You are here:
Applying High Throughput Toxicokinetics (HTTK) to Per- and Polyfluoro Alkyl Substances (PFAS)
Citation:
Wambaugh, J., R. Tornero-Velez, M. Devito, C. Lau, AND B. Wetmore. Applying High Throughput Toxicokinetics (HTTK) to Per- and Polyfluoro Alkyl Substances (PFAS). SOT, Salt Lake City, UT, March 10 - 14, 2024. https://doi.org/10.23645/epacomptox.25408537
Impact/Purpose:
N/A
Description:
Background and Purpose: Per- and polyfluoro-alkyl substances (PFAS) are a large and diverse class of organic chemicals in which all (per-) or some (poly-) carbon–hydrogen bonds have been replaced with carbon–fluorine bonds. PFAS are commonly found in human tissues. Some PFAS have been noted as having long toxicokinetic (TK) half-lives (up to several years in humans) while having comparatively short TK half-lives in toxicological model animal species. Typical allometric extrapolation methods for TK parameters of PFAS are unreliable between species and chemicals. Further, PFAS have both hydrophobic and lipophobic properties which complicates many typical TK methods that rely upon octanol:water chemical concentration ratios. There are only roughly a dozen PFAS with human in vivo measured TK half-life measurements, while new approach methodologies (NAMs) are generating data to inform toxicity for many more chemicals. Higher throughput TK (HTTK) approaches are needed to aid in the interpretation of the data from these NAMs. R package “httk” makes use of in vitro measures of toxicokinetics (hepatocyte clearance – Clint – and fraction unbound in plasma – fup) for in vitro-in vivo extrapolation (IVIVE). The HTTK tools were largely developed for typical organic chemicals – here we examine when and how well they work for PFAS. Methods: We have added new chemical-specific in vitro data for Fup and Clint for PFAS measured in Smeltz et al. (2023), Kreutz et al. (2023), and Crizer et al. (in preparation). Because many of the new chemicals are volatile, we have added new HTTK model '3compartmentss2' that solves for steady-state plasma/blood concentration resulting from elimination by metabolism, renal excretion, and exhalation. We have further added new model 'pfas1compartment' that uses the Dawson et al. (2023) machine learning (ML) model to parameterize an empirical one compartment model for PFAS chemicals. Membrane affinity – a measure of how likely a chemical is to partition into lipid bilayers – is an important property for predicting tissue partitioning in HTTK. The underlying methods of HTTK were modified for PFAS since it has been shown that they tend to have membrane affinities as much as a hundred times higher than would be expected for other organic chemicals. Combining revised membrane affinity with the newly measured fup values allows prediction of tissue-specific partition coefficients. Results: The new in vitro data allow HTTK predictions for 120 PFAS, while the ML model predictions expand HTTK to 4136 PFAS. When compared to in vivo clearances, HTTK shows wide uncertainty, while the ML model reproduces the data on which it was trained. Very recently new TK half-life estimates were developed for several PFAS, including five novel PFAS outside the ML model training set – where predictions and measurements are available for the same chemicals, the ML model has been consistent with observations. For PFAS the ratio of in vivo clearance in different species to clearance in humans ranges widely. HTTK tends to underestimate this inter-species scaling ratio, while the ML model errors are more evenly distributed. For the PFAS with NAM data but no in vivo TK data, HTTK predicts that the rat to human scaling factor for toxicokinetics is typically within a factor of 10, while the ML model predicts that the factor could be several thousand times for some chemicals. Conclusions: HTTK methods are needed for PFAS chemicals to relate NAM toxicity data to real world exposures. Due to some of the unique aspects of PFAS, the HTTK approaches for organic chemicals must be somewhat modified. New, chemical-specific human in vitro TK data have informed predictions of route of elimination and tissue distribution. However, more empirical ML approaches may be needed for inter-species extrapolation. This abstract does not necessarily represent the views or policies of the U.S. Environmental Protection Agency.
URLs/Downloads:
DOI: Applying High Throughput Toxicokinetics (HTTK) to Per- and Polyfluoro Alkyl Substances (PFAS)
POSTER.PDF (PDF, NA pp, 894.958 KB, about PDF)