ABSTRACT
Purpose
No biomarkers specific to true unexplained recurrent pregnancy loss (RPL) have yet been identified. This study aimed to investigate the lipid profile of patients with unexplained RPL using metabolomic analyses.
Methods
Comprehensive serum metabolomic profiling was performed using liquid chromatography–tandem mass spectrometry. Lipid metabolites were analyzed using orthogonal partial least squares discriminant analysis.
Results
The final analysis included serum samples from 31 healthy non‐pregnant controls, 15 healthy pregnant controls, 32 non‐pregnant patients with RPL, 19 pregnant patients with RPL and euploid miscarriage, and 30 pregnant patients with RPL and aneuploid miscarriage. Compared with controls, the non‐pregnant RPL group exhibited higher levels of inflammation‐related metabolites, including 11‐dehydro‐thromboxane B2 (11‐dehydro‐TXB2), 12‐hydroxyeicosatetraenoic acid (HETE), and prostaglandin E2 (PGE2). Pregnant controls showed increased arachidonic acid, eicosapentaenoic acid, and docosahexaenoic acid metabolites. Comparisons between pregnant controls and patients with euploid or aneuploid miscarriages showed no clear separation. Enzyme‐linked immunosorbent assay confirmed elevated serum 11‐dehydro‐TXB2 in non‐pregnant patients with RPL.
Conclusion
Patients with RPL display a distinct serum lipid signature in the non‐pregnant state, characterized by elevated levels of pro‐thrombotic and pro‐inflammatory eicosanoids, including 11‐dehydro‐TXB2, 12‐HETE, and PGE2. Further studies are warranted to clarify the potential utility of 11‐dehydro‐TXB2 as biomarkers for RPL.
Keywords: LC–MS/MS, lipidomics, metabolomics, pregnancy, recurrent pregnancy loss
1. Introduction
Pregnancy profoundly affects various metabolic processes. Recently, the relationship between recurrent pregnancy loss (RPL) and lipid metabolism has garnered increasing attention. Dyslipidaemia, particularly hypertriglyceridaemia and decreased high‐density lipoprotein cholesterol (HDL‐C) levels, is associated with endothelial dysfunction and chronic inflammation [1]. These abnormalities may impair the placental function and result in inadequate uteroplacental blood flow, thereby increasing the risk of RPL [2, 3, 4].
Lipid metabolism during pregnancy is characterized by triglycerides (TG) accumulation during the first and second trimesters, reflecting predominant anabolic activity, followed by increased circulating lipid levels during the third trimester as catabolic processes predominate [5, 6, 7]. Recent studies revealed an association between maternal blood lipid levels during pregnancy and preterm birth [8], further highlighting their clinical importance. TG consists of a glycerol backbone esterified to three fatty acids. During metabolism, TG is hydrolysed into fatty acids, some of which serve as precursors for inflammatory mediators, including eicosanoids [9], thereby establishing a mechanistic link between lipid and inflammation during pregnancy.
RPL is defined as two or more spontaneous pregnancy losses [10, 11, 12] and has an estimated prevalence of approximately 5% in Japan [13]. Established causes of RPL include antiphospholipid syndrome (APS), uterine anomalies, and parental or embryonic chromosome abnormalities. However, approximately 25% of cases are classified as unexplained RPL with normal embryonic chromosomal findings [14]. Currently, no evidence‐based treatments are available for patients with unexplained RPL. Accordingly, there is an urgent need to elucidate the pathological mechanisms underlying unexplained RPL and to develop effective therapeutic strategies.
Unexplained RPL is widely considered to result from the interaction of multiple genetic and environmental factors [15]. Metabolomics, which examines metabolites that directly reflect the physiological state of an organism, provides a powerful approach for identifying factors that contribute to RPL. Previous metabolomic studies have identified abnormalities in multiple metabolic pathways, including vitamin digestion, tryptophan metabolism, the citrate cycle, arginine biosynthesis, and glycerophospholipid and sphingolipid metabolism, in the decidua of patients with RPL [16]. However, these studies did not consider embryonic chromosomal abnormalities, and no biomarkers specific to true unexplained RPL have yet been identified [17, 18]. Therefore, this study aimed to perform a metabolomic analysis of blood samples from patients with unexplained RPL to identify aberrant metabolites and the metabolic pathways associated with RPL development.
2. Materials and Methods
2.1. Study Design and Population
The lipid profile using metabolomics analyses was compared between patients with RPL and healthy women with a history of live births and no miscarriages (control) at non‐pregnant status, between healthy pregnant women with no history of RPL (pregnant control) and the controls, and between pregnant patients with RPL whose subsequent outcomes were euploid (euploid miscarriage) or aneuploid miscarriage (aneuploid miscarriage) and the pregnant controls. Patients with RPL were recruited at Nagoya City University Hospital between 2022 and 2024. Patients with parental chromosomal translocations, uterine anomalies, APS, thyroid dysfunction, or impaired glucose tolerance were excluded. When a missed miscarriage was diagnosed, dilation and curettage were performed, followed by cytogenetic analysis of the products of conception.
