Construction of a stromal cell-related prognostic signature based on a 101-combination machine learning framework for predicting prognosis and immunotherapy response in triple-negative breast cancer.

BACKGROUND

Triple-negative breast cancer (TNBC) is a highly aggressive subtype with limited therapeutic targets and poor immunotherapy outcomes. The tumor microenvironment (TME) plays a key role in cancer progression. Advances in single-cell transcriptomics have highlighted the impact of stromal cells on tumor progression, immune suppression, and immunotherapy. This study aims to identify stromal cell marker genes and develop a prognostic signature for predicting TNBC survival outcomes and immunotherapy response.

METHODS

Single-cell RNA sequencing (scRNA-seq) datasets were retrieved from the Gene Expression Omnibus (GEO) database and annotated using known marker genes. Cell types preferentially distributed in TNBC were identified using odds ratios (OR). Bulk transcriptome data were analyzed using Weighted correlation network analysis (WGCNA) to identify myCAF-, VSMC-, and Pericyte-related genes (MVPRGs). A consensus MVP cell-related signature (MVPRS) was developed using 10 machine learning algorithms and 101 model combinations and validated in training and validation cohorts. Immune infiltration and immunotherapy response were assessed using CIBERSORT, ssGSEA, TIDE, IPS scores, and an independent cohort (GSE91061). FN1, a key gene in the model, was validated through qRT-PCR, immunohistochemistry, RNA interference, CCK-8 assay, apoptosis assay and wound-healing assay.

RESULTS

In TNBC, three stromal cell subpopulations-myofibroblastic cancer-associated fibroblasts (myCAF), vascular smooth muscle cells (VSMCs), and pericytes-were enriched, exhibiting high interaction frequencies and strong associations with poor prognosis. A nine-gene prognostic model (MVPRS), developed from 23 prognostically significant genes among the 259 MVPRGs, demonstrated excellent predictive performance and was validated as an independent prognostic factor. A nomogram integrating MVPRS, age, stage, and tumor grade offered clinical utility. High-risk group showed reduced immune infiltration and increased activity in tumor-related pathways like ANGIOGENESIS and HYPOXIA, while low-risk groups responded better to immunotherapy based on TIDE and IPS scores. FN1, identified as a key oncogene, was highly expressed in TNBC tissues and cell lines, promoting proliferation and migration while inhibiting apoptosis.

CONCLUSION

This study reveals TNBC microenvironment heterogeneity and introduces a prognostic signature based on myCAF, VSMC, and Pericyte marker genes. MVPRS effectively predicts TNBC prognosis and immunotherapy response, providing guidance for personalized treatment. FN1 was validated as a key oncogene impacting TNBC progression and malignant phenotype, with potential as a therapeutic target.