Abstract
AbstractBackgroundBreast cancer is a complex disease with diverse molecular characteristics, significantly impacting patient prognosis, outcomes, and treatment decisions. Previous studies have introduced PAM50 classifiers and claudin-low classifiers based on bulk RNA-seq samples. However, single-cell analysis has revealed the existence of distinct subtypes within the same tumor, indicating that classifiers relying on gene signatures derived from bulk samples may not accurately capture the true molecular features of breast cancer.MethodTo address this limitation, we utilized single-cell data from breast cancer patients to define the E-M ratio parameter. We identified 93 epithelial-specific genes and developed a Universal Breast cancer Subtyping 93 (UBS93). To validate the efficacy of UBS93, we conducted separate analyses using bulk RNA-seq and single-cell RNA-seq datasets of human breast cancer cell lines, as well as bulk RNA-seq data from mice. Additionally, we compared the performance of UBS93 with that of the genefu package to highlight its advantages.ResultsUBS93 demonstrated excellent performance in human and mouse datasets, including bulk RNA-seq and single-cell RNA-seq data. It exhibited higher epithelial specificity and accuracy compared to PAM50 genes. When predicting bulk RNA-seq data from breast cancer cell lines and mouse models, UBS93 outperformed the genefu package. Single-cell validation revealed the coexistence of basal and claudin-low subtypes in the HDQP1 cell line and two TNBC patients, suggesting a shared origin. Differential gene expression analysis identified ELF3 loss as a potential driver for basal-to-claudin-low differentiation. Experimental validation confirmed that the downregulation of ELF3 resulted in the downregulation of CLDN3, CLDN4, and CLDN7, facilitating the transition from basal to claudin-low cells.ConclusionOur study constructed a comprehensive breast cancer classification, UBS93, based on 93 epithelial-specific genes identified using single-cell data. By applying UBS93, we unveiled the coexistence of basal and claudin-low subtypes and illuminated the molecular mechanism underlying basal-to-claudin-low differentiation, with ELF3 loss playing a significant role in this process.BackgroundBreast cancer is a heterogeneous disease in terms of molecular alterations, cellular composition, and clinical outcomes. However, this heterogeneity poses challenges regarding clinically relevant tumor classification for prognosis and prediction [36931265] [1]. Fortunately, researchers have utilized microarray technology to develop an intrinsic breast cancer classifier called PAM50, which categorizes cancer into five subtypes: Luminal A, Luminal B, HER2-enriched, Basal-like, and Normal-like [19204204] [2]. This classification system significantly enhances the prognostic and predictive value over traditional approaches, including pathological staging, histological grading, and standard clinical biomarkers.Through in-depth investigation of gene expression profiles in breast cancer, researchers have discovered a novel subtype characterized by low expression levels of cell adhesion components such as CLDN3, CLDN4, CLDN7, and CDH1, which is associated with mesenchymal features. This new subtype exhibits increased proliferative capacity and poorer prognosis [17493263] [3]. In response, Alexi et al. developed a classification method called the nine-cell line claudin-low predictor, categorizing breast cancer into two groups: Claudin-low and Others [20813035] [4]. However, the origin of this new subtype has been a subject of ongoing debate. Some researchers suggest that the occurrence and progression of triple-negative breast cancer from luminal epithelium are driven by carcinogenic RAS signal transduction., although this conclusion has yet to be validated in human data [34145248] [5]. Another study, based on genetic, epigenetic, and gene expression analyses, found that claudin-low breast cancer originates from three subgroups, with two subgroups associated with luminal and basal-like subtypes, and the third subgroup closely related to normal human breast stem cells [32647202] [6]. In conclusion, the origin of claudin-low breast cancer is a complex process that requires further research to explore the underlying mechanisms and contributing factors.Single-cell transcriptomic analysis has provided us with deeper insights into the heterogeneity among different subtypes of breast cancer [35352511] [7]. By analyzing the intrinsic subtypes within individual malignant cells, it has been discovered that there exist cells of different subtypes within a single tumor. This indicates that the tumor subtypes defined by gene signatures obtained through bulk RNA sequencing may not always accurately reflect the true molecular phenotype of the tumor [36931265][1]. To gain a more comprehensive understanding of tumor cells, single-cell analysis is necessary. Additionally, there is a lack of standardized prediction for breast cancer subtypes. To address these issues, we have developed a novel epithelial cell-specific prediction factor called Universal Breast Cancer Subtyping 93 (UBS93). UBS93 classifies breast cancer into four subtypes: Basal, Claudin-low, HER2-amp, and Luminal. UBS93 focuses on epithelial-specific markers and aims to provide a more consistent and comprehensive approach for breast cancer subtype classification. This advancement holds the potential to improve the accuracy of classification, facilitate personalized treatment strategies, and guide clinical decision-making.
Publisher
Cold Spring Harbor Laboratory