Longitudinal metabolomics reveals dynamic associations of trimethylamine N-oxide and precursors with the progression of gestational diabetes

Trimethylamine N-oxide (TMAO) and its precursor metabolites are strongly linked to glucose and lipid metabolic disorders, yet their dynamic roles and overall effects in gestational diabetes mellitus (GDM) have not been clarified. We aimed to explore the dynamic patterns of change and the overall effects of the six TMAO-related metabolites (including Trimethylamine N-oxide, Trimethylamine, choline, L-carnitine, betaine and creatinine) in relation to GDM. A total of 180 serum samples from a nested case–control study were analyzed via targeted metabolomics covering early, middle, and late pregnancy. Conditional logistic regression, Bayesian kernel-machine regression (BKMR), weighted quantile sum regression (WQS) and quantile g-computation regression (QGC) were employed to characterize the dynamic associations and mixture patterns of TMAO-related metabolites with GDM risk. GDM patients exhibited progressive decreases in L-carnitine, alongside elevations in choline and trimethylamine. Early-pregnancy L-carnitine levels were positively associated with GDM. BKMR showed positive, U-shaped and negative associations between TMAO-related metabolites and GDM in three trimesters, respectively. L-carnitine and trimethylamine demonstrated positive associations with GDM during the first trimester. Choline exhibited a shift from negative to positive associations and showed potential interaction patterns with L-carnitine in the second and third trimesters. WQS showed L-carnitine and trimethylamine presenting higher weight in the first and third trimester, respectively. QGC further validated a mixture effect in the third trimester, primarily attributed to trimethylamine and choline. The associations of L-carnitine, trimethylamine and choline with GDM demonstrated dynamic temporal patterns throughout pregnancy.These findings may provide insight into trimester-specific metabolic variation in GDM, but should be considered exploratory and require validation in larger cohorts.