Association between serum perfluoroalkyl substances and COPD risk: population evidence and in vitro validation

Abstract

Background

Per- and polyfluoroalkyl substances are characterized by environmental persistence, bioaccumulation potential, and multi-organ toxicity. Given their ubiquitous presence in the environment and human serum, concerns regarding respiratory health risks are growing, particularly due to scarce evidence on environmental etiologies of chronic obstructive pulmonary disease in never-smokers. This study aimed to investigate the mechanistic role of these substances in chronic obstructive pulmonary disease through a population-computational-experimental paradigm, with specific focus on lipid-metabolic mediation.

Methods

Data from the National Health and Nutrition Examination Survey (2007–2018) were analyzed using weighted quantile sum regression and quantile-based g-computation to assess mixture exposure effects. Mediation analysis was performed to evaluate the triglyceride-glucose index as a metabolic intermediate pathway. Network toxicology and molecular docking analyses were conducted to identify core protein targets. Human bronchial epithelial cells were exposed to perfluorooctanoic acid to validate target gene expression and downstream pathway activation.

Results

Mixture exposure showed a significant positive association with chronic obstructive pulmonary disease risk. Perfluorooctanesulfonic acid and perfluorooctanoic acid were the primary toxicity contributors. The association remained robust among never-smokers. The triglyceride-glucose index significantly mediated the exposure-disease relationship (mediation proportion: 6.6%-7.8%). SRC, EGFR, PPARG, and MMP9 were identified as core targets. Experimental validation confirmed that perfluorooctanoic acid altered expression of these targets, activating inflammation and remodeling pathways.

Conclusions

These substances disrupt pulmonary homeostasis through concurrent molecular activation and lipid-metabolic disturbance, evidenced by triglyceride-glucose index mediation. This dual mechanism provides new evidence for chronic obstructive pulmonary disease prevention in never-smokers and identifies potential metabolic intervention targets.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12890-026-04490-x.

Introduction

Chronic obstructive pulmonary disease (COPD) is characterized by persistent, usually progressive airflow limitation. The Global Burden of Disease (GBD) 2021 study estimates that COPD affects 392 million people across 204 countries and territories, accounting for approximately 3.15 million deaths annually and ranking as the fourth leading cause of global mortality [1]. Boers et al. projected that the global prevalence of chronic obstructive pulmonary disease will rise by 23% by 2050, driven primarily by population aging and persistent exposure to high-risk environmental factors in low- and middle-income countries [2]. Furthermore, COPD imposes a substantial economic and societal burden. For instance, in the United States, the cumulative direct medical costs attributable to COPD over a 20-year period (2019–2038) are estimated to reach 800.9 billion dollars, with projected indirect societal costs amounting to an additional 101.3 billion dollars [3].

Although tobacco smoking remains the dominant risk factor for COPD, 25–45% of affected individuals have never smoked, which is particularly high in Asia and Africa, underscoring the critical role of non-tobacco environmental exposures [4]. These exposures include indoor biomass smoke, ambient air pollution, occupational dust and fumes, environmental tobacco smoke, and recurrent respiratory infections, with women often bearing a disproportionate burden [5]. Among the spectrum of environmental pollutants, emerging evidence implicates persistent organic pollutants especially per- and polyfluoroalkyl substances (PFAS) as a hitherto under-investigated environmental driver of COPD pathogenesis.

PFAS are a family of synthetic chemicals characterized by a strong carbon–fluorine bond that confers exceptional hydrophobicity and oleophobicity. These properties have led to their widespread use in industrial processes and consumer products, including fire-fighting foams, water- and stain-resistant coatings, food packaging, and non-stick cookware [6]. PFAS exhibit extreme environmental persistence, long-range transport potential, and high bioaccumulation capacity, allowing them to remain in environmental matrices and biological tissues for decades and to biomagnify through food webs, thereby posing prolonged risks to ecosystems and human health [7–9]. Their extended human half-lives result in continuous body burdens that can trigger multi-organ toxicity [10]. Consequently, the United States, the European Union, China, and other jurisdictions have intensified regulation of these “forever chemicals” [11].

Extensive studies have shown that PFAS can exert significant toxic effects on multiple organ systems in mammals, including the kidneys, liver, nervous system, reproductive system, and immune system [12–15]. Recently, emerging evidence shows that PFAS also implicate in the respiratory system. NHANES-based analyses have linked higher serum PFAS concentrations to reduced lung function in adolescents and to elevated PFAS burdens among adults with COPD [16–18]. Experimental work further indicates that PFAS impair lung function via epigenetic modification, oxidative stress, and inflammatory signaling [19]. Jabeen et al. reported that both perfluorooctanoic acid (PFOA) and perfluorooctane sulfonic acid (PFOS) provoke genome-wide DNA-methylation changes in A549 cells, whereas Zhou et al. showed that PFOS exposure induces hyper-methylation and down-regulation of HSD17B1, subsequently suppressing surfactant protein-B (SP-B) expression [20, 21]. However, most studies have examined single congeners, leaving the combined effects of real-world PFAS mixtures largely uncharacterized. Consequently, a clear mechanistic pathway linking population observations to experimental evidence is still missing.

To systematically address these knowledge gaps, we adopted a Population–Computation–Experiment triad. First, we mined the 2007–2018 NHANES cycles to quantify the association between real-world PFAS mixtures and COPD. Next, network toxicology was used to prioritize key molecular targets and signaling pathways. Finally, we established a PFOA-exposed human bronchial epithelial cell model to verify cytotoxic injury and elucidate underlying mechanisms. This integrated strategy aims to provide a robust evidence chain for PFAS-induced respiratory toxicity and to inform evidence-based updates of environmental standards and public-health interventions.

