Introduction

Metabolic dysfunction-associated steatotic liver disease (MASLD), previously known as nonalcoholic liver disease (NAFLD)1, refers to a spectrum of liver disorders, including metabolic dysfunction-associated steatohepatitis (MASH), formerly known as nonalcoholic steatohepatitis (NASH). A hallmark of MASLD is fat accumulation (steatosis) in the liver, due to chronic metabolic dysfunction2. The prevalence of MASLD in children has been growing in recent years, paralleling the rise in childhood obesity and metabolic syndrome3,4. The prevalence of MASLD is estimated to be around 5–10% in the general pediatric population in the United States5. However, among children with obesity, the prevalence of MASLD is much higher, with an estimated prevalence increase to 30–50%5. Diet restrictions, physical activity interventions, and United States Food and Drug Administration (U.S. FDA)-approved drugs, including glucagon-like peptide-1 (GLP-1) receptor agonists6 have been used with limited success in adolescents with MASLD7. This highlights the need for preventive measures such as identifying and intervening in modifiable risk factors.

Traditional risk factors for MASLD, such as excess energy intake, sedentary lifestyle, and genetics, cannot fully explain the MASLD epidemic in children8. Moreover, emerging evidence indicates that exposure to endocrine-disrupting chemicals can promote metabolic changes that result in fatty liver disease, a hypothesis referred to as the “Toxicant Fatty Liver Disease”9,10,11. Per- and polyfluorinated substances (PFAS), a large class of synthetic fluorinated organic chemicals, are ubiquitous worldwide. These chemicals have been used in industrial applications and consumer products, including water-repellent textiles, nonstick coatings, and food packaging products for over 60 years12. PFAS have been detected in the blood of over 99% of individuals in the United States (U.S.)13,14. Research has demonstrated that PFAS can accumulate in the human body with a particular predilection for the liver15,16,17. This accumulation is associated with disruptions in several hepatic functions, notably lipid metabolism18,19,20,21. A substantial body of evidence, supported by both experimental and epidemiological studies, indicates that certain PFAS are hepatotoxic in humans, with many studies specifically linking PFAS exposure to lipid dysregulation19. However, critical gaps in the literature remain. These include: (a) whether individuals with overweight or obesity are more susceptible to PFAS-induced hepatotoxicity, (b) the potential hepatotoxic effects of less-studied or replacement PFAS compounds, and (c) identification of the specific metabolic pathways impacted by PFAS that are indicative of liver damage18,19,21,22.

Production of certain PFAS, such as perfluorooctane sulfonate (PFOS) and perfluorooctanoate (PFOA), was voluntarily phased out in the U.S. during the 2000s, yet their negative health effects remain a concern because of their long half-lives (1.8–6.2 years)13,18,23,24. Consequently, newer short-chain PFAS variants (less than six fluorinated carbons), known as replacements, have been introduced, featuring shorter biological half-lives, to mitigate environmental and biological persistence23. However, growing evidence indicates that short-chain PFAS are not inherently safer as they are highly persistent, more environmentally mobile, harder to remove from drinking water, and bioaccumulate in humans, animals, and plants25,26. As these compounds have increasingly replaced long-chain PFAS, their presence in the environment has risen, yet little is known about the long-term health effects of chronic exposure27,28. Emerging research suggests they may pose similar toxicological harm as long-chain PFAS, raising concerns about the cumulative and lasting impact of short-chain PFAS27,29.

We propose a translational research framework designed to bridge scientific findings from both in vitro and in vivo epidemiologic data, with the goal of elucidating the role of PFAS in the risk and progression of MASLD (Fig. 1). The human study included in this manuscript represents a unique resource, as it involves histologically confirmed MASLD phenotypes derived from liver biopsies of adolescents with obesity. Our findings reveal a strong association between perfluoroheptanoic acid (PFHpA) exposure and the risk and severity of MASLD. PFHpA is a short-chain carboxylic (six-carbon chain) PFAS compound that remains relatively understudied and has been detected at elevated concentrations in the liver20. PFHpA emerges as the only PFAS congener measured in the Teen-LABS study to be significantly associated with MASLD, prompting a focused investigation of its potential role in metabolic liver pathology. Using an innovative approach, we assessed the impact of PFHpA on liver metabolism in vitro by employing 3D human liver spheroids coupled with single-cell transcriptomics. Our results show that PFHpA exposure mainly disrupts hepatic lipid metabolism in vitro. We then integrated multi-omic datasets from both human studies and in vitro experiments, using advanced statistical methods, and identified specific protein and metabolite signatures associated with the development of MASLD in the context of PFHpA exposure. The proteome signature is linked to markedly higher MASLD odds (OR = 7.1), whereas a metabolome profile is associated with lower odds. To the best of our knowledge, our study presents an innovative strategy to identify individuals at a high risk of developing PFAS-induced MASLD, paving the way for the development of early intervention strategies.

Fig. 1: Integrated translational framework combining in vivo and in vitro studies to identify mechanism of perfluoroheptanoic acid (PFHpA)-associated metabolic dysfunction-associated steatotic liver disease (MASLD).
figure 1

a In vivo analysis from the Teen-LABS (Longitudinal Assessment of Bariatric Surgery) cohort, including measurement of plasma-PFHpA concentrations, plasma derived metabolomic and proteomics, and histopathological assessed MASLD based on liver biopsy. b In vitro human liver spheroid model composed of primary hepatocytes and non-parenchymal cells (NPCs) exposed to 20 μM PFHpA for 7 days. Following exposure, spheroids were dissociated into single cells and underwent single-cell RNA sequencing using 10x Genomics and Illumina platforms. c Multi-omics data integration using Omics Net 2.0. A total of 156 upregulated genes from PFHpA-exposed liver spheroids were integrated with 42 plasma metabolites (MWAS) and 28 plasma proteins (PWAS) identified among those with and without MASLD. Overlapping pathways between gene expression (GEX) with metabolomic/proteomic signatures revealed 19 metabolites and 6 proteins. Latent clustering with integrated data (LUCID) methodologies were then used to integrate PFHpA exposure with the 19 identified metabolites and 6 identified proteins to determine a high risk MASLD profile of adolescents from the Teen-LABS study. Created in BioRender. Maretti Garcia, A. (2025) https://BioRender.com/n17z953

