Advancing understanding of terrestrial receptor exposure to mixtures of per- and polyfluoroalkyl substances

Abstract

Per- and polyfluoroalkyl substances (PFAS) are synthetic molecules that are generally defined by their carbon-fluorine bond(s). PFAS are used in a wide array of commercial and household products and have been in production and use around the earth for decades. Their desirable features: stability and surfactant properties, are also key drivers in their emergence as globally widespread environmental toxicants. As such, they present complex issues in estimating exposure and effects in ecological risk assessments at individual sites or screening assessments across many sites. Further, PFAS commonly occur in products and environmental media in complex mixtures that increase uncertainty in assessments. This dissertation aims to improve the understanding of PFAS mixture exposure in terrestrial ecological receptors to inform ecological risk assessment. The first question addressed is “what is the mixture of PFAS that is representative of PFAS mixtures in surface soil?” The underlying hypothesis explored is simply the hypothesis that there is such a mixture that represents surface soil PFAS concentrations from aqueous film-forming foam (AFFF) use sites. To address this question, we defined a representative mixture as made of highly sampled PFAS, from high concentration sites, and contributing to more than 90% of the sum PFAS concentration. These metrics were generated at several scales given the nested structure of the data, but the final interpretation is based on the 95th percentile samples on a site-specific basis. This arrangement captures the most potential for risk at the smallest spatial extent. The analysis demonstrated that mixtures of PFAS in surface soil on sites within military installations are of a predictable complexity. Only three PFAS are needed to represent over 90% of sum PFAS concentration at AFFF use sites. Further, perfluorooctane sulfonic acid (PFOS) was consistently the most prevalent PFAS with the highest proportional contribution to sum PFAS. The most prevalent second and third ranked PFAS were perfluorohexane sulfonic acid (PFHxS) and perfluorooctanoic acid (PFOA), respectfully. Across all sites, the median PFHxS:PFOS ratio was 0.10 and PFOA:PFOS was 0.03. These observations indicate that PFAS mixtures in surface soil on military installations can be represented by a relatively simple mixture of PFAS largely dominated by a single PFAS. The resultant conclusion is that risk assessments at sites can exclude minority PFAS and that toxicological studies to inform risk assessments can be simplified. The second questions addressed are “what are mammalian tissue concentrations when exposed to PFAS mixtures?” and “does the mixture matter when predicting tissue concentrations?” The specific PFAS mixtures of interest are those identified as representative of surface water and surface soil on military installations. The hypothesis was that exposure and potential effects observations would indicate additive relationships, but the main objective was to quantify tissue concentrations as a function of dose while considering or ignoring the mixture. To explore the hypothesis and quantify tissue concentrations, an outbred mouse was exposed to single PFAS and mixtures of PFAS. A true whole body extraction concentration was measured and in a second study, serum, liver, kidney, and brain concentrations were measured. One of the mixtures was representative of surface water and contained PFOS, PFHxS, perfluorohexanoic acid (PFHxA), and PFOA while the second mixture was representative of surface soil and contained PFOS, 6:2 fluorotelomer sulfonate (FTS), and 8:2 FTS. Mixture treatments were at varying concentrations, but a constant proportion of sum profile. The sum of the highest mixture treatment approximately overlapped the corresponding single PFAS exposure treatments’ concentrations. Interpreted in a hierarchical modeling framework, the results of this study indicate that prediction of tissue concentrations does not require knowledge of the mixture as interactive relationships between PFAS are less impactful than variation across individual mice. This simplifies models required to predict tissue concentrations given dose information, but is also indicative of dose additivity and relative tissue affinity. Further, we identified relative liver weight increases and increases in serum alanine aminotransferase (ALT). These responses can be effectively predicted by a relative potency factor (RPF) approach. Each PFAS’s RPF is relative to PFOS. The success of the RPF approach is a further signal of dose additivity in effects in addition to additivity in exposure. Our conclusions are largely highlighting the