hydroxyglutarate, average fatty acid chain length, TC in HDL, histidine, phospholipids-to-total lipids ratio in IDL, TG-to-total lipids ratio in IDL, FC-to-total lipids ratio in large LDL, TG-to-total lipids ratio in large LDL, TG-to-total lipids ratio in medium LDL, TG-to-total lipids ratio in medium VLDL, phospholipids-to-total lipids ratio in small HDL, TG-to-total lipids ratio in small LDL, TG-to-total lipids ratio in small VLDL, and TG-to-total lipids ratio in very large VLDL were linked to a increased risk of non-endometrioid EC (beta = 3.346, 2.604, 1.908, 0.4651, 0.2493, 0.7785, 0.2984, 0.2663, 0.0589, 0.266, 0.2737, 0.254, 0.2536, 0.2341, 0.2349 and 0.2774 respectively). These findings are graphically represented in Fig. 4 and summarized in Supplementary Table 6. Following adjustments for multiple comparisons, the FDR for several forms of cholesterol (including FC-to-esterified cholesterol ratio, TC in IDL, FC in IDL, TC in large LDL, cholesterol esters in large VLDL, FC in large LDL, TC in LDL, TC in medium LDL, cholesterol esters in medium LDL, TC in small LDL); phospholipids (including phospholipids in IDL, phospholipids in large LDL, phospholipids in very small VLDL), total lipids (including Total lipids in IDL, total lipids in large LDL, total lipids in small LDL); concentration of lipoprotein particles (including concentration of IDL particles, Concentration of large LDL particles, Concentration of medium LDL particles, concentration of small LDL particles) and TG in small VLDL were lower than 0.05 (Supplementary Table 5). Metabolic pathway analysis found these CMs involved in the butanoate metabolic pathway (p = 0.038), histidine metabolic pathway (p = 0.04), and the biosynthetic process of pantothenate and coenzyme A (CoA) (p = 0.049). Taken together, these analyses identified diverse CMs that appear to be associated with a risk of non-endometrioid EC. As all analyses were based on genomic data sets, it is likely that the observed changes in circulating metabolites found associated with non-endometrioid EC represent genomic traits that are associated with the risk of this EC type.
Fig. 4.The forest plot summarized the causality of circulating metabolites on risk of non-endometrioid endometrial cancer. M_LDL_C: Cholesterol in medium LDL; S_HDL_PL_pct: Phospholipids to total lipids ratio in small HDL; Est.C: Free cholesterol to esterified cholesterol ratio; HDL.C: Total cholesterol in HDL; IDL.C: Total cholesterol in IDL; IDL.FC: Free cholesterol in IDL; IDL.L: Total lipids in IDL; IDL.P: Concentration of IDL particles; IDL.PL: Phospholipids in IDL; L.LDL.C: Total cholesterol in large LDL; L.LDL.CE: Cholesterol esters in large VLDL; L.LDL.FC: Free cholesterol in large LDL; L.LDL.L: Total lipids in large LDL; L.LDL.P: Concentration of large LDL particles; L.LDL.PL: Phospholipids in large LDL; LDL.C: Total cholesterol in LDL; M.LDL.C: Total cholesterol in medium LDL; M.LDL.CE: Cholesterol esters in medium LDL; M.LDL.P: Concentration of medium LDL particles; M.LDL.PL: Phospholipids in medium LDL; S.LDL.C: Total cholesterol in small LDL; S.LDL.L: Total lipids in small LDL; S.LDL.P: Concentration of small LDL particles; S.VLDL.C.: Total cholesterol in small VLDL; S.VLDL.TG: Triglycerides in small VLDL; Serum.C: Serum total cholesterol; XS.VLDL.PL: Phospholipids in very small VLDL; ApoB_by_ApoA1: Ratio of apolipoprotein B to apolipoprotein A1; IDL_C_pct: Cholesterol to total lipids ratio in IDL; IDL_CE: Cholesteryl esters in IDL; IDL_CE_pct: Cholesteryl esters to total lipids ratio in IDL; IDL_PL_pct: Phospholipids to total lipids ratio in IDL; IDL_TG_pct: Triglycerides to total lipids ratio in IDL; L_LDL_C_pct: Cholesterol to total lipids ratio in large LDL; L_LDL_FC_pct: Free cholesterol to total lipids ratio in large LDL; L_LDL_PL: Phospholipids in large LDL; L_LDL_TG_pct: Triglycerides to total lipids ratio in large LDL; LDL_C: LDL cholesterol; LDL_FC: Free cholesterol in LDL; LDL_L: Total lipids in LDL; LDL_PL: Phospholipids in LDL; M_LDL_C_pct: Cholesterol to total lipids ratio in medium LDL; M_LDL_FC: Free cholesterol in medium LDL; M_LDL_L: Total lipids in medium LDL; M_LDL_PL_pct: Phospholipids to total lipids ratio in medium LDL; M_LDL_TG_pct: Triglycerides to total lipids ratio in medium LDL; M_VLDL_CE_pct: Cholesteryl esters to total lipids ratio in medium VLDL; M_VLDL_FC: Free cholesterol in medium VLDL; M_VLDL_TG_pct: Triglycerides to total lipids ratio in medium VLDL; S_HDL_C_pct: Cholesterol to total lipids ratio in small HDL; S_HDL_CE_pct: Cholesteryl esters to total lipids ratio in small HDL; S_LDL_C: Cholesterol in small LDL; S_LDL_C_pct: Cholesterol to total lipids ratio in small LDL; S_LDL_CE: Cholesteryl esters in small LDL; S_LDL_FC: Free cholesterol in small LDL; S_LDL_PL: Phospholipids in small LDL; S_LDL_TG_pct: Triglycerides to total lipids ratio in small LDL; S_VLDL_C_pct: Cholesterol to total lipids ratio in small VLDL; S_VLDL_FC_pct: Free cholesterol to total lipids ratio in small VLDL; S_VLDL_PL_pct: Phospholipids to total lipids ratio in small VLDL; S_VLDL_TG_pct: Triglycerides to total lipids ratio in small VLDL; XL_HDL_CE_pct: Cholesteryl esters to total lipids ratio in very large HDL; XL_VLDL_C_pct: Cholesterol to total lipids ratio in very large VLDL; XL_VLDL_TG_pct: Triglycerides to total lipids ratio in very large VLDL; XS_VLDL_FC_pct: Free cholesterol to total lipids ratio in very small VLDL.
4. Discussion
To shed light on the genomic underpinnings of EC, the risk factors evaluated here focused on circulating metabolites and their association with genetic SNPs found in clinical patient GWAS data sets. The term risk is defined here as ……………. (risk of developing EC, risk of EC progression once it exists in a patient, risk of a poor outcome, risk of lack of treatment response? At what stage of the disease were samples taken for the analyzed data sets?) Utilizing large-scale GWAS summary statistics, MR analyses were conducted, revealing associations between EC risk and several genetically predicted CMs. Notably, elevated levels of 4-androsten-3beta,17beta-diol disulfate 2 and Hexadecanedioate were associated with an increased risk of endometrioid EC. Conversely, heightened concentrations of various cholesterol forms, triglycerides, phospholipids, total lipids, and lipoprotein particles were linked to a reduced risk of non-endometrioid EC.
