PPARs, L-FABP mediate the association between Per- and polyfluoroalkyl substances and gestational diabetes: a nested case-control study

Abstract

To analyze the correlation between PFASs exposure and glucose metabolic indexes in relation to GDM risk, and to explore the mediating role of PPARs and L-FABP. Using a nested case–control study method. Blood specimens were collected during early pregnancy for quantification of PFASs, PPARs, and L-FABP. LASSO was used to select PFASs, and the selected variables were subsequently included in WQS and QGComp to analyze the relationship between PFASs and GDM. Four-way decomposition analysis and chain mediation model examined PPARs and L-FABP mediation in the PFAS-GDM association. A total of 1680 pregnant women were enrolled, of whom, 255 participants were included in the final analysis. LASSO identified seven PFASs linked to GDM risk and glucose levels. PFOA primarily drove GDM risk and elevated 2-h glucose, while PFBS predominantly influenced FPG and 1-h glucose. Mediation analysis revealed two distinct pathways: (1) a chain-mediated pathway involving FOSA-I-PPARα-L-FABP-GDM, and (2) a significant mediating effect of PPARγ in the association between HFPO-DA exposure and GDM development. Exposure to PFAS mixtures was associated with an increased risk of GDM and dysregulated glucose metabolism. Our findings highlight the mechanistic role of PPARγ in HFPO-DA-induced GDM pathogenesis, while FOSA-I appears contribute to GDM risk through sequential modulation of PPARα and L-FABP.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-36679-9.

Introduction

Gestational diabetes mellitus (GDM) is a disorder of abnormal glucose metabolism that occurs during pregnancy and affects approximately 14% of pregnant women worldwide. In China, the incidence of GDM reached 14.8% in 2017¹. Recent research suggests that by 2024, the incidence of GDM had already exceeded 20%². GDM not only predisposes women to a variety of adverse pregnancy outcomes^3–6, but also increases the risk of future obesity, insulin resistance, and type 2 diabetes mellitus in both the mother and her offspring^7,8. The etiology of GDM is complex, involving not only insulin resistance, genetics, inflammatory factors, and impaired lipid metabolism, but also many environmental pollutants^9–11.

Per- and polyfluoroalkyl substances (PFASs) are a class of synthetic chemicals widely used in industrial and consumer products^12–14. These compounds are highly persistent, have long half-lives, and are difficult to degrade¹⁵. Human exposure occurs through contaminated food, drinking water, air, and dust^16–18. PFASs exhibit diverse toxicities, including reproductive, neurological, developmental disorder, and endocrine disruption effects¹⁹.

Some studies suggest a correlation between PFASs exposure and the development of GDM, with PFASs potentially elevating blood glucose or disrupting insulin synthesis, secretion, and metabolism^20–22. Animal experiments²³ and epidemiological investigations²⁴ indicate interactive effects among PFASs, but research on mixed exposure to PFASs remains limited, with most studies focusing on model organisms (e.g. fish, algae and tolerant bacteria) rather than population studies²⁵.

In addition, the mechanisms associated with PFASs exposure and GDM remain unclear. Emerging evidence suggests peroxisome proliferator activated receptors (PPARs) may be the preferred target of PFASs²⁶. PPARs modulate insulin sensitivity indirectly by regulating lipid metabolism. Excessive accumulation of lipid intermediates can impair insulin signaling and contribute to the development of insulin resistance²⁷. Specifically, upregulation of PPARα may promote lipid accumulation, whereas increased expression of PPARγ enhances insulin sensitivity, lowers blood glucose levels, and improves lipid metabolism. In line with these glucoregulatory roles^28,29, animal studies specific to GDM have demonstrated significantly reduced PPARγ mRNA expression in GDM rats compared to normal controls³⁰. PFASs may also competitively bind liver-fatty acid binding protein (L-FABP) leading to plasma free fatty acid accumulation, which in turn stimulates phosphorylation of the insulin receptor substrate serine and subsequent glucose metabolism dysfunction³¹. In the context of hyperglycemia, studies have shown upregulation of PPARα and L-FABP alongside downregulation of PPARγ^32,33. Given that PPARs may regulate L-FABP expression³⁴, it is notable that under these conditions, the interaction between L-FABP and PPARs is also markedly enhanced³⁵. Thus, we hypothesize that PPARs and L-FABP mediate PFAS-induced GDM, athough direct evidence is still lacking.

Based on this, the present study analyze the correlation between mixed exposure to PFASs and the risk of GDM, and further explored the roles of PPARs and L-FABP in this relationship. Our findings deepen the understanding of how PFAS exposure may contribute to GDM development and provide hypotheses and preliminary evidence to guide future mechanistic research and prevention strategies.

Results

Population characteristics

The basic characteristics of the study participants are presented in Table 1.A total of 85 participants were included in the case group and 170 in the control group. No statistically significant differences were observed between the two groups for most variables, except for BMI (P = 0.013), family history of diabetes (P = 0.016), and blood glucose level (P < 0.001).

Table 1.

Basic characteristics and oral glucose tolerance test of the study population [N (%)/M (P25, P75)].

