A multifaceted workflow for advancing human risk assessment of perfluorooctane sulfonate (PFOS)

We applied an organ toxicity-oriented, sensitivity-led workflow to advance human risk assessment of perfluorooctane sulfonate (PFOS). In vitro responses across human molecular and cellular targets, focusing on EA.hy926 endothelial cells and liver spheroids, were quantified by benchmark concentration (BMC) modeling and translated to human equivalent doses (HEDs) using a physiologically based toxicokinetic approach. These HEDs were used to derive transcriptomic points of departure (PoDs), population-level exceedance percentiles, and matrix-specific threshold concentrations relevant to dietary and environmental exposure. Toxicogenomic HEDs were more sensitive than apical endpoints, with median values of 22.8 μg/kg body weight (bw)/day for endothelial cells and 29.2 μg/kg bw/day for liver spheroids. HEDs of 4-12 μg/kg bw/day produced bioactive tissue concentrations predicted to elicit gene-level responses in both targets and engage diverse molecular and biological processes. Fitted HED distributions yielded 5th percentile PoDs of 2.76 and 5.69 μg/kg bw/day for endothelium and liver, respectively. Combining these distributions with serum PFOS levels for general and occupational cohorts indicated negligible EP50 effect magnitudes in the general population and most occupational groups, while one occupational cohort showed pronounced endothelial sensitivity and a moderate liver response. Matrix-specific comparisons indicated that soil+dust, fish, and some meats exceeded PoD-based predicted PFOS concentrations for endothelial and liver responses, whereas most dairy and many produce measurements were below response-relevant ranges. This probabilistic, HED-anchored framework provides a transparent bridge from in vitro transcriptomics to population-level risk screening for PFOS and is transferable to other per- and polyfluoroalkyl substances where toxicokinetic and in vitro data are available.