Before a subsequent pregnancy, participants underwent three dimensional ultrasonography, chromosomal analysis of both partners, assessment of lupus anticoagulant (LA) using diluted activated partial thromboplastin time and diluted Russell's viper venom time, measurement of β2‐glycoprotein I dependent anticardiolipin antibodies for the diagnosis of APS, evaluation of thyroid function using thyroid‐stimulating hormone, free thyroxine (FT4) and screening for glucose metabolism abnormalities using fasting insulin and fasting glucose, with glycated hemoglobin (HbA1c) measured when indicated.
Samples were collected from the pregnant controls at Hoshigaoka Maternity Hospital and Nagoya City University Hospital during the same period. Samples were collected from the controls aged 20–43 years at the Midtown Clinic, a health check‐up centre, during the same period.
2.2. Sample Preparation for Liquid Chromatography Mass Spectrometry Analysis
Serum samples were stored at −80°C until analysis. For metabolite extraction, 30 μL of serum was mixed with 300 μL of 0.1% (v/v) formic acid in methanol (1:10, v/v) and 10 μL of an internal standard mixture. The solution was homogenized on a rotating mixer for 3 min at 4°C and then centrifuged at 13,000 rpm for 10 min at 4°C. The resulting supernatant was diluted with 900 μL of 0.1% (v/v) aqueous formic acid and applied to a solid‐phase extraction column (Strata‐X, 33 μm, 10 mg/mL, 85 Å; Phenomenex, Torrance, CA, USA). The columns were preconditioned sequentially with 1 mL 0.1% (v/v) formic acid/methanol and 1 mL of 0.1% (v/v) aqueous formic acid. After sample loading, each column was rinsed sequentially with 1 mL of 0.1% (v/v) aqueous formic acid, 1 mL of 0.1% (v/v) formic acid in 15% (v/v) ethanol, and finally 1 mL of hexane. Analytes were eluted with 300 μL of 0.1% (v/v) formic acid in methanol. The eluate was dried under vacuum, and the residue was reconstituted in 30 μL of 0.1% (v/v) formic acid in methanol for liquid chromatography–tandem mass spectrometry (LC–MS/MS) analysis.
2.3. LC–MS/MS Analysis
Chromatographic separation was performed using an ultra‐high‐performance liquid chromatography system (Nexera X2, Shimadzu) equipped with a C8 column (2.1 × 150 mm, 2.6 μm; Phenomenex), which was maintained at 40°C. The mobile phase comprised 0.1% formic acid in water (A) and acetonitrile (B), with a gradient elution at 0.4 mL/min as follows: 0–5 min, 10%–25% B; 5–10 min, 35% B; 10–20 min, 75% B; 20–20.1 min, 98% B. The injection volume was set at 5 μL.
Lipid metabolites were analyzed using a triple quadrupole mass spectrometer (LCMS‐8040, Shimadzu) equipped with an electrospray ionization source operating in multiple reaction monitoring mode. The source parameters were as follows: interface temperature, 270°C, desolvation line temperature, 250°C, heat block temperature, 400°C, nebulising gas flow rate, 3.0 L/min, and drying gas flow rate, 15 L/min. The collision‐induced dissociation gas pressure was maintained at 230 kPa. The mass‐to‐charge (m/z) values for the parent and fragment ions, along with the associated voltages, were configured using an optimisation support software (Shimadzu LabSolutions).
2.4. Measurement of Serum Levels of Total Cholesterol, Glucose, and Triglycerides
Serum levels of total cholesterol (T‐Chol), glucose, and TG were measured using LabAssay kits for cholesterol (Cholesterol Oxidase–DAOS method), glucose (Mutarotase–GOD method), and TG (GPO–DAOS method) kits (Fujifilm Wako Pure Chemicals, Osaka, Japan). All assays were performed according to the manufacturer's instructions.
2.5. Validation Using an Enzyme‐Linked Immunosorbent Assay
After identifying the metabolites, the same serum samples used for LC–MS/MS analysis, comprising 31 controls and 32 patients with RPL, were analyzed using an enzyme‐linked immunosorbent assay (ELISA) with the 11‐dehydro Thromboxane B2 Monoclonal ELISA Kit (Cayman Chemical, Ann Arbor, MI, USA).
2.6. Statistical Analysis
Lipid metabolomic data were processed and analyzed using MetaboAnalyst version 5.0. Multivariate statistical analysis was performed using orthogonal projections to latent structures‐discriminant analysis (OPLS‐DA) to maximize group separation and identify the metabolites responsible for discrimination.
For each metabolite, pairwise group comparisons were performed using the Wilcoxon rank‐sum test. For each metabolite, all pairwise comparisons were performed among the three groups: controls, pregnant women, and patients with RPL. Raw p‐values from the three pairwise comparisons were adjusted using the Bonferroni correction to account for multiple testing. Statistical analyses were performed using the R software (version 4.3.1; R Foundation for Statistical Computing, Vienna, Austria). A p < 0.05 was considered statistically significant for all analyses.
