Dynamic Metabolic Profiling of Gestational Diabetes Mellitus Differentiates Class A1 and Class A2

Introduction Gestational diabetes mellitus (GDM) is a common pregnancy complication characterized by glucose intolerance and underlying metabolic dysregulation. GDM disrupts maternal metabolism, affecting the fetal–maternal axis and potentially impairing neonatal development. Although the need for pharmacological intervention clinically distinguishes GDM A1 and A2 subtypes,1 their longitudinal metabolic trajectories and specific effects on neonatal metabolism remain poorly understood. Current management primarily focuses on glycemic control; however, adverse neonatal outcomes persist even when glucose targets are achieved.2 This suggests that broader metabolic disturbances—including alterations in lipids, amino acids, and bile acids—may influence offspring development independent of maternal glycemia. Furthermore, the lack of longitudinal data linking subtype-specific maternal metabolic profiles with neonatal outcomes limits accurate prediction of long-term metabolic risk. In this study, we use a prospective birth cohort and multi-timepoint metabolomic data to characterize temporal metabolic signatures of GDM A1 and A2. By examining the interplay between maternal subtypes and neonatal profiles, we aim to identify class-specific intergenerational patterns. These insights may clarify mechanisms linking maternal GDM to offspring outcomes, ultimately improving subtype-based risk stratification and precision clinical management. Materials and methods From a prospective Nanjing birth cohort of 5536 pregnant women, 220 participants were selected following rigorous screening. To ensure homogeneity, women with multiple gestations, chronic hypertension, pre-gestational diabetes, or fetal anomalies were excluded. Only participants with complete biospecimens and clinical data were included: 77 healthy controls (HC), 77 with GDM A1 (diet-managed), and 66 with GDM A2 (insulin-treated). The groups were comparable in maternal age, gravidity, parity, and gestational age at delivery. GDM diagnosis and classification were performed at 24–28 weeks according to the International Association of Diabetes and Pregnancy Study Groups and White’s criteria.3 The study was approved by the Ethics Committee of Nanjing Women and Children’s Healthcare Hospital (2022KY-010). In accordance with the Declaration of Helsinki, informed consent was obtained from the patient for publication and will be provided upon request. Maternal clinical data (oral glucose tolerance test (OGTT), glycated hemoglobin (HbA1c), third-trimester fasting plasma glucose (FPG)) and neonatal parameters (weight, length, sex) were obtained from electronic records. Neonatal body mass index (BMI) Z-scores and Ponderal Index were calculated using World Health Organization AnthroPlus and standard formulas.4 Maternal serum samples were collected at T1 (15–18 weeks) and T2 (32–34 weeks). Untargeted metabolomics was performed using liquid chromatography–tandem mass spectrometry (LC-MS/MS; Thermo Fisher Scientific, USA). Raw data underwent peak identification, extraction, and alignment. Metabolites with a detection rate < 50% or relative standard deviation > 30% were filtered out. Neonatal heel blood was collected 72 hours to 7 days post-birth for targeted MS/MS screening. Maternal metabolic data were normalized and log-transformed. Orthogonal partial least squares discriminant analysis (OPLS-DA) identified differential metabolites (variable importance in projection (VIP) > 1, false discovery rate (FDR) < 0.05, and log2FC > 0.25). The Kyoto Encyclopedia of Genes and Genomes (KEGG) was used for pathway enrichment analysis. Time-series and fuzzy c-means clustering grouped metabolites with similar temporal patterns.5 Weighted gene co-expression network analysis (WGCNA) assessed associations between metabolic modules and clinical variables. Multiple linear regression evaluated associations between GDM subtypes and 31 neonatal metabolites, adjusting for maternal age, pre-pregnancy BMI, and gestational weight gain. Analyses were performed using R (v4.1.2) and MetaboAnalyst (v5.0). Group differences were assessed by Kruskal–Wallis and chi-square tests; P < 0.05 was significant. Results Clinical profiles and dynamic maternal metabolic signatures across gestation A total of 220 mother–infant pairs (77 HC, 77 GDM A1, 66 GDM A2) were included. No significant differences were observed in maternal age, gravidity, parity, gestational week at delivery, or neonatal sex across the groups (Supplementary Table 1, https://links.lww.com/MFM/A131). OGTT parameters (fasting, 1-h, 2-h), third-trimester FPG (FPG-3rd), and HbA1c were significantly higher in the GDM A2 group compared with HC and A1. Notably, neonatal