Materials and methods

Study population

NHANES is a serial, cross-sectional program conducted by the National Center for Health Statistics (NCHS) to assess the health and nutritional status of the civilian, non-institutionalized U.S. population aged 6–79 years. We pooled six 2-year cycles (2007–2018) of publicly available NHANES data: 2007–2008 (n = 10,149), 2009–2010 (n = 10,537), 2011–2012 (n = 9,756), 2013–2014 (n = 10,175), 2015–2016 (n = 9,971), and 2017–2018 (n = 9,254), yielding a total of 59,842 participants examined during this period. Among these, 47,729 were excluded because serum PFAS concentrations were not measured, and an additional 3,275 were excluded due to missing covariates (n = 3,180) or pregnancy (n = 95). The present analysis therefore comprised 8,838 adults with complete data, of whom 1,021 met the pre-defined criteria for COPD (Fig. 1).A complete case analysis was applied; participants with missing serum PFAS concentrations, covariates, or pregnancy status were excluded. All study data were obtained from the publicly accessible NHANES repository (https://wwwn.cdc.gov/nchs/nhanes/Default.aspx).

Fig. 1.

Study flow diagram. Of 59,842 NHANES 2007–2018 participants, 47,729 were excluded for missing serum PFAS data and 3,275 for missing covariates or ineligibility. The final sample comprised 8,838 adults, including 1,021 with COPD

Assessment of perfluorinated compound exposure in blood

Serum samples were collected from participants, processed, and stored under appropriate frozen conditions (− 20 ℃) before being shipped to the Division of Laboratory Sciences, National Center for Environmental Health, Centers for Disease Control and Prevention (CDC) for analysis. Detailed protocols for serum collection, storage, and PFAS measurement are provided in the NHANES Laboratory Procedures Manual (https://wwwn.cdc.gov/Nchs/Data/Nhanes/Public/2017/DataFiles/PFAS_J.htm). Serum concentrations of four PFAS, perfluorohexane sulfonic acid (PFHxS), perfluorononanoic acid (PFNA), PFOA, and PFOS, were quantified on-line by solid-phase extraction coupled to high-performance liquid chromatography–Turbo Ion Spray tandem mass spectrometry (on-line SPE-HPLC-TIS-MS/MS). Values below the limit of detection (LOD) were replaced by LOD/√2. Total serum PFAS was calculated as the sum of the four measured congeners to reflect composite exposure, given their high inter-correlation in human serum.

The definition of COPD

Spirometric variables—forced expiratory volume in 1 s (FEV₁), forced vital capacity (FVC), and the FEV₁/FVC ratio—were extracted from the NHANES examination files. Data on physician-diagnosed respiratory disease were obtained from the computer-assisted personal interview. Participants were classified as having COPD if they met both of the following criteria:

1. FEV₁/FVC < 0.70 post-bronchodilator;

2. self-reported physician diagnosis of chronic bronchitis or emphysema (“Has a doctor or other health professional ever told you that you had chronic bronchitis or emphysema?”).

Evaluation of covariates

Following the approach of Huang Q et al. [22], we selected covariates potentially associated with COPD outcomes. The following variables were included as covariates in our analysis: age (years), sex (male/female), race/ethnicity (non-Hispanic White, non-Hispanic Black, Mexican American, other Hispanic, and other), education level (< high school, high school graduate, and college or above), marital status (married/living with partner, widowed, divorced, separated, and never married), family income-to-poverty ratio (FIPR: < 1.5, ≥ 1.5 to < 3.5, and ≥ 3.5), body mass index (BMI: normal weight, overweight, and obese), alcohol consumption (light, moderate, and heavy), and smoking status (never, former, and current smoker), although residual confounding from unmeasured environmental and occupational exposures (e.g., biomass fuel use, indoor air pollutants, vaping, and second- or third-hand smoke) cannot be fully excluded.

Screen potential molecular targets

Putative targets of the PFAS congeners were first retrieved from PubChem. The structures were then submitted to PharmMapper, and the resulting hits were cross-validated against TargetBank, DrugBank, BindingDB, and the Potential Drug Target Database (PDTD). COPD-related targets were compiled by querying GeneCards (https://www.genecards.org) and OMIM (https://omim.org) using the search term “chronic obstructive pulmonary disease”. Venn analysis was performed to extract the intersection between PFAS-binding proteins and COPD-associated gene products; these overlapping molecules were regarded as potential PFAS-specific targets implicated in COPD pathogenesis.

Gene Ontology (GO) enrichment analysis (biological process, cellular component, and molecular function) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis were performed using the bioinformatics online platform (https://www.bioinformatics.com.cn) on the 57 shared genes with P < 0.05 considered statistically significant. Finally, a protein–protein interaction (PPI) network was constructed and visualized using Cytoscape (v3.8.2).

Molecular docking

The three-dimensional structures of the hub target proteins were retrieved from the Research Collaboratory for Structural Bioinformatics (RCSB) Protein Data Bank (PDB). The structural data of the four PFAS congeners (PFHxS, PFNA, PFOA, and PFOS) were obtained from the PubChem database. Using AutoDockTools 1.5.6, both the PFAS compounds (ligands) and the hub proteins (receptors) were preprocessed by removing water molecules, adding polar hydrogen atoms, and assigning Gasteiger charges. Molecular docking simulations were subsequently performed using AutoDock Vina integrated with AutoDockTools 1.5.6. Binding free energies (ΔG) for all four congeners are reported in Table S6. The resulting docking poses were visualized and analyzed using PyMOL (Version 2.5.0).

Cell culture

Human bronchial epithelial Beas-2B cells were maintained in complete medium consisting of RPMI-1640 supplemented with 10% fetal bovine serum (FBS) and 1% penicillin–streptomycin, and cultured at 37 ℃ in a humidified atmosphere containing 5% CO₂. When microscopic examination indicated 80–90% confluence, the spent medium was aspirated and the monolayer was gently rinsed three times with PBS. Cells were then incubated with 0.25% trypsin-EDTA for 1–2 min until they became rounded and refractile. After removal of the enzyme, 3 mL of complete medium was added to quench trypsin activity; cells were detached by careful, gentle pipetting over the growth surface and sub-cultured at the desired ratio into fresh flasks.

Cell viability assay

Following trypsinisation, Beas-2B cells were counted and seeded into 96-well plates at 1 × 10⁵ cells per well. After adherence, the medium was replaced with fresh complete medium containing PFOA at 0, 50, 100, 200, 400, 500, 600, 700 or 800 µM (six replicate wells per concentration). Peripheral wells were filled with PBS to minimize evaporation. After 24 h exposure, 10 µL CCK-8 reagent was added to each well and the plates were incubated for 3.5 h at 37 ℃. Absorbance was measured at 450 nm with a microplate reader, and relative cell viability was calculated from the optical density (OD) values.

