PFOS Drives Pulmonary Injury via Glycolytic and Oxidative Metabolic Disruption: Multilevel Evidence Integrating Population, Computational, and Experimental Approaches

Abstract

Perfluorooctanesulfonate (PFOS), a pervasive environmental pollutant, threatens respiratory health, though its pulmonary toxicity mechanisms remain unclear. The study integrated NHANES epidemiological data (n = 1595), computational toxicology (ADMETlab 3.0 and ProTox-3.0), chronic mouse exposure models (0, 5, and 150 μg kg^–1 day^–1, 12 weeks), multidatabase bioinformatics (CTD and GeneCards), PPI network analysis, and molecular docking (CB-DOCK2). NHANES analysis revealed a significant inverse correlation between serum PFOS and the FEV1/FVC ratio, indicating PFOS-associated airflow obstruction. Crucially, restricted cubic spline regression identified a unified toxicity threshold at 6.54 ng/mL, beyond which FEV1/FVC declined disproportionately. Computational prediction indicated >80% probability of PFOS respiratory toxicity. Animal experiments demonstrated dose-dependent injury: low dose primarily induced fibrosis, while high dose triggered apoptosis (TUNEL⁺/cleaved-caspase3). Bioinformatics integration identified metabolic dysregulation as the core mechanism, with KEGG enrichment highlighting glycolysis and xenobiotic metabolism. Core target validation showed PFOS bound with high affinity to AOX1 (−9.2 kcal/mol), GPI (−9.3 kcal/mol), PKM (−8.8 kcal/mol), and CYP1A1 (−9.0 kcal/mol) and suppressed Aox1 (P < 0.01) and Gpi/Pkm mRNA (P < 0.05). Additionally, PFOS exposure was accompanied by a decrease in GSH levels and an increase in MDA concentration, thereby inducing oxidative stress. PFOS drives pulmonary injury through dual metabolic hijacking: (1) disruption of AOX1-mediated oxidative defense leading to oxidative stress and (2) suppression of the GLUT1-GPI-PKM glycolytic axis. This study establishes AOX1 enzyme activity and the serum lactate/ATP ratio as promising noninvasive biomarkers for the early detection of PFOS-induced lung damage, offering significant insights for environmental health risk assessment.

1. Introduction

Perfluorooctanesulfonate (PFOS), a persistent organic pollutant widely used in commercial products, demonstrates bioaccumulation potential with a human half-life of 5.4 years, posing long-term health risks. , As the predominant perfluoroalkyl substance (PFAS) in human serum, , PFOS exposure has been correlated with multiorgan toxicity, including hepatotoxicity and immunotoxicity, , and is classified as a high-priority carcinogen evaluation target by IARC.

Current research has made significant strides in establishing pulmonary inflammation and fibrosis as key outcomes of PFOS exposure. However, the field remains disproportionately focused on elucidating these downstream pathological end points and their associated canonical signaling pathways, often overlooking the critical question of the initiating molecular events. Limited human studies indicate PFOS accumulates in lung epithelia via phospholipid binding and associates with impaired lung function. − Yet, a fundamental gap exists in understanding how PFOS triggers these events at the cellular level. PFOS structurally mimics fatty acids, enabling direct interference with metabolic regulators like PPARs and potential disruption of core metabolic enzymes-suggesting that metabolic reprogramming may be an upstream, under-explored driver of the documented injury. Furthermore, prevailing models relying on nonphysiological high-dose/short-term exposures may fail to capture these subtle metabolic shifts, which are likely pivotal in low-dose chronic effects relevant to human environmental exposure.

Here, the study hypothesizes that chronic PFOS exposure compromises lung homeostasis through targeted disruption of core metabolic pathways. To test this, the study employed a tiered strategy: Population-level analysis to establish exposure-response relationships; Computational toxicology to predict respiratory toxicity risk; In vivo modeling of chronic low-/high-dose exposure mirroring human-relevant concentrations; Transcriptome to identify pathological hubs; Mechanistic validation of core targets via molecular docking and functional assays. This study is the first to demonstrate dose-polarized pulmonary injury (fibrosis vs apoptosis) in PFOS-exposed models; Implicate glycolytic flux suppression (“GLUT1-GPI-PKM” axis) as a driver of ATP crisis; Propose AOX1 inactivation as a sentinel event in PFOS-induced oxidative collapse. This study findings redefine PFOS pulmonary toxicity through a metabolic lens and provide biomarkers for environmental health risk assessment.

2. Materials and Methods

2.1. Data Analysis of the NHANES Population

In the 2011–2012 survey cycle, publicly available data were obtained from 7,796 participants aged 6 to 79 years. Individuals with missing data on blood PFOS metabolite measurements (n = 5,044) and spirometry tests (n = 3,104) were excluded. To address missing covariate data, multiple imputation was used under the assumption that the data were missing at random. The final analytic sample included 1,595 participants with complete data on outcomes and covariates of interest. This study was limited to data from the NHANES 2011–2012 cycle because pulmonary function parameters and biomarkers of PFOS exposure were only collected during this period.

2.2. Demographic and Lifestyle Covariates

Health-related data were collected during each examination cycle, including anthropometric measurements (e.g., height and weight) and self-reported information on demographics, medical history, and lifestyle behaviors. Race/ethnicity was categorized as non-Hispanic White, non-Hispanic Black, Hispanic, or other. Body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared (kg/m²).

