Introduction
Endometriosis (EM) is a chronic gynecological disorder marked by the presence of endometrial-like tissue outside the uterus. This condition is associated with a range of symptoms including pelvic discomfort, dysmenorrhea, and infertility [1]. Globally, approximately 10% of women of reproductive age are affected, which translates to an estimated 190 million individuals [2]. The disease significantly impacts the quality of life and often requires long-term medical management. The health burden of EM encompasses chronic pain and substantial lifetime costs, estimated at $27,855 per patient per year. This results in annual healthcare expenses on EM of approximately $22 billion in the United States and £12.5 billion in the United Kingdom, accounting for treatment, lost work, and related healthcare costs [3]. Therefore, advancing our understanding and treatment of EM are paramount priorities for women’s health. The etiology of EM is complex and encompasses multiple contributing factors, including retrograde menstruation, immune system dysfunction, benign metastasis, coelomic metaplasia, hormonal imbalances, participation of stem cells, modifications in epigenetic regulation, and various environmental factors [4-6]. Furthermore, another study highlighted a strong link between dietary patterns and the risk of endometrioma development, demonstrating that lower calcium intake is significantly associated with an increased risk of endometrioma [7]. Currently, no singular pathophysiological or molecular framework exists that sufficiently elucidates every instance of this disorder.
Recent studies have suggested that metabolic alterations play a role in the pathogenesis of EM. Women with EM have demonstrated alterations in lipid, glucose, and amino acid metabolism that may facilitate disease progression by affecting the endocrine environment, altering immune responses, and promoting lesion growth [8,9]. Extensive research has consistently demonstrated a strong correlation between the development and advancement of EM and abnormalities in lipid metabolism, as evidenced by significant differences in lipid profiles between women with EM and healthy controls [8,10-13]. Previous research has shown that individuals with EM exhibit statistically significant differences in altered amino acid levels within the tissue (eutopic endometrium), serum, follicular fluid, urine, and endometrial fluid compared to healthy controls. These findings are crucial for understanding various aspects of disease progression. Altered amino acid levels may elucidate the mechanisms of tissue injury repair in EM and the heightened energy demands of proliferative endometrial cells [8]. Glucose metabolism is significantly affected in patients with EM. Similar to tumor cells, ectopic endometrial stromal cells exhibit the Warburg effect, which is characterized by an increase in lactate production and heightened consumption of glucose [14]. Elevated levels of aerobic glycolysis and histone lactylation enhance cell proliferation and migration, thereby contributing to the pathophysiology of EM [9]. Furthermore, studies indicate that glycolysis and lactate accumulation profoundly influence the regulation of the immunomicroenvironment, with lactate acting as a crucial factor that drives M2 macrophage polarization, thereby promoting the invasion of endometriotic stromal cells both in vitro and in vivo [15]. Thus, there is a strong correlation between metabolic abnormalities and EM. A thorough investigation of the roles of relevant metabolites in EM will enhance our understanding of the underlying pathophysiological mechanisms of this complex condition and facilitate the development of novel diagnostic and therapeutic strategies.
Traditional observational studies frequently encounter challenges, such as confounding factors and reverse causation, which can impede the ability to draw definitive conclusions regarding causal relationships. Mendelian randomization (MR) is a robust alternative that employs genetic variants as instrumental variables to infer causality. This approach mitigates the issues of confounding and reverse causation, yielding more reliable estimates of causal effects. The objective of this study was to clarify the association between circulating metabolites and EM using MR analysis. This study used data from a comprehensive collection of 233 circulating metabolites derived from a genome-wide association study (GWAS) repository to examine the possible associations between EM and these metabolites. Furthermore, we sought to detect circulating metabolites that may serve as important biomarkers for the early detection of EM and contribute to the development of effective diagnostic and therapeutic strategies.
Materials and methods
1. Study design
We conducted bidirectional MR analysis to evaluate the potential influence of 233 circulating metabolites on EM risk. Following the framework established by Bowden et al. [16], our analysis was based on three key assumptions: first, the selected genetic instruments (IVs) derived from the datasets were linked to the exposure variable; second, these IVs were not associated with any hidden confounders related to exposure; and third, the IVs influenced the outcomes solely via the exposure factor without any alternative pathways. Our research involved human subjects and a reanalysis of existing publicly available data that had already received ethical approval and participant consent, thereby eliminating the need for additional ethical reviews or consent procedures. An overview of the study design and methodological flow is shown in Fig. 1.