Full size image

Methods

Study population

This study was based on data from the Teen-LABS study (ClinicalTrials.gov number, NCT00465829), a prospective, multicenter, observational study of adolescents (≤19 years of age) who underwent bariatric surgery between 2007 and 2012. Participants were enrolled at five clinical centers in the United States: Cincinnati Children’s Hospital Medical Center (Cincinnati, Ohio), Nationwide Children’s Hospital (Columbus, Ohio), the University of Pittsburgh Medical Center (Pittsburgh, Ohio), Texas Children’s Hospital (Houston, Texas), and the Children’s Hospital of Alabama (Birmingham, Alabama)30. Study inclusion criteria were (1) adolescents up to 19 years of age, (2) adolescents approved for bariatric surgery, and (3) agreement to participate in the Teen-LABS study, demonstrated through the signing of Informed Consent/Assent30. Demographics are shown in Table S1. The Teen-LABS steering committee, which included a site principal investigator from each participating center, collaborated with the data coordinating center and project scientists from the National Institute of Diabetes and Kidney Disease (NIH-NIDDK) to design and implement the study30. All bariatric procedures were performed by surgeons who were specifically trained in study data collection (Teen-LABS-certified surgeons)30,31,32,33. The present study included 136 participants whose plasma was collected at the time of surgery. The study protocol, assent/consent forms, and data and safety monitoring plans were approved by the Institutional Review Boards of each institution, including Cincinnati Children’s Hospital Medical Center, Baylor College of Medicine, Children’s Hospital of Alabama, University of Pittsburgh, and Nationwide Children’s Hospital, and by the independent data and safety monitoring board prior to study initiation30. Written informed consent or assent, as appropriate for age, was obtained from all parents/guardians and adolescents30. This study was approved by the University of Southern California Review Board.

Data collection

Standardized methods for data collection have been described previously30,31,32,33. Fasting blood samples were obtained preoperatively. Liver histology and liver biopsy methodology have been previously detailed33, however, liver biopsies were obtained using the core needle technique after anesthesia induction and before performing the bariatric surgery procedure. Owing to the observational study design and lack of published consensus on whether intraoperative liver biopsies should be the standard of care at the time of bariatric surgery, the decision to perform a liver biopsy was deferred to the surgical teams at each site. Accordingly, 99% of all biopsies were performed at sites where intraoperative biopsy is the standard of care. Liver biopsies were stained using hematoxylin-eosin and Masson’s trichrome techniques, reviewed, and scored centrally by an experienced hepatopathologist using the validated MASH Clinical Research Network scoring system34. MASLD was defined based on the histopathological diagnosis (Table S2). Detailed descriptions of study methods, comorbidity, and other data definitions, case report forms, and laboratory testing have been included in previous publications30,31,32. For this analysis, covariates, including participants’ age, sex assigned at birth, race, and parents’ income, were obtained at the time of surgery by trained study personnel30,31. The collected data were maintained in a central database by the data-coordinating center.

Plasma-PFAS laboratory analysis

The samples were transported on dry ice with temperature logging by a World Courier (AmerisourceBergen Corporation, Conshohocken, PA) and stored at −80 °C until analysis. The samples were analyzed by online solid-phase extraction followed by liquid chromatography and mass spectrometry (LC-MS/MS), as previously described20. The limit of detection (LOD) is 0.03 ng/mL for all reported PFAS compounds. Values below the LOD were imputed as ½ LOD. The batch imprecision for the quality control samples was less than 6% for all measured compounds. Plasma concentrations of PFAS measured in Teen-LABS participants have been previously published20.

Plasma metabolomics

Untargeted metabolomics was performed on plasma samples collected at the time of bariatric surgery. Liquid chromatography and high-resolution mass spectrometry (LC-HRMS) were used as described by Liu et al.35, with dual-column and dual-polarity approaches and both positive and negative electrospray ionization. This resulted in four analytical configurations: reverse-phase (C18)-positive, C18 negative, hydrophilic interaction (HILIC) positive, and HILIC negative. Unique features were identified using mass-to-charge ratio (m/z), retention time, and peak intensity. Features were adjusted for batch variation36 and excluded if they were detected in <20% of the samples or if there was a >30% coefficient of variability of the quality control samples after batch correction. After processing, there were 3716 features from the C18 negative mode, 5069 from the C18 positive mode, 7444 from the HILIC negative mode, and 6944 from the HILIC positive mode for a total of 23,173 features included in the analyses. The raw intensity values from LC-HRMS were scaled to a standard normal distribution and log2 transformed. The details of the analytical process have been described previously37. Confirmed annotations with a confidence level of 1 were available for 358 metabolomic features38. For confirmed metabolites, the find.Overlapping.mzs function in the XMSanalyzer v2.0.6.1 R package was used39. Matching was completed using a 5 ppm mass error and either 15 or 30 s time error. Metabolites were identified by comparing them with authentic chemical standards under identical analytical conditions, and peaks were matched to annotations using m/z and retention time. In instances where multiple annotations were possible because more than one molecule had retention times within the allowable error, the annotation with the closest retention time to the known standard was chosen. The measured m/z and retention times, theoretical m/z and retention times, adducts, possible annotations, and additional analytical details are listed in Table S3.

Plasma proteomics

Proteins were measured in fasting plasma samples using the proximity extension array (PEA) method from the Olink Explore 384 Cardiometabolic panel and Olink Explore 384 Inflammation panel40. These panels measure the relative abundance of 731 proteins, reported as normalized protein expression (NPX) levels after log2 transformation41. After excluding proteins with over 50% of observations below the LOD and duplicate proteins, 702 proteins were retained from the initial 731 offered after processing.

Liver spheroid assay

We used a 3D InSight™ Human Liver Model (cat# MT-02-302-04, InSphero Inc., Schlieren, Switzerland) to test the impact of PFHpA on human liver metabolism. This model is composed of human primary hepatocytes from 10 donors (five males and five females) and non-parenchymal cells from one donor. A PFHpA stock solution (CAS 375-85-9, cat# 342041, Sigma-Aldrich, St Louis, MO, USA) was prepared in dimethyl sulfoxide -DMSO- (cat# D8418, Sigma-Aldrich). The final working solution was diluted in 3D InSight™ Human Liver Lean Maintenance Medium (cat# CS-07-305B-01, InSphero Inc.) to a final non-cytotoxic concentration of 20 µM PFHpA and 0.1% DMSO42. Liver spheroids were continuously exposed to PFHpA for 7 days, and culture medium was replaced every 2–3 days with media containing freshly diluted PFHpA. For the control, liver spheroids were exposed to 0.1% DMSO diluted in lean spheroid media and cultured for 7 days, following the same regimen of media replacement. Spheroids were cultured in a 96-well format, with a single spheroid per well, under sterile conditions and incubated at 37 °C and 5% CO2 following the manufacturer’s instructions. We used 96 spheroids per condition.

