Gestational diabetes mellitus (GDM) is a prevalent metabolic complication among pregnant women, affecting a significant proportion of pregnancies worldwide. The global prevalence rate is approximately 14.0% (1), and it is specifically 14.7% in the Western Pacific region, including China (1). This condition is known to disrupt various metabolic pathways in pregnant women, resulting in abnormal placental modifications that can negatively affect both short- and long-term health outcomes for both the mother and the child (2–5). GDM also elevates the risk of complications, such as spontaneous abortion, gestational (6) hypertension, polyhydramnios, macrosomia, and neonatal hypoglycemia. Furthermore, it predisposes both mothers and their offspring to lifelong health challenges, such as diabetes, metabolic syndromes, and cardiovascular diseases (7).

The etiology of GDM is complex and multi-factorial, with advanced maternal age, obesity, and a history of adverse obstetrical outcomes being prominent risk factors. Previous studies have indicated a correlation between elevated serum triglyceride (TG) levels early in pregnancy and GDM (8, 9). Elevated thyroid-stimulating hormone (TSH) levels during early pregnancy have been linked to a higher risk of GDM (10). Moreover, the potential role of endocrine-disrupting chemicals (EDCs), particularly bisphenol A (BPA), in the pathogenesis of gestational diabetes mellitus (GDM) has attracted attention, given the concurrent rise in GDM incidence and BPA usage. This chemical, which is a widespread environmental contaminant found in food packaging, thermal receipts, and various consumer products, has been associated with GDM, although study findings have been inconsistent (11).

Beyond environmental factors, individual genetic profiles significantly influence GDM development. A multitude of single-nucleotide polymorphisms (SNPs) associated with GDM have been identified through candidate gene and genome-wide association studies, many of which are linked to type 2 diabetes (T2D) genes (12–15). Peroxisome proliferator-activated receptor-γ coactivator-1α (PPARGC1A/PGC-1α) is a coactivator of peroxisome proliferator-activated receptor-γ (PPARγ). Its expression products can regulate mitochondrial biogenesis, fatty acid metabolism, and insulin sensitivity (16). Glucokinase (GCK) encodes the key enzyme of the hexokinase family (hexokinase IV), which catalyzes the first step of glycolysis in liver and pancreatic islet beta cells (17). This enzyme regulates insulin secretion by sensing changes in glucose concentration. The missense variation at the 1444th locus (rs8192678) in exon 8 of PPARGC1A has attracted significant attention. However, epidemiological studies have produced inconsistent findings regarding its association with the risk of T2D (18–21). In vitro studies suggest that this polymorphism may impact PGC-1α stability, thereby influencing glucose and fat metabolism (22). Despite speculation, no significant association between rs8192678 and gestational diabetes mellitus has been observed in Scandinavian, Austrian, or Italian populations (23–25). Similar research in Chinese women remains limited.

Furthermore, elevated studies suggest that BPA exposure during pregnancy may influence glucose metabolism in both mothers and their children (26). However, the potential modification of genetic effects on GDM by BPA exposure or endocrine biomarkers, such as TSH, during pregnancy remains understudied. To address these knowledge gaps, we conducted a nested case–control study embedded within a prospective cohort of Chinese pregnant women. This study aimed to validate genetic variants identified through whole exome sequencing (WES) and prior research, examined the association between these variants, early gestational BPA exposure, TSH levels, and GDM, and investigated whether the genetic impact on GDM was modified by BPA and TSH levels during pregnancy.

This nested case–control study underscores the potential roles of genetic variants in GDM and reveals an association between GDM and the missense variant of PPARGC1A rs8192678 and the variant of GCK rs2971672. Upon sub-population analysis, PPARGC1A rs8192678, urinary BPA concentration, and serum TG and TSH levels were related to GDM. The difference in GDM probability between the TT and CC genotype of PPARGC1A rs8192678 varied with changes in urinary BPA and serum TSH levels, following a U-shaped distribution for BPA and monotonic attenuation with increasing TSH levels.

Gestational diabetes mellitus (GDM) is a complex condition influenced by the interplay of genetic and environmental factors. Research has shown that GDM shares genetic characteristics with other forms of diabetes, such as type 1 diabetes (T1DM), type 2 diabetes (T2DM), and maturity-onset diabetes of the young (MODY). Numerous polymorphisms in genes associated with these diabetes types have been identified or confirmed to be linked to GDM through candidate gene studies, meta-analyses, and genome-wide association studies (GWAS) (36). This genetic overlap highlights the importance of understanding the underlying genetic predispositions when assessing the risk and management of GDM. In our study of 2,884 Chinese pregnant women, we identified 308 missense variants associated with GDM through WES. Further validation confirmed a significant association between the PPARGC1A rs8192678 (C > T) variant and GDM. The missense variant of PPARGC1A, located in the exonic region, has been linked to elevated BMI, higher total cholesterol levels, and increased fasting plasma glucose (37). It also appears to be associated with reduced insulin sensitivity and heightened insulin resistance, which can influence the development of T2DM (38, 39). However, we observed that T carriers were less likely to develop GDM. The T allele’s protective effect in GDM may arise from compensatory adaptation of the placenta during pregnancy. As far as we know, the variant may affect the differentiation of human white adipocytes, lipogenesis, and the content and turnover of PGC-1α, as indicated by recent findings (40). Previous research suggests that Gly482Ser missense polymorphism in PGC-1α has metabolic consequences on lipid metabolism that could influence insulin secretion (20), while lipid and lipid metabolism play an important role in diabetic complications (41).

