Longitudinal untargeted maternal metabolomics identifies potential metabolic signatures of pregnancy failure

Comprehensive metabolomic profiling in reproductive medicine is sought to clarify the specific mechanisms underlying potential exposome-metabolome interactions in adverse pregnancy outcomes. Taking the advantages of longitudinal data, untargeted metabolomics, and machine learning coupled with traditional analysis, we aimed to study the associations between altered metabolome in the first and third trimesters of pregnancy and subsequent implications to explore causal associations. Totally, 201 pregnant women from a low- and middle-income community (LMIC), known for high levels of environmental pollution, were enrolled during their first trimester, 13 ended in pregnancy failure. Gas chromatography-mass spectrometry (GC-MS) was used to obtain untargeted metabolic profiles and to quantify relative levels of metabolome signatures in serum samples. Data processing and analysis were conducted to select features associated with adverse pregnancy outcomes (including miscarriage, stillbirth, preterm birth, and infant death), adjusting for participants' occupational status, education level, smoking, and the season of conception. Metabolic network and pathway enrichment analyses were then conducted to explore metabolome-associated pregnancy failure. Statistical and machine learning methods were used to visualize the associations between metabolomic features and the risk of adverse pregnancy and neonatal outcomes, accounting for other covariates. The pattern of associations between maternal metabolome during pregnancy and birth outcomes revealed a clear separation of pregnancy failure cases from medically approved healthy-term births (p < 0.05). L-alanine, dioctyl phthalate, L-phenylalanine, L-threonine, cholesterol, L-serine, proline, L-isoleucine, L-valine, arabinofuranose and gluconic acid were upregulated in the pregnancy failure participants, while glycine, L-lactic acid, arachidonic acid, L-tryptophan, creatinine, palmitic acid, L-tyrosine, ornithine, glutamic acid, phosphate, 1,5-anhydrosorbitol, taurine, 3-hydroxybutyric acid, oxoproline, D-glucose, oleic acid and linoleic acid were less abundant. Specific metabolite patterns linked to pregnancy failure were discovered by machine learning methods over the course of pregnancy. Our analysis identified L-alanine, cholesterol, D-glucose, and urea as potential biomarkers for the early detection of pregnancy failure. While promising, further studies are needed to validate these findings and assess their clinical applicability, particularly in populations highly exposed to environmental pollutants.

The online version contains supplementary material available at 10.1007/s40201-025-00963-z.