EC represents the most prevalent gynecological malignancy in developed nations and ranks second globally. In contrast to other gynecological cancers, which have exhibited declining incidence rates over the past two decades, the global incidence of EC has been on the rise. The pathogenetic mechanisms underlying the onset and progression of EC largely remain elusive. Considering the robust association between metabolic syndrome components, such as obesity, hypertension, diabetes and EC risk, as identified by observational epidemiological studies, it is hypothesized that circulating metabolites might bear a causal relationship with EC and potentially contribute to its pathogenesis. CMs often mirror the genetic profiles of individuals, acting as functional intermediates following environmental exposures, and can potentially predict or influence disease development. The absence of large RCTs to confirm causal links between CMs and EC underscores the potential of MR analysis of CMs as a valuable approach to discern risk factors and pathogenic pathways of EC. Thus, the results from our MR based studies reveal much needed new information pertaining to cause-and-effect relationships between CMs and EC development and progression. EC has been less extensively researched in relation to metabolomics compared to other gynecological malignancies. However, recent investigations have uncovered intriguing associations. Several metabolites and metabolic pathways have been implicated as dysregulated in EC. The prospective analysis of Dossus et al. [21] found the levels of glycine, serine, sphingomyelin C18:0 and free carnitine may represent specific metabolic pathways linked to EC development. Knific et al. [22] found the plasma levels of sphingomyelins and phosphatidylcholines could served as diagnostic and prognostic biomarkers of EC. Shi et al. [23] found the level of phenylalanine, indoleacrylic acid (IAA), phosphocholine and lyso-platelet-activating factor-16 (lyso-PAF) were differentially detected in early stage EC patients from healthy volunteers, and functional analyses suggested that they may play roles in modulation of tumor cell behavior. Troisi et al. [24] found the serum metabolomics signature of EC patients is peculiar because it differs from that of healthy controls and from that of benign endometrial disease and from other gynecological cancers. Audet-Delage et al. [25] identified putative serum biomarkers useful for detecting EC and predicting recurrence following initial surgery, to ultimately improve patient survival based on better stratification and informed treatment decisions. Nevertheless, conventional observational studies are vulnerable to residual confounding, reverse causation, and other biases that compromise robust causal inference. Consequently, the causal nature of these risk factors and their suitability as intervention targets for EC prevention remain ambiguous. However, the identification of CMs associated with EC, particularly Type II EC, the non-estrogen dependent variant of the disease sheds new light into the complexity of likely genomic underpinnings of this cancer type.
Recent MR studies have hinted at potential causal links between various metabolites, including cholesterol in LDL, insulin, total and bioavailable testosterone, sex hormone-binding globulin (SHBG), and endometrial cancer risk, affirming the causal role of BMI in endometrial cancer risk [26, 27, 28, 29, 30]. Yet, numerous molecular risk factors previously associated with EC from traditional observational studies have not been examined in population-scale MR studies, leaving their causal relevance in disease initiation unclear. Notably, no MR studies have attempted a comprehensive exploration of the potential causal influence of the circulating metabolome on EC risk. Hence, the present MR study holds significant clinical implications.
In the present study, we utilized three of the largest scales of circulating metabolome to date, enabling more comprehensive and reliable conclusions. Through meticulous analysis, out of 1893 exposure-outcome tests, 140 surpassed the significance threshold of 0.05. After applying the Benjamini-Hochberg correction for multiple testing, 23 associations persisted, affirming the causal influence of certain metabolites on EC risk.
The MR analysis substantiated the causal roles of 4-androsten-3beta,17beta-diol disulfate 2 and Hexadecanedioate in endometrioid EC risk. Sex steroids originate from the intracrine conversion of steroid precursors in adipose tissue and skin. The biosynthesis pathway of androgens is intricate, involving compounds such as Dehydroepiandrosterone (DHEA), which serves as a precursor for androgens and estrogens. This process is catalyzed by enzymes like 5α-reductase, leading to epiandrosterone, androsterone, testosterone, among other steroids. Notably, 4-androstene-3beta,17beta-diol disulfate 2, recognized as androstenediol and an intermediate in testosterone synthesis from DHEA, exhibits estrogenic activity due to its structural similarity to estrogen and ability to bind estrogen receptors. Prior MR analyses have associated 4-androstene-3beta,17beta-diol disulfate 2 with reduced risks of lacunar stroke [31], Graves’ disease [32] and dry eye disease [33], while indicating an elevated risk of breast cancer [34]. Our study further validates 4-androstene-3beta,17beta-diol disulfate 2’s significant biological function, potentially establishing it as a novel target for further mechanism exploration and drug selection.