Characteristics	All subjects (N = 255)	Non-GDM (N = 170)	GDM (N = 85)	P
Maternal age (years)	30.00 [27.00, 33.00]	30.00 [27.00, 33.00]	30.00 [27.00, 33.00]	0.981
Gestational weeks at sampling	10.00 [8.00, 12.00]	10.00 [8.00, 12.00]	9.00 [8.00, 12.00]	0.596
Maternal educational level				0.781
Junior high school and below	50 (19.6)	31 (18.2)	19 (22.4)
Secondary or high school	71 (27.8)	46 (27.1)	25 (29.4)
College or Bachelor	125 (49.1)	87 (51.2)	38 (44.7)
Postgraduate	9 (3.5)	6 (3.5)	3 (3.5)
BMI (kg/m2)	23.05 [20.43, 26.03]	22.85 [20.17, 25.82]	24.22 [21.83, 26.57]	0.013
Underweight	19 (7.5)	16 (9.4)	3 (3.5)	0.048
Normal weight	128 (50.2)	90 (52.9)	38 (44.7)
Overweight or obese	108 (42.3)	64 (37.7)	44 (51.8)
Location				0.674
City	59 (23.1)	38 (22.4)	21 (24.7)
Rural	196 (76.9)	132 (77.6)	64 (75.3)
Monthly per capita household income (yuan)				0.413
< 1000	13 (5.1)	9 (5.3)	4 (4.7)
1001–3000	56 (22.0)	32 (18.8)	24 (28.2)
3001–5000	109 (42.7)	76 (44.7)	33 (38.8)
> 5000	77 (30.2)	53 (31.2)	24 (28.3)
Smoking status				0.430
No	233 (91.4)	157 (92.4)	76 (89.4)
Yes	22 (8.6)	13 (7.6)	9 (10.6)
Parity				0.130
0	115 (45.1)	71 (41.8)	44 (51.8)
1+	140 (54.9)	99 (58.2)	41 (48.2)
Family history of diabetes				0.016
No	227 (89.0)	157 (92.4)	70 (82.4)
Yes	28 (11.0)	13 (7.6)	15 (17.6)
OGTT
FPG (mmol/L)	4.63 [4.37, 5.06]	4.49 [4.30, 4.73]	5.16 [4.90, 5.46]	< 0.001
1-h glucose (mmol/L)	7.79 [6.59, 9.36]	7.26 [6.33, 8.14]	10.06 [8.61, 10.65]	< 0.001
2-h glucose (mmol/L)	6.40 [5.70, 7.46]	6.12 [5.52, 6.78]	7.95 [6.38, 8.83]	< 0.001

Parity is categorized as 0 (no previous delivery) and 1+ (one or more previous deliveries). BMI, Body mass index; OGTT, Oral glucose tolerance test; FPG, fasting plasma glucose.

Maternal serum PFASs levels

Serum concentrations of 19 PFASs were measured. PFASs with detection rates below 60% were excluded from further analysis, resulting in 15 PFASs being included. Statistically significant differences were observed in the levels of PFBS (P = 0.042) and PFOA (P = 0.003) between the case and control groups (Table 2). Spearman’s correlation analysis revealed significant correlations among PFAS congeners, with correlation coefficients ranging from − 0.15 to 0.87.Notably, PFNA, PFDA, PFOS, PFHpS, PFHxS, PFUdA, and 6: 2 Cl-PFESA showed strong correlations, with r ≥ 0.6 (Fig S1).

Table 2.

Per- and polyfluoroalkyl substances (PFAS) levels (ng/mL) in early pregnancy.

PFAS	All subjects (N = 255)			Non-GDM (N = 170) Median [IQR]	GDM (N = 85) Median [IQR]	P
	LOD (ng/ml)	> LOD (%)	Median [IQR]
PFBA	0.0025	77.3	0.02 [0.00, 0.03]	0.02 [0.00, 0.03]	0.02 [0.01, 0.03]	0.141
PFHpA	0.0021	85.1	0.15 [0.04, 0.32]	0.16 [0.03, 0.33]	0.14 [0.04, 0.31]	0.778
PFUdA	0.0100	99.6	0.12 [0.08, 0.16]	0.11 [0.08, 0.17]	0.12 [0.08, 0.15]	0.963
FOSA-I	0.0010	99.6	0.01 [0.00, 0.01]	0.01 [0.00, 0.01]	0.01 [0.00, 0.01]	0.866
PFBS	0.0050	100	0.05 [0.04, 0.07]	0.05 [0.04, 0.07]	0.06 [0.04, 0.08]	0.042
PFDA	0.0061	100	0.12 [0.08, 0.16]	0.11 [0.08, 0.16]	0.12 [0.09, 0.15]	0.244
PFHpS	0.0007	100	0.02 [0.01, 0.03]	0.02 [0.01, 0.03]	0.02 [0.01, 0.02]	0.586
PFNA	0.0052	100	0.18 [0.14, 0.25]	0.18 [0.13, 0.24]	0.20 [0.16, 0.27]	0.108
PFOA	0.0045	100	1.72 [1.20, 2.64]	1.54 [1.14, 2.43]	2.21 [1.45, 2.87]	0.003
PFPeS	0.0010	100	0.01 [0.00, 0.01]	0.01 [0.00, 0.01]	0.01 [0.00, 0.01]	0.686
PFTeDA	0.0090	100	0.14 [0.11, 0.17]	0.14 [0.11, 0.18]	0.14 [0.11, 0.17]	0.496
PFOS	0.0035	100	1.06 [0.80, 1.38]	1.01 [0.76, 1.38]	1.10 [0.84, 1.41]	0.158
PFHxS	0.0035	100	0.15 [0.11, 0.22]	0.15 [0.10, 0.21]	0.16 [0.12, 0.24]	0.126
6:2 ClPFESA	0.0008	100	0.17 [0.13, 0.26]	0.17 [0.12, 0.26]	0.18 [0.14, 0.27]	0.772
HFPO-DA	0.0010	100	0.03 [0.02, 0.04]	0.03 [0.03, 0.04]	0.03 [0.02, 0.04]	0.221

GDM, gestational diabetes mellitus; LOD, the limit of detection; IQR: inter-quartile rang.

Identify key PFAS associated with GDM and blood glucose levels

To address potential collinearity among PFASs, a risk prediction model was developed using LASSO penalized regression, incorporating 15 PFASs. The trajectories of regression coefficients were examined by dynamically adjusting the penalty parameter (λ) in the LASSO model. The optimal λ value was determined using 12-fold cross-validation (Fig. 1). Ultimately, seven PFASs were selected as being associated with GDM or blood glucose levels: PFOA, PFBA, PFBS, PFDA, PFTeDA, FOSA-I, and HFPO-DA (Table S1).