Physical activity was assessed using age-specific NHANES questionnaires. For children aged 6–11 years, screen time was used as a proxy indicator, based on self-reported average daily hours spent watching television, videos, or using a computer. Screen time was dichotomized as <2 h/day or ≥2 h/day. For participants aged ≥12 years, physical activity status was determined based on self-reported engagement in moderate and/or vigorous recreational activities and classified as either “physically active” or “inactive”. In the current study, children with ≤2 h/day of screen time and adolescents/adults reporting regular physical activity were jointly classified as physically active.

Smoking status was categorized following National Health Interview Survey definitions: participants who had smoked fewer than 100 cigarettes in their lifetime were considered never smokers; those who had smoked more than 100 cigarettes but were not currently smoking were defined as former smokers; and current smokers were those who had smoked more than 100 cigarettes and reported active smoking at the time of the survey.

Asthma status was determined based on responses to the questions: “Has a doctor or other health professional ever told you that you have asthma?” and “Do you still have asthma?” Participants answering “Yes” to both were classified as having persistent asthma. Those who answered “No” to both were considered nonasthmatic, while those responding “Yes” to the first question but “No” or “Don’t know” to the second were classified as having a history of asthma.

2.3. Animal Experiment and Design

To elucidate the pulmonary toxicity effects of PFOS and validate alterations in key target gene expression, an in vivo animal model was established. A total of 30 male C57BL/6 mice (8 weeks old) were obtained from Vital River Laboratory Animal Technology Co., Ltd. (Jiangsu, China; license number: SCXK [Jiangsu] 2003–0009) and housed in the Animal Facility of Southeast University. All experimental procedures were approved by the Institutional Animal Care and Use Committee of Southeast University.

While inhalation adds to the local exposure in the lungs as well as to the total PFAS level in blood, the main contribution to the blood levels of the different PFASs in the general population is from food and drinking water and then accumulate in the lung tissue, , we administered PFOS to mice by water. Thus, mice were randomly allocated into three groups (n = 10 per group): a control group, a low-dose PFOS group (5 μg kg^–1 day^–1), and a high-dose PFOS group (150 μg kg^–1 day^–1). Mice in the control group received no treatment, while the PFOS groups were exposed to PFOS via drinking water for a duration of 12 weeks. At the end of the treatment period, all animals were anesthetized and euthanized in accordance with ethical guidelines. Lung tissues were collected for subsequent molecular and histological analyses. PFOS (CAS: 2795–39–3) was purchased from J&K Scientific (Beijing, China).

2.4. Immunohistochemistry

Fresh lung tissue samples were collected, fixed in 4% paraformaldehyde, and then embedded in paraffin. For immunohistochemistry (IHC), the paraffin sections were deparaffinized and rehydrated using a graded ethanol series. Subsequently, the tissues were permeabilized, endogenous peroxidase activity was blocked, and antigen retrieval was performed to expose the antigenic epitopes. After blocking with serum, the sections were incubated overnight at 4 °C with the primary antibody α-SMA (Abcam, 1:100 dilution). After removing the primary antibody, the sections were incubated with the secondary antibody for 45 min. The HRP-conjugated polymer detection system (Golden Bridge International, Washington, USA) was used for the colorimetric reaction, and the sections were counterstained with hematoxylin. Finally, the sections were dehydrated, cleared with xylene, and mounted with neutral resin. The slides were air-dried in a fume hood and subsequently examined under a microscope. IHC DAB staining was quantified using ImageJ. Regions of interest (ROIs) were manually delineated, and mean optical density (MOD) was measured from at least three randomly selected fields per sample. All imaging and analysis parameters were kept constant across experimental groups.

2.5. Terminal Deoxynucleotidyl Transferase dUTP Nick End Labeling (TUNEL) Staining

TUNEL staining was used to evaluate cell apoptosis. The cells were incubated with the cell death detection kit (Servicebio, China) according to the manufacturer’s instructions. Nuclei were costained for 5 min with 0.1 g/mL DAPI (Servicebio, China), and images were captured under a microscope.

2.6. Total RNA Extraction and Quantitative RT-PCR

Lung tissue (30 mg) was weighed and homogenized in 1200 μL TRI reagent, followed by centrifugation at 12,000 rcf for 15 min at 4 °C. The supernatant (1000 μL) was transferred to a new tube, and 150 μL of 1-bromo-3-chloropropane (BCP) was added for extraction. RNA was precipitated using isopropanol and washed with prechilled 75% ethanol. After discarding the supernatant, the RNA pellet was air-dried and dissolved in an appropriate volume of nuclease-free water to achieve a concentration of 1000 ng/μL. The total RNA was then treated with RNase-free DNase and reverse transcribed using AMV Reverse Transcriptase (Promega). Quantitative RT-PCR was performed using gene-specific primers and the LightCycler 480 SYBR Green I kit (Roche Diagnostics GmbH). Amplification was carried out on the LightCycler 480 Instrument (Roche Diagnostics GmbH) with an initial step at 95 °C for 5 min, followed by a three-step PCR protocol. β-actin mRNA was used as an internal control for mouse-related mRNA. Primer sequences used in this study are listed in Table S1.

