Fig. 2: Resistant and sensitive tumors have distinct TME compositions, with sensitive tumors having greater immune cell infiltration.

A Heatmap showing relative abundance of high-quality (HQ) cells of each cell type in tumor biopsies of the discovery and validation cohorts. Tumors samples (y-axis) are clustered into three compositional archetypes based on pairwise distance of compositional similarity (UMAP and Gaussian mixture model (GMM)). Cell types (x-axis) are clustered using hierarchical clustering to show correlation of pairwise abundance, with immune cell abundances being highly correlated with one another. Equivalent read depth cutoff applied to select HQ cells from both cohorts (see methods). B Distinction of three archetypal tumor compositions shown by UMAP dimension reduction using logit-Euclidean distance of compositional similarity (points = tumor samples, color = archetype, shape = cohort). Tumor data points close together have high compositional similarity. Distinct archetypal tumor compositions (top panel colors) were identified by applying the GMM to the UMAP composition space coordinates of the discovery cohort data. Then, validation cohort tumor compositions were projected into the same UMAP model space and compositional archetypes were classified by the parameterized GMM. Tumor response outcomes were associated with their compositional archetype pre, during and post treatment with combination ribociclib (middle panel) or letrozole alone (bottom panel) (shape = timepoint). Post treatment, sensitive tumors more frequently exhibited an immune hot archetype (logistic generalized linear model predicting tumor response probability by archetype compared responsiveness of immune hot to other archetypes: Immune hot response rate relative to other archetypes: Combination ribociclib estimate = 1.7,se = 0.8, df = 29,z = 2.12, p = 0.033; Letrozole alone estimate = 3.3,se = 1.3, df = 17,z = 2.47, p = 0.013). C Tumor archetypes differ in immune cell type abundance and diversity (Shannon diversity) (color = relative abundance, shape = archetype). Comparing tumor archetypes, tumors in the immune hot and diverse archetype show increased: i) immune cell abundance (logistic generalized linear model: Increased logit immune fraction estimate = 1.47,se = 0.01, df = 166,z = 101.3, p = 2e-16), ii)T cell abundance (logistic generalized linear model: Increased logit T cell fraction estimate = 2.62,se = 0.04, df = 166,z = 62.4, p = 2e-16) and iii) Shannon diversity (ANOVA log diversity increase: estimate = 0.87,se = 0.15, df = 166, t = 5.82, p = 2.9e-8). Sample size = 422,635/424,581 annotated cells (ensuring equivelant HQ between cohorts) from 173 biopsy samples of 62 patient tumors at 3 timepoints. All statistical tests two-sided. Source data are provided as a Source Data file.