Fig. 1.

Screening map of PFASs based on LASSO regression. (A), (C), (E) and (G): Coefficient paths of the 15 PFASs across log (λ) for models with the four glycemic outcomes. The trajectory of each PFAS coefficient is represented by a distinct colored line. (B), (D), (F) and (H): A 12-fold cross-validation of the LASSO regression model. All models were adjusted for BMI and family history of diabetes. LASSO, least absolute shrinkage and selection operator; GDM, gestational diabetes mellitus; FPG, fasting plasma glucose.

WQS regression analysis of the relationship between PFAS, GDM and blood glucose levels

WQS regression analysis was conducted after adjusting for BMI and family history of diabetes. The results showed that exposure to the mixture of seven PFASs identified through LASSO regression was significantly associated with all four outcome measures. These included GDM (P = 0.016), FPG (P = 0.023), 1-h glucose level (P = 0.044), and 2-h glucose level (P = 0.044).

Table 3 presents the associations between the WQS index and both GDM risk and blood glucose levels. After adjusting for potential confounders, the WQS index showed a statistically significant positive association with GDM risk (β = 0.824, 95% CI: 0.152, 1.500). Among the mixture components, PFOA contributed the most, accounting for 31.1% of the overall effect. In the adjusted models assessing blood glucose levels, each quartile increase in the WQS index was associated with a 0.165 mmol/L increase in FPG (95% CI: 0.025, 0.305), a 0.573 mmol/L increase in 1-h glucose level (95% CI: 0.019, 1.126), and a 0.374 mmol/L increase in 2-h glucose level (95% CI: 0.014, 0.735). Notably, PFBS had the greatest relative contribution to the effects on FPG (31.2%) and 1-hour glucose levels (24.0%). PFOA showed the strongest influence on 2-hour glucose levels (32.3%). These findings suggest that PFOA and PFBS are key drivers of the mixture effects of PFAS exposure on GDM risk and blood glucose regulation.

Table 3.

Associations of WQS regression model with GDM risk and blood glucose levels.

Model		β-WQS (95%CI)	Weight
GDM	Positive	0.824 (0.152, 1.500) P = 0.016	PFOA	0.311
			PFTeDA	0.225
			PFBS	0.208
			PFDA	0.106
			PFBA	0.096
			FOSA-I	0.030
			HFPO-DA	0.024
	Negative	0.310 (-0.375, 0.995) P = 0.375	N/A
FPG	Positive	0.165 (0.025, 0.305) P = 0.023	PFBS	0.312
			PFOA	0.207
			PFDA	0.190
			PFBA	0.119
			FOSA-I	0.104
			PFTeDA	0.054
			HFPO-DA	0.014
	Negative	0.044 (-0.115, 0.203) P = 0.589	N/A
1-h glucose level	Positive	0.573 (0.019, 1.126) P = 0.044	PFBS	0.240
			PFBA	0.222
			PFOA	0.158
			FOSA-I	0.152
			PFTeDA	0.132
			PFDA	0.095
			HFPO-DA	0.001
	Negative	0.255 (-0.093, 0.602) P = 0.152	N/A
2-h glucose level	Positive	0.374 (0.014, 0.735) P = 0.044	PFOA	0.323
			PFBS	0.244
			PFDA	0.193
			FOSA-I	0.119
			PFBA	0.101
			HFPO-DA	0.019
			PFTeDA	0.000
	Negative	-0.146 (-0.461, 0.170) P = 0.367	N/A

Data are presented as weights for individual PFAS and β (95% CI) estimates for PFAS mixtures obtained from the WQS model, representing the single-component weights and the combined effects associated with each quartile increase in the PFAS mixture on GDM risk. The model was adjusted for BMI and family history of diabetes. WQS, weighted quantile sum; GDM, gestational diabetes mellitus; FPG, fasting plasma glucose; CI, confidence interval.

The relationship between PFAS, GDM and blood glucose levels in the QGComp model

The QGComp analysis demonstrated significant associations between PFAS mixture exposure and GDM risk (P = 0.024), fasting plasma glucose (FPG) (P = 0.010), and 1-h glucose level (P = 0.027) in the study population (Table S2). A marginal association was also observed with 2-h glucose level (P = 0.056).

After adjusting for confounders, PFAS mixture exposure was positively associated with GDM risk (ψ = 0.344, 95% CI: 0.045, 0.643). Each quartile increase in exposure corresponded to increases of 0.123 mmol/L in FPG (95% CI: 0.030, 0.218), 0.574 mmol/L in 1-h glucose level (95% CI: 0.051, 1.097), and 0.266 mmol/L in 2-h glucose level (95% CI: -0.005, 0.538).

Analysis of individual PFAS contributions revealed that HFPO-DA and PFTeDA were consistently negatively associated with all outcomes, suggesting potential protective effects. In contrast, PFOA, PFBA, and PFBS exhibited positive associations in the model, indicating potential risk-enhancing effects within the PFAS mixture.

Maternal PPARs, L-FABP levels

The levels of PPARs and L-FABP among study participants are presented in Table 4.Compared with the control group, PPARα levels were significantly higher in the case group (P = 0.048). In contrast, PPARγ (P < 0.001) and L-FABP (P = 0.019) levels were significantly lower in the case group.

Table 4.

PPARs, L-FABP levels in study subjects [M (P25, P75)].