3. Results
3.1. Study Design
Age‐stratified analyses revealed marked differences in metabolite profiles between participants in their 20s and those in their 40s when patients with RPL were compared with the controls and pregnant women. Therefore, participants aged 40 years or older were excluded from each group to minimize the potential confounding effect of age. Finally, serum samples were obtained from 32 patients with RPL, 31 controls, 15 pregnant controls, 19 euploid miscarriages, and 30 aneuploid miscarriages (Figure 1). Background information such as age, body mass index, basic blood test data (blood glucose, cholesterol, and TG levels), and the number of previous miscarriages were presented in Table 1. Two patients experienced RPL, comprising one early miscarriage and one intrauterine foetal death. The mean gestational ages of the healthy pregnant controls, euploid miscarriage, and aneuploid miscarriage groups at the time of sample collection were 9.13 ± 1.13 weeks, 8.21 ± 1.27 weeks, and 8.33 ± 1.24 weeks, respectively (mean ± SD).
FIGURE 1.
Flow diagram of participant selection and metabolomic analysis. Flowchart showing participant selection, exclusion criteria, and the final study cohorts included in the liquid chromatography–tandem mass spectrometry‐based lipid metabolomic analysis.
TABLE 1.
Characteristics of patients and controls included in the metabolomic analysis.
| Factor | Control | Pregnant control | Patients with RPL | p† | p‡ | |||
|---|---|---|---|---|---|---|---|---|
| n = 31 | n = 15 | n = 32 | ||||||
| n (%) | n (%) | n (%) | ||||||
| Age (years), mean (SD) | 35.1 (3.5) | 33.7 (2.7) | 33.8 (3.3) | 0.067 | 0.085 | |||
| BMI (kg/m2), mean (SD) | 20.6 (2.8) | 22.0 (2.6) | 20.9 (2.3) | 0.563 | 0.079 | |||
| Glucose level (mg/dL), mean (SD) | 78.9 (14.1) | 82.2 (27.7) | 72.6 (21.7) | 0.074 | 0.064 | |||
| Total cholesterol level (mg/dL), mean (SD) | 159.3 (32.0) | 101.5 (32.9) | 144.0 (40.5) | 0.155 | < 0.001 | |||
| Triglyceride level (mg/dL), mean (SD) | 41.5 (30.2) | 71.0 (21.0) | 79.2 (20.5) | < 0.001 | < 0.01 | |||
| Number of previous pregnancy losses | ||||||||
| 0 | 31 | 100 | 12 | 80.0 | 0 | 0 | ||
| 1 | 0 | 0 | 3 | 20.0 | 0 | 0 | ||
| 2 | 0 | 0 | 0 | 0 | 17 a | 53.1 | ||
| 3 | 0 | 0 | 0 | 0 | 12 | 37.4 | ||
| ≥ 4 | 0 | 0 | 0 | 0 | 3 | 9.4 | ||
| Number of previous live births | ||||||||
| 0 | 0 | 0 | 9 | 60.0 | 20 | 62.5 | ||
| 1 | 13 | 41.9 | 2 | 13.3 | 12 | 37.5 | ||
| 2 | 14 | 45.2 | 3 | 20.0 | 0 | 0 | ||
| ≥ 3 | 4 | 12.9 | 1 | 6.7 | 0 | 0 | ||
| Antiphospholipid antibodies | ||||||||
| Presence | 0 | 0 | 0 | |||||
| Absence | 0 | 0 | 32 | |||||
| Unknown | 31 | 15 | 0 | |||||
Note: The p† represents the statistical difference between the control and pregnant women groups. The p‡ represents the statistical difference between the control and patients with RPL groups.
Abbreviation: BMI, body mass index.
Two patients experienced two pregnancy losses, consisting of one early miscarriage and one intrauterine foetal death.
3.2. Metabolomic Profiling of Serum Samples
Targeted serum metabolomic profiling was performed to investigate systemic metabolic alterations associated with RPL and pregnancy status. All analyses were based on normalized metabolite area ratios derived from LC–MS/MS data.
3.2.1. Metabolic Changes Associated With Normal Pregnancy
The pregnant controls showed distinct serum lipid metabolomic profiles compared with the controls (Figure 2A). The OPLS‐DA revealed a clear separation between the groups, demonstrating high explanatory and predictive capabilities (R 2 Y = 0.916, Q 2 = 0.992, p < 0.001; Figure 2B). The volcano plot shows all identified metabolites (Figure 2C). Some metabolites exhibited significant differences. Variable importance in the projection (VIP) analysis identified several key metabolites, including 12‐hydroxyheptadecatrienoic acid (12‐HHT), arachidonic acid (AA), 12‐hydroxyeicosatetraenoic acid (12‐HETE), eicosapentaenoic acid (EPA), and docosahexaenoic acid (DHA) (Figure 2D). A heat map of these differentially abundant metabolites (Figure 2E) showed elevated levels of thromboxane B1 (TXB1), thromboxane B3 (TXB3), 12‐HHT, AA, EPA, and DHA in the pregnant control group, whereas 13‐hydroxydocosahexaenoic acid (13‐HDHA) and prostaglandin D1 (PGD1) were more abundant in the control group.
FIGURE 2.
Serum lipid profiles of the pregnant controls compared with the controls. (A) Orthogonal projections to latent structures‐discriminant analysis (OPLS‐DA) score plot showing the distinction between pregnant women and the controls. (B) Permutation test (1000 iterations) of the model in (A), validating its predictive reliability (Q 2 = 0.916, R 2 Y = 0.992, p < 0.001). (C) Volcano plot of the pregnant women and the controls. (D) Variable importance in projection plot from OPLS‐DA comparing the pregnant women and the controls. (E) Heatmap comparing the pregnant women and the controls.