Hexadecanedioate, considered biomarkers of transporter function, are metabolic byproducts of fatty acids’ ω—oxidation. They partake in the cytoplasmic synthesis of long-chain fatty acids through enzymatic catalysis, β—oxidation decomposes fatty acids into shorter acyl-CoA molecules, with hexadecanedioate gradually transformed into shorter fatty acyl-CoA, ultimately generating pyruvic acid and acetyl-CoA. As the terminal product of fatty acid metabolism, acetyl-CoA can enter the citric acid cycle for oxidative metabolism, yielding energy and carbon dioxide. The precise mechanisms linking Hexadecanedioate to EC remain elusive due to limited research elucidating their roles in cancers. The scarcity of comprehensive studies hinders our understanding of underlying mechanisms. However, potential mechanisms involved in the association between Hexadecanedioate and endometrioid EC could include various factors.
Experimental studies [35, 36, 37] have demonstrated the physiological functions of Hexadecanedioate, implicating it in hypolipidemic, anti-obesity and anti-diabetogenic properties. Multi-omic association studies [38] have revealed an inverse relationship between altered abundances of maternal fecal Hexadecanedioate and maternal hyperglycemia, offering a partial explanation for a causal link between hexadecanedioate and endometrioid EC. Furthermore, hexadecanedioate acts as an inhibitor for organic anion-transporting polypeptide (OATP) transporters in preclinical species and humans, essential for DHEA circulation. Elevated levels of OATP-family transporters in patients with endometrial hyperplasia suggest a functional role for OATP in endometrioid EC pathogenesis [39, 40]. Previous MR studies [41] have identified robust causal associations between hexadecanedioate levels and breast cancer, aligning with our analytical findings.
The efficacy of Myo-inositol in addressing insulin resistance and metabolic syndrome is increasingly becoming the subject of substantial scholarly attention. Accumulating evidences [42, 43, 44] suggest that Myo-inositol significantly enhances insulin sensitivity by optimizing the cellular response to insulin, effectively alleviating insulin resistance, which is fundamental to metabolic syndrome. This function of Myo-inositol not only facilitates the reduction of blood glucose levels but also exhibits potential for the amelioration of lipid profiles, often abnormal in metabolic syndrome. Moreover, given the intricate connection between metabolic syndrome and EC, it is reasonable to infer that the beneficial impacts of Myo-inositol may extend to include EC. The present MR studies also demonstrated an association between Myo-inositol and a diminished risk of endometrioid EC. While direct evidence linking Myo-inositol to EC remains limited, its prospective role in modulating cancer-associated metabolites necessitates further scientific exploration. Consequently, Myo-inositol emerges as a promising therapeutic pathway for metabolic syndrome, with potential applications in the prevention or adjuvant treatment of EC.
Comparatively, the risk factors for non-endometrioid EC remain largely unidentified due to its rarity and hormone independence. However, observational studies have also associated non-endometrioid EC with obesity. The estrogen-rich environment, hyperinsulinemia, and alterations in inflammatory factors and cytokines related to obesity, likely play significant roles [45, 46, 47]. Our MR analysis has uncovered robust associations between various forms of cholesterol, triglycerides, phospholipids, total lipids, and concentrations of lipoprotein particles in the circulating metabolome with the risk of non-endometrioid EC.
Altered lipid metabolism has been previously linked to cancer, as cancer cells require lipids for growth, energy and survival. Upregulation of lipids has been associated with cancer progression, including metastasis [48, 49]. Cholesterol, a precursor to steroid hormones, is integral to the synthesis of corticosteroids, sex hormones, vitamin D, among others. These hormones perform critical roles in human physiological processes, such as metabolism and immunity, potentially heightening certain cancer risks. Recent MR studies have hinted at potential causal relationships between cholesterol in low-density lipoprotein (LDL) and EC risk [50]. Phospholipids significantly influence tumor initiation and progression by altering cell membrane structure and function and participating in signaling pathways. Triglycerides (TGs), typically stored in adipocytes and peripheral tissues as the body’s primary energy source, have been implicated in ovarian cancer cell invasion and metastasis. A study by Yan et al. [51] found an association between EC and elevated plasma TG levels. In cancer cells, polyunsaturated fatty acids (PUFAs) and their downstream metabolites regulate diverse processes such as cell signaling, neurotransmission, cell growth and protection, and inflammation [52, 53]. A lipidome analysis of serum from EC patients by Cheng et al. [54] showed a trend towards increased total TG levels, while specific TGs such as TG (20:5_34:0), TG (20:5_36:2), TG (22:6_32:1) and TG (22:6_34:2) decreased, consistent with our analysis.
