Abstract

Background Ferroptosis plays a critical role in immune regulation and tumor microenvironment remodeling. However, its therapeutic potential in enhancing immune checkpoint inhibitor (ICI) efficacy remains incompletely understood and warrants further investigation. Methods To investigate the potential of ferroptosis in improving ICI response, we constructed a machine learning-based predictive model using ferroptosis-related genes and analyzed large-scale single-cell RNA sequencing datasets. Mechanistic studies were performed to examine the role of interferon (IFN)-γ signaling in ferroptosis sensitization, including functional validation in vitro and in vivo. Lipidomic, transcriptomic, chromatin Immunoprecipitation sequencing (ChIP-seq) and Cleavage Under Targets and Tagmentation analyses were employed to dissect downstream pathways, focusing on IRF1 and AGPAT3. Results Our model successfully predicted ICI response based on ferroptosis-related gene signatures, identifying IFN-γ as a key enhancer of ferroptosis sensitivity in tumor cells. IFN-γ treatment induced activation of the transcription factor IRF1, which in turn upregulated AGPAT3 expression, driving lipid remodeling and accumulation of polyunsaturated ether phospholipids. This lipid remodeling significantly increased tumor cell susceptibility to ferroptosis and enhanced ICI efficacy. Loss of AGPAT3 impaired IFN-γ-mediated tumor elimination both in vitro and in vivo. Clinically, higher AGPAT3 expression in tumors was associated with increased immune activation and improved overall survival in ICI-treated patients. Conclusion The IFN-γ-IRF1-AGPAT3 axis represents an important antitumor mechanism that promotes ferroptosis. Targeting this pathway in combination with our ferroptosis-driver model prediction may improve ICI efficacy and patient outcomes.