3.2.2. Distinct Metabolic Signature in Non‐Pregnant Patients With RPL
OPLS‐DA revealed a clear separation between non‐pregnant patients with RPL and the controls, indicating distinct metabolic profiles (Figure 3A). The model demonstrated high explanatory and predictive capabilities, with R 2 Y = 0.895 and Q 2 = 0.793 (p < 0.001), as confirmed by permutation testing (Figure 3B). The volcano plot shows all identified metabolites (Figure 3C). Some metabolites exhibited significant differences. VIP analysis identified several metabolites with VIP scores greater than 1, including 12‐HHT, TXB1, 9‐hydroxy‐octadecadienoic acid (9‐HODE), 12‐hydroxyeicosapentaenoic acid (12‐HEPE), and 11‐dehydro‐TXB2 (Figure 3D). Among the significantly altered metabolites, 11‐dehydro‐TXB2, 12‐hydroxyeicosatetraenoic acid (12‐HETE), and prostaglandin E2 (PGE2) each met the fold change > 4 (see the comprehensive differential metabolite plot, Table S1). Heatmap visualization further illustrated clear differentiation between the two groups based on these discriminative, differentially abundant metabolites (Figure 3E). Collectively, these metabolites define a distinct metabolic signature that effectively separates non‐pregnant patients with RPL from controls.
FIGURE 3.
Serum lipid profiles of patients with RPL compared with the controls. (A) Orthogonal projections to latent structures‐discriminant analysis (OPLS‐DA) score plot demonstrating the separation between patients with recurrent pregnancy loss (RPL) and the controls. (B) Permutation test (1000 iterations) of the model in (A), confirming its ability and non‐overfitting (Q 2 = 0.793, R 2 Y = 0.895, p < 0.001). (C) Volcano plot of patients with RPL and the controls. (D) Variable importance in projection plot from OPLS‐DA comparing patients with RPL and the controls. (E) Heatmap comparing patients with RPL and the controls.
3.2.3. Integrated Three‐Group Metabolic Comparison
We visualized the distribution of previously identified discriminative metabolites across all three groups to contextualize the metabolic state associated with RPL. Figure 4 presents the normalized peak area ratios displayed as side‐by‐side metabolic pathway maps. The Wilcoxon rank‐sum test identified specific patterns of significant differences, with pronounced alterations observed particularly in the cyclooxygenase (COX) and lipoxygenase (LOX) pathways. This integrated analysis confirmed that metabolites driving separation in the pairwise models also exhibited consistent variation across a broader clinical spectrum. All metabolites exhibiting statistically significant differences are shown in Figure S1.
FIGURE 4.
Metabolic pathway map, which shows significant differences among the controls versus pregnant women versus patients with RPL. Metabolic pathways showing significant differences among the three groups. Wilcoxon rank‐sum tests reveal specific patterns of significant differences, particularly in metabolites within the cyclooxygenase and lipoxygenase pathways (*p < 0.05, **p < 0.01, ***p < 0.001).
3.2.4. Lipid Profiles and Embryonic Chromosomal Status
Additional OPLS‐DA comparisons were performed to assess the influence of embryonic chromosomal abnormalities. No significant metabolic separation was observed between pregnant controls and patients with RPL experiencing euploid miscarriages, nor between pregnant controls and patients with RPL experiencing aneuploid miscarriages, nor between patients with euploid miscarriages and patients with aneuploid miscarriages.
3.3. ELISA‐Based Confirmation of 11‐Dehydro‐TXB2 Elevation in Non‐Pregnant Patients With RPL
Furthermore, we focused on 11‐dehydro‐TXB2, a metabolite known for its stability in the blood, and performed quantitative analysis using ELISA on serum samples. The results are presented in Table 2. The ELISA findings were consistent with trends observed in the OPLS‐DA and differential expression analyses, with significantly higher concentrations of 11‐dehydro‐TXB2 observed in non‐pregnant patients in the RPL group than in the control group.
TABLE 2.
Comparison of 11‐dehydro‐TXB2 of serum samples.
| Control (n = 31) | Patients with RPL (n = 32) | p | |
|---|---|---|---|
| 11‐dehydro‐TXB2, mean ± SD (pg/mL) | 171.4 ± 113.7 | 4025 ± 9236 | < 0.001 |
Abbreviation: 11‐dehydro‐TXB2 = 11‐dehydro‐Tromboxane B2.
4. Discussion
4.1. Strengths of This Study
This study elucidated characteristic features of the serum lipid metabolome in non‐pregnant patients with RPL, pregnant controls, and controls. The OPLS‐DA model demonstrated clear separation between the controls and the pregnant controls, as well as between the controls and patients with RPL, indicating significant overall metabolic differences. Some metabolites within the COX and LOX pathways showed significant differences among the three groups, enabling comparison of metabolic profiles while explicitly accounting for pregnancy status. Furthermore, 11‐dehydro‐TXB2, as validated by ELISA, was significantly different in non‐pregnant patients with RPL, supporting the potential of 11‐dehydro‐TXB2 as a promising biomarker associated with RPL.