birthweight and length were highest in the GDM A2 group. OPLS-DA revealed distinct temporal signatures among subtypes. At T1, no differential metabolites were identified between GDM A1 and HC (Supplementary Fig. 1, https://links.lww.com/MFM/A131), whereas GDM A2 exhibited 10 differential metabolites, including decreased acetylglycine (Supplementary Table 2, https://links.lww.com/MFM/A131). Linoleic acid metabolism was significantly enriched in GDM A2 at T2 versus HC (FDR = 0.002; Supplementary Fig. 2, https://links.lww.com/MFM/A131). At T1, eight differential metabolites (4 upregulated and 4 downregulated) distinguished GDM A1 and A2 (Supplementary Table 3, https://links.lww.com/MFM/A131). By T2, metabolic disruptions intensified, with 117 (111 upregulated and 6 downregulated) and 156 (145 upregulated and 11 downregulated) differential metabolites identified in GDM A1 and A2, respectively (Supplementary Tables 4 and 5, https://links.lww.com/MFM/A131). Between A1 and A2 at T2, 13 metabolites (12 upregulated and 1 downregulated) differed (Supplementary Table 6, https://links.lww.com/MFM/A131). Eleven metabolites, including sphingosine-1-phosphate, 2-hydroxybutyric acid, and 3-methylhistidine, were uniquely elevated in GDM A2. Four metabolites (Cohibin B, LysoPC(14:0/0:0), butyl 3-hydroxybutyrate arabinosyl-(1→6)-glucoside, LysoPA(16:0/0:0)) remained consistently elevated in GDM A2 across T1 and T2. Time-series clustering identified eight metabolite clusters with gestational age–dependent patterns (Supplementary Fig. 3, https://links.lww.com/MFM/A131). GDM A2 showed stronger alterations in organoheterocyclic compounds (orange), whereas GDM A1 was more associated with phenylpropanoids and polyketides (gray). Linoleic acid metabolism was significantly enriched in both subtypes (FDR = 0.002), while GDM A2 was uniquely associated with pantothenate and CoA biosynthesis. At T1, enrichment highlighted primary bile acid biosynthesis and porphyrin metabolism (Supplementary Fig. 4, https://links.lww.com/MFM/A131), whereas T2 showed broader pathway disruption, including purine, histidine, alanine/aspartate/glutamate, and nicotinate/nicotinamide metabolism. Overall, metabolic dysregulation was more pronounced in GDM A2, particularly in late gestation. WGCNA-based intergenerational associations and neonatal metabolic profiles WGCNA identified metabolite modules associated with clinical phenotypes at T2 (Supplementary Fig.5, https://links.lww.com/MFM/A131). In GDM A1, five modules were detected; the brown module correlated with the neonatal ponderal index (PI), while the turquoise module—enriched in valine, leucine, and isoleucine (BCAA) biosynthesis—was associated with neonatal sex. In GDM A2, module–trait associations were broader. The brown module correlated with maternal FPG-3rd, while the red and turquoise modules were associated with maternal FPG-3rd and neonatal growth indicators, including weight, PI, and BMI Z-score. KEGG analysis revealed that the GDM A2 turquoise module was enriched in BCAA, aromatic amino acid (phenylalanine, tyrosine, and tryptophan) biosynthesis, and linoleic acid metabolism. Multiple linear regression further quantified intergenerational effects (Supplementary Fig.6, and Supplementary Table 7, https://links.lww.com/MFM/A131). Maternal GDM status was positively associated with neonatal phenylalanine in a dose-dependent manner: GDM A1 (β = 0.029, 95% confidence interval (CI): 0.00500–0.05400, P = 0.018) and GDM A2 (β = 0.058, 95% CI: 0.01700–0.09800, P = 0.006). Regarding neonatal acylcarnitines, GDM A2 was associated with higher C0 (β = 0.055, 95% CI: 0.01300–0.09700, P = 0.011), C16:1OH (β = 0.001, 95% CI: 0.00010–0.00300, P = 0.033), C18:1 (β = 0.024, 95% CI: 0.00200–0.04500, P = 0.031), and C18:2 (β = 0.009, 95% CI: 0.00050–0.01800, P = 0.037). In contrast, GDM A1 only showed a negative association with neonatal C3DC:C4OH (β = –0.006, 95% CI: –0.01300– –0.00001, P = 0.049). Collectively, these findings underscore that GDM A2 exerts a stronger influence on neonatal metabolic programming, particularly affecting amino acid- and lipid-related pathways associated with neonatal overgrowth. Discussion This study revealed distinct longitudinal metabolic profiles between GDM subtypes, with GDM A2 exhibiting earlier and more pronounced dysregulation. Compared with GDM A1, the A2 group showed significantly higher OGTT, glucose levels FPG-3rd, and HbA1c, as well as increased neonatal birthweight. These findings indicate that GDM A2 represents a more severe metabolic phenotype in which dietary intervention alone is insufficient to mitigate fetal overgrowth. The elevation of sphingosine-1-phosphate and 2-hydroxybutyric acid in GDM A2 at both time points is consistent with their roles as markers