2. Data sources
The circulating plasma metabolite dataset used in this study was obtained from the GWAS Catalog database (ID: GCST90301941-GCST90302173). This dataset comprises 233 metabolic traits, including 213 lipid and lipoprotein parameters or fatty acids, along with 20 non-lipid traits, including amino acids, ketone bodies, metabolites related to glycolysis/gluconeogenesis, fluid balance, and inflammation. Following variant filtering and quality control, 13,389,637 imputed autosomal single-nucleotide polymorphisms (SNPs) were included in the meta-analysis involving 136,016 participants [17]. EM data were obtained from the R10 release dataset of the FinnGen Consortium (https://r10.finngen.fi/), which comprised 16,588 case samples and 111,583 control samples, all of which were of European ancestry [18].
3. Instrumental variables selection
In accordance with methodologies utilized in prior research, a diverse array of IVs was systematically selected for each circulating metabolite and EM condition within the framework of our MR analysis [19-22]. We identified SNPs that exhibited statistically significant associations with circulating metabolites, adhering to a genome-wide significance threshold (P<1×10−5; r2=0.001; genetic distance=10,000 KB). To ensure sufficient SNP availability for sensitivity assessments, relaxed selection criteria were applied to circulating metabolites. For EM analyses, however, stringent parameters were adopted (P<5×10−8; r2=0.001) with preserved 10,000 KB genetic distancing. Linkage disequilibrium pattern characterization leveraged the European reference dataset of the 1,000 Genomes Project using clumping procedures. The variance explanation capacity of the IVs was determined through R2 computation coupled with rigorous F-statistic filtering (F>10) to ensure adequate instrument strength.
4. Statistical analysis
In order to assess the associations between 233 circulating metabolites and EM, we employed the Mendelian Randomization package version 0.4.3 (Stephen Burgess, Cambridge, UK) to perform various analyses, including inverse-variance weighted (IVW), MR Egger, weighted median, simple mode, and weighted mode approaches [16,23-28]. These analytical methodologies were carefully selected to reduce possible bias and enhance the reliability of our research outcomes. In the MR analysis, the IVW method was predominantly employed because of its capacity to offer a dependable assessment of the exposure-outcome relationship under the condition that the IVs exhibited no pleiotropic effects. Cochran’s Q statistics were used to evaluate heterogeneity among individual SNPs. When no significant heterogeneity was found (P<0.05), a fixed-effects model was used; however, when notable heterogeneity was present, a random-effects model was used. To mitigate potential pleiotropic bias, we implemented MR-Egger regression to evaluate the systematic bias from pleiotropic effects through an intercept term analysis. Complementary sensitivity assessments included the MR-pleiotropy residual sum and outlier approach to systematically identify and remove genetic variants that exhibited pleiotropic distortions that might compromise causal estimates. A leave-one-out sensitivity analysis was conducted to assess whether any single SNP introduced bias affecting the overall causal conclusions. Scatter plots showed that no outliers significantly impacted the findings, and funnel plots confirmed the robustness of the association in the absence of heterogeneity. All statistical analyses were performed with a two-sided significance threshold of 0.05 and were executed using R software (R Foundation for Statistical Computing, Vienna, Austria). A multiple-testing-adjusted threshold of P<1.07×10−4 (0.05/466) was established based on the Bonferroni correction, to identify a statistically significant association [29]. Furthermore, metabolites with P<0.05, which exceeded the Bonferroni-corrected threshold, were reported as suggestive risk predictors for EM. The combination of these complementary approaches led to strong and dependable assessments of the association between circulating metabolites and EM.
Results
Two-sample bidirectional MR analysis was conducted to investigate the association between circulating metabolites and EM. The IVW method served as the primary analytical framework and was supplemented by the MR-Egger regression, simple mode, weighted mode, and weighted median approaches.