These findings align with known pathophysiological mechanisms involving inflammation and pro‐thrombotic tendencies in RPL [19]. Importantly, this difference was evident even when patients with RPL were not pregnant, suggesting that the presence of a persistent pro‐inflammatory and pro‐thrombotic metabolic state in the circulation extends beyond the conventional view of inflammation and thrombosis as pregnancy‐limited phenomena [20, 21].
Methodologically, the use of serum samples is critical because serum collection avoids interference from anticoagulants during platelet activation. The clotting process ensures maximal platelet activation, resulting in the release of thromboxane A2 (TXA2) and its rapid, stable conversion to thromboxane B2 (TXB2). This approach ensures highly reproducible thromboxane measurements and enables direct comparability across all samples due to standardized in vitro activation [22, 23, 24].
Among the metabolites showing the largest differences, 11‐dehydro‐TXB2, a stable TXA2 metabolite of TXA2, was markedly elevated in non‐pregnant patients with RPL. This finding strongly suggests enhanced platelet activation and turnover in RPL [25, 26, 27, 28]. Overproduction of 12‐HETE, a product of the 12/15‐lipoxygenase pathway, may contribute to reproductive pathology [29]. 12‐HETE functions as a transient lipid signaling molecule that increases during embryonic implantation. In mouse models, uterine levels of 12‐HETE and 15‐HETE increased sharply on the day of implantation, whereas inhibitory lipids, such as 9‐HODE, remained low [30]. In a pathological context, excess 12‐HETE has been associated with placental dysfunction in conditions such as preeclampsia [31, 32]. In the present study, PGE2 was also significantly elevated in patients with RPL compared with non‐pregnant controls, consistent with findings of previous studies showing higher PGE2 concentrations in the cervical mucus and endometrial tissue of women with recurrent implantation failure [33]. The lipid metabolic abnormalities observed in the present study are consistent with previous reports linking maternal lipid dysregulation to pregnancy‐related disorders. Abnormal maternal lipid profiles have been associated with gestational diabetes mellitus, hypertensive disorders of pregnancy, preterm birth, and adverse perinatal outcomes [34].
These metabolic alterations may be associated with the development of cardiovascular diseases in patients with RPL [35].
Notably, the three‐group comparative design revealed that the RPL metabolic signature associated with RPL was distinct from the physiological metabolic reprogramming observed in healthy pregnancies. Healthy pregnant women showed upregulation of AA, EPA, DHA, and 12‐HHT, indicative of active polyunsaturated fatty acid metabolism that supports foetal development and prostaglandin synthesis during processes such as angiogenesis. By contrast, the RPL metabolic profile was characterized by elevated levels of 11‐dehydro‐TXB2 and 12‐HETE. These findings demonstrate that RPL represents a distinct pathological state rather than a simple deficiency in normal pregnancy adaptation.
Collectively, these metabolomic findings provide new pathophysiological insights, indicating that patients with RPL maintain a biochemical environment skewed towards inflammation and thrombosis even in the non‐pregnant state. This persistent abnormality may predispose affected women to placental dysfunction during the earliest stages of a subsequent pregnancy. The identification of a distinct serum lipid profile represents a substantial advance, particularly in the absence of validated clinical biomarkers for unexplained RPL risk. These findings lay the groundwork for improved risk stratification and underscore the urgent need to define predictive biomarkers capable of identifying at‐risk individuals before pregnancy loss occurs [36].
The three‐group comparative metabolomics approach employed in this study provides a key advantage over conventional two‐group analyses. The inclusion of both non‐pregnant and pregnant control groups enabled effective differentiation between pregnancy‐induced metabolic changes from those specific to RPL pathology, thereby allowing more precise dissection of underlying mechanisms and more reliable identification of RPL‐specific metabolic signatures.
4.2. Clinical Implications
The clinical importance of the metabolomic analysis lies in its potential to support the development of novel serum biomarkers for RPL risk assessment. Currently, clinicians have limited tools to predict whether women are at high risk of miscarriage. If validated in large‐scale cohorts or prospective studies, metabolites, such as 11‐dehydro‐TXB2, could serve as objective markers of latent thrombotic and inflammatory predispositions underlying pregnancy loss. For example, during pre‐conception screening, the measurement of urinary or serum 11‐dehydro‐TXB2—a stable terminal metabolite that is readily measurable in clinical laboratories—may provide a practical means of identifying women at high risk [37, 38].
4.3. Comparison of the Pregnant Controls, Euploid Miscarriages, and Aneuploid Miscarriages
Since chromosomal aneuploidy itself is an independent cause of miscarriage, metabolic differences were not expected between the pregnant patients with aneuploid miscarriage and the pregnant controls. By contrast, chromosomal euploid miscarriage represents a form of unexplained RPL, and differences were therefore expected between these patients and the pregnant controls. However, no significant metabolic separation was observed in either comparison. Several interconnected factors may account for this finding.
Unexplained RPL includes heterogeneous causes, including genetic variants and epigenomic abnormalities [15, 39]. Therefore, it is difficult to explain all causes solely by abnormalities in lipid metabolism.
Pregnancy induces profound systemic metabolic adaptations. These changes, particularly in lipid and eicosanoid pathways (e.g., increased levels of arachidonic acid, EPA, DHA, and various prostaglandins), may mask subtle RPL‐specific metabolic abnormalities in the circulating metabolome.
