Development and Verification of a Prognostic Stemness-Related Gene Signature in Triple-Negative Breast Cancer

Abstract

Background. It is well known that cancer stem cells can induce cancer metastasis, which causes the majority of cancer-related death, especially in triple-negative breast cancer (TNBC). TNBC features a high metastatic rate and low metastasis-free survival and is regarded as the most malignant subtype of breast cancer. The purpose of this study is to explore prognostic biomarkers that can predict metastasis of triple-negative breast cancer. Methods. The human triple-negative breast cancerSUM149PT cells were used for the study. The cancer stem cell spheres (sum149-Stem) and paired adherent cancer cells (sum149-Tumor) were collected to extract total RNAs. RNA-seq was used to analysis the mRNA expression of cancer stem cells and paired adherent cancer cells. Two different gene expression omnibus datasets (https://www.ncbi.nlm.nih.gov/gds), GSE58812 and GSE33926, were used to explore the mechanism of different expression genes between stem cells and adherent cancer cells. Seven genes showed prognostic function in all datasets. The STITCH database (https://www.stitchdata.com/) was used to explore the possible metastasis-inhibiting drugs that can target the seven genes. Each gene expression was compared by Pearson analysis. The receiver operating characteristic curve (ROC) and Kaplan–Meier survival curve were performed to assess the metastasis prognostic ability of the seven-gene modeling two different GEO datasets. Results. A subset of 7 stemness-related genes (SRGs) containing UCN, ST3GAL5, FDPS, HK2, MALL, LMTK3, and CRHR2 were identified in three independent cohorts. Univariate Cox analysis showed that ST3GAL5 plays an antitumor role in TNBC metastasis, and the other 6 genes promote the metastatic progression of TNBC. The ability of the 7-SRGs gene Cox model to predict TNBC metastasis was constructed with the GSE58812 dataset. Most of the genes showed significant expression in patients with different risk levels. Additionally, the model showed predictive value in another GEO dataset of TNBC patients. ROC curves indicated that the seven-gene model has a significant predictive value of TNBC metastasis. Conclusions. Expression analysis of the 7-SRGs signature model at diagnosis has predictive value for metastasis in TNBC patients.