Abstract

Objective: As a dominant component within the tumor microenvironment, macrophages exert essential roles in nearly all aspects of triple-negative breast cancer (TNBC). This work explored macrophage-associated signature genes for prognostication and treatment for TNBC.Methods: Single-cell (GSE180286) and bulk transcriptome profiles (TCGA-TNBC, GSE96058 and GSE45255) were analyzed by multiple computational approaches. The expression of signature genes was verified in tumor tissues and paracancerous normal tissues from patients with TNBC (n=5), HER2+ breast cancer (n=5), and HR+ breast cancer (n=5) through immunofluorescence and Western blot. Additionally, gene expression was examined in breast cancer cells (MDA-MB-231, and MCF-7) and mammary epithelial cells (MCF10A) using RT-qPCR and Western blot. Following RNA interference or overexpression, CCK-8, wound scratch and Transwell assays were performed. To assess model robustness, 1000 iterations of Bootstrap resampling were performed to calculate optimism-corrected performance metrics; calibration curves were generated via the rms package. Decision Curve Analysis (DCA) was conducted to evaluate the clinical decision-making value.Results: A single-cell map of the microenvironment in non-TNBC and TNBC was depicted. At both the single-cell and bulk levels, macrophages exhibited a higher abundance in TNBC versus non-TNBC. A macrophage-based gene signature (CTSD, CTSL, ELK4, HSPA8, XRCC4) was developed, with a high-risk score predicting poorer outcomes. This signature demonstrated reliable performance in external validation, particularly for one-year survival (AUC > 0.9). Bootstrap analysis corrected the original AUC from 0.706 to 0.739 (optimism=-0.033, difference <5%), and AUC values from 1000 resamplings concentrated in 0.70-0.75 (standard deviation=0.018). External validation confirmed the signature's ability to reliably pr