of insulin resistance and oxidative stress. Notably, acetylglycine was already reduced in GDM A2 at T1, prior to clinical diagnosis. Given its role in adipose tissue regulation, this early reduction may reflect gut microbiota dysbiosis and altered energy metabolism, potentially contributing to accelerated metabolic deterioration in Class A2.6 In addition, the concurrent activation of linoleic acid, pantothenate, and primary bile acid pathways suggests a compounded metabolic burden in GDM A2. These pathways are key regulators of inflammation and glucose homeostasis, thereby potentially impairing the maternal–fetal metabolic environment.7 A critical finding is the intergenerational impact of amino acid dysregulation. In GDM A2, BCAA and aromatic amino acid pathways (phenylalanine, tyrosine, and tryptophan) were strongly associated with maternal glycemia and neonatal growth. While BCAAs are known to impair insulin signaling via mTOR activation, our results highlight phenylalanine as a particularly important mediator.8 We observed a clear phenylalanine gradient from GDM A2 to HC, with maternal levels at T2 significantly predicting neonatal overgrowth. As an essential amino acid, fetal phenylalanine availability depends on placental transport, suggesting that maternal hyperglycemia may enhance placental amino acid flux via transporter upregulation. This may promote fetal protein synthesis and adiposity, contributing to increased birthweight. These results provide longitudinal evidence that maternal phenylalanine dysregulation in GDM A2 contributes to altered fetal metabolic programming. Furthermore, neonates born to GDM A2 mothers show elevated free carnitine (C0) and long-chain acylcarnitines (C16:1OH, C18:1, C18:2). Accumulation of long-chain acylcarnitines reflects mitochondrial overload and incomplete β-oxidation, indicating reduced metabolic flexibility. These results suggest that GDM A2 imposes a greater lipid metabolic burden on the fetus than A1, potentially inducing mitochondrial stress and impairing early-life metabolic function. Such disturbances may increase susceptibility to metabolic disease later in life. To our knowledge, this is the first study to systematically compare metabolic trajectories across GDM subtypes during pregnancy. Strengths include the prospective cohort design and comprehensive metabolomic profiling. However, the relatively small sample size and single-center design may limit the generalizability. Larger multicenter studies are needed to validate these findings and clarify their implication for long-term outcomes. Collectively, our results emphasize the need for intensified metabolic monitoring and personalized intervention strategies in GDM A2 to improve intergenerational health. Conclusion This study identifies distinct metabolic trajectories between GDM subtypes, with GDM A2 exhibiting more pronounced dysregulation. Both subtypes demonstrated significant enrichment of linoleic acid metabolism, with metabolic imbalances intensifying in late pregnancy. Notably, BCAA and phenylalanine dysregulation in GDM A2 were associated with maternal hyperglycemia and neonatal overgrowth, highlighting their role in adverse intrauterine metabolic programming. These findings suggest that GDM A2 poses a greater intergenerational metabolic risk than A1. Early subtype-specific identification and personalized management are essential to improve maternal and offspring outcomes. Acknowledgements We would like to thank all technical staff involved in this project for their meticulous efforts in data collection, data entry, and sample collection and processing. Funding This study was supported by the Noncommunicable Chronic Diseases-National Science and Technology Major Project (2024ZD0532100, 2024ZD0532103); the National Natural Science Foundation of China (82371697, 81971410); the Natural Science Foundation of Jiangsu (BK20251773); the Key Project of Changzhou Medical Center of Nanjing Medical University (CMCM202312); Changzhou Key Laboratory of Maternal and Child Health Medicine (CM20240013). Author Contributions Shanshan Wang and Jiamin Zhu contributed to the preparation of the original manuscript. Jiamin Zhu was responsible for data curation and statistical analysis. Shiman Hu contributed to data visualization and validation. Zhonghua Shi and Mingzhi Zhang contributed to the study conception, design, and project supervision. All authors contributed to the critical revision of the manuscript and reviewed and approved the final version for submission. Zhonghua Shi serves as the guarantor of this work and takes full responsibility for the integrity of the data and the accuracy of the data analysis. Conflict of Interest None. Data Availability The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.