Abstract:
Per- and polyfluoroalkyl substance (PFAS) are man-made chemicals that are widely used in commercial, industrial and military products. PFAS are ubiquitously found in environmental media and organisms including human beings due to their persistence and bioaccumulation potential. Many studies have shown the correlation of PFAS exposure and diseases. There are more than 14,000 PFAS chemicals in the CompTox Chemicals Dashboard but only fraction of them have toxicological information. Cell-based high-throughput screening (HTS) bioassays may inform risk assessment of large numbers of PFAS provided that quantitative in vitro to in vivo extrapolation (QIVIVE) can be developed. Challenges of QIVIVE lie in the predictive accuracy of the bioavailability of PFAS, species difference, as well as the specificity of cellular responses that may lead to potential adverse outcomes.
This thesis aimed to evaluate the toxic effects and exposure of PFAS in cell-based bioassays to facilitate the use of in vitro bioassay data for risk assessment. The first objective was to measure in vitro exposure and determine the binding constants to plasma proteins of different species. The bioavailability of PFAS can be presented as the free concentrations (Cfree) in bioassays and plasma. PFAS bound to proteins and lipids usually result in lower Cfree compared to their nominal (i.e. dosed) concentrations (Cnom). Solid-phase microextraction (SPME) combined with liquid chromatography mass spectrometry (LCMS) was used to measure the Cfree of PFAS among biomaterials, including bioassay media, cell homogenates and blood plasmas. The binding constants were derived from different binding models. Binding isotherms of 16 PFAS with human and trout (fish) plasmas were compared. Anionic PFAS showed higher binding affinities to human plasma in the low concentration ranges compared to the trout plasma, because there were more proteins in human plasma, which led to very strong and specific binding of PFAS with proteins. Partitioning of PFAS to plasma was also predicted correctly by mass balance models (MBMs) that were parameterized with the protein-water and lipid-water binding constants (chemical characteristics) as well as the protein and lipid contents of the plasma (species characteristics). The second objective was to inform risk assessment with a simple form of QIVIVE, which is the ratio of in vivo human plasma concentrations (Cplasma) and in vitro cell-based effect concentrations (EC), either based on nominal or free concentrations. The Cfree can be measured experimentally or predicted by validated MBMs. Cnom,plasma of PFAS were collected from literature and Cfree,plasma were predicted by the MBMs parameterized with plasma binding constants. A cell-based reporter gene assays targeting peroxisome proliferator-activated receptor gamma (PPARγ) was selected to measure the effect concentrations ECnom or ECfree of PFAS, because PPARs have been shown to be specific targets of PFAS. QIVIVEfree ratios, which are the ratios between freely dissolved concentration of PFAS in blood and their ECfree, of some hydrophobic PFAS were up to 1000 times lower than their corresponding QIVIVEnom ratios. This was caused by a strong affinity to proteins and human plasma contained 50 times more proteins than in the bioassay medium, leading highly specific binding at low PFAS concentrations in human plasma, contrasted by nonspecific partitioning to proteins at high concentrations that were required to trigger an effect in bioassays. The proteins and lipids in plasma may act as reservoirs of PFAS in human bodies that pose a risk of chronic exposure. The case study using PPARγ is a demonstration of the importance of using Cfree for the QIVIVE ratios, but for a comprehensive risk assessment, a large set of specific responsive in vitro cell-based bioassays needs to be applied. The third objective was to identify the specific from nonspecific effects among different cell-based HTS bioassays, where baseline toxicity can be a reference. Baseline toxicity is the minimal toxicity that is caused by nonspecific accumulation of chemicals in the cellular membranes. Exceedance of the critical membrane burden leads to cell death. Separate baseline toxicity prediction models were developed for anionic PFAS and neutral chemicals, which were used to define the specificity of cell response of 30 PFAS on six target effects (activation of PPARγ, aryl hydrocarbon receptor, oxidative stress response, and neurotoxicity in own experiments, and literature data for activation of several PPARs and the estrogen receptor). HFPO-DA showed high specificity for PPARs, while the majority of PFAS acted as baseline toxicants. This implicates a heightened need for the risk assessment of PFAS mixtures, because nonspecific effects, i.e., baseline toxicity, behave concentration-additive in mixtures.
In brief, QIVIVE is suggested to consider the bioavailability using Cfree. It will not always be necessary to measure Cfree of PFAS but existing data of plasma concentration and bioassay effect based on Cnom can be converted to Cfree by using MBMs. The differences of plasma binding can be predicted by amounts of proteins and lipids in plasmas, which is influenced by many factors not only between species but also between individuals, and also depends on the concentration range because specific binding at low concentrations is very strong while nonspecific binding at higher concentration has lower binding constants. The baseline toxicity prediction model may be able to re-evaluate the specificity of existing bioassay results, as well as provide a testing strategy in future studies. The identification of high specificity of targets may advance the development of adverse outcome pathways related to single PFAS, which may also be applied for PFAS mixtures. The combined assessment in vitro cellular responses of PFAS and organism exposure levels (e.g. human and fish plasma) can facilitate the comprehensive human and environmental risk assessment of PFAS in the near future.
Plasma binding