The present study is characterized by several strengths, including the aggregation of the most extensive metabolite collection to date, the involvement of the largest participant sample size hitherto, and the application of multiple testing corrections to reinforce the reliability of the results. The utilization of MR methodology effectively mitigates confounding factors frequently encountered in epidemiological studies, achieving an evidentiary parity with randomized controlled trials (RCTs). Additionally, the SNPs selected demonstrated pronounced associations with circulating metabolites. Sensitivity analyses revealed no indications of pleiotropy or heterogeneity, underscoring the statistical resilience of the conclusions drawn.
Despite its contributions, this study is not exempt from limitations. Primarily, the genetic markers derived from European populations may impose limitations on the generalizability of the findings to other ethnicities due to potential variations induced by racial and geographical factors as well as mitochondrial haplogroups in the investigated populations that determine contributions to energy metabolism. Secondly, while MR is considered analogous to RCTs, it is not entirely devoid of bias and confounding. Inadequate control for population stratification or the utilization of genetic markers associated with multiple biological pathways can introduce confounding. Thirdly, attaining sufficient statistical power in MR studies necessitates large sample sizes, particularly when estimating modest effect sizes. The challenge lies in acquiring and analyzing expansive datasets while concurrently minimizing Type II errors (false negatives) and governing Type I errors (false positives). Fourthly, the possibility of selection bias emerges when genotype and outcome measurements derive from disparate populations. Lastly, MR studies are constrained to examining heritable exposures and offer limited capacity to assess complex environmental exposures or behavioral influences.
Future investigations should endeavor to encompass larger and more diverse participant samples across various racial and geographic backgrounds to ascertain more substantiated causal relationships. Simultaneously, deeper mechanistic explorations are necessary to elucidate the intricate interactions between circulating metabolites and endometrial cancer.
5. Conclusions
This MR study provides novel genetic evidence suggesting potential causal roles of specific circulating metabolites in EC, particularly highlighting subtype-specific associations. While no metabolites survived multiple testing corrections for overall EC, robust signals emerged in stratified analyses: notably, 4-androsten-3beta,17beta-diol disulfate 2 and Hexadecanedioate showed significant associations with endometrioid EC, and a prominent group of lipid metabolites (lipometabolomics) demonstrated strong, statistically robust links to non-endometrioid EC risk. These findings significantly advance our understanding of the metabolic underpinnings of EC heterogeneity and identify high-priority candidate metabolites for future investigation into mechanisms, biomarker development, and targeted preventive or therapeutic strategies.
Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request.
Author contributions
HY—wrote, discussed, reviewed the manuscript, and designed and created figures.
Ethics approval and consent to participate
This study utilized data from publicly available databases, therefore ethical approval was not required for this specific analysis. It is important to note that all original studies included in the analysis have secured ethical approval and obtained informed consent from participants.
Acknowledgment
The author thanks the participants of all GWAS cohorts included in the present work and the investigators of the IEU Open GWAS project, Metabolomics GWAS Server, and UK Biobank for sharing the GWAS summary statistics.
Funding
This work was supported by the Guangdong Foundation for Basic and Applied Fundamental Research (23202104030001207).
Conflict of interest
The author declares no conflict of interest.
Supplementary material
Supplementary material associated with this article can be found, in the online version, at https://oss.ejgo.net/files/article/1933401730292039680/attachment/Supplementary%20material.xls.