Lipid accumulation assay and analysis

After 7 days of treatment, the spheroids were fixed with 4% paraformaldehyde (cat#1004960700, Sigma-Aldrich) diluted in phosphate-buffered saline–PBS (cat#10010023, Gibco, Grand Island, NY USA) for 1 h, permeabilized with 0.2% Triton-X100 (cat#85111, Thermo-Scientific, Waltham, MA USA) in PBS for 30 min, and blocked with 1% bovine serum albumin–BSA (cat#A2153, Sigma-Aldrich) in PBS for 1 h at room temperature. Spheroids were then stained with Nile Red (cat#72485, Sigma-Aldrich) at 1 µg/mL, and DAPI (cat#D9542, Sigma-Aldrich) at 2 µg/mL in 1% BSA/PBS for 30 min and washed 3x with PBS. Spheroids were mounted with Prolong Gold Mounting Medium (cat#P36934, Invitrogen, Waltham, MA, USA) and images were acquired using a Leica TCS SP8 confocal microscope. Lipid accumulation was quantified using ImageJ (v1.49)43.

Single-cell RNA library preparation, sequencing, and data analysis

Spheroids were dissociated using 0.25% trypsin/EDTA (cat#0103, ScienCell, Carlsbad, CA, USA) for 15 min, and dead cells were removed using a Dead Cell Removal Kit (cat#130-090-101, Miltenyi Biotec, Bergisch Gladbach, Germany). Viable cells were partitioned with the Chromium Next GEM Single Cell 3ʹ Kit (cat#1000269, 10X Genomics, Pleasanton, CA, USA). Libraries were sequenced on a USC Molecular Genomics Core using the Illumina platform. Raw data were processed using Cell Ranger count pipeline (10X Genomics) with low-quality cell removal according to sample-specific quality control (QC) using the R package Seurat44. Comparison between PFHpA-treated and control samples was performed by integrating samples into a unified dataset using the SCTransform integration workflow implemented in Seurat45. Cell annotation was based on the expression of known cell type marker genes. Differentially expressed genes between clusters from treatment groups were detected using Wilcoxon Rank Sum test in the Seurat FindMarkers, and genes with a Bonferroni adjusted p value < 0.1 were considered significant. Biological data interpretation was performed using Ingenuity Pathways Analysis (IPA)46.

Multi-omics data integration

We performed knowledge-driven integration of the differentially expressed genes obtained from the transcriptomic PFHpA/liver spheroids in vitro assay (GEX), and the plasma proteome wide association study (PWAS) and metabolome wide association study (MWAS) data obtained from the Teen-LABS cohort using the online platform OmicsNet 2.0 (www.omicsnet.ca)47. Briefly, OmicsNet identified the significant canonical pathways in each dataset and separately overlapped the results from GEX, PWAS, GEX, and MWAS. We then identified the proteins and metabolites in the overlapping pathways and used them as omics signatures for further analysis using latent unknown clustering integrating multi-omics data (LUCID).

Statistics and reproducibility

Plasma-PFHpA and MASLD

We first evaluated the associations of eight PFAS measured in plasma with MASLD (yes/no) using logistic regression and controlling for multiple comparisons using Bonferroni correction. Plasma-PFAS concentrations were log2-transformed. PFHpA was the only congener found to be associated with MASLD in our study, and the robust association between PFHpA and MASLD phenotypes prompted the focus of investigation of its potential role in MASLD pathology. Therefore, we focused on subsequent analyses to further explore the association between plasma PFHpA and MASLD and its features using either multinomial logistic regression models or logistic regression models, based on the number of categories in each outcome. The outcomes of interest included the progression of MASLD (No MASLD, MASLD not MASH, MASH, multinomial regression), hepatocellular ballooning (None: No hepatocyte ballooning observed; Few: Few ballooned hepatocytes [1–4 ballooned hepatocytes per 20x field]; Many: Many cells with prominent ballooning [≥5 ballooned hepatocytes per 20x field])48; multinomial logistic regression), grade of steatosis (none, 5–33%, 34–67%; multinomial logistic regression), fibrosis (none, present; logistic regression), and NAS activity score (none, 1, 2, ≥3; multinomial logistic regression). Covariates were selected a priori based on their relevance to both PFHpA exposure and MASLD. For all epidemiological models these included age (continuous; years), sex (binary; male or female), race (binary; white or other), parental income (categorical; <25,000 United States Dollars (USD) per year, 25,000–74,999 USD per year, ≥75,000 USD per year), and study site (categorical; Baylor College of Medicine [BCM], Cincinnati Children’s Hospital [CIN], Nationwide Children’s Hospital [NCH], University of Alabama at Birmingham [UAB]). Parental income was used as a proxy for socioeconomic status. Although dietary and other environmental exposure data were not uniformly available, adjustment for study site may partially account for regional differences in environmental and lifestyle factors. To test the concentration-dependent relationship between plasma-PFHpA and MASLD, we used quantiles of PFHpA exposure with a continuous exposure value to determine the trend P value across quantiles, depicting a readily interpretable dose-response relationship. All statistical analyses were performed using R version 4.4.1.

Metabolome-wide association study (MWAS) —linking PFHpA exposure with disease pathways

For metabolomics analysis, we utilized linear regression to model log2-transformed PFHpA concentrations with log2-normalized metabolite intensity. Models were adjusted for participants’ age (continuous; years), sex (binary; male or female), race (binary; white or other), parental income (categorical; <25,000 USD per year, 25,000–74,999 USD per year, ≥75,000 USD per year), and study site (categorical; BCM, CI, NCH, UAB). As our primary objective was not to investigate individual feature-level significance but rather system-level insights, we conducted an enrichment analysis for features that were altered by PFHpA exposure (nominal p < 0.05) using canonical pathways and diseases and biofunctions curated from Qiagen Knowledge Base using QIAGEN Ingenuity Pathway Analysis (IPA, Qiagen Inc.). For the MWAS, only metabolites confirmed through accurate mass, retention time, and MS/MS spectral matching against the laboratory’s internal reference library were included in the final analyses. The confirmation library was restricted to endogenous human metabolites and does not include xenobiotics compounds. This approach was used to ensure high-confidence identification and facilitate pathway-based interpretation.

Proteome-wide association study (PWAS)—linking PFHpA exposure with disease pathways

For the proteomics analysis, the 702 proteins retained after quality control filtering were log2-transformed. These NPX values were modeled as dependent variables in multiple linear regression models, with log2-transformed PFHpA concentration as the independent variable. The same covariates as the MWAS—age, sex, race, parental income, and study site—were included to adjusted for potential confounding. To identify functional implications beyond individual protein associations, we performed enrichment analysis on proteins associated with PFHpA nominal p value < 0.05, using canonical pathways and disease/function annotations curated from the Qiagen Knowledge Base in IPA.