Variables	All subjects (N = 255)	Non-GDM (N = 170)	GDM (N = 85)	P
PPARα (ng/ml)	7.69 [4.97, 11.39]	6.71 [4.63, 11.38]	8.67 [6.29, 11.56]	0.048
PPARγ (ng/ml)	7.70 [6.07, 9.78]	8.40 [6.47, 10.79]	6.78 [5.32, 8.50]	< 0.001
L-FABP (pg/ml)	840.66 [673.05, 1001.13]	851.93 [692.90, 1048.11]	826.79 [652.80, 932.53]	0.019

GDM, gestational diabetes mellitus; PPAR, peroxisome proliferator activated receptors; L-FABP, liver-fatty acid binding protine.

Mediating and interacting roles of PPAR, L-FABP in PFAS-induced GDM

Using four-way decomposition analysis, the total effects of PFASs on GDM mediated through PPARs and L-FABP were partitioned into four components: CDE, INTref, INTmed, and PIE.

In the model with PPARα as the mediator, the total effect of PFTeDA on GDM risk was primarily driven by the reference interaction effect (P < 0.001). In contrast, the association between FOSA-I and GDM risk was mainly explained by the pure indirect effect (P = 0.043). For PFDA (P = 0.001) and HFPO-DA (P = 0.033), the controlled direct effect played a dominant role. These results suggest that PPARα may partially mediate the relationship between FOSA-I exposure and GDM risk during pregnancy (Table S3).

In the model with PPARγ as the mediator, the total effect of PFDA on GDM risk was mainly driven by the mediated interaction effect (P = 0.000). The total effect of HFPO-DA on GDM risk was primarily attributable to the pure indirect effect (P = 0.027). For PFBA (P < 0.001) and FOSA-I (P = 0.039), the reference interaction effect was dominant. These findings suggest that PPARγ may mediate the associations of PFDA and HFPO-DA with GDM risk (Table S4).

In the model with L-FABP as the mediator, the effects of PFDA, PFOA, PFBA, and PFBS on GDM risk were predominantly explained by the controlled direct effect. The total effect of PFTeDA on GDM risk was mainly attributable to the mediated interaction effect (P = 0.020). For FOSA-I, the total effect was primarily driven by the pure indirect effect (P = 0.002). These results suggest that L-FABP may mediate the associations of PFTeDA and FOSA-I with GDM risk (Table S5).

Mediation and interaction of L-FABP in the association between PPAR and GDM

By four-way decomposition analysis, the total effect of PPAR with L-FABP on GDM risk was decomposed into four components, CDE, INTref, INTmed, and PIE (Table 5). The results demonstrated that the reference interaction effect (P = 0.015), the mediated interaction effect (P = 0.015), and the pure indirect effect (P = 0.004) of PPARα and L-FABP on GDM risk were all statistically significant. In contrast, PPARγ predominantly exhibited the reference interaction effect (P = 0.021), with an attributable proportion of 85.03%. These findings suggest that L-FABP mediates the association between PPARα and GDM risk, whereas PPARγ exerts a direct effect on GDM risk.

Table 5.

Decomposition of the effect of PPAR on GDM due to mediation and interaction with L-FABP.

Mediation	Component	Effect	95%CI	P	Proportion (%)	P
PPARα	CDE	-0.058	(-0.569, 0.454)	0.825	11.07	0.802
	INTref	-0.512	(-0.922, -0.102)	0.015	98.04	0.067
	INTmed	0.522	(0.104, 0.940)	0.015	-99.90	0.107
	PIE	-0.474	(-0.796, -0.153)	0.004	90.78	0.071
	Total	-0.522	(-1.174, 0.129)	0.116
PPARγ	CDE	142.030	(-181.96, 466.02)	0.390	85.03	0.021
	INTref	68.617	(-280.65, 417.88)	0.700	41.08	0.533
	INTmed	-43.578	(-257.32, 170.16)	0.690	-26.09	0.574
	PIE	-0.032	(-0.383, 0.319)	0.859	-0.02	0.878
	Total	167.040	(-284.11, 618.18)	0.468

Adjusted for BMI and family history of diabetes. PPAR, peroxisome proliferator activated receptors; L-FABP, liver-fatty acid binding protine; CI, confidence interval.

Analysis of the chain-mediated effects of PPAR, L-FABP in PFAS-induced GDM

Based on the findings of the four-way decomposition analysis, PFAS potentially influencing the occurrence of GDM via pathways involving PPARα, PPARγ, or L-FABP were further investigated for chain-mediating effects. Using GDM as the dependent variable and PFAS as the independent variable, chain mediation analyzes were conducted after adjusting the potential confounders. Two distinct mediation models were tested: (1) PPARα and L-FABP as sequential mediators. (2) PPARγ and L-FABP as sequential mediators.

As illustrated in Fig. 2, FOSA-I were positively correlated with PPARα (β = 0.187, P < 0.01), PPARα was negatively correlated with L-FABP (β=-0.133, P < 0.05) and L-FABP (β=-0.458, P < 0.01) was negatively correlated with GDM risk. The mediating role of L-FABP in FOSA-I to GDM was established [0.126 (95% CI: 0.043, 0.259)] with the proportion of -293.02%. The chain mediating role of FOSA-I-PPARα-L-FABP-GDM was established [0.011 (95% CI: 0.001, 0.029)] with the proportion of -25.58% (Table S6).

Fig. 2.

Chain-mediated effects of PPARα, PPARγ and L-FABP in PFAS on GDM. *P<0.05, **P<0.01, ***P<0.001.

HFPO-DA was negatively correlated with PPARγ (β=-0.123, P < 0.05), PPARγ (β=-0.965, P < 0.001) and L-FABP (β=-0.433, P < 0.01) with GDM risk. PPARγ mediated the effect of HFPO-DA on GDM risk, with an estimated indirect effect of 0.119 (95% CI: -0.001, 0.255), accounting for − 67.23% of the total association (Table S6).

Sensitivity analysis

Sensitivity analysis showed that the positive associations between PFAS exposure and both glucose levels and GDM risk remained robust after excluding participants with a family history of diabetes (Table S7). Furthermore, the main contributing PFASs driving these associations were also consistent with the primary findings (Fig. S3).