2.7. Total Protein Extraction and Western Blotting

Lung tissue (50 mg) was homogenized in a glass homogenizer with 500 μL of lysis buffer (containing protease inhibitors) to prepare the mouse lung tissue lysate. Protein concentration was measured using the BCA assay according to the manufacturer’s instructions. The protein samples were separated by 10% SDS-PAGE and transferred to a polyvinylidene fluoride (PVDF) membrane (Millipore, Billerica, MA). The membrane was incubated overnight at 4 °C with the following primary antibody: caspase3 (Abways, 1:2000 dilution) and cleaved caspase-3 (Abways, 1:1000 dilution). β-Actin was used as the internal control for total protein. The membrane was then incubated with the corresponding secondary antibody and detected using the enhanced chemiluminescence (ECL) reagent kit.

2.8. Detection of GSH and MDA Levels in Lung Tissue

The animal tissue was accurately weighed, and the weight-to-volume ratio (mg: μL) was maintained at 1:9. A homogenization medium (recommended 0.86% or 0.9% saline solution) was added in a 9-fold volume. The homogenate was then centrifuged at 2500–3000 rpm for 10 min, and the supernatant was collected for the subsequent measurement of GSH (Servicebio, China) and MDA (Servicebio, China) levels, following the manufacturer’s instructions.

2.9. Functional Genomics and Pathway Enrichment Analysis for Target Proteins

We utilized the Comparative Toxicogenomics Database (CTD, https://ctdbase.org) to search for PFOS-related targets using “PFOS” as the keyword. After standardizing target names through STRING and UniProt, and removing duplicates, a total of 8,866 potential toxicological targets for PFOS were identified (accessed March 2025). Lung injury targets were primarily retrieved from the GeneCards database (https://www.genecards.org/), using “lung injury” and “lung disease” as keywords (accessed March 2025). Targets with a relevance score ≥1 were selected, resulting in the identification of 10,123 lung injury-related targets and 18,668 lung disease-related targets. Venn diagram analysis was performed to obtain the intersection of PFOS, lung injury, and lung disease-related targets, resulting in 4,732 common targets (https://www.bioinformatics.com.cn/). Enrichment analysis of the intersected genes was carried out using the analytical platform Sangerbox 3.0 (http://vip.sangerbox.com/). Gene Ontology (GO) analysis encompassed biological processes (BP), cellular components (CC), and molecular functions (MF). Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis highlighted significant pathways associated with the intersected genes, enhancing our understanding of their potential mechanisms in PFOS.

2.10. Protein–Protein Interaction (PPI) Network Construction and Core Target Screening

A protein–protein interaction (PPI) network was constructed using the STRING database (https://string-db.org/). We further applied the Cytoscape (version 3.10.0) to visualize the PPI network. To identified the core genes, the study used five algorithms MCC (Maximum Clique Centrality), MNC (Maximum Neighborhood Component), DEGREE, Closeness (Closeness centrality) and Betweenness (Betweenness centrality) in CytoHubba. Kyoto Encyclopedia of Genes and Genome (KEGG) analyses of core genes were performed using the Database for Annotation Visualization and Integrated Discovery (DAVID, https://davidbioinformatics.nih.gov/).

2.11. Molecular Docking

Protein–ligand docking is an effective method in toxicology studies. CB-Dock2 (https://cadd.labshare.cn/cb-dock2/php/index.php) is an advanced protein–ligand blind docking technology that uses AutoDock Vina to automatically identify potential binding sites, calculate their center coordinates and size, and customize the size of the docking box according to the ligand structure, ultimately achieving molecular docking. Extensive benchmark test results show that focusing on the protein cavity regions for docking can significantly improve the hit rate and accuracy of blind docking. Therefore, CB-Dock2 combines the advantages of the curvature-based cavity detection algorithm (CurPocket) and AutoDock Vina, enabling accurate prediction of the binding sites of target proteins, thereby accelerating the docking process and improving the accuracy of the predictions. The active ingredients were obtained in SDF format from the PubChem database (https://pubchem.ncbi.nlm.nih.gov/, COMPOUND_CID_74483), while the core proteins were sourced in PDB format from Protein Data Bank (PDB, http://www.rcsb.org/): AOX1 (PDB 8EMT, Homo sapiens, Chains A and B), CYP1A1 (PDB 6UDM, Homo sapiens, Chains A–D), GPI (PDB 7w72, Homo sapiens, Chain A), PKM (PDB 5X0I, Homo sapiens, Chains A–D). CB-Dock2 was used for protein–ligand blind docking, and the cavity size and vina score were evaluated. A binding energy of ≤−6.0 kcal/mol was considered as significant binding.

2.12. Data Analysis

The Student’s t test was used to compare statistical differences between two groups. For multiple group comparisons, analysis of variance (ANOVA) followed by the Student–Newman–Keuls test was applied. All quantitative data are expressed as the mean ± standard error of the mean (SEM). Statistical significance was accepted at P-value < 0.05.

3. Results

3.1. Correlation between PFOS and Pulmonary Function Based on NHANES

In this study, we analyzed data from the NHANES 2011–2012 cycle to explore the relationship between PFOS exposure and various pulmonary function parameters. A total of 1595 participants with complete data on PFOS exposure and pulmonary function tests were included in the analysis (Figure A).

1.

Relationship between serum PFOS and lung function parameters: analyses of NHANES 2011–2012. (A) Flowchart of participant inclusion criteria. (B–F) Association between serum PFOS and lung function parameters (FEV1, FVC, FEV1/FVC, PEF, and FEF 25–75%) evaluated by restricted cubic splines. The horizontal axis represents the PFOS (ng/mL). The adjusted effect estimates (solid lines) and 95% confidence intervals (shadow area) were computed using the 25th percentile of the ln-transformed PFOS levels as the reference. Test results for nonlinear association were represented as P nonlinear. The test results for overall association were analyzed if none of the tests for nonlinear association were significant (P nonlinear ≥ 0.05), and their linear association was determined if test for overall association and linearity was significant (P < 0.05).

