Integrated pan-cancer analysis reveals a cancer-associated fibroblast oxidative stress response signature predicting immunotherapy response and prognosis.

Oxidative stress plays a significant regulatory role in tumor immune responses and can influence the efficacy of immunotherapy. Accordingly, therapeutic interventions targeting oxidative stress-related mechanisms, whether alone or in combination with other modalities, represent a compelling strategy to enhance cancer therapy. In this study, we first profiled the landscape of oxidative stress responses within the tumor microenvironment by integrating pan-cancer single-cell RNA sequencing datasets, which revealed that cancer-associated fibroblasts (CAFs) possessed the highest oxidative stress response score. Based on this finding, we developed a fibroblast-derived oxidative stress-related signature (FOSR.Sig) by screening for genes most correlated with oxidative stress responses in CAFs. Furthermore, with a machine learning framework, our model achieved exceptional accuracy in predicting ICI response, and its robustness was subsequently validated. Importantly, a prognostic model incorporating the FOSR.Sig was developed using TCGA pan-cancer datasets and LASSO regression analysis, which provides novel prognostic biomarkers applicable across diverse cancer types. Mechanistic investigation of TFG, the top risk score gene, revealed its critical role in the tumor microenvironment through comprehensive in vitro and in vivo experiments and RNA-seq assays. Our study highlights the therapeutic potential of targeting oxidative stress in cancer-associated fibroblasts as a novel strategy to empower antitumor immunity and prevent immune escape. And provide a promising powerful tool for predicting responses to tumor immunotherapy and patient outcomes.