A comprehensive single-cell breast tumor atlas defines cancer epithelial and immune cell heterogeneity and interactions predicting anti-PD-1 therapy response

Abstract

We present an integrated single-cell RNA-seq resource of the breast tumor microenvironment consisting of 236,363 cells from 119 biopsy samples across 8 publicly available datasets. In this computational study, we first leverage this novel resource to define cancer epithelial cell heterogeneity based on two clinically relevant markers and identify six new and distinct subsets of natural killer cells. We then illustrate how cancer epithelial cell heterogeneity impacts immune cell interactions. We develop T cell InteractPrint, which considers how cancer epithelial cell heterogeneity shifts the predicted strength of T cell interactions. We use InteractPrint to predict response to immune checkpoint inhibition (ICI) in two clinical trials testing immunotherapy in patients with breast cancer. T cell InteractPrint was predictive in both trials (AUC = 0.81 and 0.84), versus PD-L1 expression (AUC = 0.54 and 0.72). This result provides an alternative predictive biomarker to PD-L1 to select patients who should receive ICI. STATEMENT OF SIGNIFICANCEWe developed a novel integrated single-cell atlas of the breast tumor microenvironment to interrogate breast tumor cell heterogeneity and define how heterogenous cancer epithelial cell and immune cell interactions predict response to anti-PD-1 therapy.