Chemotherapy Response in Triple-negative breast cancer defined by single cell sequencing

Triple-negative breast cancer (TNBC) is an aggressive subtype of breast cancer that is often treated with chemotherapy. While about half of the patients completely respond, the remaining patients develop drug resistance and progress to metastatic disease, leading to poor survival rates. Currently, it remains unclear which tumor cell expression programs and cell states in the tumor microenvironment (TME) are associated with response to chemotherapy. To investigate this question, we conducted single-cell RNA sequencing (scRNA-seq) on fresh core biopsies collected from 101 treatment-naive TNBC patients. These patients were subsequently treated with neoadjuvant chemotherapy, in which 45 patients achieved complete pathological response (pCR) and 37 had residual disease (RD). The scRNA-seq data included a total of 427,857 cells encompassing eight major cell types. Unsupervised pseudo-bulk analysis of the tumor cells revealed four primary expression archetypes at the patient level: luminal secretory-related (LSR), basal-related (BR), immunoreactive (IR), and luminal androgen receptor (LAR). BR and IR patients were more likely to exhibit RD and pCR, respectively. At the single-cell level, we discovered 13 phenotypic expression programs shared across patients, such as cell cycling, stress, hypoxia, interferon response, HLA, epithelial-mesenchymal transition, and others. Within the TME, we identified 14 T/NK cell states, 6 B cell states, 15 myeloid cell states, 4 fibroblast cell states, 4 pericyte cell states, and 7 endothelial cell states. These results, resolved at single-cell resolution, allowed us to depict the TNBC ecosystem as a network of co-occurring tumor cell expression programs and TME cell states, grouped into eight eco-traits. Importantly, we found that the presence of specific eco-traits was associated with archetypes and chemotherapy response. For instance, an eco-trait featuring high co-occurrences between interferon-related immune cells, actively cycling tumor cells, and interferon-responding tumor cells was predominantly present in the IR and pCR patients. Last, we developed a gene expression-based logistic regression model to predict chemotherapy response. Overall, this study delivers a comprehensive single-cell atlas of treatment-naive TNBC and pinpoints the association between distinct archetypes, eco-traits, and chemotherapy response.