Abstract
Background: Radiotherapy resistance in breast cancer remains a major clinical challenge. The key molecular determinants and cellular populations driving this resistance are not fully understood. Methods: A radiotherapy resistance (RR) gene panel was identified from TCGA-BRCA and GSE120798 cohorts. Single-cell and spatial transcriptomics characterized RRhigh epithelial cells (RRhighepi). A prognostic model, named SuperPC and StepCox-based Radiotherapy Resistance model (SSRR), was built via machine learning and Mendelian randomization. Functional roles of Prolyl 4-Hydroxylase Subunit Alpha 2 (P4HA2) were validated in vitro. Results: The RR gene panel was upregulated in tumors and enriched for cell cycle pathways. RRhighepi cells exhibited elevated stemness, activated cell cycle and metabolic programs, and enhanced DNA damage repair. RRhighepi represented a developmental origin and communicated with endothelial cells. The SSRR model stratified patients into high-risk groups with poorer survival and distinct therapeutic responses. P4HA2, a key model gene, was upregulated in multiple cancers. P4HA2 knockdown suppressed proliferation, invasion, and colony formation, and synergized with radiotherapy to reduce stemness and enhance DNA damage. WGCNA confirmed co-module membership of P4HA2 and the RR panel. Conclusions: This study, through multi-omics analysis, proposes a potential mechanistic model associated with radiotherapy resistance in breast cancer. P4HA2 is a potential therapeutic target that sensitizes breast cancer to radiotherapy. The RR gene panel and SSRR model provide insights into resistance mechanisms and prognostic stratification.