Discussion

In this study, we used a nested case-control study method to investigate the association of PFASs with GDM risk and blood glucose levels. Based on the LASSO regression analysis, we selected seven PFAS that were associated with GDM or blood glucose levels: PFOA, PFBA, PFBS, PFDA, PFTeDA, FOSA-I, and HFPO-DA.

Among them, PFOA has been the most extensively studied and has been linked to altered lipid and glucose homeostasis, oxidative stress, and dysregulation of PPARs, which may influence insulin sensitivity and β-cell function^36,37. PFBA and PFBS, as short-chain PFAS, can upregulate key adipogenic transcription factors at both mRNA and protein levels. This triggers a cascade of lipid metabolism disorder, vascular endothelial dysfunction, and reduced insulin sensitivity, ultimately contributing to GDM development^20,38. PFDA, a long-chain PFAS, has been associated with altered metabolic profiles and interference with fatty acid oxidation, potentially through PPARα–mediated pathways^21,39. In experimental models, the replacement compound HFPO-DA (GenX) has been shown to cause hepatic injury and metabolic disturbance, primarily through inducing PPARα targets and oxidative stress⁴⁰. Five of the selected PFAS have been reported in previous studies to be associated with similar health endpoints, which indirectly supports part of the model’s selection rationale. The remaining two PFAS (PFTeDA, FOSA‑I) currently lack independent studies investigating their association with glucose metabolism. They may represent previously under‑recognized risk substances or key contributors identified by LASSO in the context of co‑exposure, and their specific roles require further investigation.

In recent years, PFASs have become one of the important substances in environmental pollution, which has triggered widespread concern among the public, scholars and relevant authorities⁴¹. Combined with previous studies, the results suggest that the implementation of policies related to the control of substances with PFASs can effectively reduce the concentration of PFASs in human serum. A study in the United States that recruited participants from 2005 to 2009⁴² had a lower concentration of serum PFASs than that in the Chinese cohort from 2013 to 2016⁴³. Precisely because the United States had already begun to phase out PFOA and its related products, while China had not yet imposed restrictions on PFASs during that period. However, the serum PFAS levels in pregnant women from both cohorts were higher than those observed in the present study. This finding suggests that population exposure to PFASs has improved since China began focusing on PFAS control in 2019. Due to the widespread dissemination, persistence, and significant hazards of PFASs, it is expected that more regulations and restrictions will be implemented in various countries to reduce PFASs exposure.

This study found that mixed exposure to PFASs was positively associated with blood glucose levels and GDM risk. Among them, the effects of PFASs on GDM risk and 2-h glucose levels were mainly driven by PFOA. The effects on FPG and 1-h glucose levels were mainly driven by PFBS. This phenomenon has also been previously reported in some studies. For example, a prospective cohort study conducted in the United States found that GDM increased nearly twofold with each standard deviation increase in serum PFOA level⁴². A repeat measurement-based prospective study in China examined the serum PFOA in 358 pregnant women in early pregnancy showed a positive correlation between PFOA levels and GDM risk⁴³. A multicentre prospective cohort study in the United States also found a significant association between PFOA exposure and GDM risk among pregnant women with a family history of type II diabetes⁴⁴. A cross-sectional study in China found a 0.09 mmol/L increase in 1-h glucose level (95% CI: 0.02, 0.17) for every doubling of PFBS⁴⁵. However, no correlation was found between PFASs and GDM risk by Wang et al. ’s study⁴⁶. Even PFASs were found to be negatively correlated with maternal blood glucose levels by Anne et al. ’s study⁴⁷. Differences in the results of these studies may be due to differences in sample size, genetic background, concentration and source of PFASs, living environment and lifestyle.

In real-world settings, population exposure to PFASs is typically characterized by mixed exposures. Therefore, investigating the impact of PFAS mixtures on blood glucose levels and GDM risk may provide more comprehensive evidence. However, we have only found two studies that have examined PFASs mixture exposure with GDM risk and changes in blood glucose levels. A prospective cohort study with Bayesian kernel machine regression (BKMR) modelling found that PFASs mixture exposure was positively associated with the incidence of GDM and 1-h and 2-h glucose levels²⁰. The Viva project assessed the PFASs mixture exposure on GDM using BKMR model and also found a positive correlation with blood glucose levels⁴⁸. In this study, both WQS regression and QGComp suggested a significant positive overall association of the PFAS mixture with blood glucose levels and GDM risk. Among the PFAS compounds, PFOA and PFBS contributed most to these associations. Moreover, after excluding participants with a family history of diabetes, these associations remained significant. These findings hold significant implications for elucidating the role of key PFASs and guiding strategies for PFASs control and health promotion. Although PFOA has been banned in most countries, our study found that its levels remain relatively high among pregnant women in China. These findings highlight the need for continued attention to its potential adverse effects on blood glucose levels and GDM risk.

The most important contribution of this study is the discovery of the mediating role of PPARs and L-FABP in the elevation of blood glucose levels by PFASs exposure and the increased risk of GDM. Only a limited number of previous studies have examined the associations of PPARs and L-FABP with PFASs or blood glucose levels. However, these studies did not consider them as mediators. Moreover, most of the existing evidence is based on animal or cellular experiments, with a lack of population-level data.

PPARα regulates insulin sensitivity by modulating lipid accumulation. Evidence from mouse models shows that suppression of PPARα expression alleviated insulin resistance⁴⁹. PPARα is predominantly expressed in the liver and skeletal muscle, where its expression is closely associated with that of L‑FABP. As the only fatty acid‑binding protein specifically expressed in the liver, L‑FABP promotes fatty acid β‑oxidation, reduces lipid accumulation, and improves insulin sensitivity⁵⁰. In patients with Type 2 diabetes mellitus (T2DM), L-FABP levels showed a positive correlation with Homeostatic Model Assessment of β-cell function (HOMA-β)⁵¹. Studies in diabetic transgenic mice supported this clinical association⁵². Collectively, these findings indicate that reduced L-FABP expression impairs β-cell function, thereby promoteing insulin deficiency and accelerating diabetes progression.