The mean values and standard errors (SE) for key pulmonary function parameters, including FEV1 (forced expiratory volume in 1 s), FVC (forced vital capacity), FEV1/FVC ratio, PEF (peak expiratory flow), and FEF25–75% (forced expiratory flow at 25–75% of FVC), were calculated across various demographic groups (Table ). The mean values for FEV1, FVC, FEV1/FVC, PEF, and FEF25–75% exhibited significant differences (P < 0.001) between age groups, suggesting that age is a strong determinant of pulmonary function. Males showed significantly higher values for FEV1, FVC, and PEF compared to females (P < 0.001 for all), whereas the FEV1/FVC ratio was lower in males. Significant differences in pulmonary function were observed across different racial/ethnic groups, with Non-Hispanic White participants showing the highest values for FEV1 and FVC, and Non-Hispanic Black participants having the lowest values for PEF (P values ranged from 0.005 to 0.924 across the parameters). Pulmonary function was significantly associated with BMI for FEV1 and FEV1/FVC (P values of 0.023 and 0.025, respectively), but not for FVC or PEF. Current smokers had significantly lower values for FVC, PEF and FEF 25–75% compared to never-smokers (P < 0.001). The analysis also revealed significant associations between pulmonary function and various covariates such as physical activity and asthma status. Physical activity was positively associated with FEV1, FVC, and PEF (P < 0.05 for all parameters), while asthma history was negatively associated with pulmonary function (P = 0.003 for FVC and P = 0.006 for PEF).

1. Study Population’s Clinical Characteristics.

Characteristic	N	%	FEV1 (mL)	FVC (mL)	FEV1/FVC	PEF (mL/s)	FEF25–75% (mL/s)
Total	1595	100
Age (years)
	1595	100	2692. 76 (812. 88)	3546.2 (1140.32)	0.84 (0.06)	3936.25 (1113.47)	7362.14 (2057.27)
p-Value			<.001	<.001	<.001	<.001	<.001
Sex
Male	827	49.13	3734.63 (45.75)	4791.13 (60.98)	0.78 (0.01)	9518.96 (136.72)	3439.81 (63.53)
Female	768	50.87	2729.21 (25.91)	3417.69 (35.59)	0.80 (0.00)	7029.86 (96.95)	2717.09 (43.82)
p-Value			<.001	<.001	<.001	<.001	<.001
BMI
	1595	100	2484.24 (796.23)	3368.1 (940.58)	0.89 (0.06)	3434.06 (1216)	6806.35 (2102.53)
p-Value			0.023	0.156	0.025	0.203	0.175
Race
Mexican American	184	8.21	3296.04 (51.39)	3998.62 (62.92)	0.83 (0.01)	8252.09 (155.72)	3527.40 (92.68)
Other Hispanic	162	(6.60)	3154.62 (45.57)	3842.07 (54.72)	0.82 (0.01)	8094.78 (152.25)	3356.70 (77.28)
Non-Hispanic White	529	(66.05)	3278.80 (39.03)	4239.45 (46.32)	0.77 (0.01)	8316.65 (129.11)	2965.18 (69.08)
Non-Hispanic Black	428	(11.55)	2955.43 (42.09)	3658.46 (52.62)	0.81 (0.00)	7996.64 (84.49)	3044.73 (56.02)
Other	292	(7.58)	3127.80 (62.38)	3793.51 (78.91)	0.83 (0.01)	8225.76 (166.35)	3305.72 (86.83)
p-Value			0.005	0.014	0.924	0.554	0.028
Education
Less than high school	515	(21.75)	3117.44 (46.14)	3822.41 (45.94)	0.82 (0.01)	7555.15 (134.27)	3196.24 (88.00)
High school or equivalent	277	(17.80)	3197.35 (79.82)	4085.55 (84.86)	0.78 (0.01)	8240.34 (225.21)	3037.82 (101.79)
College or above	803	(60.45)	3268.88 (38.11)	4191.74 (45.92)	0.78 (0.01)	8507.53 (100.84)	3037.69 (80.75)
p-Value			0.037	<.001	0.002	<.001	0.325
Poverty index ratio (PIR)
	1595	100	2923.3 (937.29)	3426.114 (145.37)	0.84 (0.08)	3029.02 (1336.07)	7108.43 (2518.2)`
p-Value			0.663	0.046	0.023	<.001	0.086
Smoking status
Never	970	(58.95)	3277.84 (31.28)	4064.70 (39.01)	0.81 (0.00)	8335.84 (93.25)	3285.62 (58.49)
Former	302	(21.45)	3127.37 (65.10)	4109.77 (88.06)	0.76 (0.01)	8356.14 (206.67)	2720.88 (96.23)
Now	323	(19.60)	3127.37 (65.10)	4157.27 (88.70)	0.76 (0.01)	7890.10 (145.14)	2814.78 (55.35)
p-Value			0.063	<.001	<.001	0.006	<.001
Drink
Mild drinking	705	(45.40)	3175.41 (46.91)	4072.70 (64.28)	0.78 (0.01)	8342.16 (110.70)	2957.37 (74.14)
Moderate drinking	549	(34.84)	3146.68 (54.19)	3972.39 (72.77)	0.79 (0.01)	7998.41 (142.86)	3047.44 (69.78)
Heavy drinking	341	(19.76)	3467.95 (58.20)	4349.81 (55.98)	0.80 (0.01)	8496.28 (127.21)	3379.60 (132.68)
p-Value			0.021	0.110	0.041	0.864	0.016
Asthma status
Former	812	(51.20)	3278.69 (34.90)	4156.84 (41.65)	0.79 (0.01)	8417.78 (122.08)	3132.59 (77.47)
Now	783	(48.80)	3165.00 (37.47)	4025.01 (47.28)	0.79 (0.00)	8079.77 (92.76)	3008.81 (46.08)
p-Value			0.052	0.003	0.730	0.006	0.213
Physical activity
Low physical activity	493	(27.88)	2906.18 (47.36)	3717.5 (60.75)	0.78 (0.00)	8417.78 (122.08)	2770.87 (68.79)
High physical activity	1102	(72.12)	3345.75 (34.75)	4237.46 (37.92)	0.79 (0.00)	8079.77 (92.76)	3188.66 (66.65)
p-Value			<.001	0.045	0.122	<.001	0.002