Multi-omics integration- Latent Unknown Clustering by Integrating multi-omics Data (LUCID)

LUCID49 is a quasi-mediation analysis approach of multi-omics data that estimates the joint associations between the environmental exposure \({E}\) (PFHpA), the multi-omics data \(Z\) (19 metabolites and 6 proteins that were identified through prior pre-screening procedures), and the outcome \(Y\) (MASLD, yes/no) if supervised via the latent cluster variable \(X\). The Expectation-Maximization (EM) algorithm was implemented to iteratively estimate \(X\) and update the parameters until convergence. For an unsupervised LUCID model, the parameters of interest include (1) \(\beta\), representing PFHpA-to-cluster associations; (2) \(\mu\), representing the cluster-specific means of omics features; and (3) the individual-level inclusion probability (IP) for each latent cluster. In the unsupervised LUCID approach, \(X\) integrates information from both \(E\) and \(Z\), effectively delineate distinct risk profiles among the subjects. The original LUCID framework was initially proposed for the early integration of multi-omics data, entailing concatenation of all omics layers into a single matrix. Detailed descriptions of the original LUCID have been previously introduced49. As an extension of the original LUCID, LUCID in parallel utilizes an intermediate integration strategy of multi-omics data to estimate separate latent clusters \(X\) within each omic layer by assuming no correlations across different omics layers50. In our study, there were two layers of multi-omics data, \(Z\), metabolites, and proteins, resulting in two individually estimated latent cluster variables, \({X}_{{{{\rm{metabolome}}}}}\) and \({X}_{{{{\rm{proteome}}}}}\). We fitted the optimal unsupervised LUCID in parallel model using PFHpA as \(E\) and pre-selected 19 metabolites and 6 proteins as \(Z\). Based on model selection procedures using Bayesian information criterion (BIC), the number of latent clusters per omic layer of the optimal model was chosen to be 2. We extracted the IPs to \({X}_{{{{\rm{metabolome}}}}}\) and \({X}_{{{{\rm{proteome}}}}}\,\)(\({{IP}}_{{{{\rm{metabolome}}}}}\) and \({{IP}}_{{{{\rm{proteome}}}}}\)) from the converged optimal LUCID in the parallel model. \({{IP}}_{{{{\rm{metabolome}}}}}\) and \({{IP}}_{{{{\rm{proteome}}}}}\) are continuous probabilities indicating the likelihood of being included in each level of the cluster \({X}_{{{{\rm{metabolome}}}}}\) and \({X}_{{{{\rm{proteome}}}}}\), respectively. These probabilities were determined by the subjects’ exposure levels and the presence of metabolites and proteins, respectively. In follow-up analyses to explore how \({{IP}}_{{{{\rm{metabolome}}}}}\) and \({{IP}}_{{{{\rm{proteome}}}}}\) were associated with the outcome of interest, MASLD, we fitted a logistic regression model using \({{IP}}_{{{{\rm{metabolome}}}}}\) and \({{IP}}_{{{{\rm{proteome}}}}}\) as predictors and MASLD as the binary response variable, while adjusting for covariates, age (continuous; years), sex (binary; male or female), race (binary; white or other), parental income (categorical; <25,000 USD per year, 25,000–74,999 USD per year, ≥75,000 USD per year), and study site (categorical; BCM, CI, NCH, UAB).

Lipid accumulation in liver spheroids

Lipid quantification was performed using ImageJ (v1.49). The average lipid concentration of each spheroid (n = 4) was determined by the average Corrected Total Cell Fluorescence (CTCF) from the five images as follows:

$$ {{{\rm{CTCF}}}}={{{\rm{Integrated}}}}\; {{{\rm{Density}}}}\,{{{-}}}\,\\ ({{{\rm{Area}}}}\; {{{\rm{of}}}}\; {{{\rm{selected}}}}\; {{{\rm{cell}}}}\; {{{\rm{X}}}}\; {{{\rm{Mean}}}}\; {{{\rm{fluorescence}}}}\; {{{\rm{of}}}}\; {{{\rm{background}}}}\; {{{\rm{readings}}}})$$

Lipid quantification significance was calculated by GraphPad Prism (v10.2.3) using the ordinary one-way ANOVA test with Tukey multiple comparisons correction.

Results

Human study: PFHpA increases the risk for MASLD in adolescents–insights into steatosis severity and disease progression

This study included 136 adolescents with severe obesity who underwent bariatric surgery (Table S1). Based on the liver biopsy results, participants were categorized into three groups: 55 (40%) were classified as non-MASLD, 51 (38%) MASLD not MASH, and 30 (22%) as MASH (Table S2). Demographic characteristics were generally similar across MASLD status groups. Participants with MASH had slightly higher baseline BMI (mean = 55.1 kg/m²) and were more likely to be male (40.0%) compared to those with MASLD without MASH (23.5%) and those without MASLD (21.8%). The distribution of race and parental income was comparable across groups, with a majority identifying as White and reporting annual household incomes below $75,000 (Table S1). As expected, and shown in Table S2, histologic features of MASLD differed by MASLD status. Participants with MASH had the highest prevalence of severe steatosis (≥ 34%, 43.3%), fibrosis (56.7%), and hepatocellular ballooning (63.4% with few or many ballooned cells), along with a higher NAFLD Activity Score (NAS ≥ 3 in 73.3%) compared to those with MASLD without MASH and those without MASLD, who exhibited minimal or no histologic abnormalities. Among the 8 PFAS congeners analyzed, PFHpA plasma concentration among MASLD severity: No MASLD median = 0.6 ng/mL, Interquartile Range (IQR) = 0.09 ng/mL; MASLD no MASH median = 0.12 ng/mL, IQR = 0.13 ng/mL; MASH median = 0.17 ng/mL, IQR = 0.16 ng/mL) was the only congener significantly associated with increased MASLD risk and disease progression (Figs. 2a and S1 and Table S3). Each doubling of PFHpA plasma levels was associated with an 80% increase in the risk of developing MASLD (Odds Ratio (OR), 1.8; 95% Confidence Interval (CI): 1.3, 2.5) (Fig. S1 and Table S4). A significant dependent-response relationship between PFHpA levels and MASLD risk was observed (ptrend < 0.0001), indicating a biological gradient across the exposure quantiles (Fig. 2b). Additionally, PFHpA was significantly associated with several indicators of disease severity, including the degree of steatosis (OR = 1.6, 95% CI: 1.2, 2.2 for mild steatosis and OR = 1.9, 95% CI: 1.2, 2.9, moderate steatosis), fibrosis (OR = 1.49, 95% CI: 1.0, 2.1), hepatocellular ballooning (OR = 1.6, 95% CI: 1.0, 2.6 for few balloon cells and OR = 2.8, 95% CI: 1.1, 7.6 for many balloon cells), and the NAFLD activity score (NAS) (OR = 3.0, 95% CI: 1.8, 5.2) (Fig. 2c and Table S5).