In this study, PPARα was found to be negatively correlated with L-FABP levels. However, some animal studies have demonstrated that intestinal PPARα signaling modulates dietary fatty acid uptake through upregulation of L-FABP, thereby influencing glucose and lipid metabolism⁴⁹. Experimental studies using primary mouse hepatocytes also revealed an increase in L-FABP expression when stimulated by PPARα⁵³. The occurrence of different results from the present study may be due to physiological differences, exposure pathways, genetic diversity, and experimental cycle limitations. Current research on the association between PPARα and L-FABP remains limited. Further mechanistic investigations and population-based studies are required to elucidate their specific roles in mediating PFASs-induced GDM pathogenesis.

The results of the four-way decomposition analysis showed that the total effect of FOSA-I on GDM in the models mediated by PPARα and L-FABP were mainly due to the pure indirect effect. Chain-mediated models also showed that FOSA-I exposure was positively associated with PPARα, PPARα was negatively associated with L-FABP, and L-FABP was negatively associated with GDM. This suggests that FOSA-I may increase the risk of GDM by affecting PPARα and L-FABP. FOSA-I is a precursor compound of PFOS that can be converted to PFOS in vivo or in the environment. Cellular experiments by Vanden et al. revealed that PFOS can activate PPARα in mice, rats and humans⁵⁴. Shipley et al. also found that both mouse and human PPARα were able to be activated by FOSA⁵⁵. This is consistent with our findings. However, no activation of PPARα by FOSA was found in Rosenmai’s study⁵⁶. With the restriction of PFOS use, there has been increasing attention toward precursors such as FOSA in recent research. However, existing studies remain limited with inconsistent findings, and no investigations have specifically explored the association between FOSA-I and L-FABP to date.

PPARγ is a major modulator of insulin sensitivity and glucose homeostasis, controlling the expression of numerous genes involved in peripheral glucose metabolism⁵⁷. Evidence from pharmacological and genetic studies underscores its critical role: synthetic agonists such as thiazolidinediones improve glucose tolerance by enhancing insulin sensitivity and restoring β-cell function⁵⁸, while dominant-negative mutations in the PPARγ gene are associated with severe hyperglycemia in patients⁵⁹. However, findings regarding the relationship between PPARγ and GDM remain inconsistent. Several human studies, including the present one, have reported downregulated PPARγ expression in adipose and placental tissues from GDM patients⁶⁰ and in mildly hyperglycemic rat models⁶¹. In contrast, upregulated PPARγ expression has been observed in placentas from severely hyperglycemic diabetic mice⁶². These discrepancies may partly reflect tissue-specific, species-specific, or glycemia-dependent variations in PPARγ regulation. Further studies examining PPARγ expression across different species, tissues, and degrees of hyperglycemic conditions are needed to resolve these conflicting observations.

Based on molecular docking and dynamics simulations⁶³, PFASs can bind to two specific sites on PPARγ. Notably, several emerging PFAS alternatives-such as ADONA, GenX, and 6:2 FTOH-exhibit relatively high binding affinities. HFPO-DA (trade name GenX) has been introduced as a polymerization aid in the production of high-performance fluoropolymers following the phase-out of PFOA. Nevertheless, its safety in the human body remains uncertain compared to PFOA. Its toxic effects in glycolipid metabolism have now been investigated. One study found that PPARγ and mRNA were highly expressed in placental tissue in rats exposed to HFPO-DA⁶⁴Evans et al. also identified HFPO-DA and HFPO-DA-AS as among the most potent activators in human and rat PPARγ assays⁶⁵. However, the study by Houck et al. showed no significant effect of HFPO-DA on PPARγ⁶⁶. Our findings demonstrate an inverse association between HFPO-DA exposure and PPARγ levels. While the effects of HFPO-DA on PPARγ regulation remain controversial in current literature, no population-based studies beyond the present investigation have specifically examined this relationship in human populations. The results of the four-way decomposition analysis showed that the total effect of HFPO-DA on GDM was mainly attributable to the pure indirect effect. The chain mediation model also showed that the mediating role of PPARγ was established and that PPARγ was negatively associated with GDM risk. Animal studies by Hosni et al. have demonstrated downregulated PPARγ expression in GDM rat models, suggesting its potential involvement in GDM pathogenesis³⁰. These findings align with the results observed in our current investigation.

The receptor-binding preference of HFPO-DA and FOSA-I can be partly explained by their distinct structural features. HFPO-DA is a short-chain ether PFAS with a branched perfluorinated backbone and a highly flexible ether linkage, which allows a better steric fit within the larger ligand-binding pocket of PPARγ. Molecular docking and dynamics simulations have shown that HFPO-DA demonstrates relatively high binding affinity for two specific sites on PPARγ, supporting its selective activation of this receptor^65,67. By contrast, FOSA-I contains a longer perfluorinated carbon chain and a sulfonamide functional group, which increases hydrophobic interactions with the PPARα binding pocket. PPARα preferentially binds ligands with elongated hydrophobic domains that resemble native fatty acids, and previous in vitro studies have found that sulfonamide PFASs exhibit stronger transactivation of PPARα than PPARγ. In addition, PPARα is predominantly expressed in the liver, which is also the primary site of accumulation for sulfonamide-containing PFASs due to their higher lipophilicity and hepatic uptake. This supports the notion that FOSA-I may primarily exert effects through PPARα-associated lipid oxidation pathways⁶⁸. Together, these structural and toxicokinetic differences provide a biologically plausible explanation for the differential receptor activation observed in our study.