Real-time quantitative PCR (RT-qPCR)

Total RNA was extracted from cells with TRIzol reagent, quantified on a NanoDrop 2000 spectrophotometer, and reverse-transcribed into cDNA using the Reverse Transcription Kit (EG15133S, Yugong Biotech, Lianyungang, China). RT-qPCR amplifications were subsequently performed on a QuantStudio 6 Flex system with Taq SYBR Green qPCR Premix (EG20117M, Yugong Biotech). Primer sequences are listed in Table S1. All five shared hub genes were subjected to RT-qPCR primer design and pre-validation; ESR1 was excluded due to consistently low expression below the detection limit, and the remaining four genes were carried forward to formal analysis.

Statistical analysis

All analyses were conducted with R software (v4.4.1). Missing data were addressed via complete case analysis. Statistical analyses were performed using the following R packages: “survey” for complex survey design and weighted logistic regression, “gWQS” or weighted quantile sum regression, “qgcomp” for quantile-based g-computation, “rms” for restricted cubic spline regression, and “mediation” for causal mediation analysis. Descriptive statistics were used to characterize participants’ demographic and clinical profiles, Table 1 presents unadjusted group comparison; categorical variables are presented as absolute numbers (n) or percentages (%). The association between individual serum PFAS concentrations and COPD was evaluated with survey-weighted univariable and multivariable logistic regression models. The lowest quartile (Q1) served as the reference category, and results are reported as odds ratios (ORs) with corresponding 95% confidence intervals (CIs). Three sequential models were constructed: (i) Crude model; (ii) Model 1, adjusted for age, sex, and race/ethnicity; and (iii) Model 2, additionally adjusted for family income-to-poverty ratio (FIPR), BMI, education level, marital status, smoking status, and alcohol consumption. Restricted cubic spline (RCS) regression was employed to explore potential dose–response relationships between serum PFAS and asthma. Weighted quantile sum (WQS) regression and quantile-based g-computation (QGComp) were used to estimate the overall effect of the PFAS mixture and the relative contribution of each congener. Two-sided P < 0.05 was considered statistically significant for all tests. To explore the mediating role of metabolic disturbance in the association between PFAS exposure and COPD, we conducted causal mediation analysis with the triglyceride-glucose (TyG) index—a well-established surrogate marker of insulin resistance—as the mediator. The analysis followed a three-step procedure: first, we constructed a multivariable regression model of PFAS exposure on TyG, adjusting for all covariates; second, we constructed a multivariable regression model of PFAS exposure on COPD, additionally adjusting for TyG to estimate the direct effect; and finally, we calculated the average causal mediation effect (ACME), the average direct effect (ADE), and the proportion of mediation. All mediation analyses were performed using the “mediation” package in R software.

Table 1.

Characteristics of the study participants: NHANES 2007–2018

Characteristics	Overall (n = 8838)1	Non-COPD (n = 7817, 88.4%)1	COPD (n = 1021, 11.6%)1	P-value
Age (years)	45 (31, 59)	43 (30, 57)	57 (45, 66)	< 0.001
Sex, n (%)				0.500
Female	4546 (51.4%)	4039 (51.7%)	507 (49.7%)
Male	4292 (48.6%)	3778 (48.3%)	514 (50.3%)
Race, n (%)				< 0.001
Non-Hispanic White	3714 (42.0%)	3116 (39.9%)	598 (58.6%)
Non-Hispanic Black	1799 (20.4%)	1606 (20.5%)	193 (18.9%)
Mexican American	1330 (15.0%)	1254 (16.0%)	76 (7.4%)
Other race	1110 (12.6%)	1033 (13.2%)	77 (7.5%)
Other Hispanic	885 (10.0%)	808 (10.3%)	77 (7.5%)
Education level, n (%)				< 0.001
Below High School	1990 (22.5%)	1710 (21.9%)	280 (27.4%)
High School Grad/GED	1834 (20.8%)	1577 (20.2%)	257 (25.2%)
College or above	4456 (50.4%)	3973 (50.8%)	483 (47.3%)
Missing	558(6.3%)	557 (7.1%)	1 (< 0.1%)
Marital status, n (%)				< 0.001
Married/living with partner	4949 (56.0%)	4361 (55.8%)	588 (57.6%)
Missing	557 (6.3%)	557 (7.1%)	0 (0%)
Widowed/divorced/separated/never married	3332 (37.7%)	2899 (37.1%)	433 (42.4%)
Family income ratio (FIR)	2.90 (1.43, 5.00)	2.94 (1.45, 5.00)	2.70 (1.27, 4.83)	0.035
FIR level, n (%)				0.043
Low (<1.5)	3389 (38.3%)	2947 (37.7%)	442 (43.3%)
Medium (≥ 1.5, <3.5)	2814 (31.8%)	2502 (32.0%)	312 (30.6%)
High (≥ 3.5)	2635 (29.8%)	2368 (30.3%)	267 (26.1%)
BMI, kg/m2	28 (24, 32)	28 (24, 32)	28 (24, 33)	0.200
BMI level, n (%)				0.400
Normal (<25 kg/m2)	2758 (31.2%)	2455 (31.4%)	303 (29.7%)
Overweight (≥ 25, <30 kg/m2)	2783 (31.5%)	2471 (31.6%)	312 (30.5%)
Obese (≥ 30 kg/m2)	3297 (37.3%)	2891 (37.0%)	406 (39.8%)
Alcohol, n (%)				0.001
Mild	4801 (54.3%)	4180 (53.5%)	621 (60.8%)
Moderate	1167 (13.2%)	1030 (13.2%)	137 (13.4%)
Major	1463 (16.6%)	1291 (16.5%)	172 (16.8%)
Missing	1407 (15.9%)	1316 (16.8%)	91 (8.9%)
Smoking, n (%)				< 0.001
Never smoker	4776 (54.0%)	4450 (56.9%)	326 (31.9%)
Former smoker	1948 (22.0%)	1605 (20.5%)	343 (33.6%)
Current smoker	1687 (19.1%)	1335 (17.1%)	352 (34.5%)
Missing	427 (4.8%)	427 (5.5%)	0 (0%)
PFHxS, ng/ml	1.40 (0.80, 2.50)	1.40 (0.80, 2.40)	1.60 (1.00, 2.80)	0.002
PFNA, ng/ml	0.81 (0.50, 1.31)	0.80 (0.50, 1.30)	0.98 (0.60, 1.39)	< 0.001
PFOA, ng/ml	2.27 (1.40, 3.67)	2.20 (1.37, 3.57)	2.77 (1.71, 4.27)	< 0.001
PFOS, ng/ml	7.00 (4.00, 12.00)	7.00 (4.00, 12.00)	8.00 (5.00, 15.00)	< 0.001
PFAS, ng/ml	12.00 (7.00, 20.00)	12.00 (7.00, 20.00)	15.00 (9.00, 23.00)	< 0.001