Based on in silico structural predictions, the Gly482Ser variant may alter the residue’s flexibility, binding dynamics, protein size, and backbone conformation (42). Such alterations may affect the binding of PGC-1αwith RNF34, theoretically altering its ubiquitin-mediated degradation pathways and modifying oxidative stress and regulating lipid metabolism in brown fat cells, as well as affecting placental mitochondrial function during pregnancy (43, 44). Notably, further prediction indicated that the protein coded by PPARGC1A may undergo phase separation, with quantiles for both self-assembling and partner-dependent phase-separation predictors exceeding 0.8.

Although this mechanistic framework is supported by computational evidence showing altered binding interfaces and prior reports of RNF34-mediated PGC-1α regulation, it remains a hypothesis, requiring empirical validation in pregnancy-relevant models. This validation is necessary to confirm our predictions about the protein’s phase separation capability and to ascertain whether the mutation affects the protein’s structure and function in these models, as well as its potential impact on related biological processes, such as lipid and glucose metabolism.

While the interaction between the PPARGC1A variant rs8192678 and the G6PC2 variant rs16856187 on GDM was not significant, with an FDR greater than 0.1, this interaction may still be considered exploratory due to the limited power of our study to detect such interactions at FDR thresholds with the current sample size. G6PC2 encodes the islet-specific glucose-6-phosphatase catalytic subunit-related protein (IGRP) and is crucial for blood glucose regulation and diabetes pathophysiology (43, 45). As a member of the G6PC protein family, G6PC2 catalyzes the conversion of glucose 6-phosphate into glucose and phosphate. Overexpression of full-length G6PC2 could increase glucose-6-phosphatase activity (46). PGC-1α regulates hepatic gluconeogenesis by co-activating HNF4α or Foxo1, which are key transcription factors for G6PC (47, 48). The rs8192678 (Gly482Ser) variant may alter PGC-1α stability, potentially suppressing G6PC2 expression and hepatic glucose output. However, rs16856187 is a downstream variant for G6PC2 and unlikely to play a functional role, though it has been linked to T2DM (49, 50). It may tag functional variants in ABCB11 (bile salt export pump) through LD. ABCB11 dysfunction may impair glucose homeostasis by altering bile acid-mediated FXR signaling (51).

The protective effect of the T allele of rs8192678 against GDM was not observed in populations from Scandinavia, Austria, and Italy (23–25). This indicates potential ethnic or population-specific differences in genetic factors affecting GDM risk. The varying minor allele frequencies of rs8192678 among populations likely explain the inconsistent associations with GDM, with the T allele frequencies being 44.25% in the East Asian population compared to 36.08% in the European population and 25.65% in the American population.⁷ Beyond the difference in T allele frequencies between Asian and European populations, the protective effect of rs8192678 (Gly482Ser) in the Asian population might be influenced by co-inherited variants in mitochondrial biogenesis pathways, such as TFAM rs1937, which are more prevalent in the Asian population than in the European population.⁸ These variants may affect PGC-1α’s role in placental metabolism. Environmental factors, such as BPA or dietary elements, may also modulate this effect. Our study reveals a more than 2-fold higher median BPA level (4.9 ng/mL) compared to the European population (1.8 ng/mL) (52). Higher BPA exposure may unmask genetic protection in contexts of metabolic stress (53, 54). Dietary modifiers, such as soy isoflavones (high in Asian diets), can compete with BPA for ERβ binding, potentially altering the protective effect of the rs8192678-T allele (55). Another possibility involves linkage disequilibrium (LD). In LD-based indirect correlation analysis, if a disease-causing locus and genetic markers (polymorphic alleles) exhibit strong LD, they can be compared to those in healthy individuals to determine the relative risk of disease-causing loci in the affected population. If the LD between the SNP and the causal loci is weaker in the European population, it may result in a less detectable association (56). If the effect size is very low in the European population, it could only be identified by increasing the statistical power through a larger sample size.