The timing of blood sampling is also an important factor. Blood samples were obtained after the diagnosis of missed miscarriage but before surgical intervention, while the pregnancy was still ongoing. At this stage, the maternal metabolic system continues to respond to the presence of the conceptus, and the metabolic alterations induced by the failing pregnancy may not fundamentally differ from those of a normal early pregnancy in the peripheral circulation.
Another important point is the distinction between local and systemic metabolic abnormalities. The pathogenesis of euploid miscarriage likely involves localized abnormalities at the maternal–foetal interface, including defective decidualization, impaired spiral artery remodeling, or aberrant trophoblast invasion. These local processes may not be sufficiently reflected in the systemic blood metabolome [40, 41, 42]. In the present study, circulating metabolites were measured, which integratively reflect whole‐body metabolism but may not fully reflect localized metabolic alterations at the maternal–foetal interface. Therefore, the absence of differences in peripheral blood does not exclude the presence of RPL‐related metabolic disturbances within the uterine microenvironment.
Additionally, the biological heterogeneity of euploid miscarriage itself (different underlying aetiologies) may have further obscured consistent metabolic patterns.
Taken together, these considerations do not contradict the main conclusion of the present study that non‐pregnant patients with RPL exhibit a distinct metabolic signature compared with healthy controls. Rather, they suggest that physiological metabolic adaptations during pregnancy may obscure RPL‐specific metabolic abnormalities, making these signatures more readily detectable in the non‐pregnant state.
4.4. Limitations
This study has some limitations. The small sample size may limit the generalisability of the findings and the statistical power to detect subtle metabolic changes. Since implantation and placentation are highly dependent on local immune, vascular, and inflammatory dynamics, future studies integrating paired analyses of circulating and uterine (endometrial or decidual) metabolomes are essential for clarifying the mechanistic links between systemic metabolic profiles and intrauterine reproductive failure. Further, this was a cross‐sectional study, which limits causal inference. Further validation is therefore required to determine whether the identified metabolites can predict pregnancy outcomes. Across the patients and the controls, non‐pregnant and pregnant women differed in several respects. It was hypothesized that alterations in lipid metabolism due to pregnancy, as well as individual differences, would be affected. This heterogeneity represents an additional limitation of the study.
5. Conclusion
Patients with RPL display a distinct serum lipid signature in the non‐pregnant state, characterized by elevated levels of pro‐thrombotic and pro‐inflammatory eicosanoids, including 11‐dehydro‐TXB2, 12‐HETE, and PGE2. These findings suggest that these metabolites may serve as potential biomarkers for RPL risk stratification, particularly when assessed before conception. However, the absence of substantial metabolic differences during ongoing pregnancy, likely due to masking by pregnancy‐induced physiological changes, indicates that further large‐scale, prospective studies are needed to validate these observations and to determine their clinical utility.
Funding
This study was supported by the Ministry of Education, Culture, Sports, Science and Technology (MEXT) Promotion of the Distinctive Joint Research Center Program (grant number JPMXP0621467963) and JSPS KAKENHI (grant number JP22H03227).
Ethics Statement
This study was approved by the Ethics Committee of Nagoya City University (approval number 60‐22‐0046), Okayama University, and all participating hospitals (approval number 2211‐007).
Consent
Written informed consent was obtained from each participant before participation in the study.
Conflicts of Interest
The authors declare no conflicts of interest.
Supporting information
Figure S1: Distribution of discriminative lipid metabolites across three study groups. This figure presents normalized peak area ratios of previously identified discriminative metabolites across the healthy controls, the pregnant controls, and patients with recurrent pregnancy loss. Statistical comparisons were performed using the Wilcoxon rank‐sum test. Significant alterations were mainly observed in metabolites involved in the cyclooxygenase and lipoxygenase pathways.
Table S1: Differentially expressed lipid metabolites identified by metabolomic analysis.
Data Availability Statement
Data are not publicly available. However, on reasonable request, data can be obtained from the corresponding author.