3.2. Association between PFOS Exposure and Pulmonary Function Parameters

To further characterize the dose–response relationship between PFOS exposure and pulmonary function, restricted cubic spline regression was performed. The models revealed significant inverse associations between serum PFOS levels and several key pulmonary parameters.

For FVC, a significant nonlinear association was detected (P for overall <0.05; P for nonlinearity <0.05), indicating a biphasic response: compensatory elevation (<6.54 ng/mL) followed by accelerated volume reduction. A similar pattern was observed for FEV1 (P = 0.002). PEF also showed significant dose-dependent suppression (P = 0.003), reflecting compromised large airway flow. Of particular interest, FEF 25–75%, a sensitive marker of small airway function, exhibited both a significant inverse association (P = 0.005) and marked nonlinearity (P = 0.006), with a clear turning point at 6.54 ng/mL. Notably, an inflection point was observed consistently at a PFOS concentration of approximately 6.54 ng/mL across all five lung function indicators, suggesting a threshold beyond which pulmonary impairment becomes more pronounced. These findings indicate that 6.54 ng/mL may represent a critical exposure threshold, beyond which PFOS exerts progressively deleterious effects on lung function. The observed nonlinearity and early response of small airways underscore the need to re-evaluate current safety limits for low-level PFOS exposure (Figure B).

3.3. Chronic PFOS Exposure Induces Dose-Dependent Pulmonary Injury via Divergent Fibrotic and Apoptotic Pathways

An integrated analysis using the ADMETlab 3.0 and ProTox-3.0 platforms yielded a comprehensive toxicological profile of PFOS. Both platforms demonstrated respiratory toxicity for PFOS, with predicted probabilities of 80% and 54%, respectively (Table , Table S1).

2. Prediction Results of ADMETlab 3.0.

Property	Probability
Genotoxicity	0.981272459
Respiratory	0.804656863
Nephrotoxicity-DI	0.709412873
H-HT	0.622040093
Ototoxicity	0.552112699
Hematotoxicity	0.470266819
Carcinogenicity	0.448093027
FDAMDD	0.361899704
Ames	0.252026379
Neurotoxicity-DI	0.210817233
HEK293	0.059145354
DILI	0.053897295
hERG	0.02442502

To elucidate mechanisms of PFOS-induced respiratory toxicity, we established a translational mouse model simulating human exposure through PFOS-administrated drinking water at environmentally relevant doses: Low-dose (5 μg kg^–1 day^–1, based on national human safety thresholds) and high-dose (150 μg kg^–1 day^–1). After 12 weeks of exposure, HE result of low-dose group showed mild granulocyte infiltration in alveolar walls diffuse interstitial thickening with compensatory alveolar dilatation; occasional alveolar macrophage exudation; bronchiolar eosinophilic aggregates and epithelial sloughing; high-dose group showed aggravated architectural distortion; prominent perivascular edema; significant lymphocytic infiltration (Figure A). Then, fibrosis levels were assessed through α-SMA immunostaining, which revealed positive staining in both dose groups. Notably, the low-dose group exhibited higher α-SMA⁺ cell density than the high-dose group (Figure B and C). TUNEL⁺ cells significantly elevated and apoptosis-associated proteins (caspase-3 and cleaved-caspase3) expression increased in the high-dose group (Figure D–F). These findings indicate that chronic PFOS exposure triggers pulmonary injury, with low dose promoting fibrotic responses and high-dose dominating apoptotic cascades.

2.

Lung damage was induced by long-term low-dose exposure to PFOS. (A) H&E staining of the lungs was conducted. Representative scans are shown (200×). (B–C) Representative images of α-SMA-positive cells in lung sections (200×) and the corresponding quantitative analysis. (D) Representative images of TUNEL analysis in different group. The nuclei of cells were stained by DAPI. Magnification: 200×. (E) Representative Western blot images and quantification of caspase-3 protein levels. (F) Representative images and quantification of cleaved -caspase3. Error bars represent standard deviations (n = 4). Blue arrows indicate neutrophil infiltration, light green arrows show occasional macrophage exudate in the alveolar space; green arrows indicate eosinophilic material and shed epithelial cells, gray arrows indicate perivascular edema, purple arrows indicate scattered lymphocyte infiltration, and black arrows indicate infiltration of cells positive for α-SMA. L-pfos: low-dose PFOS group (5 μg kg^–1 day^–1); H-pfos: high-dose PFOS group (150 μg kg^–1 day^–1) group. **P < 0.01 versus control; ****P < 0.0001 versus control.