¹Median (Q1, Q3); n (unweighted)

²Design-based KruskalWallis test; Pearson’s X^2: Rao & Scott adjustment

Results

Characteristics of the NHANES population

This study analyzed 8,838 adults aged 31–59 years (1,021 COPD cases) from the 2007–2018 NHANES cycles. Compared with non-COPD controls, COPD participants were older and differed significantly with respect to race/ethnicity, education, marital status, family income-to-poverty ratio, alcohol consumption, and smoking status (all P < 0.05); no differences were observed for sex or BMI. Serum concentrations of PFHxS, PFNA, PFOA, and PFOS were all markedly higher in COPD cases than in controls (P < 0.05). Detailed characteristics of the study participants are presented in Table 1.

Logistic regression analysis of the association between PFAS exposure and COPD prevalence

As shown in Table 2, in the Crude model, positive dose-dependent associations with COPD prevalence were observed across Q2-Q4 for total serum PFAS, ln-PFHxS and ln-PFOS, peaking in Q4 [OR (95%CI) = 2.03 (1.49, 2.76), 1.73 (1.22, 2.44), and 2.08 (1.55, 2.80)], whereas ln-PFNA and ln-PFOA exhibited significant elevations in Q3 and Q4, with maxima at Q3 for PFNA [OR (95%CI) = 1.81 (1.44, 2.29)] and Q4 for PFOA [OR (95%CI) = 2.08 (1.58, 2.74)]. These relationships remained robust and statistically significant after progressive multivariable adjustment, with total PFAS retaining significance only in Q4 in both Model 1 and Model 2 [OR (95%CI) = 1.42 (1.02, 1.97), 1.50 (1.08, 2.09)], PFHxS significant solely at Q2 [OR (95%CI) = 1.42 (1.09, 1.86), 1.46 (1.09, 1.95)], and the highest effect estimates for PFNA, PFOA and PFOS in the fully adjusted model being 1.63 (1.28, 2.08), 1.54 (1.15, 2.05), and 1.61 (1.16, 2.24), respectively.

Table 2.

Association between Per- and polyfluoroalkyl substances exposure and the prevalence of COPD

Characteristics	Crude		Model 1		Model 2
	OR1(95% CI1)	P-value	OR1(95% CI1)	P-value	OR1(95% CI1)	P-value
lnPFHxS.quantile
Q1	—		—		—
Q2	1.65(1.26, 2.15)	< 0.001	1.42(1.09, 1.86)	0.011	1.46(1.09, 1.95)	0.013
Q3	1.57(1.20, 2.03)	0.001	1.22(0.91, 1.62)	0.200	1.20(0.88, 1.63)	0.200
Q4	1.73(1.22, 2.44)	0.002	1.30(0.90, 1.88)	0.200	1.32(0.90, 1.93)	0.200
lnPFNA.quantile
Q1	—		—		—
Q2	1.14(0.87, 1.50)	0.300	1.01(0.76, 1.36)	> 0.900	1.10(0.83, 1.47)	0.500
Q3	1.81(1.44, 2.29)	< 0.001	1.53(1.20, 1.94)	< 0.001	1.63(1.28, 2.08)	< 0.001
Q4	1.74(1.35, 2.25)	< 0.001	1.41(1.08, 1.84)	0.013	1.52(1.16, 1.99)	0.003
lnPFOA.quantile
Q1	—		—		—
Q2	1.29(0.97, 1.71)	0.074	1.11(0.82, 1.49)	0.500	1.08(0.79, 1.48)	0.600
Q3	1.72(1.35, 2.20)	< 0.001	1.33(1.03, 1.71)	0.032	1.38(1.06, 1.81)	0.018
Q4	2.08(1.58, 2.74)	< 0.001	1.57(1.17, 2.10)	0.003	1.54(1.15, 2.05)	0.004
lnPFOS.quantile
Q1	—		—		—
Q2	1.51(1.12, 2.03)	0.007	1.29(0.95, 1.75)	0.100	1.31(0.95, 1.81)	0.094
Q3	1.85(1.44, 2.38)	< 0.001	1.44(1.12, 1.85)	0.005	1.58(1.22, 2.06)	< 0.001
Q4	2.08(1.55, 2.80)	< 0.001	1.45(1.06, 2.00)	0.021	1.61(1.16, 2.24)	0.006
lnPFAS.quantile
Q1	—		—		—
Q2	1.43(1.04, 1.96)	0.028	1.22(0.88, 1.68)	0.200	1.22(0.88, 1.69)	0.200
Q3	1.67(1.28, 2.19)	< 0.001	1.25(0.94, 1.65)	0.120	1.32(0.99, 1.74)	0.055
Q4	2.03(1.49, 2.76)	< 0.001	1.42(1.02, 1.97)	0.036	1.50(1.08, 2.09)	0.017

¹OR Odds Ratio, CI Confidence Interval. Model 1 was adjusted for age, sex, and race/ethnicity; Model 2 was further adjusted for family income-to-poverty ratio (FIR), body mass index (BMI), education level, marital status, smoking status, and alcohol consumption.