1. Exploration of the association effect of circulating metabolites on EM
The results of the MR analysis are summarized in Fig. 2. The IVW method suggested that genetically predicted higher levels of diacylglycerol were associated with an increased risk of EM (odds ratio [OR], 1.225; 95% confidence interva [CI], 1.136-1.321; P=1.16×10−7), corresponding to a 22.5% increase in risk per standard deviation increase in genetically predicted diacylglycerol levels, and this association remained significant after applying a strict Bonferroni correction. Several other metabolites showed nominally significant associations. Genetically predicted lower levels of 3-hydroxybutyrate (OR, 0.863; 95% CI, 0.767-0.972; P=0.015), a lower ratio of saturated fatty acids to total fatty acids (OR, 0.865; 95% CI, 0.759-0.986; P=0.030), and a lower ratio of total cholesterol to total lipids in very small very low-density lipoproteins (VLDL) (OR, 0.855; 95% CI, 0.789-0.927; P=1.51×10−4) were also associated with a higher EM risk. Furthermore, supplementary methodologies validated our results and demonstrated a consistent direction of the effect (Fig. 2). The stability of the identified association links was confirmed using various alternative methods and sensitivity analyses, as outlined in Supplementary Table 1. Visual representations, including scatter and funnel plots, further confirmed the consistency and reliability of the findings (Supplementary Fig. 1).
2. Exploration of the association effect of EM on circulating metabolites
The results of the reverse MR analysis are shown in Fig. 3. Using the IVW method, we found evidence suggesting that a genetic predisposition to EM was associated with several circulating metabolites. Specifically, EM showed nominally significant associations with higher levels of conjugated linoleic acid (CLA) (OR, 1.026; 95% CI, 1.000-1.053; P=0.043), ratios of CLA to total fatty acids (OR, 1.026; 95% CI, 1.000-1.053; P=0.045), and tyrosine levels (OR, 1.036; 95% CI, 1.016-1.057; P=3.33×10−4). To strengthen the associations identified in our study, we used multiple supplementary analytical approaches, along with sensitivity analyses (Supplementary Table 2). We also generated scatter and funnel plots to illustrate the robustness and validity of the results (Supplementary Fig. 2), further enhancing the credibility of the conclusions.
Discussion
In this study, we elucidated the association between 233 genetically predicted serum metabolites and EM using genetic variation as an instrumental variable within a two-sample MR framework. Our findings indicate associations between two metabolites (diacylglycerol and 3-hydroxybutyrate) and two metabolite ratios (the ratio of saturated fatty acids to total fatty acids and the total cholesterol to total lipid ratio in very small VLDL) in EM, while also demonstrating that EM influences two metabolites (CLA and tyrosine) and one metabolite ratio (the ratio of CLA to total fatty acids). Notably, one of these associations was statistically significant after correcting for multiple tests, suggesting a strong relationship. In this exploratory study, we found that multiple metabolites were associated with EM. This study contributes to the understanding of the metabolic factors involved in the pathogenesis of EM and provides a foundation for future studies and potential therapeutic targets.
Alterations in lipid metabolism have been associated with the onset and progression of EM, and previous studies have confirmed the dysregulation of various lipids, including phosphatidylcholines, sphingomyelins, phosphatidylethanolamines, and triglycerides [13,30,31]. Diacylglycerol is a significant lipid molecule involved in various cellular signaling pathways, including those regulating inflammation and cell proliferation [32]. Research has demonstrated significant differences between the endometriotic and endometrial tissues in these patients [33]. However, few studies have examined the relationship between diacylglycerol and EM. The findings of this study demonstrated a significant correlation between higher concentrations of diacylglycerol and elevated susceptibility to EM onset. These findings are consistent with those of previous studies. This study also found that decreased ratios of saturated to total fatty acids and total cholesterol to total lipids in very small VLDL were associated with an increased risk of EM. Therefore, our findings indicate a close relationship between lipid metabolism and EM, although further investigation of the underlying mechanisms is warranted.
3-hydroxybutyrate is one of the primary ketone bodies produced during fatty acid metabolism and serves not only as an intermediate metabolite but also as an important regulatory molecule. Research indicates that 3-hydroxybutyrate plays significant biological roles in the regulation of energy metabolism, as well as in antioxidant and anti-inflammatory responses [34]. Furthermore, 3-hydroxybutyrate can influence cellular survival and function by modulating intracellular signaling pathways and metabolic processes [34,35]. However, studies on the association between 3-hydroxybutyrate and EM are limited. Angioni et al. [36] analyzed 22 serum samples from patients with symptomatic EM and 10 from those without EM using gas chromatography-mass spectrometer and revealed a significant increase in 3-hydroxybutyric acid levels among the patients with EM. In our MR analysis, a nominally significant association was observed between lower genetically predicted levels of 3-hydroxybutyrate and an increased risk of EM. This exploratory finding suggests that altered 3-hydroxybutyrate metabolism may be involved in the pathogenesis of EM but requires verification in larger and more comprehensive studies. Our findings offer new insights into and perspectives on this issue.