potential increased efficiency of exposure and effect prediction using additive models that do not require knowledge of the mixture. This is a key addition to risk assessments on sites where mixtures of PFAS will be observed, but regulations will generally be focused on single PFAS. The demonstration of relative affinity and potency also allows for relative ranking of PFAS in their potential for exposure and effects. So, while the list of PFAS addressed here is from a narrow applicability context, there is concurrence between our observations and the literature that legacy PFAS likely have similar potential for exposure and effects. Those PFAS with reduced potential for exposure and effects are generally smaller PFAS or those PFAS that appear to have excretion pathways. The third questions addressed is “how do mixture profiles change with increasing trophic levels?” and “can toxicokinetic data inform estimates of potential for trophic transfer?” The underlying hypothesis is that the two food chains studied—soil to worms to toads and soil to plants to rabbits—would have differing PFAS profiles in the diet, but knowledge of the internal kinetics would be required to explain the PFAS profile in the upper trophic level consumers. To explore these questions, two studies were performed. In the first study, adult, terrestrial lifestage toads were fed a diet of worms that had been grown in PFAS-spiked soil and their liver and pooled remaining tissue PFAS concentrations were quantified. In the second study, adult rabbits were fed a diet of plants that had been spiked to match concentrations in plants grown in PFAS-spiked soil. Rabbit livers, kidneys, and muscle tissue PFAS concentrations were quantified. Toxicokinetic models were fit on a tissue- and PFAS-specific basis for both organisms. In the toads, a differential equation system was also developed to evaluate the utility of physiological realism. The uptake and elimination rates from the definitive per-PFAS models of the estimated whole animal concentrations were used to generate trophic transfer coefficients (TTCs). Interpretation of the TTCs along with the toxicokinetic data suggests that some PFAS (i.e. PFOS) are consistently likely to be trophic magnifiers regardless of food chain, while others (i.e. perfluoroheptane sulfonic acid (PFHpS), 8:2 FTS) vary from trophic magnifiers to trophic diluters depending on food chain. The downstream conclusions, in context to field data in literature and the above observations of additivity, are that food chain specific trophic transfer data may be needed to predict upper trophic level exposure levels, but that, for PFOS specifically, uniform potential for trophic magnification is likely. A further observation is that kinetic data is informative towards relative estimates of potential for trophic magnification and explanatory towards general mechanisms. As an example, we note that elimination rate variation likely drives the difference in potential for trophic magnification in toads, while in rabbits, both uptake and elimination rates appear to be equivalent drivers of potential for trophic magnification. Overall, this dissertation identifies priority PFAS (Chapter 2), clarifies additivity of PFAS mixtures (Chapter 3), and advances the understanding of potential for trophic transfer in terrestrial food webs (Chapters 4 and 5). As a whole, these conclusions can simplify the data and approaches needed to perform an ecological risk assessment, support the desire to generalize across common PFAS, and provide laboratory confirmation of some field observations. Areas of improvement or continued research broadly fall under refining or increasing resolution. The representativeness of the soil and mouse mixtures are spatiotemporally static and are assumed to be consistently postprocessing of fate and transport dynamics. If this assumption were to be replaced by an understanding of spatiotemporally dynamics, site-specific risk assessments and exposure assessments could be of increased spatiotemporal resolution and increase their accuracy in regards to the where and when PFAS were released on the site and where and when receptors are present on the site. This refinement could influence the entire suite of study methods in this dissertation. Additionally, an increased number of PFAS analyzed is a desire of all PFAS studies. In this case, increased quantification of ‘ultra-short’ PFAS such as trifluoroacetic acid (TFA), volatile PFAS (i.e. alcohol head groups), and known degradation pathway member PFAS could increase the capacity to extend these observations to a greater range of PFAS, but also address concerns about missed biotransformations. In conclusion, this dissertation suggests that exposure to mixtures is not as complicated as feared.