Turning up a new pattern: Identification of cancer-associated fibroblast-related clusters in TNBC

Abstract

Growing evidence indicates a connection between cancer-associated fibroblasts (CAFs) and tumor microenvironment (TME) remodeling and tumor progression. Nevertheless, how patterns of CAFs impact TME and immunotherapy responsiveness in triple-negative breast cancer (TNBC) remains unclear. Here, we systematically investigate the relationship between TNBC progression and patterns of CAFs. By using unsupervised clustering methods in the Molecular Taxonomy of Breast Cancer International Consortium (METABRIC) dataset, we identified two distinct CAF-associated clusters that were related to clinical features, characteristics of TME, and prognosis of patients. Then, we established a CAF-related prognosis index (CPI) by the least absolute shrinkage and selection operator (LASSO)-Cox regression method. CPI showed prognostic accuracy in both training and validation cohorts (METABRIC, GSE96058, and GSE21653). Consequently, we constructed a nomogram with great predictive performance. Moreover, the CPI was verified to be correlated with the responsiveness of immunotherapy in three independent cohorts (GSE91061, GSE165252, and GSE173839). Taken together, the CPI might help us improve our recognition of the TME of TNBC, predict the prognosis of TNBC patients, and offer more immunotherapy strategies in the future.

Keywords: cancer-associated fibroblasts; machine learning; prognostic model; triple-negative breast cancer; tumor microenvironment.