Although previous studies have documented associations between PFASs and GDM, the underlying biological mechanisms remain unclear. Based on the regulatory roles of PPARs and L-FABP in glucose metabolism, we conducted a mediation analysis to explore their potential involvement. The results revealed a direct negative association between FOSA-I and HFPO-DA with GDM, whereas a positive indirect effect was observed through the PPARs and L-FABP pathways. This suggests that PPARs and L-FABP represent only one of several potential pathways contributing to PFAS-induced GDM pathogenesis. It is important toemphasize that our study is based on population-level statistical analysis; the specific biological mechanisms of PPARs and L-FABP in PFAS-induced GDM require further experimental research. Overall, our findings suggest the existence of a PFASs-PPARs-L-FABP-GDM axis and offer initial clues to the potential biological pathways underlying PFAS-related GDM development.

However, this study has several limitations. First, while we assessed PFAS exposure levels by measuring their concentrations in maternal serum, data on primary exposure pathways during pregnancy were still lacking: dietary habits, use of non-stick cookware, and consumption of disposable food packaging. Second, the study did not control for structurally similar persistent organic pollutants (POPs). These POPs may exert synergistic or analogous effects on glucose metabolism regulation. Third, our study relied on PFAS measurements from a single time point using early pregnancy serum samples. This limits the ability to capture changes in exposure levels throughout pregnancy. Fourth, the causal interpretation of our findings is inherently limited, as PFASs, PPARs, and L-FABP were all measured at the same time point. Nevertheless, the results consistently demonstrate a significant association between PFASs exposure and changes in PPARs and L-FABP levels during early pregnancy. Finally, an important methodological limitation is that our mediation analysis examined only single PFAS compounds as the exposure. To reflect the reality of combined exposure, future studies should develop analytical frameworks capable of assessing mediation pathways for complex PFAS mixtures as the holistic exposure variable.

Conclusion

Our study confirms the association between PFAS exposure and GDM risk and advances the field by revealing distinct underlying mechanisms. We provide evidence that HFPO-DA may influence GDM development primarily by dysregulating PPARγ. Additionally, we identify a potential pathway for FOSA-I, which involves the sequential modulation of PPARα and L-FABP. Our findings contribute to a more refined understanding of how specific PFAS congeners may disrupt glucose homeostasis during pregnancy. Importantly, they indicate a clear need for future environmental and public health risk assessments to consider the distinct mechanisms of action of different PFAS compounds.

Methods

Study population

This study recruited participants between January 2022 and June 2023 at Jiexiu Maternal and Child Healthcare Hospital in Shanxi Province, China. Pregnant women who had resided in the city for at least three years and planned to deliver at the hospital were enrolled as study participants. All participants provided written informed consent after being informed of the study’s purpose and procedures (Ethical Review No. 2021GLL077).

Inclusion criteria were as follows: (1) age 18 years or older; (2) in good health prior to pregnancy, with no history of diabetes, hypertension, coronary heart disease, or other cardiovascular diseases; (3) in early pregnancy, defined as within 12 weeks of gestation; and (4) having established a pregnancy health record at the city’s Maternal and Child Health Care Center, with all prenatal examinations conducted at this center.

Exclusion criteria were as follows: (1) unwillingness to participate or to provide biological samples; (2) twin or multiple pregnancies; and (3) incomplete records or adverse pregnancy outcomes, including preeclampsia, preterm birth (< 37 weeks), and fetal growth disorders.

This study adopted a nested case-control design. Oral glucose tolerance test (OGTT) results were obtained from 1, 680 participants. Among them, 247 were diagnosed with gestational diabetes mellitus (GDM), corresponding to a prevalence rate of 14.7% in the study population. Sample size estimation was conducted using PASS 15 software for nested case-control studies. Based on previous literature¹², the estimated exposure rate in the control group was 30%, with an expected odds ratio (OR) of 2.01.The β value was set at 0.10 and the α value at 0.01.Cases and controls were matched at a 1: 2 ratio, with a matching criterion of age ± 2 years. The estimated required sample size was 85 cases and 170 controls. Figure S4 presents the flow chart outlining the exclusion of participants from the study population.

Data collection

Before 12 weeks of gestation, data were collected through face-to-face interviews using a structured health questionnaire. Participants’ height and weight were measured by uniformly trained nurses. A 75 g OGTT was administered at 24–28 weeks of gestation to obtain fasting plasma glucose (FPG), 1-h plasma glucose, and 2-h plasma glucose levels. Potential confounding variables that could bias the association between PFAS exposure and blood glucose levels were identified based on previous literature^20,25,42,69 and directed acyclic graphs (Fig. S5).

Variable definition

GDM was diagnosed based on a 75 g OGTT conducted between 24 and 28 weeks of gestation, according to the criteria of the International Association of Diabetes and Pregnancy Study Groups (IADPSG)⁷⁰. A diagnosis of GDM was made if any one of the following thresholds was met: FPG ≥ 92 mg/dL (≥ 5.1mmol/L), 1-h plasma glucose ≥ 180 mg/dL (≥ 10.0mmol/L), or 2-h plasma glucose ≥ 153 mg/dL (≥ 8.5mmol/L).

Body mass index (BMI) was calculated as weight (kg) divided by height squared (m²). According to the weight classification criteria for Chinese adults recommended by WS/T 428–2013, BMI was categorized as underweight (BMI < 18.5 kg/m²), normal weight (18.5 ≤ BMI < 24 kg/m²), and overweight or obese (BMI ≥ 24 kg/m²). Alcohol consumption was defined as drinking at least once per week for a duration of six consecutive months. Smoking or passive smoking was defined as smoking more than one cigarette per day for over six months, or exposure to secondhand smoke for at least 15 min on at least one day per week⁷¹.