Potential dose–response relationship between serum PFAS concentrations and COPD

The preceding analyses indicated a dose-dependent manner between PFAS exposure and COPD risk, we therefore used RCS regression to characterize the dose–response curves. As shown in Fig. 2, total ln-PFAS exhibited a positive, essentially linear relationship with COPD prevalence (P_non−linear = 0.2043). Similar positive linear trends were observed for ln-PFHxS, ln-PFNA, ln-PFOA and ln-PFOS (P_non−linear = 0.2043, 0.6591, 0.6385 and 0.3412). Notably, increasing exposures to PFNA, PFOA and PFOS were each significantly associated with progressively higher COPD risk (P = 0.0013, 0.0075, and 0.0004).

Fig. 2.

Restricted cubic spline (RCS) curves depicting the dose–response relationships between per- and polyfluoroalkyl substances (PFAS) and the risk of COPD. (A) ln-PFHxS versus COPD risk. (B) ln-PFNA versus COPD risk. (C) ln-PFOA versus COPD risk. (D) ln-PFOS versus COPD risk. (E) Total ln-PFAS versus COPD risk

Association between serum PFAS mixture exposure and COPD

We next applied QGComp and WQS regression to quantify the joint effect of the four PFAS on COPD prevalence and to estimate the relative contribution of each congener (Fig. 3). After adjustment for all covariates, the combined PFAS mixture was positively associated with COPD. In the WQS model (Fig. 3A), PFOS contributed most to the overall effect (weight = 55.3%), followed by PFOA (35.0%), PFNA (8.5%), and PFHxS (1.2%). The QGComp analysis revealed a different pattern (Fig. 3B): PFOS, PFOA, and PFNA exerted positive weights, whereas PFHxS showed a modest negative weight. PFOA displayed the strongest positive association with COPD in this framework.

Fig. 3.

Weighted quantile sum (WQS) and quantile-based g-computation (QGComp) analyses of the joint effect of per- and polyfluoroalkyl substance (PFAS) mixtures on COPD risk. (A) Relative contribution (%) of individual PFAS to the WQS index for COPD. (B) Positive or negative weight of each PFAS in the QGComp mixture–COPD association

Subgroup analysis of the association between serum PFAS exposure levels and COPD

Subgroup analyses were performed to evaluate whether the association between serum PFAS exposure and COPD varied by individual characteristics (Table 3). Total ln-PFAS was positively associated with COPD among participants with ≤ high-school education [OR (95%CI) = 1.27 (1.01, 1.59)], normal BMI [OR (95%CI) = 1.27 (1.00, 1.61)] and never-smokers [OR (95%CI) = 1.28 (1.02, 1.61)]. ln-PFHxS was associated with COPD only in the low-FIPR subgroup [OR (95%CI) = 1.18 (1.02, 1.36)]. ln-PFNA showed positive associations among participants aged ≥ 50 years [OR (95%CI) = 1.25 (1.03, 1.51)], males [OR (95%CI) = 1.35 (1.14, 1.59)], those married/living with a partner [OR (95%CI) = 1.20 (1.01, 1.42)], low FIPR [OR (95%CI) = 1.20 (1.02, 1.40)], normal or overweight BMI [OR (95%CI) = 1.32 (1.03, 1.67), 1.31 (1.06, 1.61)], light drinkers [OR (95%CI) = 1.21 (1.03, 1.41)], and never-smokers [OR (95%CI) = 1.44 (1.16, 1.79)]. ln-PFOA was positively linked among participants with ≤ high-school education [OR (95%CI) = 1.34 (1.06, 1.69)], low FIPR [OR (95%CI) = 1.24 (1.01, 1.51)], overweight BMI [OR (95%CI) = 1.25 (1.01, 1.55)], and light alcohol consumption [OR (95%CI) = 1.22 (1.04, 1.43)]. Finally, ln-PFOS exhibited positive associations among participants aged < 50 years [OR (95%CI) = 1.22 (1.03, 1.44)], low FIPR [OR (95%CI) = 1.19 (1.03, 1.38)], normal BMI [OR (95%CI) = 1.23 (1.01, 1.49)], heavy drinkers [OR (95%CI) = 1.27 (1.00, 1.62)], and never-smokers [OR (95%CI) = 1.23 (1.01, 1.50)].

Table 3.