Previous studies have demonstrated that individuals diagnosed with EM display statistically significant alterations in amino acid levels across various biological matrices, including the eutopic endometrium, serum, follicular fluid, urine, and endometrial fluid, compared to healthy controls [8]. However, the conclusions drawn from these studies are inconsistent. A study conducted by Pocate-Cheriet et al. [37] indicated that the concentrations of amino acids such as tyrosine are lower in women with deep-infiltrating EM than in control participants. In contrast, Li et al. [38] reported that the metabolomic profile of the eutopic endometrium in patients with EM is marked by a significant increase in L-tyrosine concentration. Our exploratory reverse MR analysis suggested that a genetic predisposition to EM was associated with higher circulating tyrosine levels. This finding is consistent with the observation by Li et al. [38] of increased L-tyrosine levels in the eutopic endometrium of patients with EM. Although this inverse association does not imply causality, it may reflect metabolic alterations secondary to the disease state or related pathophysiology. Thus, tyrosine metabolism could be an area of interest in EM, and its role warrants further investigation to determine whether it represents a compensatory mechanism, biomarker of disease activity, or contributor to progression.
CLA is a polyunsaturated fatty acid that previous studies have identified as playing significant roles in various biological processes. CLA possesses anti-inflammatory, antitumor, and immunomodulatory activities [39]. Furthermore, research indicates that CLA exerts its anticancer effects through mechanisms such as regulation of cell signaling pathways, inhibition of tumor cell proliferation, and induction of apoptosis [40]. Research on the relationship between CLA and EM is limited. Our reverse MR analysis generated a novel hypothesis by revealing a nominally significant association between a genetic predisposition to EM and elevated levels of CLA (and its ratio to total fatty acids). Given the established roles of local inflammation and immune dysregulation in EM [5,6], one speculative interpretation is that the body may upregulate CLA as a counter-regulatory response to the inflammatory milieu of EM. Alternatively, this association may highlight a dysregulated metabolic pathway in patients with EM. These findings suggest that CLA is a candidate molecule whose relationship with EM warrants further investigation to clarify its specific biological role.
This study had some limitations that warrant further investigation. First, a significant limitation of this study was its reliance on publicly available genetic datasets, which may pose constraints regarding the cohort scale, demographic representation, and spectrum of genetic variations associated with circulating metabolites and EM. Second, our study was subject to limitations inherent to metabolomic GWAS data. Metabolite measurements are platform-specific and may affect the coverage and comparability of certain metabolites. Additionally, potential heterogeneity in sample collection, processing, and quantification across original studies could influence the precision of our genetic instrument variables and the generalizability of the findings. Third, the metabolite data predominantly originated from European populations, and all EM-related data were derived from individuals of European ancestry. Differences in genetic determinants across ethnic groups (e.g., allele frequencies of genes involved in metabolic pathways), dietary habits, gut microbiome composition, and environmental exposures could contribute to variations in metabolite profiles, thus limiting the generalizability of our findings to different ethnic groups. Future studies incorporating large-scale metabolomic data and EM GWAS from diverse populations are crucial for validating and refining these findings. Fourth, although the study encompassed a relatively broad spectrum of metabolites, the functions and mechanisms of certain metabolites in relation to the disease remain inadequately understood, which limits the interpretation of our results from this MR analysis. Consequently, further validation is necessary by repeating the study in different groups and performing functional investigations to reinforce our findings and clarify the underlying mechanisms.
In summary, this study employed bidirectional two-sample MR to generate hypotheses regarding the association between circulating metabolites and EM. While one association survived strict Bonferroni correction and prioritized a metabolic pathway for future mechanistic investigations, all findings, including those that are suggestive, must be considered hypothesis-generating. They collectively proposed novel metabolic risk factors for EM that require substantial functional validation and replication in larger independent cohorts before clinical translation can be considered. Finally, this study provides a foundation for future research exploring the underlying biology and potential therapeutic targets of EM.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Full text via PMC
Download Citation
Supplement
Print
