Breast cancer-subtype and infiltrating macrophage signatures predict chemotherapy-response and survival

Triple-Negative Breast Cancer (TNBC) is a heterogeneous collection of cancers where personalized treatment is difficult and chemotherapy and immunotherapy combinations are the main treatment options. Many attempts to tackle patient heterogeneity have focused on defining cancer-intrinsic subtypes based on differential tumor mRNA-expression across patient cohorts. While these multi-gene diagnostics have shown success in hormone receptor- positive cancers (e.g. OncotypeDX), no TNBC classifiers have shown clinical utility in predicting patient survival or treatment response. We hypothesize that TNBC-infiltrating immune-cells both affect mRNA-based classification and contribute to treatment response variability. To evaluate this hypothesis, we benchmarked the performance of common TNBC-classification (TNBC-type) and infiltrating immune (CIBERSORT) algorithms on the same underlying datasets. Encouragingly, we found that – as with OncotypeDx– highly proliferative TNBC-subtypes (BL1) show the strongest evidence of response to cytotoxic chemotherapies. Interestingly, this cancer-proliferative signature (BL1) is strongly correlated with enrichment in tumor-infiltrating lymphocyte signatures (TIL) which show superior prognostic and predictive power. In addition, Tumor Associated Macrophage (TAM) signatures show independent predictive and prognostic power for both patient survival and response to anthracycline- and taxane-based chemotherapies. These gene signature-based correlations were validated in a new independent cohort of 67 TNBC-patients treated with neoadjuvant chemotherapy. Overall, these results argue for the independent contributions of both cancer-intrinsic and -extrinsic factors in predicting treatment response in the neoadjuvant setting. Overall design: To assess the variability of the tumor microenvironment in triple negative breast cancer patients, we sequences 42 pre-treatment biopsies and 31 post-treatment surgical resected tumor samples. We performed various enrichment analysis and cellular deconvolution methods to bulk-RNA-seq samples.