Subgroup analysis

Characteristics			lnPFHxS	lnPFNA			lnPFOA	lnPFOS	lnPFAS
Age (years)
<50			1.06(0.91, 1.24)	1.12(0.92, 1.35)			1.16(0.99, 1.37)	1.22(1.03, 1.44)	1.17(0.98, 1.40)
≥ 50			1.08(0.91, 1.29)	1.25(1.03, 1.51)			1.19(0.95, 1.48)	1.15(0.97, 1.38)	1.20(0.97, 1.49)
Sex
Male			0.94(0.78, 1.13)	1.35(1.14, 1.59)			1.19(0.99, 1.44)	1.12(0.93, 1.33)	1.12(0.92, 1.37)
Female			1.12(0.95, 1.32)	1.05(0.89, 1.24)			1.12(0.91, 1.37)	1.14(0.98, 1.32)	1.14(0.96, 1.37)
Race
Non-Hispanic White			1.01(0.86, 1.18)	1.14(0.97, 1.35)			1.10(0.93, 1.30)	1.10(0.94, 1.28)	1.09(0.92, 1.30)
Non-Hispanic Black			1.04(0.88, 1.22)	1.16(0.99, 1.37)			1.12(0.95, 1.33)	1.15(0.98, 1.34)	1.14(0.95, 1.36)
Mexican American			1.12(0.78, 1.60)	1.35(0.99, 1.84)			1.21(0.83, 1.76)	1.20(0.87, 1.64)	1.27(0.87, 1.86)
Other race			0.96(0.71, 1.30)	1.04(0.79, 1.37)			1.12(0.80, 1.57)	1.16(0.87, 1.55)	1.12(0.79, 1.59)
Education
Below High School	1.13(0.94, 1.36)			1.22(0.98, 1.51)	1.34(1.06, 1.69)			1.21(0.99, 1.48)	1.27(1.01, 1.59)
High School Grad/GED	1.08(0.85, 1.39)			1.15(0.95, 1.39)	1.21(0.97, 1.50)			1.21(0.99, 1.48)	1.20(0.95, 1.51)
College or above	0.97(0.83, 1.12)			1.18(0.98, 1.43)	1.06(0.87, 1.30)			1.08(0.92, 1.27)	1.07(0.89, 1.29)
Marital status
Married/living with partner	1.05(0.89, 1.25)			1.20(1.01, 1.42)	1.11(0.94, 1.31)			1.11(0.95, 1.30)	1.12(0.94, 1.33)
Widowed/divorced/separated/Never married	0.97(0.82, 1.14)			1.09(0.90, 1.33)	1.22(0.98, 1.50)			1.16(0.99, 1.36)	1.15(0.95, 1.40)
FIR
Low (<1.5)	1.18(1.02, 1.36)			1.20(1.02, 1.40)	1.24(1.01, 1.51)			1.19(1.03, 1.38)	1.15(0.95, 1.40)
Medium (≥ 1.5,<3.5)	0.93(0.75, 1.15)			1.20(0.98, 1.47)	1.15(0.93, 1.42)			1.09(0.90, 1.32)	1.07(0.85, 1.33)
High (≥ 3.5)	0.97(0.77, 1.22)			1.13(0.84, 1.52)	1.07(0.80, 1.43)			1.11(0.88, 1.41)	1.10(0.84, 1.44)
BMI (kg/m 2 )
Normal (<25)	1.12(0.92, 1.36)			1.32(1.03, 1.67)	1.24(0.97, 1.59)			1.23(1.01, 1.49)	1.27(1.00, 1.61)
Overweight (≥ 25,<30)	1.12(0.91, 1.38)			1.31(1.06, 1.61)	1.25(1.01, 1.55)			1.16(0.95, 1.43)	1.20(0.95, 1.50)
Obese (≥ 30)	0.89(0.75, 1.06)			0.99(0.80, 1.23)	1.01(0.78, 1.31)			1.04(0.87, 1.25)	1.01(0.81, 1.25)
Alcohol
Mild	1.06(0.90, 1.26)			1.21(1.03, 1.41)	1.22(1.04, 1.43)			1.13(0.96, 1.33)	1.16(0.96, 1.39)
Moderate	0.98(0.76, 1.26)			1.01(0.66, 1.55)	1.04(0.60, 1.81)			1.06(0.77, 1.45)	1.06(0.71, 1.56)
Major	0.89(0.68, 1.17)			1.16(0.90, 1.49)	1.01(0.78, 1.31)			1.27(1.00, 1.62)	1.14(0.86, 1.50)
Smoking
Never smoker	1.12(0.94, 1.35)			1.44(1.16, 1.79)	1.20(0.91, 1.58)			1.23(1.01, 1.50)	1.28(1.02, 1.61)
Characteristics	lnPFHxS			lnPFNA	lnPFOA			lnPFOS
Former smoker	0.97(0.77, 1.21)			1.02(0.83, 1.25)	1.02(0.80, 1.29)			1.05(0.87, 1.27)	1.03(0.83, 1.29)
Current smoker	1.00(0.83, 1.20)			1.07(0.87, 1.33)	1.22(1.01, 1.47)			1.12(0.94, 1.34)	1.10(0.90, 1.34)

Bold value indicates significance

TyG index mediates the association between PFAS exposure and COPD

Causal mediation analysis revealed that the TyG index significantly mediated the associations between individual PFAS exposures and COPD risk (Fig. 4). For PFHxS, the ACME was 0.0013 (95% CI: 0.0004, 0.0022; P = 0.002), the ADE was 0.0169 (95% CI: 0.0059, 0.0282; P < 0.001), and the mediation proportion was 6.96% (Fig. 4A). For PFNA, the ACME was 0.0013 (95% CI: 0.0004, 0.0022; P < 0.001), the ADE was 0.0156 (95% CI: 0.0016, 0.0309; P = 0.036), and the mediation proportion was 6.96% (Fig. 4B). For PFOA, the ACME was 0.0015 (95% CI: 0.0005, 0.0025; P = 0.002), the ADE was 0.0199 (95% CI: 0.0081, 0.0308; P = 0.002), and the mediation proportion was 6.86% (Fig. 4C). For PFOS, the ACME was 0.0009 (95% CI: 0.0003, 0.0016; P = 0.002), the ADE was 0.0128 (95% CI: 0.0048, 0.0182; P = 0.004), and the mediation proportion was 6.55% (Fig. 4D). These findings indicate that insulin resistance, as reflected by elevated TyG, partially mediates the pathway from PFAS exposure to COPD risk.

Fig. 4.

Causal mediation analysis of the TyG index in the associations between PFAS exposure and COPD risk. A PFHxS, B PFNA, C PFOA, D PFOS. ACME, average causal mediation effect; ADE, average direct effect; CI, confidence interval. *P <0.05, **P < 0.01, ***P < 0.001

Multi-Database Integration Decodes the Molecular Mechanisms Underlying COPD Driven by Common PFAS Targets

Using PharmMapper and integrating TargetBank, DrugBank, BindingDB, and PTDLD, we identified 196 putative targets for PFHxS, 206 for PFNA, 234 for PFOA, and 202 for PFOS (Table S2). Querying GeneCards and OMIM yielded 2,724 genes linked to COPD (Table S3). Venn analysis revealed 58 COPD-related targets for PFHxS, 64 for PFNA, 69 for PFOA, and 63 for PFOS (Fig. 5A, Table S4). Intersection of the four PFAS-specific COPD target sets identified 57 shared genes (Fig. 5B, Table S5). GO enrichment indicated that these common targets are involved in biological processes such as intracellular receptor signaling and transcription initiation from RNA polymerase II promoters; localize to membrane microdomains including caveolae and membrane raft; and exhibit molecular functions related to nuclear receptor activity and ligand-activated transcription factor activity (Fig. 5C). KEGG pathway analysis further revealed over-representation in fluid shear stress and atherosclerosis, PPAR signaling pathway, and Th17 cell differentiation (Fig. 5D).

Fig. 5.

