Closing classification gaps in luminal breast cancer with single-cell RNA-seq insights from normal breast lineages.

Abstract

e13168 Background: Almost two decades ago, invasive breast cancer was classified into five subgroups based on expression patterns, each with distinct biological and prognostic characteristics. Advances in single-cell RNA sequencing offers insights into the expression profiles of normal luminal epithelial cells. Here, we investigate the connection between breast cancer (BC) classification and luminal expression profiles. Methods: We applied single-cell data from recent studies (Table) to compile gene signatures corresponding to expression profiles of luminal epithelial cell clusters in normal breast tissue. Normalized enrichment scores (ESs) of these signatures were calculated for bulk RNA-seq gene expression data from open source cohorts (TCGA-BRCA, SCAN-B, METABRIC). The BC molecular subtype of each tumor sample was predicted based on a modified PAM50 classification using gene expression profiling (1); LumA and LumB subtypes were combined into Luminal. ESs of signatures were compared between the BC subtypes (Table). Results: The ESs of all signatures differed significantly across BC subtypes (p < 0.001, Kruskal-Wallis test, Table 1). The ESs of signatures for LumSec-major and Luminal progenitor clusters are the highest in subtypes HER2-low and Basal-like (p < 0.001, Dwass-Steel-Critchlow-Fligner test [DSCF]). Basal-like also had the highest enrichment of the ductal signature (p < 0.001, DSCF). Luminal subtype showed the highest ESs of signatures corresponding to HR+ clusters of mature luminal cells and of the TDLU signature (p < 0.001, DSCF). HER2-low subtype showed the highest ES of the LumSec-lac signature (p < 0.001, DSCF). Conclusions: Our analyses challenge traditional classifications by showing that luminal breast cancers are less similar to most luminal epithelial clusters in normal breast tissue than other breast cancer subtypes. These findings suggest traditional expression classification may not fully capture the complexity of this largest subset of luminal-type breast cancers. 1. Turova et al., SABCS, 2023. [Table: see text]