Fig. 5: Senescence defects account for treatment failure.

a Heatmap of differentially expressed genes (p value < 0.01, log-FC > 1) comparing in vitro-ADR-treated control;bcl2 or ADR/4OHT-double-treated Suv39h1:ER;bcl2 senescent (TIS) to ADR-treated Suv39h1⁻ or Suv39h1:ER;bcl2 senescence-incapable, or untreated non-senescent lymphomas (collectively 47 lymphomas). b Heatmap and associated probability of TIS (black line) vs. no senescence (i.e., proliferating, grey line) in CTX-treated lymphomas based on LPS classification using data in (a), which had 85.1% overall accuracy in a tenfold cross validation of the in vitro-training data. c Heatmap of GSEA Normalized Enrichment Scores (NES) compiled of overall 47 5d ADR-in vitro-treated senescent control;bcl2 lymphomas vs. non-senescent untreated and Suv39h1⁻;bcl2 or non-4OHT-exposed Suv39h1:ER;bcl2 senescence-incapable lymphomas compared to 54 4 h-in vivo-CTX-treated vs. 55 untreated lymphomas. Clustering was performed using Euclidean distance. Senescence-related gene sets are shown, some of them specifically indicated⁷⁰. d Probability of a TIS phenotype according to the LPS classification (cf. panel b, black line) of 4 h-CTX-treated or untreated lymphomas belonging to the indicated response groups (n = 20 CTX NR, n = 19 CTX RP, n = 15 CTX RES, n = 20 native NR, n = 19 native RP, n = 16 native RES). Higher TIS scores represent a stronger senescence likelihood, while lower scores indicate a proliferating phenotype. Box-and-whisker plot showing the median (black line inside the box), upper and lower quartile (box edges), minima and maxima (whiskers), and outliers (dots). The p value was calculated using the Kruskal–Wallis rank-sum test. e Tumor-free survival (TTR) of mice bearing senescence responder (TIS score > 0.8; n = 42), non-responder (TIS score < 0.2; n = 38) or unclassified (0.8 ≤ TIS score ≥ 0.2; n = 22) Eµ-myc lymphomas as determined from the LPS classification (based on the TIS status in the 4 h-CTX-exposed condition [cf. panel d]). Two mice per lymphoma were used. f GSEA NES results comparing ABC-like to GCB-like subtypes in DLBCL patients (ABC/GCB training set from GSE10846, n = 150) and Eµ-myc lymphomas (dataset as used in Fig. 1, ABC/GCB classified, n = 95).