Fig. 2: Associations between plasma perfluoroheptanoic acid (PFHpA) concentrations and histopathological assessed metabolic dysfunction-associated steatotic liver disease (MASLD) and related liver phenotypes in the Teen-LABS (Longitudinal Assessment of Bariatric Surgery) study (n = 136).
figure 2

This figure depicts the odds ratios and 95% confidence intervals (CIs) for a the association between PFHpA and MASLD severity categories (no MASLD, MASLD without metabolic dysfunction-associated steatohepatitis (MASH), MASH), b Concentration-dependent relationship between PFHpA exposure quantiles and the presence of MASLD (yes/no), c the associations between PFHpA and specific MASLD phenotypes: hepatocellular ballooning (None: No hepatocyte ballooning observed; Few: Few ballooned hepatocytes [1–4 ballooned hepatocytes per 20x field]; Many: Many cells with prominent ballooning [≥5 ballooned hepatocytes per 20x field]), steatosis grade (None, 5–33%, 34–67%), fibrosis (yes/no), and Nonalcoholic Fatty Liver Disease (NAFLD) Activity Score (NAS, None, 1, 2, ≥3). All models are adjusted for the following variables: age (continuous; years), sex (binary; male or female), race (binary; white or other), parental income (categorical; <25,000 United States Dollars (USD) per year, 25,000–74,999 USD per year, ≥75,000 USD per year), and study site (categorical; Baylor College of Medicine [BCM], Cincinnati Children’s Hospital [CIN], Nationwide Children’s Hospital [NCH], University of Alabama at Birmingham [UAB]). PFHpA (ng/mL) concentrations were log2 transformed and interpreted as per doubling of exposure.

Full size image

Human study: PFHpA affects lipid metabolism, oxidative stress, and inflammation in adolescents

Using Metabolome-Wide Association Study (MWAS) and Proteome-Wide Association Study (PWAS) approaches, we identified 48 metabolites and 55 proteins measured in plasma that were significantly altered by PFHpA exposure (Tables S6 and S7). Ingenuity Pathway Analysis (IPA) revealed the biological pathways affected by PFHpA, with the top 10 pathways depicted in Fig. 3a, b (nominal p < 0.05). The metabolites predominantly affected pathways related to amino acid metabolism, including L-carnitine biosynthesis, arginine, β-alanine, phenylalanine degradation, proline catabolism, and glyoxylate metabolism, as well as lipid metabolism, which involves the transport of bile salts, organic acids, metal ions, and amine compounds, along with Farnesoid X receptor/Retinoid X receptor (FXR/RXR) activation. Notably, the elevated concentrations of cholate, glucuronate, ursocholic acid, murideoxycholic acid, glycine, and glycocholic acid among the top 10 metabolites suggested a potential dysfunction in bile acid synthesis. In the plasma proteomic analysis, we used the Olink Explore® cardiometabolic and inflammatory panels to identify proteins relevant to MASLD. Overall, the canonical pathways enriched by these proteins were related to chronic inflammation, cytokine signaling, and hepatic fibrosis (Fig. 3b). Among the top 10 proteins from PWAS, five were directly associated with MASLD severity (C-C Motif Chemokine Ligand 20 (CCL20)51, C-C Motif Chemokine Ligand 25 (CCL25)52, Alcohol Dehydrogenase 4 (ADH4)53, Interleukin 1 Receptor Antagonist (IL1RN)54, and Triggering Receptor Expressed on Myeloid Cells 2 (TREM2)55).

Fig. 3: Omics-Wide Association Study (OWAS) of perfluoroheptanoic acid (PFHpA) in adolescents in the Teen-LABS (Longitudinal Assessment of Bariatric Surgery) cohort (n = 131).
figure 3

This figure presents enriched biological pathways identified through: a MWAS of plasma PFHpA (ng/mL) concentrations and metabolite levels b PWAS for plasma PFHpA (ng/mL) concentrations and proteins. MWAS and PWAS analyses adjust for the following variables: age (continuous; years), sex (binary; male or female), race (binary; white or other), parental income (categorical; <25,000, 25,000–74,999, ≥75,000 USD/yr), and study site (categorical; Baylor College of Medicine [BCM], Cincinnati Children’s Hospital [CIN], Nationwide Children’s Hospital [NCH], University of Alabama at Birmingham [UAB]). PFHpA concentration is log2 transformed. Abbreviations: FXR/RXR Farnesoid X Receptor/Retinoid X Receptor activation, IL interleukin, MWAS metabolome-wide association study, PWAS proteome-wide association study.

Full size image

In vitro study: PFHpA disturbs lipid metabolism in human liver spheroids

To determine the role of PFHpA in MASLD progression, we conducted an in vitro assay, where we exposed human 3D liver spheroids composed of primary hepatocytes and non-parenchymal cells (NPCs) to PFHpA for 7 days in a normal-glucose medium (Fig. 1b). Subsequent analysis using single-cell RNA sequencing (scRNA-seq) revealed significant transcriptomic alterations in the liver spheroid cells. We identified hepatocytes, T lymphocytes (T cells), Kupffer cells, and Natural Killer cells (NK cells) (Fig. 4a). B lymphocytes (B cells) were present in minimal numbers and were therefore excluded from further analysis. Exposure to PFHpA resulted in the alteration of 472 genes in liver spheroids, with 156 upregulated and 316 downregulated genes (Fig. 4b). Hepatocytes and T cells displayed the highest number of differentially expressed genes (DEGs), with 263 DEGs in hepatocytes (137 upregulated and 126 downregulated) and 383 DEGs in T cells (98 upregulated and 285 downregulated) (Fig. 4b). The top differentially expressed genes in whole spheroids, hepatocytes and T cells are shown in Tables S8, S9, and S10. NK cells exhibited 26 DEGs (19 upregulated and seven downregulated), whereas Kupffer cells showed only the upregulation of a single gene. IPA pathway analysis highlighted the notable impact of PFHpA on liver metabolism. In the whole liver spheroids, hepatocytes, and T cells, we observed a substantial upregulation of pathways involved in lipid metabolism (48%, 45%, and 40%, respectively), amino acid metabolism (21%, 18%, and 20%, respectively), and detoxification pathways (14%, 14%, and 10%, respectively), indicating a significant disturbance induced by PFHpA exposure (Fig. 4c–e). Additionally, PFHpA exposure led to activation of peroxisome proliferator-activated receptor alpha (PPAR-α) and enhanced fatty acid oxidation in human hepatocytes (Fig. 4f). Most pathways related to hepatocyte lipid metabolism are associated with lipid anabolism (lipogenesis), with upregulation of several lipid biosynthesis pathways, including cholesterol. This was corroborated by Nile Red imaging, which revealed a significant increase of lipid accumulation in PFHpA-exposed liver spheroids compared to that in controls (p = 0.01) (Fig. 4g, h). Although PFHpA exposure also affected lipid metabolism in T cells, the upregulated pathways were primarily associated with nuclear hormone receptor activation and did not directly affect lipid biosynthesis, unlike in hepatocytes (Fig. S2). These findings emphasize the critical role of PFHpA in disrupting lipid metabolism, particularly in hepatocytes, and suggest the potential contribution of PFHpA to MASLD pathogenesis.