Laboratory tests

Fasting peripheral venous blood samples were collected before 12 weeks of gestation. The serum was separated by centrifugation, aliquoted into cryopreservation tubes, and stored at − 80 °C. Serum PFAS concentrations were measured using solid-phase extraction coupled with ultra-performance liquid chromatography-tandem mass spectrometry (SPE-UPLC-MS/MS; Agilent 1290–6490, Agilent Technologies Inc., USA). The 19 measured PFAS congeners were perfluorooctane sulfonate (PFOS), perfluorooctanoic acid (PFOA), perfluorononanoic acid (PFNA), perfluorodecanoic acid (PFDA), perfluorohexane sulfonate (PFHxS), perfluorododecanoic acid (PFDoA), perfluorobutane sulfonate (PFBS), perfluorobutanoic acid (PFBA), perfluoroheptanoic acid (PFHpA), perfluoroundecanoic acid (PFUdA), perfluoroheptane sulfonate (PFHpS), perfluoropentane sulfonate (PFPeS), perfluorotetradecanoic acid (PFTeDA), perfluoropentanoic acid (PFPeA), perfluorotridecanoic acid (PFTrDA), 6:2 Chlorinated polyfluoroalkyl ether sulfonic acid (6:2 ClPFESA), perfluorooctane sulfonamide (FOSA-I), hexafluoropropylene oxide dimer acid (HFPO-DA), 4:2 fluorotelomer sulfonic acid (4:2FTS). PFAS concentrations below the LOD were replaced by the LOD/√2^20,72,73. Levels of PPARα, PPARγ, and L-FABP were quantified by double-antibody sandwich ELISA (Wuhan Fine Biotech Co., Ltd).

Statistical analysis

Statistical analyzes were performed using R 4.2.2, SPSS 27.0, and SAS 9.4 (M7). Normally distributed variables were presented as mean ± standard deviation (), while non-normally distributed variables were expressed as median and interquartile range (IQR). Categorical variables were described as frequencies or percentages. Differences in baseline characteristics and related indicators between the gestational diabetes group and the control group were assessed using the χ² test and Mann–Whitney test. Correlation heatmap analysis was conducted to illustrate the relationships among individual PFAS compounds.

A least absolute shrinkage and selection operator (LASSO) regression model was employed to select PFAS variables that best fit the model. The analysis was performed using the glmnet package (version 4.1-8) in R, with the optimal tuning parameter (λ) determined via 12-fold cross-validation and using the default L1 penalty. The selected PFASs were subsequently incorporated into a weighted quantile sum (WQS) regression model⁷⁴ and a quantile-based g-computation (QGComp) models⁷⁵. Analyses were implemented using the gWQS package (version 3.0.5) and the qgcomp package (version 2.18.4). Individual PFAS concentrations were categorized into quartiles for the mixture analyses, which were constructed using 1000 bootstrap samples. These models were used to estimate the combined effect of mixed PFAS exposure on blood glucose levels and the risk of gestational diabetes. All models were adjusted for BMI and family history of diabetes. A two-sided p-value below 0.05 was considered statistically significant.

Causal mediation analyzes were conducted using the four-way decomposition analysis⁷⁶ and a chained mediation effects model. These analyzes aimed to explore whether changes in PPARs and L-FABP levels mediated the association between PFAS exposure and the risk of GDM. The four-way decomposition analysis partitions the total effect (TE) of the exposure on the outcome into four components (Fig. S6): (1) the controlled direct effect (CDE), which represents the direct effect of the exposure variable (X) when the mediator (M) is not present; (2) the reference interaction effect (INTref), representing the effect attributable solely to the interaction between X and M; (3) the mediated interaction effect (INTmed), which captures the effect due to the combined influence of X and M via mediation; and (4) the pure indirect effect (PIE), reflecting the effect mediated exclusively by M.

The PROCESS macro for SPSS 27.0 was used to examine whether the chained mediation effect involving the exposure variable (X) - mediator variable (M1) - mediator variable (M2) - outcome (Y) was statistically significant. The relationships within the chained mediation model are illustrated in Fig. S7. In this model, three types of mediation effects are identified: a₁-b₁ represents the mediating effect of M1, a₂-b₂ represents the mediating effect of M2, and a₁-a₃-b₂ represents the mediation effect of M1 through M2. The path c denotes the direct effect of the exposure variable X on the outcome Y.

To mitigate the potential confounding influence of genetic factors on GDM risk and glucose metabolism, we conducted a sensitivity analysis by excluding participants with a family history of diabetes and re-evaluated the associations between PFASs exposure and GDM and dysregulated glucose metabolism.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Qikai Xiang and Lijian Lei did the background research and conceptualization. Qikai Xiang, Pu Guo and Yufen Liang contributed to the data curation. Qikai Xiang conducted the formal analysis. Pu Guo and Qinghua Tian helped to validate and interpret the results. Qinghua Tian, Yufen Liang and Fan Zhang did the Indicator detection, and took responsibility for the integrity and accuracy of the data. Lingyun Zhuo, Yuebin Yang and Chengzhen Zhao helped to the data collection. Qikai Xiang drafted the manuscript. Yingying Zhang and Lijian Lei supervised the project, carefully reviewed and edited the manuscript.

Funding

This work was supported by the National Natural Science Foundation of China (82103938), as well as the Ten Billion Project of Shanxi Medical University (Project No. BYBLD002) and the Shanxi Provincial Graduate Education Innovation Program(2025SJ224).

Data availability

The datasets generated during and analyzed in the current study are available from the corresponding author upon reasonable request.

Declarations

Competing interests

The authors declare no competing interests.

Ethics approval and consent to participate

This study adhered to the principles outlined in the Declaration of Helsinki. Approval was obtained from the Research Ethics Committee of the Shanxi medical University (approval number [No. 2021GLL077]), and written consent was obtained from all participating women.