Integrated multi-omics analysis identifies ELF1 as a B-Cell-associated prognostic biomarker in breast cancer

Breast cancer progression is influenced not only by tumor-intrinsic alterations but also by interactions within the tumor immune microenvironment (TIME). We aimed to identify genetically regulated susceptibility genes associated with immune-related cellular states in breast cancer. We first performed a transcriptome-wide association study (TWAS) using FinnGen summary statistics to identify candidate genetically regulated genes associated with breast cancer susceptibility. Their transcriptional activity was subsequently mapped onto the GSE176078 single-cell atlas using AUCell, UCell, and AddModuleScore. Because B cells consistently showed the highest enrichment, we further applied hdWGCNA to characterize co-expression modules associated with the high TWAS score (HTS) state. We then integrated LASSO, Random Forest, Boruta, and XGBoost analyses to prioritize candidate hub genes. The prognostic and immunological relevance of ELF1 was further evaluated in GSE96058 (n = 3,273) and METABRIC (n = 2,509). Functional validation was further performed using ELF1-modulated B-cell models, including Raji and GM12878 cells, together with direct co-culture assays with MDA-MB-231 breast cancer cells. TWAS analysis identified 65 candidate genetically regulated genes, and their activity appeared to be preferentially enriched in B cells. hdWGCNA further linked HTS-associated modules to RNA splicing and B-cell activation pathways. Among the candidate genes, ELF1 was the only candidate feature consistently prioritized across all four machine learning algorithms. Low ELF1 expression was associated with inferior overall survival in the GSE96058 cohort, while a consistent survival trend was observed in the METABRIC cohort. ELF1 expression positively correlated with naive B-cell infiltration but showed negative correlations with plasma cells and M0/M1 macrophages. Functional experiments demonstrated that ELF1 modulation affected B-cell proliferation, activation marker expression, RNA splicing-related molecules, and altered the proliferative and invasive behaviors of breast cancer cells in direct co-culture systems. By integrating TWAS, single-cell transcriptomics, machine learning, and preliminary experimental validation, we identified ELF1 as a B-cell-associated prognostic biomarker in breast cancer. These findings suggest that ELF1 may represent a molecular link between genetically regulated expression patterns and B-cell-related immune states within the tumor microenvironment.