Abstract
Esophageal cancer (ESCA) is a significant malignancy with rising global incidence rates and considerable impacts on patient survival and quality of life. Current diagnostic and therapeutic strategies face limitations, necessitating research into its underlying mechanisms and potential biomarkers for early diagnosis. This study aims to investigate the role of perfluorooctane sulfonate (PFOS), an environmental toxicant, in the development of ESCA through a comprehensive bioinformatics approach. Using the TCGA-ESCA dataset, we identified differentially expressed genes (DEGs) and intersected them with PFOS-related toxicity targets predicted via Comparative Toxicogenomics Database (CTD) and SuperPred. Machine learning (Random Forest, XGBoost, LASSO, SVM) were applied to prioritize core targets. Survival analysis, in vitro qPCR (ESO-26/FLO-1 cells), and molecular docking were performed. Immune infiltration and pathway activity (GSVA) were assessed. We identified 98 PFOS-related DEGs in ESCA, enriched in hypoxia response, epithelial migration, and cancer-associated pathways (e.g., AGE-RAGE, PI3K-Akt). Machine learning highlighted three core targets: PLAU, TOP2A, and BAX. High expression of these genes correlated with poor survival (PLAU, p = 0.047) and was upregulated in ESCA tissues. PFOS exposure significantly elevated their expression in esophageal cancer cells. Molecular docking revealed strong binding affinities between PFOS and core targets. GSVA linked PLAU/TOP2A/BAX to oncogenic pathways (angiogenesis, DNA repair), while immune analysis showed PLAU's association with stromal infiltration and TOP2A's negative correlation with CD8 + T cells. PFOS exacerbates ESCA by dysregulating PLAU, TOP2A, and BAX, which drive tumor progression via immune modulation, genomic instability, and oncogenic signaling. These targets may serve as biomarkers and therapeutic vulnerabilities for PFOS-associated ESCA, underscoring the need for environmental regulation and targeted therapies.