Fig. 4: Perfluoroheptanoic acid (PFHpA) exposure disturbs lipid metabolism in human liver spheroids.
figure 4

This figure presents a Uniform Manifold Approximation and Projection (UMAP) plot of integrated single-cell transcriptomic data from control (CTR) and PFHpA-exposed liver spheroids. Identified cell types include hepatocytes, Kupffer cells (liver-resident macrophages), T lymphocyte cells (a type of adaptive immune cell), natural killer (NK) cells (a type of innate immune cell), and B cells (antibody-producing immune cells). b Differentially expressed genes (DEGs) identified in whole spheroids (global) and individual cell clusters following PFHpA exposure. c–e Biological process categories significantly upregulated in response to PFHpA in c whole liver spheroids, d T cells, and e hepatocytes f Lipid metabolism-related pathways upregulated in hepatocytes, including activation of the peroxisome proliferator-activated receptor alpha (PPAR-α). g Confocal imaging of liver spheroids stained with Nile Red to visualize lipid accumulation (red). CTR Control group, PFHpA exposed group. h Quantification of lipid accumulation in liver spheroids using ImageJ (version 1.49), presented as correlated total cell fluorescence (CTCF).

Full size image

Unveiling biomarker signatures for PFHpA-induced MASLD through multiomic integration

We employed the OmicsNet 2.0 platform to integrate data from human and in vitro studies to identify plasma biomarkers associated with PFHpA-induced MASLD. Our multiomic analyses combined 156 upregulated genes (GEX) in whole liver spheroids with 42 metabolites (MWAS) and 28 proteins (PWAS) found at higher concentrations in the plasma of human subjects (Fig. 1c). We identified a metabolomic signature for PFHpA-induced MASLD that included 19 metabolites involved in lipid and amino acid metabolism (Fig. 5a and Table S11). Additionally, a proteomic signature was defined consisting of six proteins linked to lipid degradation and immune response pathways (Fig. 5b and Table S11).

Fig. 5: Integration of in vitro and in vivo datasets to identify shared proteomic and metabolomic signatures.
figure 5

This figure illustrates the pathways commonly found in a GEX (in vitro) and metabolites (in vivo), and b GEX (in vitro) and proteins (in vivo) datasets. Identified pathways reflect coordinated molecular changes across experimental systems. GEX gene expression, PWAS Proteome-wide association study.

Full size image

To further analyze the relationship between PFHpA exposure, multiomic signatures, and MASLD risk, we utilized a Latent Unknown Clustering with Integrated Data (LUCID) model49. This model grouped individuals based on similarities in PFHpA exposure, proteome and metabolome signatures, and disease outcomes, focusing on disease risk rather than on stratification. Figure 6a illustrates the associations between PFHpA and two latent clusters derived from each omic layer (metabolome and proteome) that characterize the low and high MASLD risk groups of individuals. PFHpA was more strongly associated with the high-risk MASLD cluster characterized by specific proteins (OR = 2.73) than with the low-risk cluster characterized by metabolites (OR = 1.05). Specifically, individuals in proteome profile 1 had significantly higher odds of MASLD (OR = 7.08) than those in proteome profile 0, while individuals in metabolome profile 1 had lower odds of MASLD (OR = 0.51) than those in metabolome profile 0. This analysis identified metabolome profile 0 and proteome profile 1 as high-risk multiomic profiles for MASLD.

Fig. 6: Multi-omics integration of (perfluoroheptanoic acid) PFHpA, proteomics, and metabolomics to identify high-risk clusters for metabolic dysfunction-associated steatotic liver disease (MASLD) (n = 131).
figure 6

Using LUCID (Latent Unknown Clustering Integrating multi-omics Data) framework, we identified a distinct multi-omics risk profile associated with increased odds of MASLD. The profile links elevated PFHpA plasma concentration with increased plasma protein expression and decreased plasma metabolite expression. a Associations between predicted individual profile scores (PIPS) from unsupervised LUCID models and MASLD status (yes/no). The model incorporates both omics layers (proteins and metabolites) as well as PFHpA concentration and is adjusted for the following variables: age (continuous; years), sex (binary; male or female), race (binary; white or other), parental income (categorical; <25,000 United States Dollars (USD) per year, 25,000–74,999 USD per year, ≥75,000 USD per year), and study site (categorical; Baylor College of Medicine [BCM], Cincinnati Children’s Hospital [CIN], Nationwide Children’s Hospital [NCH], University of Alabama at Birmingham [UAB]). PFHpA is log2 transformed and interpretated as per doubling of exposure. b Classification of individuals into high- and low-risk MASLD profiles based on protein, metabolite, and PFHpA exposure. Abbreviations: PFHpA perfluoroheptanoic acid, MASLD metabolic dysfunction-associated steatotic liver disease, LUCID Latent Unknown Clustering Integrating multi-omics Data, PIPS predicted individual profile scores, n number of participants, HYAL1 hyaluronidase-1, F7 coagulation factor VII, CA5A carbonic anhydrase 5A, C2 complement component 2, ADH4 alcohol dehydrogenase 4, ACY1 aminoacylase-1.