Our study found that GCK rs2971672 (A > C) was related to GDM, with the CC genotype linked to a reduced risk of GDM. Although no prior research directly linked this variant to GDM, this variant has been previously associated with elevated blood glucose and lipid metabolism (57–59). This intron variant has a minor C allele frequency that varies across populations, ranging from 41.15 to 61.88%.⁹ GCK encodes a hexokinase family protein essential for glucose-stimulated insulin secretion and glycogen synthesis (60). It employs multiple promoters and alternative splicing during its transcription process, leading to distinct isoforms of its coding protein. These isoforms exhibit tissue-specific expression in the pancreas and liver, allowing for precise regulation of glucose metabolism and insulin secretion in response to physiological needs. Although the exact functional implications of the rs2971672 variant are still unclear, it may affect gene transcription and enzyme activity, influencing metabolic processes (61). Although the interaction between GCK rs2971672 and the GPI missense variant rs8191371 T > C on GDM was not significant with an FDR greater than 0.1, this interaction may still be considered exploratory due to our limited power to detect interactions at FDR thresholds with the current sample size. As far as we know, the encoded product of GPI functions as a glycolytic enzyme (glucose-6-phosphate isomerase) that interconverts glucose-6-phosphate and fructose-6-phosphate. Further researches are needed to elucidate these relationships and their potential impact on GDM risk. GCK catalyzes the initial step of glycolysis, while GPI (glucose-6-phosphate isomerase) catalyzes the second step (17, 62). Altered GPI activity may lead to the accumulation of G6P (63), inhibiting hexokinase and disrupting insulin secretion (64). However, the function of rs2971672 remains unknown, as does that of rs8191371. Therefore, future studies should test PGC-1α binding to the G6PC2 promoter in hepatocyte models and measure glycolytic flux in β-cells co-expressing GCK/GPI variants. This corresponding text has been added to the discussion section.

In addition to genetic factors, environmental health factors significantly influence the occurrence and development of GDM. Exposure to endocrine disruptors has been linked to metabolic diseases, such as diabetes, and can disrupt glucose homeostasis. These substances can interfere with the synthesis, activity, and elimination of natural hormones that regulate glucose metabolism. By altering hormonal balance, endocrine disruptors may impact insulin sensitivity and glucose regulation, contributing to the development of metabolic disorders (65). Of concern, it was linked to altered metabolic pathways and an increased risk of GDM. These disruptors can impact the immunological and metabolic status of women during pregnancy by inducing cellular and molecular changes in maternal biological fluids and at the maternal–fetal interface (66). In this study, we examined urinary BPA concentrations in early pregnancy among 155 cases and controls to evaluate their association with GDM, considering potential genetic effects. Our findings revealed that the odds of GDM more than doubled with each unit increase in the natural square root-transformed urine BPA concentration, even after adjusting for maternal genetic effect, triglycerides, TSH, parity, gravidity, and age. This association remained consistent across different genetic models based on the PPARGC1A rs8192678 variant. Differently, previous studies have produced inconsistent results regarding the link between first-trimester BPA levels—whether in urine or serum—and GDM. Some cohort studies found no association (67–69), while a nested case–control study identified a positive correlation between first-trimester BPA and GDM risk among non-Asian/Pacific Islanders (70). Maternal early-pregnancy BPA exposure has been associated with glucose level, potentially exhibiting a non-linear relationship, although findings have been inconsistent (68, 69, 71–73). Previous research suggests that BPA may act through several pathways, such as the receptor pathways, disruption of the neuroendocrine system, modulation of immune and inflammatory responses, and epigenetic mechanisms (74). Low doses of BPA intake during gestation and early development have been shown to cause islet insulin hypersecretion in rat offspring for up to 1 year after exposure (75). Differentiated mature adipocytes exposed to BPA exhibited insulin resistance, with approximately a 25% reduction in insulin-stimulated glucose uptake (76). Ariemma et al. reported a 3.5-fold increase in the expression of peroxisome proliferator-activated receptor gamma (PPARγ) and a hyperregulation of inflammatory factors after 3 weeks of BPA exposure in 3 T3-L1 adipocytes, noting that BPA led to lipid accumulation and impaired insulin function, significantly reducing insulin-stimulated glucose utilization (77). LaRocca et al. found a correlation between placental miR-142-3p levels and first-trimester urinary phenols (78), while women later diagnosed with GDM exhibited higher first-trimester serum levels of miR-142-3p (79).

The subsequent marginal effect analysis revealed that the rs8192678-T allele’s protective effect exhibited a quasi-U-shaped distribution at varying BPA levels by primary marginal analysis and bootstrap validation (Supplementary Table S16). This association may be due to BPA’s non-monotonic dose–response effects (80). BPA exposure may influence hepatic ERα/ERβ homeostasis. While at a lower dose it promotes Akt phosphorylation, BPA at the higher dose attenuates ERK1/2 phosphorylation, suggesting potential alteration in insulin sensitivity (81). In this context, the rs8192678 (Gly482Ser) may change PGC-1α protein’s conformation and stability, synergistically altering placental mitochondrial function and modifying oxidative stress. In large cohorts, U-shaped BPA–T2DM/children’s blood pressure associations have been reported, which supports our finding (82).

