Water consumption contributes to maternal PFAS exposure: From source to metabolic perturbations.

Per- and polyfluoroalkyl substances (PFAS) are widespread persistent pollutants and pose a risk to human health. However, the transfer efficiencies (TEs) of PFAS from source water into maternal serum, and its metabolic pathways linking PFAS exposure and human disease are remain unclear. Here, we present an integrative study combining multi-water and serum PFAS analysis, lipidomics, and machine learning to decipher TEs and PFAS-metabolism interactions. Analyzing 30 PFAS across source water (N = 18), tap water (N = 18), water during drinking water treatment process (DWTP, N = 12), and serum of pregnant women alongside 297 lipid species, we developed a SHAP (Shapley Additive Explanations) framework to quantify PFAS origins and serum-specific lipid responses. We found PFAS were prevalent in source water samples, with PFBA, PFPeA, 6:2 FTS, PFOA, and PFOS being predominant chemicals. In addition, DWTP may not be effective in removing PFAS, with most of target chemicals tested exhibiting removal below 50 %. Moreover, source water outperforms tap water in predicting maternal serum PFAS concentrations, suggesting source water contamination reflects maternal exposure more directly. Machine learning further showed that maternal serum lipid metabolism was influenced by PFAS, such as HFPO-DA and N-MeFOSAA mainly disrupted glycerphospholipid homeostasis (e.g., PC and LPC), highlighting risks to maternal metabolism. Our findings pioneer AI-driven tracing of PFAS transfer dynamics and lipidomic disruptions, warranting a comprehensive strategy beyond drinking water to handle water PFAS contamination.