Fig. 3: Cell type communication differences between resistant (growing) and sensitive (shrinking) tumors pre- and post-treatment.

A Box plot showing cell type contribution to signaling with each cell type (points) across combination ribociclib or letrozole alone treated tumors of the discovery/validation cohorts (cohort = shape). Cancer cells were higher signal contributors than non-cancer cells (Linear model: Ribociclib:estimate = 0.67,se = 0.11, df = 970, t = 6.26, p = 5.6e-10; Letrozole:estimate = 0.37,se = 0.13, df = 970, t = 2.89, p = 0.0039). B Box plot showing signal strength received by each cell type (points = sender types) across tumors within treatment groups in the discovery/validation cohorts (cohort = shape). Cancer cells received fewer signals across communication pathways (Linear model: Ribociclib:estimate = −2.45,se = 0.076, df = 970, t = −32.0, p = 2e-16; Letrozole:estimate = −2.47,se = 0.10, df = 970, t = −24.3, p = 2e-16), despite receiving stronger growth factor signaling via ERBB receptors. Sample size for A/B = 108 cell type communication measurements for each of 9 cell types (x-axis). Measurements quantify mean signal sent/received by a cell type to/from another across resistant /sensitive tumors and sample days (0,14,180) in the discovery/validation cohort. Box elements in A/B represent median signal sent/received across cell type communication measurements (center line), upper/lower quantiles (hinges), 1.5*inter-quartile range (whiskers). C Differences in cell type communication (x-axis = sending cell type, y-axis = receiving cells type) between tumors resistant and sensitive to combination ribociclib or letrozole alone (top/bottom subpanels) pre- and post-treatment (left/right subpanels). Cell types communicating significantly more strongly in resistant tumors (red) and sensitive tumors (blue) were identified using permutation-based bootstrap randomization tests in the discovery and validation cohorts (left/right panel). Z-scores (color intensity) quantify resistant-sensitive tumor communication differences (white = no significant difference). Black boxes indicate: i) strengthened cancer/epithelial-myeloid communication in ribociclib-resistant tumors pre-treatment, ii) reduced signaling to CD8 + T cells from across the TME in ribociclib-resistant tumors throughout treatment and iii) increased cancer signaling to diverse cell types in post-treatment letrozole-resistant tumors. D Network graphs showing divergence in cell type communication between resistant and sensitive tumors before treatment (top/bottom panel) in the discovery and validation cohorts (left vs right panels). Nodes represent cell types; arrow width indicates average directed communication strength across tumors. Strong communications (black) exceed 1sd above mean (other communications = gray). Sample size = 424,581 annotated cells, 1444 LR communication pathways from 173 biopsy samples of 62 patient tumors. All statistical tests two-sided. Source data are provided as a Source Data file.