References
- 1. Chung Y., Melo P., Pickering O., Dhillon‐Smith R., Coomarasamy A., and Devall A., “The Association Between Dietary Patterns and Risk of Miscarriage: A Systematic Review and Meta‐Analysis,” Fertility and Sterility 120 (2023): 333–357. [DOI] [PubMed] [Google Scholar]
- 2. Ball E., Bulmer J. N., Ayis S., Lyall F., and Robson S. C., “Late Sporadic Miscarriage Is Associated With Abnormalities in Spiral Artery Transformation and Trophoblast Invasion,” Journal of Pathology 208 (2006): 535–542. [DOI] [PubMed] [Google Scholar]
- 3. Brosens I., Pijnenborg R., Vercruysse L., and Romero R., “The “Great Obstetrical Syndromes” Are Associated With Disorders of Deep Placentation,” American Journal of Obstetrics and Gynecology 204 (2011): 193–201. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Pijnenborg R., Vercruysse L., and Brosens I., “Deep Placentation,” Best Practice & Research. Clinical Obstetrics & Gynaecology 25 (2011): 273–285. [DOI] [PubMed] [Google Scholar]
- 5. Mulder T., Louwen F., van Poppel M., and Faas M., “Lipid Metabolism During Pregnancy: Consequences for Mother and Fetus,” Current Opinion in Lipidology 35 (2024): 123–131. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Bartels Ä. and O'Donoghue K., “Cholesterol in Pregnancy: A Review of Knowns and Unknowns,” Obstetric Medicine 4 (2011): 147–151. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Brizzi P., Tonolo G., Esposito F., et al., “Lipoprotein Metabolism During Normal Pregnancy,” American Journal of Obstetrics and Gynecology 181 (1999): 430–434. [DOI] [PubMed] [Google Scholar]
- 8. Lv Y., Xu L., He Z., Liu X., and Guo Y., “The Association Between Pregnancy Levels of Blood Lipids and the Risk of Preterm Birth,” Scientific Reports 14 (2024): 10800. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Dichlberger A., Schlager S., Maaninka K., Schneider W. J., and Kovanen P. T., “Adipose Triglyceride Lipase Regulates Eicosanoid Production in Activated Human Mast Cells,” Journal of Lipid Research 55 (2014): 2471–2478. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Quenby S., Gallos I. D., Dhillon‐Smith R. K., et al., “Miscarriage Matters: The Epidemiological, Physical, Psychological, and Economic Costs of Early Pregnancy Loss,” Lancet 397 (2021): 1658–1667. [DOI] [PubMed] [Google Scholar]
- 11. ESHRE Guideline Group on RPL , Bender Atik R., Christiansen O. B., et al., “ESHRE Guideline: Recurrent Pregnancy Loss: An Update in 2022,” Human Reproduction Open 2023 (2023): hoad002. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Practice Committee of the American Society for Reproductive Medicine , “Definitions of Infertility and Recurrent Pregnancy Loss: A Committee Opinion,” Fertility and Sterility 99 (2013): 63. [DOI] [PubMed] [Google Scholar]
- 13. Sugiura‐Ogasawara M., Ebara T., Yamada Y., et al., “Adverse Pregnancy and Perinatal Outcome in Patients With Recurrent Pregnancy Loss: Multiple Imputation Analyses With Propensity Score Adjustment Applied to a Large‐Scale Birth Cohort of the Japan Environment and Children's Study,” American Journal of Reproductive Immunology 81 (2019): e13072. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Sugiura‐Ogasawara M., Ozaki Y., Katano K., Suzumori N., Kitaori T., and Mizutani E., “Abnormal Embryonic Karyotype Is the Most Frequent Cause of Recurrent Miscarriage,” Human Reproduction 27 (2012): 2297–2303. [DOI] [PubMed] [Google Scholar]
- 15. Sonehara K., Yano Y., Naito T., et al., “Common and Rare Genetic Variants Predisposing Females to Unexplained Recurrent Pregnancy Loss,” Nature Communications 15 (2024): 5744. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Wang L. L., Liu H., Zhao S. J., et al., “The Metabolic Landscape of Decidua in Recurrent Pregnancy Loss Using a Global Metabolomics Approach,” Placenta 112 (2021): 45–53. [DOI] [PubMed] [Google Scholar]
- 17. Wang X. H., Xu S., Zhou X. Y., et al., “Low Chorionic Villous Succinate Accumulation Associates With Recurrent Spontaneous Abortion Risk,” Nature Communications 12 (2021): 3428. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Li M., Haixia Y., Kang M., et al., “The Arachidonic Acid Metabolism Mechanism Based on UPLC‐MS/MS Metabolomics in Recurrent Spontaneous Abortion Rats,” Frontiers in Endocrinology (Lausanne) 12 (2021): 652807. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Grandone E. and Piazza G., “Thrombophilia, Inflammation, and Recurrent Pregnancy Loss: A Case‐Based Review,” Seminars in Reproductive Medicine 39 (2021): 62–68. [DOI] [PubMed] [Google Scholar]
- 20. Patil R., Ghosh K., Satoskar P., and Shetty S., “Elevated Procoagulant Endothelial and Tissue Factor Expressing Microparticles in Women With Recurrent Pregnancy Loss,” PLoS One 8 (2013): e81407. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Lu X., Liu Z., Zhang X., Kang X., Shen W., and Zhao A., “Prothrombotic State of Patients With Unexplained Recurrent Spontaneous Abortion,” International Journal of Gynaecology and Obstetrics 131 (2015): 161–165. [DOI] [PubMed] [Google Scholar]
- 22. Good R. I. S., McGarrity A., Sheehan R., et al., “Variation in Thromboxane B2 Concentrations in Serum and Plasma in Patients Taking Regular Aspirin Before and After Clopidogrel Therapy,” Platelets 26 (2015): 17–24. [DOI] [PubMed] [Google Scholar]