Full size image

Figure 6b illustrates the differentiation between the high- and low-risk groups for MASLD based on clusters of metabolites and proteins. Key metabolites, including tryptophan, glycochenodeoxycholic acid, and deoxycarnitine, were identified as significant features that distinguished these risk groups. The combined presence of high concentrations of the metabolites Trans-4-Hydroxy-L-Proline and 5-Aminovaleric acid, and proteins aminoacylase 1 (acy1), alcohol dehydrogenase 4 (adh4), complement component 2 (c2), carbonic anhydrase 5A (ca5a), coagulation factor VII (f7), and hyaluronidase 1 (hyal1) was indicative of a higher risk MASLD group. Notably, the proteins hyal1, f7, ca5a, c2, adh4, and acy1 exhibited similar scaled values ranging from 0.21 to 0.25 for the high-risk group, while the values for low-risk group ranged from −0.63 to −0.52.

Discussion

To the best of our knowledge, this study presents a novel translational framework that integrates human epidemiologic data with mechanistic in vitro models to evaluate the role of PFHpA, an unregulated short-chain PFAS congener56,57,58 that accumulates in high concentrations in the liver relative to other PFAS congeners20, in the development of metabolic dysfunction-associated steatotic liver disease (MASLD) among adolescents with obesity. Using histologically confirmed liver outcomes from the Teen-LABS cohort, we found that each doubling in plasma PFHpA concentration was associated with an 80% increase in MASLD risk and progression across several markers of liver pathology. Integrative analysis revealed overlapping dysregulated pathways in both human and spheroid models, particularly those related to innate immunity, inflammation, and lipid metabolism. Multi-omics analysis of plasma metabolomics and proteomics further revealed that PFHpA exposure disrupts lipid metabolism, bile acid transport, and immune response pathways—biological mechanisms that were also mirrored in human liver spheroid models. Our results identified a proteome profile with approximately 600% higher odds of having MASLD, whereas a distinct metabolome profile was associated with a 49% reduction in the odds of MASLD. These findings indicate specific multi-omic profiles as high-risk factors for MASLD and underscore the crucial role of protein dysregulation in disease pathogenesis. Specifically, PFHpA exposure induced lipid accumulation in hepatocytes and activated PPAR-α signaling and downstream fatty acid oxidation pathways. Latent unknown clustering of integrated omics data (LUCID) identified distinct multi-omics profiles associated with elevated MASLD risk, highlighting a proteomic signature enriched in inflammatory and fibrotic markers. This translational framework, which integrates findings from in vitro spheroid studies with human data, can be applied to future research aimed at uncovering the molecular mechanisms of PFAS-induced liver disease and guiding the development of targeted prevention and treatment strategies for MASLD.

Among the eight PFAS congeners included in this study, only PFHpA was found to be associated with MASLD. We were particularly interested in PFHpA as it is understudied in the literature, and we found associations with MASLD (y/n) and MASLD disease severity (no disease/MASLD, MASH), and we observed a concentration-dependent relationship between PFHpA and MASLD. Notably, David et al. observed similar findings; PFHpA was the only association observed and was found to be associated with MASLD severity, including advanced steatosis and fibrosis59. The PFHpA-associated perturbed canonical pathways among Teen-LABS participants indicate the role of innate immunity and inflammatory markers, and when we integrate the human study with the spheroid study, we see the importance of inflammation and dysregulation of lipid-related pathways in disease progression. The PWAS analysis in Teen-LABS highlights the dysregulation of pathways related to innate immunity by PFHpA, showing enrichment of inflammatory cytokines, including Interleukin-6 (IL-6), Interleukin-1 (IL-1), Interleukin-10 (IL-10), Interleukin-33 (IL-33), and Toll-like receptor (TLR) signaling pathways. In addition to promoting inflammation, these pathways have important metabolic effects on lipid metabolism60,61,62,63,64. Activation of the innate immune system plays a crucial role in initiating and intensifying liver inflammation in MASH65. Multiomic integration using LUCID revealed that proteins drive a high-risk MASLD profile. These results may be particular to our cohort, as it is comprised of youth with obesity. While we do see that nearly 60% of the cohort has been diagnosed with MASLD, the large majority of participants had Grade 1 steatosis, reflecting an earlier stage disease66. The LUCID results highlighted the role of chronic inflammation in the etiology of PFHpA exposure and MASLD. Studies have shown that several proteins included in the high-risk profile play a role in the development of fibrosis, a late stage of liver disease progression seen in MASLD, including carbonic anhydrase67,68, coagulation factor VII69,70,71, and hyaluronidase72,73,74,75. In particular, hyaluronidase, an enzyme that breaks down hyaluronic acid, is relevant, as increased hyaluronic acid can contribute to the degradation of the extracellular matrix, which may affect liver fibrosis progression72,73,74,75. Although a direct link between complement 2 (C2) and MASLD has not yet been firmly established, the involvement of the complement system in inflammation and immune responses suggests that C2 may play a role in disease’s progression76,77,78. Chronic inflammation and immune-mediated liver injury, both influenced by complement activation, are key aspects of MASLD pathogenesis77,78,79,80. Because the liver is a primary site of complement protein synthesis, progression of MASLD may influence hepatic production of complement components, thereby modifying innate immune responses. This dynamic may contribute to the chronic inflammation and immune dysregulation observed in MASLD. Further research is needed to clarify the role of these mechanisms in MASLD and determine whether they could serve as therapeutic targets or biomarkers.

To delve deeper into the molecular mechanisms underlying PFHpA-associated MASLD, we conducted in vitro experiments to examine how PFHpA affects liver metabolism, using a human 3D liver spheroid co-culture model. This model has been previously validated for hepatotoxicity screening using over 150 FDA-approved small-molecule drugs, consistently demonstrating high predictive accuracy for human outcomes81. Our study used single-cell transcriptomics to evaluate the impact of PFHpA on hepatic cell populations in liver spheroids. Our findings reveal that PFHpA primarily affects hepatic lipid metabolism, leading to a notable upregulation of lipogenesis pathways in human primary hepatocytes, which was reflected by the significant lipid accumulation observed after PFHpA exposure. Among the pathways involved in lipid metabolism in hepatocytes, the “Regulation of lipid metabolism by PPAR-α” and “Fatty Acid β-oxidation I” pathways showed the strongest activation. This suggests that PFHpA promotes hepatocyte lipogenesis likely through peroxisome proliferator-activated receptor (PPAR)-α signaling. PPARs are members of the nuclear hormone receptor superfamily that act as ligand-activated transcription factors82. In the liver, PPAR-α plays a crucial role in regulating fatty acid oxidation and lipid and lipoprotein metabolism83. Previous research has shown that various PFAS, including PFHpA, can activate PPAR-α in cell lines and animal models84,85,86. Recently, Yang et al. proposed that the hepatic lipid metabolism disruption caused by PFOA and PFOS depends on the PPAR-α/ACOX1 (acyl-CoA oxidase 1) axis56. Our in vitro analysis indicated that this pathway is disrupted in hepatocytes, but not in immune cells. We found that ACOX1 expression was upregulated and integrated the pathway “Regulation of lipid metabolism by PPAR-α” in hepatocytes from spheroids exposed to PFHpA. In contrast, while PPAR-α signaling was also upregulated in T cells exposed to PFHpA, ACOX1 was not differentially expressed, and no lipid biosynthesis pathways were detected (Fig. S2). Our findings indicate that PFHpA alter signaling of PPAR-α/ACOX1 axis in hepatocytes but not T cells, to instigate abnormal hepatic lipid metabolism in humans. Once steatosis is established, it can induce lipotoxicity, oxidative stress, and cellular injury, which contribute to the activation of pro-inflammatory cells and ultimately promote hepatic inflammation87.

