Co-occurrence of transcriptionally distinct persister cell states underpins neoadjuvant therapy resistance in triple‑negative breast cancer.

backgroundAlthough the addition of neoadjuvant immune checkpoint blockade to chemotherapy has improved patient outcome in early triple‑negative breast cancer (TNBC), some patients continue to have poor disease outcomes.

methodsTo characterise neoadjuvant chemotherapy (NAC) resistant cell populations that persist post-NAC, we created a high-resolution single-cell atlas of 129,433 cells from fourteen TNBC patient-derived xenograft (PDX) models with residual disease post-NAC. We identified transcriptionally distinct cancer subpopulations or cell states using unsupervised clustering and characterised‑ their regulatory network as well as clinical associations with treatment response, metastatic progression and survival. Findings were validated using multiple independent TNBC cohorts.

resultsFour major transcriptionally distinct cancer cell states were identified that were shared across the PDX models. These cell states were characterised by enrichment of developmental, hypoxia, interferon signalling, chromosomal instability, and DNA-damage pathways. Hypoxia and interferon signalling cell states shared transcriptional features of drug-tolerant persister cells. Hypoxia cell states contained both cycling and non-cycling cells and were not associated with mutation status. Extensive validation in TNBC patient cohorts confirmed that dysregulated pathways enriched in residual disease were pre-existing in chemotherapy-naïve patients, maintained in distant metastases and were both prognostic and predictive of treatment response. In vitro functional validation highlighted that inhibition of the lysine demethylase KDM5B suppressed the emergence of drug‑tolerant persister cells.

conclusionsNeoadjuvant therapy selects for epigenetically regulated, hypoxia-related cycling and non-cycling drug‑tolerant persister cells in TNBC. These cell states are maintained in metastatic disease and are prognostic/predictive of therapy response.