Toxicity Assessment of Mixed Exposure of Nine Perfluoroalkyl Substances at Concentrations Relevant to Daily Intake

Abstract

Per- and poly-fluoroalkyl substances (PFAS) exhibit high persistence in the environment and accumulate within the human body, warranting a thorough assessment of their toxicity. In this study, we exposed mice (male C57BL/6J mice aged 8 weeks) to a composite of nine PFAS, encompassing both long-chain PFAS (e.g., perfluorooctanoic acid and perfluorooctanesulfonic acid) and short-chain PFAS (e.g., perfluorobutanoic acid and perfluorobutanesulfonic acid). The exposure concentrations of PFAS were equivalent to the estimated daily human intake in the composition reported (1 µg/L (sum of the nine compounds), representing the maximum reported exposure concentration). Histological examination revealed hepatocyte vacuolization and irregular hepatocyte cord arrangement, indicating that exposure to low levels of the PFAS mixture causes morphological changes in liver tissues. Transcriptome analysis revealed that PFAS exposure mainly altered a group of genes related to metabolism and chemical carcinogenesis. Machine learning analysis of the liver metabolome showed a typical concentration-independent alteration upon PFAS exposure, with the annotation of substances such as glutathione and 5-aminovaleric acid. This study demonstrates that daily exposure to PFAS leads to morphological changes in liver tissues and alters the expression of metabolism- and cancer-related genes as well as phospholipid metabolism.

Keywords: environmentally relevant exposure; liver toxicity; multi-omics analysis; poly-fluoroalkyl substances.

Figure 1

Histology of the liver determined via hematoxylin and eosin (HE) staining of samples in the control, per- and polyfluoroalkyl substances (PFAS)-Low (100 ng/kg/day), and PFAS-High (5000 ng/kg/day) groups. Diffuse vacuolation and hypertrophy of hepatocytes with eosinophilic granules were observed in all mice in the PFAS-Low and -High groups (N = 5, each). Representative images of livers stained with periodic acid–Schiff (PAS) stain with/without amylase digestion. Scale bars = 100 μm.

Figure 2

Transcriptome analysis of the liver. (A) Shows the number of upregulated and downregulated genes based on the comparison pair’s fold change (FC). (B) The high-expression similarities were grouped together using each sample’s normalized value. (Distance metric = Euclidean distance; Linkage method = Complete Linkage.) (C) Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis between the PFAS-Low and control groups. (D) KEGG enrichment analysis between the PFAS-High and control groups. ***; p < 0.001 detected by false discovery rate-adjusted t-tests.

Figure 3

Real-time quantitative polymerase chain reaction. Representative genes of pathways with high variability in RNA sequencing were quantified (N = 5, each group). The y-axis shows the gene expression value relative to the control using β-actin (Actb) as the calibration gene. The data are represented as means ± SEM (n = 5). All data were measured in four replicates for each mouse. Asterisks (*) represent significant differences compared with control using Tukey–Kramer’s HSD test—*; p < 0.05, **; p < 0.01, and ***; p < 0.001.

Figure 4

Metabolomics of the liver. (A) Heatmap of all 189 metabolites detected in the liquid chromatography–mass spectrometry analysis and hierarchical clustering. (B) Quantitative enrichment analysis compared with PFAS-High or PFAS-Low and control groups (N = 5, each). (C) An example of a decision tree in a random forest classifier. (D) Top 10 features of importance for the random forest classifier (11 compounds were tied for 10th place).

Figure 5

(A) Theoretical binding pose obtained via molecular docking simulation for peroxisome proliferator-activated receptor (PPAR) with perfluorooctanoic acid or their known ligands. The protein is shown in ribbon representation with the binding residues shown in stick representation, and orange dashed lines in the 3D diagram indicate ligand–protein interactions. These interactions are visualized as 2D figures using LigPlot+. (B) Regression of docking scores and carbon chain length of PFAS. The known ligands are indicated by filled shapes. Linear regression was performed using GraphPad Prism 9. The dashed lines represent the 95% confidence interval of the regression shown as colored lines. The docking scores are shown in Table S4.

Figure 6

Plasma concentration of hepatic biomarkers, including ALP (alkaline phosphatase), AST (aspartate aminotransferase), and cholesterol in plasma (N = 5, each group). These concentrations were measured using colorimetric methods. The data are graphically represented as box-and-whisker plots, where the boxes represent each quartile along the median, + indicates the mean value, and the whiskers extend to the maximum and minimum values. All data points were measured in duplicate. * denotes significant differences detected using Tukey–Kramer’s HSD test (p < 0.05).