3.4. Integrated Multidatabase Analysis Identifies Metabolic Pathways as Central Mechanisms of PFOS Pulmonary Toxicity

To elucidate the mechanisms of PFOS-induced lung injury, we performed a systematic bioinformatic analysis. First, we identified 8,866 PFOS toxicity targets from the Comparative Toxicogenomics Database (CTD) and retrieved 10,123 lung injury-related targets and 18,668 lung disease-associated targets from GeneCards. After integration and deduplication, we obtained 4,732 overlapping targets, which are potential mediators of PFOS-induced lung injury (Figure A); Subsequently, overlapping targets were analyzed using Sangerbox. Significantly enriched GO terms (FDR < 0.25) included: Biological Processes (BP): chemical response, stress response, transport, etc (Figure B); Cellular Components (CC): mitochondrial inner membrane, organelle inner membrane, respiratory chain complex, etc (Figure C); Molecular Functions (MF): catalytic activity, oxidoreductase activity, cofactor binding, etc (Figure D). KEGG Analysis showed among 211 significantly enriched pathways (FDR < 0.25), metabolic pathways ranked first with 666 genes and other top pathways included Pathways in cancer, MAPK signaling, etc (Figure E–F). This multidatabase integration robustly implicates metabolic dysregulation as the core mechanism of PFOS pulmonary toxicity.

3.

Target of PFOS-induced lung toxicity identified and enrichment analysis. (A) Venn diagram of PFOS exposure, lung injury, and lung disease. (B–D) Bubble charts of GO enrichment analysis of overlapping molecular targets (BP, CC, and MF). (E,F) Histogram and bubble chart of KEGG enrichment analysis of core target genes. BP: Biological Process; CC: Cellular Component; MF: Molecular Function.

3.5. Integration of PPI Network Topology and Tissue-Specific Expression Identifies Core Targets for PFOS-Induced Toxicity

To identify key genes through which PFOS induces lung injury via metabolic pathways, we performed protein–protein interaction (PPI) analysis on the 666 genes using STRING (Homo sapiens, confidence ≥ 0.9). The resulting network was imported into Cytoscape software, where targets with a degree less than 20 were filtered out, yielding a network comprising 228 nodes and 3504 edges (Figure A). Using five algorithms (MCC, MNC, Degree, Closeness, and Betweenness) in CytoHubba, we identified the top 10% of genes ranked by each algorithm (Figure B). From these, we selected 10 core targets, identified at the intersection of the top 10% targets from each algorithm: CYP2E1, CYP4A11, CYP1A2, CYP1A1, AOX1, CYP26A1, CYP3A4, CYP2A6, GPI, and PKM (Figure A). DAVID-based KEGG enrichment of these targets revealed three key pathological axes: Xenobiotic metabolism dysregulation (CYP1A2, CYP2A6, and CYP3A4) in drug metabolism (hsa 00982), xenobiotic metabolism (hsa 00980), and DNA adduct formation (hsa 05204); Energy metabolism hijacking (GPI and PKM), specifically enriched in glycolysis (hsa 00010); Carcinogenic transformation involving receptor activation (hsa05207) and ROS-mediated pathways (hsa 05208). Retinol metabolism and DNA adduct pathways exhibited the highest enrichment significance (Figure B). Notably, CYP4A11 was absent from the top 20 pathways, suggesting possible indirect regulation.

4.

Identification of the key targets. (A) PPI network constructed by 666 potential targets. (B) Concentric circles represent the calculation of key genes using the “MCC”, “MNC”, “DEGREE”, “Closeness”, and “Betweenness” algorithms. The top 10% key genes are highlighted within these circles. MCC: Maximum Clique Centrality; MNC: Maximum Neighborhood Component; DEGREE: Degree; Closeness: Closeness centrality; Betweenness: Betweenness centrality.

5.

Potential biological functions of core genes. (A) Venn diagram of top 10% genes of five algorithms. (B) Sankey diagram and bubble chart of KEGG enrichment analysis for 10 core targets.

Via The Human Protein Atlas (https://www.proteinatlas.org/) to obtain the target-specific localization and expression profiling in lung tissue, the results demonstrated that GPI (TPM = 89.1) and PKM (TPM = 72.5) served as universal energy metabolism hubs; AOX1 (TPM = 34.6) showed apical polarity in bronchial epithelium, functioning as an aldehyde detoxification barrier; CYP1A1 (TPM = 0.3, protein-negative baseline) acted as an inducible sentinel linked to five carcinogenic pathways. Other targets displayed negligible expression (TPM < 1) (Figure S1). This integrated analysis establishes a tripartite core target system including AOX1 (oxidative defense), GPI/PKM (glycolytic energy flux), and CYP1A1 (carcinogenic switch), providing mechanistic anchors for pulmonary toxicity intervention.

3.6. Chronic PFOS Exposure Impairs Detoxification and Glycolytic Flux via Core Target Suppression

To validate the above findings, molecular docking simulations between PFOS and the predicted core targets were conducted using CB-DOCK2. The results demonstrated strong binding affinity between PFOS and four core targets: AOX1 (−9.2 kcal/mol), CYP1A1 (−9.0 kcal/mol), GPI (−9.3 kcal/mol), and PKM (−8.8 kcal/mol) (Figure A,B).

6.

