Spatial immunophenotypes orchestrate prognosis in triple-negative breast cancer with Miller-Payne grade 4 following neoadjuvant chemotherapy

Abstract

Some triple-negative breast cancer (TNBC) patients evaluated as Miller-Payne 4 with ypN0 after neoadjuvant chemotherapy (NACT) who have better prognoses should avoid escalation of therapy. We aim to identify these patients by evaluating pretherapeutic spatial distributions of immunophenotypes. Our retrospective study in patients with TNBC assessed as Miller-Payne grade 4/5 with ypN0 showed that Miller-Payne 4 with ypN0 group had poorer 5-year disease-free survival (DFS, 63.8% vs. 83.0%, p = 0.003) and the 5-year overall survival (OS, 71.0% vs. 85.5%, p = 0.007) than Miller-Payne 5 with ypN0 group. High TILs were significantly associated with better DFS and OS in patients with Miller-Payne 4 and ypN0 (both p = 0.016). Spatially, detected by multiplexed ion beam imaging by the time of flight combined with proteomics, tumors assessed as Miller-Payne 4 and ypN0 with good prognosis exhibited an inflamed phenotype, with dominant CD8+ T cells on tumor center, few scattered CD68+ myeloid-derived cells far away from T cells, and deposit of increased activated molecules of lymphocyte. While those with poor prognoses presented excluded phenotypes, with few CD8+ T cells restricted to invasive margins and a high density of CD14+CD68+CD11c+ myeloid cells. A good classifier model based on 29 spatial immunophenotypes was established by the random forest algorithm (AUC = 0.975), for identifying patients with Miller-Payne 4 and ypN0 who had favorable prognoses. We also observed similar signatures in patients with Miller-Payne 5 and ypN0. Taken together, spatial immunophenotypes may assess the prognosis in TNBC patients with Miller-Payne 4 and ypN0 after NACT.