Putative targets and signaling pathway of PFAS in COPD. (A) Intersection of PFHxS, PFNA, PFOA, and PFOS targets with COPD-related genes. (B) Venn diagram showing the overlap of targets among the four PFAS and COPD. (C) GO enrichment analysis of the common targets. (D) KEGG enrichment analysis of the common targets

Core targets of PFHxS, PFNA, PFOA, and PFOS and their molecular docking in COPD

Intersection analyses yielded 58, 69, 64, and 63 COPD-related targets for PFHxS, PFNA, PFOA and PFOS, respectively. A protein–protein interaction network constructed in Cytoscape revealed that MMP9, PPARG and other hubs exhibited the closest connectivity with COPD (Fig. 6A-D). Notably, CASP3 was unique to the PFHxS network, whereas AKT1 was absent from PFHxS but present in the other three. The five common hubs were PPARG, MMP9, EGFR, ESR1, and SRC; ESR1 was excluded from validation due to undetectable mRNA, leaving four targets for molecular docking and RT-qPCR. Molecular docking simulations indicated that PFOA exhibits strong binding affinities to SRC, EGFR, and MMP9, with predicted binding free energies (ΔG) < -7 kcal/mol, whereas its interaction with PPARG is of moderate affinity (ΔG = -5.3 kcal/mol, Fig. 7A). Comparable docking analyses for PFHxS, PFNA, and PFOS revealed variable binding profiles across the four targets (Table S6).

Fig. 6.

Protein–protein interaction networks of key PFAS targets in relation to COPD. (A) Core target interactions for PFHxS and COPD. (B) Core target interactions for PFNA and COPD. (C) Core target interactions for PFOA and COPD. (D) Core target interactions for PFOS and COPD

Fig. 7.

Molecular docking of PFOA with key COPD targets and validation in BEAS-2B cells. (A) Binding mode of PFOA–SRC, PFOA–PPARG, PFOA–EGFR, and PFOA–MMP9. (B) BEAS-2B cells were exposed to different concentrations of PFOA (0, 50, 100, 200, 400, 600, 700, 800 μM) for 24 h, and cell viability was evaluated using the CCK-8 assay. (C) BEAS-2B cells were treated with PFOA (500 μM) or DMSO (vehicle control) for 24 h; relative mRNA levels of SRC, PPARG, EGFR, and MMP9 were detected by quantitative real-time polymerase chain reaction (RT-qPCR). Data are presented as mean ± SD (n = 3). *P < 0.05

Cytotoxic effects of PFOA on Beas-2B cells and associated gene expression

Guided by network-toxicology predictions and docking simulations, we next exposed BEAS-2B bronchial epithelial cells to PFOA to experimentally validate the in-silico findings. The CCK-8 assay (Fig. 7B) revealed a classic hormetic response: cellular viability remained ≥ 80% at 50–500 µM, whereas 700 µM reduced viability to 50.2%—approximating the IC₅₀—after 24 h. Therefore, we selected 500 µM PFOA for further gene expression analysis. The q-PCR data revealed that treatment with PFOA significantly upregulated the expression of SRC and PPARG genes (P < 0.05), while it significantly downregulated the expression of EGFR and MMP9 genes compared to the DMSO control group (P < 0.05, Fig. 7C). These findings suggest that PFOA may modulate the function of BEAS-2B cells by regulating the expression of these critical genes, potentially linking to the pathogenesis of COPD.

Discussion

COPD represents a worldwide public-health challenge whose etiology extends far beyond traditional tobacco smoke. Although smoking remains the dominant risk factor, approximately 25% of all COPD cases and up to 41% in China occur in never-smokers, underscoring the critical contribution of non-tobacco environmental exposures to disease initiation and progression [23, 24]. Among emerging environmental pollutants, PFAS represent a potentially serious long-term health threat because of their extreme environmental persistence and high bioaccumulation capacity [25]. Although preliminary studies have suggested a potential link between PFAS and respiratory health, critical knowledge gaps remain regarding causality, the effects of mixed exposure, and the specific pathogenic mechanisms underlying the association with COPD. Here, we innovatively implemented a population-computation-experiment triad to systematically elucidate the relationship between PFAS exposure and COPD risk and to clarify its biological basis.

Leveraging nationally representative data from NHANES, our study furnishes epidemiological evidence for a modest positive association between PFAS and COPD. After comprehensive adjustment for traditional risk factors including smoking status, individual PFAS congeners (PFHxS, PFNA, PFOA, and PFOS) as well as their mixture were positively and significantly associated with COPD prevalence. RCS modeling further revealed dose–response relationships, supporting biological plausibility. Importantly, we provide the first systematic application of WQS regression and QGComp in respiratory epidemiology to quantify the joint effect of PFAS mixtures. These findings are consistent with, and substantially extend, the observations reported by Zhang and Shao et al. [24, 26]. The associations observed in never-smokers suggest that PFAS may contribute to COPD risk independent of active smoking. However, this does not preclude residual confounding from environmental tobacco smoke, indoor air pollution, or occupational exposures not captured in NHANES. Thus, these findings reflect independence from active smoking rather than complete independence from all tobacco-related or environmental pathways. Moreover, the pronounced effects evident in low-socioeconomic-status populations underscore the stark reality of environmental health inequities and identify a precise target for public-health intervention.

This study quantifies the mediating role of insulin resistance in the PFAS-COPD association. Using the TyG index as a surrogate marker for metabolic dysfunction, we found that TyG significantly mediated 6.55%–6.96% of the effects of PFHxS, PFNA, PFOA, and PFOS on COPD risk. Previous evidence indicates that PFAS exposure is associated with insulin resistance and metabolic dysfunction [27]. Meanwhile, elevated TyG has been linked to systemic inflammation, lung function decline, and airway damage [28]. Our findings bridge population-level PFAS exposure with internal metabolic markers, suggesting that PFAS may exacerbate airway injury partially through insulin resistance. Notably, cross-sectional mediation analysis cannot establish definitive temporal causality. Future prospective studies are needed to validate this cascade. These results provide a rationale for our subsequent exploration of receptor-mediated mechanisms in cellular models.