- 23. Breier M., Wahl S., Prehn C., et al., “Targeted Metabolomics Identifies Reliable and Stable Metabolites in Human Serum and Plasma Samples,” PLoS One 9 (2014): e89728. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Agueusop I., Musholt P. B., Klaus B., Hightower K., and Kannt A., “Short‐Term Variability of the Human Serum Metabolome Depending on Nutritional and Metabolic Health Status,” Scientific Reports 10 (2020): 16310. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Ellis E. F., Oelz O., L. J. Roberts, 2nd , et al., “Coronary Arterial Smooth Muscle Contraction by a Substance Released From Platelets: Evidence That It Is Thromboxane A2 ,” Science 193 (1976): 1135–1137. [DOI] [PubMed] [Google Scholar]
- 26. Salzman P. M., Smith J. B., and Lefer A. M., “Effects of Prostaglandins and Thromboxane A2 on Vascular Reactivity,” Journal of Pharmacology and Experimental Therapeutics 215 (1980): 240–247. [PubMed] [Google Scholar]
- 27. Hamberg M., Svensson J., and Samuelsson B., “Thromboxanes: A New Group of Biologically Active Compounds Derived From Prostaglandin Endoperoxides,” Proceedings of the National Academy of Sciences of the United States of America 72 (1975): 2994–2998. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Flood K., Peace A., Kent E., et al., “Platelet Reactivity and Pregnancy Loss,” American Journal of Obstetrics and Gynecology 203 (2010): 281.e1–281.e5. [DOI] [PubMed] [Google Scholar]
- 29. Dobrian A. D., Lieb D. C., Cole B. K., Taylor‐Fishwick D. A., Chakrabarti S. K., and Nadler J. L., “Functional and Pathological Roles of the 12‐ and 15‐Lipoxygenases,” Progress in Lipid Research 50 (2011): 115–131. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Li Q., Cheon Y. P., Kannan A., Shanker S., Bagchi I. C., and Bagchi M. K., “A Novel Pathway Involving Progesterone Receptor, 12/15‐Lipoxygenase–Derived Eicosanoids, and Peroxisome Proliferator‐Activated Receptor γ Regulates Implantation in Mice,” Journal of Biological Chemistry 279 (2004): 11570–11581. [DOI] [PubMed] [Google Scholar]
- 31. Lee N., Spencer S. K., LaMarca B., Wallace K., and Murphy S., “Alterations in Placental Eicosanoids Contribute to Impaired Vascular Function in Preeclampsia,” American Journal of Obstetrics and Gynecology 213 (2015): 856–864. [Google Scholar]
- 32. Walsh S. W., Reep D. T., Alam S. M. K., et al., “Placental Production of Eicosanoids and Sphingolipids in Women Who Developed Preeclampsia on Low‐Dose Aspirin,” Reproductive Sciences 27 (2020): 2158–2169. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Ye Y., Vattai A., Ditsch N., et al., “Prostaglandin E2 Receptor 3 Signaling Is Induced in Placentas With Unexplained Recurrent Pregnancy Losses,” Endocrine Connections 7 (2018): 749–761. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Jin W. Y., Lin S. L., Hou R. L., et al., “Associations Between Maternal Lipid Profile and Pregnancy Complications and Perinatal Outcomes: A Population‐Based Study From China,” BMC Pregnancy and Childbirth 16 (2016): 60. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Kyriacou H., Al‐Mohammad A., Muehlschlegel C., et al., “The Risk of Cardiovascular Diseases After Miscarriage, Stillbirth, and Induced Abortion: A Systematic Review and Meta‐Analysis,” European Heart Journal Open 2 (2022): oeac065. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. El Hachem H., Crepaux V., May‐Panloup P., Descamps P., Legendre G., and Bouet P. E., “Recurrent Pregnancy Loss: Current Perspectives,” International Journal of Women's Health 9 (2017): 331–345. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Szczeklik W., Stodółkiewicz E., Rzeszutko M., et al., “Urinary 11‐Dehydro‐Thromboxane B2 as a Predictor of Acute Myocardial Infarction Outcomes: Results of Leukotrienes and Thromboxane in Myocardial Infarction (LTIMI) Study,” Journal of the American Heart Association 5 (2016): e003702. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. Monni G., Atzori L., Corda V., et al., “Metabolomics in Prenatal Medicine: A Review,” Frontiers in Medicine (Lausanne) 8 (2021): 645118. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39. Matsumoto Y., Shinjo K., Mase S., et al., “Characteristic DNA Methylation Profiles of Chorionic Villi in Recurrent Miscarriage,” Scientific Reports 27, no. 12 (2022): 11673. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40. Tokareva A., Frankevich N. A., Chagovets V., et al., “A Machine Learning Model Based on First‐Trimester Lipidomic Signatures for Predicting Metabolic Pregnancy Complications,” International Journal of Molecular Sciences 26 (2025): 11824. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41. D'Ippolito S., Di Nicuolo F., Pontecorvi A., et al., “Glucose/Insulin Metabolism and Vitamin D in Women With Recurrent Pregnancy Loss,” American Journal of Reproductive Immunology 87 (2022): e13523. [DOI] [PubMed] [Google Scholar]
- 42. Ticconi C., Nicastri E., D'Ippolito S., et al., “Diagnostic Factors for Recurrent Pregnancy Loss: An Expanded Workup,” Archives of Gynecology and Obstetrics 308 (2023): 127–142. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Figure S1: Distribution of discriminative lipid metabolites across three study groups. This figure presents normalized peak area ratios of previously identified discriminative metabolites across the healthy controls, the pregnant controls, and patients with recurrent pregnancy loss. Statistical comparisons were performed using the Wilcoxon rank‐sum test. Significant alterations were mainly observed in metabolites involved in the cyclooxygenase and lipoxygenase pathways.
Table S1: Differentially expressed lipid metabolites identified by metabolomic analysis.
Data Availability Statement
Data are not publicly available. However, on reasonable request, data can be obtained from the corresponding author.