CELL INTELLIGENCE · 肿瘤细胞治疗研究

The immune microenvironment influences the sensitivity of patients to radiotherapy (RT), yet determinants of therapeutic resistance remain elusive. This study integrates single-cell transcriptomics and machine learning to delineate immune predictors of RT outcomes. Comprehensive analysis reveals reduced epithelial cell numbers, accompanied by enhanced apoptosis, complement activation, and inflammatory responses. RT triggers macrophage accumulation, particularly an RT-responsive M1-like HSPA1B+ subset with elevated antigen-presenting capacity. While T and NK cell cytotoxicity increases, their exhaustion markers (e.g., PDCD1, TIGIT) are exacerbated. CellChat analysis identifies robust epithelial-myeloid crosstalk mediated by the C3/C3AR1 axis. In murine models, C3AR1 antagonism diminishes RT efficacy, impairing macrophage infiltration and M1 polarization. Leveraging 25 single-cell-derived immune features, an 8-feature multilayer perceptron model: Cervical Cancer Radiotherapy Immune-Response Model (CCRTIM) is developed. CCRTIM robustly predicts prognosis (AUC = 0.76) and exhibits risk stratification. These findings unveil dynamic immune remodeling post-RT and establish actionable biomarkers for precision radiotherapy strategies.