A machine learning approach to elucidating PFOS-induced alterations of repressive epigenetic marks in kidney cancer cells with single-cell imaging

Perfluorooctane sulfonate (PFOS) is a synthetic chemical and an environmental pollutant widely recognized for its industrial use, given its thermal and chemical stability. Growing evidence suggests that PFOS exposure is linked with epigenomic alteration and disease progression. Whether PFOS toxicity could lead to a phenotypical epigenetic state is of interest in toxicology research. Given the limited studies on kidney cancer and disease, the molecular mechanisms underlying the epigenetic modulated state upon PFOS exposure remains unclear. This study explores epigenetic indicators associated with PFOS exposure in A498 kidney cancer cells. We demonstrate a data-driven approach to identify critical features by classifying the epigenetic states upon PFOS exposure. By combining single-cell super-resolution imaging and machine learning analysis, we show that the repressive epigenetic markers are distinctively different upon PFOS exposure. Further we note that, PFOS exposure altered the spatial distribution of H3K9me3 and H3K27me3 with nanoscale imaging. Quantification of gene expression related to tumorigenesis indicates altered expression levels upon PFOS exposure. Our results suggest that the epigenetic changes partake in establishing PFOS-induced toxicity development. Furthermore, the distribution of H3K9me3/H3K27me3 alteration could act as one of the drivers for PFOS-induced phenotype development in disease progression.