Our multi-omics analysis provides a significant insight into the biological mechanisms linking PFHpA exposure to metabolic dysfunction and steatotic liver disease. In particular, the high-risk MASLD group was characterized by consistent reductions in amino acids (e.g., tryptophan, proline, glycine), purine metabolites (e.g., uric acid, hypoxanthine), and bile acids (e.g., glycocholic acid, glycochenodeoxycholic acid). These perturbations suggest impairments in key hepatic and systemic processes, including mitochondrial function, hepatic detoxification, and gut-liver axis communication—each of which plays a critical role in maintaining metabolic homeostasis88,89,90,91,92,93,94. Reduced levels of tryptophan and its microbiome-derived metabolite indole-3-acetate may indicate altered gut microbial signaling and impaired immune regulation95. Concurrently, lower concentrations of glycine and eicosapentaenoate suggest diminished anti-inflammatory capacity, which may increase susceptibility to hepatic injury96,97. The observed downregulation of purine and urea cycle intermediates (e.g., uric acid, sarcosine, beta-alanine) is consistent with impaired nitrogen handling and oxidative stress response in the liver98. These findings are further supported by transcriptomic data from PFHpA-exposed liver spheroids, which demonstrated dysregulation in pathways related to lipid metabolism, amino acid catabolism, and PPAR signaling. Proteomic profiling of the high-risk group—marked by increased expression of proteins involved in inflammation and metabolic stress (e.g., HYAL1, F7, and CA5A)—reinforces this narrative. Together, the integrated omics approach reveals a coherent pattern of hepatic and systemic dysregulation that underpins MASLD pathogenesis in the context of environmental chemical exposure.

A few challenges and limitations are worth noting, although we adjusted for several key sociodemographic variables, we acknowledge the potential for residual confounding form unmeasured factors such as diet, physical activity, and co-exposure to other environmental toxicants. Future studies with more detailed exposure and lifestyle data will be important to further disentangle these relationships. The relatively small sample size of the Teen-LABS cohort may limit our ability to detect small but biologically meaningful associations, particularly in high-dimensional omics analyses. Additionally, while our PWAS and MWAS involved a large number of comparisons, we did not apply formal multiple testing correction to individual analytes, as our aim was to identify broad pathway-level signals rather than definitive biomarker associations. This exploratory framework prioritizes biological interpretability and cross-validation across omics layers. Another challenge of utilizing high-dimensional datasets is that they carry an inherent risk of model overfitting, especially when the number of variables far exceeds the number of observations. To mitigate this, we employed rigorous quality control measures, focused on biologically informed pathway-level analyses, and integrated results across omic layers to enhance robustness. While our MWAS leveraged an untargeted metabolomics platform, the final analysis was restricted to confirmed endogenous metabolites using the laboratory’s internal reference library. This introduces a semi-targeted aspect to our study, which may bias findings toward more well-characterized compound and potentially overlook novel or exogenous metabolites that could also be relevant to PFHpA exposure or MASLD pathogenesis. The proteomic and metabolomic datasets differed in their design and inclusion criteria. Proteomic profiling was performed using the Olink® Explore 384 Cardiometabolic and Inflammation panels, which provide targeted quantification of 731 proteins relevant to metabolic and inflammatory pathways—offering broad but predefined proteomic coverage. The differences in metabolomics and proteomics datasets should be considered when comparing the depth and scope of each omics layer. Future validation in larger, independent cohorts is needed to confirm our findings as well as studies that incorporate broader metabolite coverage to capture the full spectrum of exposure-related metabolic perturbations. Additionally, future studies—particularly those incorporating experimental components—should aim to evaluate the effects of PFAS mixtures and define population-relevant mixture profiles in the context of MASLD and its progression. Addressing PFAS mixtures in translational research remains a significant challenge, especially when bridging experimental findings with human epidemiological data.

The Teen-LABS cohort consists of adolescents with obesity undergoing bariatric surgery, a highly specific clinical population that may differ from the general adolescent population. The transportability of these findings to all adolescents, is therefore, limited. Additionally, while the cross-sectional design of our study limits our ability to infer causality or rule out reverse causation—whereby MASLD may influence PFHpA metabolism or retention—we address this limitation by integrating transcriptomic data from PFHpA-exposed liver spheroids which also strengthens the biological plausibility of the role of PFHpA in MASLD severity and progression and it supports the translational value of the findings within this high-risk population. This experimental evidence, which reveals PFHpA-induced gene expression changes consistent with MASLD-related pathways, enhances the biological plausibility and supports the potential role of PFHpA as a contributing factor in MASLD severity and progression. While the integration of human and in vitro data provides mechanistic insights into the PFHpA-associated MASLD relationship the constraints of the integration should be considered. The liver spheroid model reflects an acute, liver-specific response to PFHpA in a controlled environment, which differs from the chronic and multifactorial exposures experienced by adolescents with obesity in the Teen-LABS cohort. Additionally, systemic physiological processes, immune interactions, and other organ systems are not captured in this model. Therefore, while overlapping pathways between models support a hepatic role in MASLD development, these findings should be interpreted within the context of each system’s inherent constraints.

Our study has a number of strengths, including the integration of human epidemiologic data which includes histologically determined MASLD with in vitro liver spheroid providing great novelty to our study. We developed a translational research framework that bridges population-based and experimental approaches, enabling identification of molecular signatures and high-risk profiles associated with PFHpA and MASLD risk. The integrative approach not only advances our understanding of PFHpA-related liver toxicity but also provides actionable insights to inform public health policies aimed at reducing PFAS-associated disease burden. In conclusion, while our study established a robust association between PFHpA exposure and MASLD in Teen-LABS adolescents, our study additionally elucidates key molecular pathways—particularly those related to innate immunity, inflammation, and lipid metabolism, highlighting potential therapeutic targets for MASLD.