Abstract 515: Single-cell transcriptomic analysis of HR+/HER2- breast cancer identifies gene signatures that predict outcomes of luminal A and B subtypes

Abstract Introduction: Patients with luminal A and B early-stage ER+/HER2- breast cancer (BrCa) are uniformly treated with adjuvant endocrine therapy (ET) (±chemotherapy). A better understanding of drivers of ET resistance is required as subsequent lethal metastatic disease remains a major clinical problem. Therefore, there is an unmet clinical need to identify biomarkers that select low- and high-risk patients who may benefit from de-escalation of current treatments or alternative therapeutic interventions. In the present study, we hypothesized that while luminal A and B tumors both arise from normal hormone-responsive cells, they are transcriptionally distinct, hence allowing the identification of unique gene signatures that can predict outcomes in each subtype. Methods: Tumors from 10 early-stage ER+/HER2- BrCa patients were subjected to single-cell RNA-sequencing (scRNA-seq) analysis (10X Genomics); 6/10 tumors were classified as luminal A and 4/10 as luminal B based on combined PAM50 and immunohistochemical classification (Ki67 cut-off=20%). We performed a direct transcriptional comparison between luminal A and B tumors, using well-established signatures and unbiased differential gene expression analysis. To identify unique luminal A and B tumor-specific genes, we compared the gene expression profile of each luminal subtype with 10 non-neoplastic breast tissues. A predictive model (LASSO) was applied to select genes with the highest frequency using a training dataset. This resulted in 5- and 4-gene signatures for luminal A and B, respectively, which were used to calculate risk scores that divided each subtype into low- and high-risk groups. The prognostic value of the above signatures was validated in an independent dataset. Results: The integrated scRNA-seq analysis of luminal A and B tumors revealed transcriptionally distinct tumor cell clusters while tumor microenvironment (TME) clusters were well intermixed. Luminal B tumors had higher cell cycle and BrCa-specific scores, low ER pathway-gene expression scores, and increased 8q amplifications. IFNγ, OXPHOS, p53, hypoxia and MYC targets were the most upregulated pathways in the luminal B subtype. The TME of luminal B tumors was comprised of lower CD4+ and CD8+ T cell but higher Treg levels. Comparison with normal breast tissues revealed that early-stage ER+ BrCa arises from hormone-responsive epithelial cells and provided a number of tumor-specific genes that were used to generate prognostic signatures. These signatures were capable of differentiating high- from low-risk patients within each subtype and predicting survival outcomes in two large-scale training and validation cohorts. Conclusions: We developed a novel prognostic tool that can be used to determine duration of adjuvant ET and/or new therapeutic strategies for high-risk luminal A and B patients in the early-stage ER+/HER2- setting. Citation Format: Sofia Mastoraki, Jerome Lin, Xiayu Rao, Sophie R. Liu, Harsh Batra, Maria G. Raso, Edwin R. Parra Cuentas, Akshara S. Raghavendra, Komal S. Rasaputra, Min Yi, Jing Wang, Aysegul Sahin, Debasish Tripathy, Kelly K. Hunt, Nicholas E. Navin, Khandan Keyomarsi. Single-cell transcriptomic analysis of HR+/HER2- breast cancer identifies gene signatures that predict outcomes of luminal A and B subtypes [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 515.