Expression of core targets in human lung tissues and the molecular docking of PFOS with highly expressed core genes. (A,B) The molecular docking and binding energy of PFOS with “AOX1”, “CYP1A1”, “GPI”, and “PKM”. (C,D) The relative expression of “Aox1” and “Cyp1a1” mRNA in lung tissues. (E,F) MDA and GSH content of lung. (G–I) The relative expression of “Gpi”, “Pkm”, and “Glut1” mRNA in lung tissues. Error bars represent standard deviations (n = 3); L-pfos: low-dose PFOS group (50 μg kg^–1 day^–1); H-pfos: high-dose PFOS group (150 μg kg^–1 day^–1) group. *P < 0.05 versus control; **P < 0.01 versus control.

Subsequently, mRNA expression levels of Aox1 and Cyp1a1 were quantified in lung tissues in the mice models. We found chronic PFOS exposure significantly downregulated Aox1 expression (Figure C). While Cyp1a1 expression showed an upward trend, it did not reach statistical significance (Figure D). This mild elevation potentially reflects slight activation of the aryl hydrocarbon receptor (AhR) pathway, which is typically induced by higher exposure doses. Alternatively, it may represent a compensatory response, in which the organism attempts to upregulate alternative enzymatic systems to counter the impaired detoxification capacity resulting from AOX1 suppression. Given that AOX1 downregulation is often associated with compromised detoxification and increased oxidative stress, we further assessed relevant oxidative stress markers. The analysis revealed a significant decrease in glutathione (GSH) levels and an increase in malondialdehyde (MDA) levels in lung tissues (Figure E,F). These results indicate not only a diminished capacity to metabolize and eliminate PFOS and its metabolites but also an impaired ability to counteract PFOS-induced oxidative stress. Consequently, these effects may lead to the accumulation of toxic substances in the lungs, increasing susceptibility to chronic injury.

Validation of the predicted core targets Gpi and Pkm confirmed significant downregulation of their mRNA levels (Figure G,H). Furthermore, mRNA expression of the glucose transporter Glut1 showed a downward trend in the PFOS-exposed group (Figure I). These results demonstrate that chronic low-dose PFOS exposure suppresses glucose supply mediated by GLUT1 and disrupts glycolytic flux regulated by GPI and PKM. This results in a three-tiered cascade inhibition along the GLUT1-GPI-PKM axis, ultimately leading to energy depletion and pulmonary injury.

4. Discussion

This study systematically revealed the pulmonary toxic effects of PFOS and its molecular mechanisms through a combination of population epidemiology, computational toxicology, animal models, and multiomics integration analysis. The integrated analysis reveals a previously unrecognized mechanism for PFOS-induced pulmonary injury: “dual metabolic hijacking”. This model posits that PFOS simultaneously disrupts two pillars of cellular homeostasis-the AOX1-GSH oxidative defense axis and the GLUT1-GPI-PKM glycolytic axis. While prior research has firmly established PFOS-induced inflammation and fibrosis as terminal outcomes in various organs, including the lung, and these pathological processes are recognized as hallmarks of its toxicity, our study reveals an upstream initiating event, the study findings position metabolic reprogramming as the upstream driver of these pathological events, representing a significant shift from the prevailing toxicological paradigm.

Based on the association analysis from the NHANES database, it demonstrates for the time a significant negative correlation between elevated PFOS serum concentrations and impaired pulmonary function. These computational predictions align with prior empirical evidence on PFOS-mediated toxicity in humans, providing a theoretical basis for systematic investigation of its pathophysiological mechanisms in the respiratory system. , Consistent with previous epidemiological evidence, higher PFOS exposure has been associated with impaired pulmonary function. Kung et al. found that intrauterine PFOS exposure was inversely associated with FEV1, FVC, and PEF in 8-year-old children, suggesting that prenatal PFOS exposure may impair subsequent lung development. Another study reported that serum PFOS levels were significantly inversely correlated with FEV1, FVC, and FEF25–75% among asthmatic children, but not among nonasthmatic controls. Together, these findings strengthen the epidemiological evidence linking PFOS exposure with functional impairment of both developing and mature lungs.

Our experimental data reveal a dose-polarized pulmonary response to PFOS, characterized by low-dose upregulation of α-SMA-indicative of a compensatory fibrotic response-and high-dose induction of TUNEL-positive apoptotic cells. This biphasic injury pattern aligns with the concept of hormesis, wherein low-level stressors activate adaptive programs that are overwhelmed at higher intensities. The observed nonmonotonic response is consistent with emerging evidence of complex PFAS toxicodynamics, as exemplified by Gazowska et al., who reported that low-dose PFOS promotes TGF-β/Smad-driven epithelial-mesenchymal transition (EMT), facilitating fibrotic repair, whereas high-dose exposure culminates in ATP exhaustion and mitochondrial apoptosis. We further posit that the apoptosis-prone phenotype of metabolically active mesenchymal cells under energy-deficient conditions explains the reduction in α-SMA expression at high PFOS doses. − Collectively, these findings reinforce the nonlinear nature of the PFOS exposure-response relationship and underscore the limitation of conventional linear toxicological models in capturing such dynamic adaptive-to-degenerative transitions.

