UC Davis

An increasing numbers of trace organic contaminants (TrOCs) are being produced, used, and discharged into the environment. Over 17,000 pesticides and 4,730 per- and polyfluoroalkyl substances (PFAS) were reported by the Organisation for Economic Co-operation and Development (OECD) to be in circulation. Further TrOC diversity is generated when commercially produced compounds transform through myriad pathways. Understanding the public health impacts of environmental exposures to such a large number and wide structural diversity of TrOCs remains a significant challenge. Targeted chemical assays, the most common approach to bioaccumulation and risk assessment, provide a limited understanding of contaminant profiles in biological tissues and associated risks. More comprehensive analytical pipelines are needed. In this dissertation, I develop and apply new methods to assess and predict bioaccumulation and toxicity of complex mixtures of TrOCs, using in vivo, in vitro and in silico approaches. I present the three complementary strategies that together aim to improve management strategies for broad classes of TrOCs.First, I develop and demonstrate an in vivo screening approach to assess the bioaccumulation and risk of TrOCs in complex mixtures. I apply the approach to assess pesticide and PFAS bioaccumulation in edible insect larvae (i.e., H. illucens) reared on agricultural by-products (i.e., almond hulls). Rather than targeting specific TrOCs from the onset, as is typical in chemical risk assessment, my approach broadly screens substrates and organism tissues for TrOCs using large databases of known contaminants. Multiple chemical extracts, obtained using methods that capture TrOCs with a broad range of physiochemical properties, are subjected to one or more of four analytical pipelines (two operational modes each for liquid and gas chromatography coupled with time-of-flight mass spectrometry). Semi-quantitative analysis of the substrates screening revealed that bioaccumulative, persistent, and toxic chemicals are abundant in the agricultural by-products. Initial substrate screening was also used to guide targeted and non-targeted substrate and tissue analyses from in vivo bioaccumulation assays. Using this approach, I found that bifenthrin and a novel PFAS class bioaccumulate in the larvae tissue, initiating a pathway for contaminant transfer into the food chain. Second, I use in vitro and in silico approaches to assess bioaccumulation potential of PFAS in a commercially available class-B firefighting foam—aqueous film forming foams (AFFF). PFAS accumulation in serum and organ tissues are caused by PFAS precursor metabolisms and protein bindings. It is challenging to identify and quantitatively assess the contribution of each bioaccumulation pathway. However, mechanistic understanding in PFAS bioaccumulation is essential for prediction model and bioaccumulation potential assessment across the chemicals. In this dissertation, I collect binding data from equilibrium dialysis experiments using human serum albumin (HSA), the most abundant serum protein in human blood, and AFFF. This in vitro approach eliminates the influence of biotransformation to bioaccumulation and help discover ultra-strong binding or potentially covalent binding PFAS to HSA. I further use experimental data to test the effectiveness of molecular docking predicted HSA-binding and -nonbinding compounds in AFFF. The combination of in vitro and in silico approaches provide replicable, high-throughput workflows for assessing bioaccumulation potentials of chemicals in commercial products that are structurally diverse like PFAS. Third, I use in silico simulations validated in Part 2 to generate noncovalent binding scores of over 4,760 PFAS structures to human carrier and receptor proteins. In this dissertation, I hypothesize that protein binding scores generated from molecular docking can improve model performance for bioactivity/toxicity prediction of PFAS. Moreover, the prediction model can be used to predict bioactivity/toxicity of PFAS with similar chemical structure. I test this theory by correlating protein binding scores along with 45 other chemical descriptors to negative health endpoints using linear discrimination analysis and machine learning algorithms. This quantitative structure-activity relationship (QSAR) uses state-of-the-art knowledge to predict toxicity of PFAS with known structure. Altogether, the combination of experimental and modeling techniques provided value in assessing the bioaccumulation and toxicity of organic contaminants.