Serum TG and TSH levels have been linked to an elevated risk of GDM (9, 10), while our study found consistent results for TG but inconsistent observations regarding TSH. Meanwhile, in the marginal effect analysis, the disparity in GDM probability between the TT genotype and the CC genotype of PPARGC1A rs8192678 decreased as serum TSH levels increased. As far as we know, adipose tissue TSH played a role in the maintenance of adipocyte mitochondrial function and regulating energy balance and adiposity by inhibiting the browning of white fat (83, 84). Both visceral and subcutaneous adipose tissue TSHB gene expression was positively correlated with the expression of mitochondrial function, such as PPARGC1A (84).

The protective effect of the PPARGC1A rs8192678-T allele decreases with higher TSH levels (Primary: dy/dx = −0.119 (−0.258 ~ 0.018), p = 0.09; validation: dy/dx = −0.119 (−0.277–0.037), p = 0.133) (Supplementary Table S16). This suggests that TSH-mediated thyroid hormone suppression counteracts genetic benefits. Elevated TSH can lower circulating free thyroxine (FT4) through the hypothalamic–pituitary–thyroid axis and downregulate thyroid receptor β-dependent PPARGC1A expression, potentially affecting tissue-specific mitochondrial function (85). The decreased FT4 would change thermal efficiency during energy conversion by regulating the expression of uncoupling proteins (UCPs) to improve (86, 87). Nevertheless, further external validation and the establishment of precise cut-offs for BPA and TSH levels to modulate the protective effect are necessary, ideally through studies with larger cohorts.

The primary strength of our study is revealing how the genetic effect of PPARGC1A rs8192678 on GDM slightly changes with urinary BPA and serum TSH level in a nested case–control design, which is based on a prospective cohort of nearly 3,000 pregnant women and highly efficient for avoiding reverse causation. Additionally, we accessed each participant’s OGTT data through the medical records system, with GDM diagnoses made by doctors according to IADPSG criteria. Another key strength is our use of both WES and candidate gene strategies to screen for potential SNPs, allowing us to visualize the molecular structures of the original and mutant amino acids of PPARGC1A and predict their functions. We also assessed the protein’s phase separation capability, finding evidence suggesting it may possess this ability, though further validation is needed.

While this study provides novel insights into gene–environment interactions in GDM pathogenesis, several limitations warrant acknowledgment. First, although we expanded adjustment to include pre-pregnancy BMI, gravidity, and parity—strengthening causal inference—residual confounding by unmeasured factors (e.g., dietary patterns and socioeconomic status) remains possible (52, 88, 89). Second, thyroid hormone interpretation is constrained by unavailable data on iodine supplementation or medications. However, all participants resided in iodine-sufficient regions (Shanghai) with universal salt iodization, minimizing population-level confounding. Third, exploratory and subgroup analyses were limited by sample size in genotype-exposure strata. We addressed this through rigorous bootstrap validation, confirming robust primary associations: rs8192678 TT vs. CC aOR = 0.417 (0.220–0.790), rs8192678 CT + TT vs. CC aOR = 0.593 (0.376–0.935), rs8192678 TT vs. CC + CT aOR = 0.524 (0.198–0.919) for exploratory analysis; BPA aOR = 2.295 (95% CI: 1.361–3.867) for subgroup analysis (Supplementary Table S17). Fourth, focusing solely on BPA overlooks mixtures of endocrine disruptors. Reassuringly, coexposure to BPA alternatives (BPS/BPF) is low in Chinese pregnant women (<10% detection; Spearman’s r = 0.10–0.17 vs. BPA), reducing confounding potential (89). Finally, the proposed RNF34-PGC-1α mechanism for rs8192678’s protective effect requires experimental validation. Future studies should construct the Gly482Ser mutant and validate changes in binding with RNF34 using co-immunoprecipitation (40).

In summary, our nested case–control study highlights the potential roles of genetic variants in GDM and identifies associations between GDM and the missense variant of PPARGC1A rs8192678, as well as the variant of GCK rs2971672. Sub-population analysis indicates that PPARGC1A rs8192678, BPA concentration, and serum TG and TSH levels are correlated with GDM. Marginal effect analysis further indicates the protective effect of PPARGC1A rs8192678 on GDM mildly varied with urinary BPA and serum TSH levels, even after controlling for potential confounders. However, further external validation and in vitro experiments are needed to confirm our findings.