Beyond these morphological end points, the study establishes metabolic dysregulation as a central pathogenic mechanism in PFOS-induced lung injury by computational toxicology. In contrast to the established focus on inflammatory and fibrotic end points in prior studies, our analysis reveals that PFOS extensively perturbs core metabolic nodes, with significant enrichment of genes involved in oxidoreductase activity and mitochondrial respiratory complexes. ,, This metabolic model reframes PFOS pulmonary toxicity not merely as an inflammatory condition, but as a state of acquired metabolic insufficiency, where in the disruption of energy and redox homeostasis serves as the upstream driver of downstream pathology. Such a perspective not only provides a mechanistic basis for the observed nonlinear dose–response but also positions metabolic stability as a vulnerable axis in environmental lung health, suggesting that future risk assessments for PFAS and related pollutants incorporate metabolic end points to better detect and interpret subtoxic effects.

Our integrated analysis suggests that AOX1, GPI/PKM, and CYP1A1 may represent potential key nodes underlying PFOS-induced pulmonary toxicity, collectively indicating a disturbance of redox and energy homeostasis. Specifically, PFOS exposure resulted in an approximately 80% reduction of AOX1 expression in lung tissue-an important peroxisomal enzyme involved in alveolar aldehyde detoxification. , This pattern contrasts sharply with the induction of AOX1 observed in hepatic tissue under similar exposure conditions, possibly reflecting a lung-specific vulnerability attributable to its limited antioxidant reserves. We propose that the marked suppression of AOX1 may initiate a self-amplifying oxidative injury cycle, including glutathione depletion and lipid peroxidation, thereby positioning AOX1 downregulation as an early contributing event in PFOS-associated oxidative stress. , Nevertheless, the precise physiological role of AOX1 in pulmonary defense and its direct causal contribution to PFOS toxicity remain to be experimentally confirmed. The modest induction of CYP1A1 observed in our study likely reflects compensatory activation of the AhR pathway for detoxification, consistent with previous data demonstrating AhR-mediated Cyp1a1 upregulation following PFAS exposure. The enrichment of the carcinogenic DNA adduct pathway (hsa05204) further supports the hypothesis that chronic PFOS exposure may increase carcinogenic potential via oxidative DNA damage.

In parallel, molecular docking analysis revealed high-affinity interactions between PFOS and GPI/PKM, accompanied by significant transcriptional downregulation of both enzymes. These findings imply that PFOS may impair glycolytic flux in the lung, leading to reduced ATP production and cellular energy crisis. Although previous studies have reported PFOS-induced glycolytic inhibition in hepatic and renal systems, , our work extends this observation to lung tissue. In contrast, reports of transient glycolytic activation under acute exposure conditions highlight the possibility that PFOS-mediated metabolic alterations may vary in a dose- and time-dependent manner. It should be noted that these conclusions are primarily based on computational modeling and expression-level correlations; whether PFOS directly binds to and inhibits GPI/PKM enzymatic activity, and whether such inhibition is sufficient or necessary for pulmonary toxicity, requires further functional validation.

Based on the integration of phenotypic and molecular data, we propose a dose-dependent framework of PFOS-induced lung injury: low-dose exposure may partially inhibit PKM and related targets, activating the AMPK/TGF-β axis and promoting pro-fibrotic remodeling, whereas high-dose exposure triggers severe ATP depletion and mitochondrial dysfunction, leading to apoptosis. Although this framework is consistent with several existing studies , and offers a plausible mechanistic explanation for the nonlinear dose–response relationships observed in our data, its completeness and generalizability will require validation in additional independent investigations.

5. Conclusions

In conclusion, our multilevel analysis points to a previously unrecognized mechanism in PFOS-induced lung injury, involving concurrent disruption of the AOX1-GSH oxidative defense system and the GLUT1-GPI-PKM glycolytic axis. Nevertheless, several limitations of this study must be acknowledged. First, the NHANES analysis was limited to the 2011–2012 cycle due to data availability, and broader longitudinal data sets are needed to confirm temporal stability. Second, although our integrative analyses consistently point to AOX1 suppression and glycolytic disruption as key features of PFOS-induced lung injury, the evidence is associative rather than causal; targeted genetic or pharmacological rescue studies will be required to establish causality. Lastly, the proposed metabolic biomarkers remain preliminary and must be validated in prospective human cohorts and controlled animal studies to assess their clinical utility. While direct causal validation remains for future study, our work provides a compelling, evidence-based hypothesis and lays the groundwork for a metabolism-driven framework in environmental health risk assessment.

All data sets supporting this studyincluding raw experimental measurements, processed analytical results, and computational scriptsare permanently archived in in the SEUdata repository under anonymous access: https://anonymous.4open.science/r/SEUdata/.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsomega.5c10653.

• Additional Figure S1 and Tables S1 and S2 (PDF)

#.

Li Ma and Shiyi Tan have contributed equally to this work. Li Ma: Writing, original draft, visualization and methodology. Shiyi Tan: Software, data curation and methodology. Xulei Zuo: Methodology and review. Wei Wang: Review and editing. Hao Gao: Conceptualization. Rongli Sun: Writing, review and editing. Yuepu Pu: Writing review and editing. Juan Zhang: Writing, review and editing, supervision.

Ethics Approval and Consent to Participate: The data analyzed in this study were sourced from publicly available databases, including the National Health and Nutrition Examination Survey (NHANES) and the GeneCards. As these databases contain anonymized and freely accessible data, no additional ethical approval or informed consent was required for their use. All data were utilized strictly for research purposes in compliance with ethical guidelines governing publicly available data sets. Animal experiments were conducted in accordance with the ARRIVE guidelines. All procedures followed relevant animal welfare regulations and were approved by the Animal Experimentation Ethics Committee of Southeast University.

The authors declare no competing financial interest.