To elucidate the molecular mechanisms through which PFOA—a major contributor to the PFAS mixture effect identified in our population analysis—may drive COPD, we constructed an unbroken evidence chain that progressed from in-silico prediction to experimental validation. Network toxicology screening revealed significant enrichment of the PPAR signaling pathway, fluid shear stress and atherosclerosis, among others, implying that PFAS may contribute to COPD by modulating lipid metabolism, inflammatory responses and endothelial function. Molecular-docking analyses further demonstrated that PFOA stably binds to the core target proteins EGFR and SRC with binding energies below − 7.0 kcal/mol, indicating its potential to directly perturb these signaling nodes. These docking and cellular findings are specific to PFOA and should not be extrapolated to other PFAS congeners (PFHxS, PFNA, PFOS) or their mixture effects. PFOA exposure significantly regulated the mRNA expression of EGFR and SRC in human bronchial epithelial cells. EGFR is a central regulator of airway mucus secretion and remodeling, whereas SRC serves as an upstream hub for multiple pro-inflammatory and pro-fibrotic pathways. Their coordinate activation provides clear molecular evidence that PFAS can trigger airway inflammation, epithelial–mesenchymal transition, and airway wall remodeling [29, 30].

EGFR activation not only promotes airway smooth-muscle proliferation and mucus hypersecretion via the RAS/RAF/MEK/ERK cascade, but also enhances cell survival and anti-apoptotic capacity through the PI3K/AKT/mTOR axis, thereby amplifying tissue remodeling under chronic inflammatory conditions [31, 32]. Once activated, SRC phosphorylates multiple downstream substrates including focal adhesion kinase (FAK) and paxillin thereby enhancing cell migration and invasion. Concurrently, it modulates transcription factors such as NF-κB and STAT3 to amplify the release of pro-inflammatory cytokines (e.g., IL-6, TNF-α), establishing a positive-feedback loop that further compromises epithelial barrier integrity and promotes fibroblast activation [33–36]. Notably, EGFR and SRC engage in reciprocal activation: SRC can amplify EGFR signaling by phosphorylating specific tyrosine residues on the receptor, whereas EGFR facilitates SRC kinase activity by recruiting it to membrane microdomains [37, 38]. This positive crosstalk implies that once PFAS initiate the network, the signal becomes self-amplifying, perpetuating airway remodeling [39]. Moreover, PFAS-mediated expression regulation of EGFR and SRC may imprint long-term transcriptional memory via epigenetic modifications—such as DNA methylation and histone post-translational changes—thereby providing a mechanistic explanation for the persistent decrements in lung function observed even after low-dose environmental exposures [40, 41].

MMP9 and PPARG transcript levels are also changed in our in-vitro model, this implies their functional relevance in PFAS-induced lung injury. MMP9 activity is primarily governed by post-translational modifications and micro-environmental cues. PFAS may indirectly potentiate its proteolytic capacity by activating upstream kinases or altering the balance of endogenous inhibitors such as TIMP-1, thereby facilitating airway structural damage [42]. As a key node in the crosstalk between metabolism and inflammation, PPARG activity is dually regulated by ligand binding and epigenetic modifications. Through its ligand-receptor interaction (e.g., ligand-mimetic binding) with PPARG, PFAS interferes with its transcriptional regulatory function at the transcriptional level, ultimately leading to dysregulation of lipid metabolism and accumulation of proinflammatory mediators [43, 44].This apparent transcriptional silence may reflect alternative regulatory strategies. Comprehensive elucidation of their true roles in PFAS-induced lung injury will require future proteomic and post-translational modification analyses.

The principal strength of this work lies in its pioneering integrative design. By deploying the “population-computation-experiment” triad, we moved seamlessly from statistical association to mechanistic dissection, constructing a self-validating, stepwise scientific narrative. Use of the nationally representative NHANES database ensures broad external validity, while the application of mixture-exposure modeling coupled with multi-level mechanistic validation confers novelty at the intersection of environmental epidemiology and toxicology.

Nevertheless, several limitations should be acknowledged. First, the cross-sectional nature of NHANES precludes definitive temporal inference; although our in-vitro findings provide biological plausibility, longitudinal cohorts are required to establish causality. Second, a single serum PFAS measurement may not capture complete exposure history, and complete case analysis may introduce selection bias if missingness is not completely at random. Residual confounding from unmeasured environmental or occupational exposures cannot be excluded. Third, while the cellular model affords precise molecular insights, it cannot fully recapitulate the complex pathophysiology observed in vivo. Fourth, the in vitro PFOA concentration (500 µM) exceeds typical serum levels and should be interpreted as proof-of-concept rather than direct environmental extrapolation. Finally, the cellular and docking findings are specific to PFOA and do not capture real-world mixture toxicity or toxicological interactions among congeners.

Conclusions

Using a “population-computation-experiment” triad, we provide evidence suggesting that PFAS may serve as independent environmental risk factors for COPD, with PFOS and PFOA being the main drivers of the mixture effect. Mechanistically, PFOA coordinately regulates core genes to orchestrate airway inflammation, mucus hypersecretion, and epithelial–mesenchymal transition, thereby driving airway remodeling in COPD. These findings offer a new etiological perspective on non-smoking COPD and furnish critical scientific support for tightening PFAS regulations and developing targeted interventions.

Authors’ contributions

Yansong Hu: Conceptualization, Methodology, Writing original draft, Writing-review & editing. Huanyu Cui: Conceptualization, Methodology, Writing-original draft. Yakun Wang: Formal analysis, Methodology, Writing-review & editing. Ye Tang: Formal analysis, Methodology, Validation. Yuhua Chen: Formal analysis, Methodology. Yanyan Qin: Data curation, Formal analysis. Min Tang: Data curation, Formal analysis. Siying Pang: Data curation, Formal analysis. Yaqin Pang: Supervision, Project administration, Writing-review & editing. Zhengbao Zhang: Conceptualization, Methodology, Supervision, Writing-review & editing. Xiaonian Zhu: Conceptualization, Writing-original draft, Writing-review & editing. Di Li: Validation, Writing-original draft, Writing-review & editing.

Funding

This work was supported by the Key Science and Technology Research and Development Program Project of Guangxi (AB22035017), and Guangxi Bagui Young Leading Talent Program.

Data availability

The datasets analyzed during the current study are publicly available in the NHANES repository, https://www.cdc.gov/nchs/nhanes/index.html.The analysis scripts are available from the corresponding author upon reasonable request.

Declarations

Competing interests

The authors declare no competing interests.