AML senescence capacity is a unifying outcome predictor and plastic APL classifier

BACKGROUND Acute myeloid leukemia (AML) is a highly heterogeneous disease. The predictive role of therapy-induced senescence (TIS) in cancer treatment in general and newly diagnosed (nd) AML patients in particular is undetermined. METHODS We established a flow cytometry-based senescence assay to detectsenescence in primary patient blasts to ex vivo chemotherapy (daunorubicin or cytarabine) in patients with newly diagnosed (nd) non-M3 AML. Bulk and single-cell analyses assessed gene expression and recurrent AML mutations. TIS-related gene signatures were interrogated in additional AML patient cohorts. Computational analyses, chromatin immunoprecipitation sequencing and other functional assays characterized TIS-related leukemia biology. RESULTS Ex vivo-determined TIS capacity predicted superior disease-free and overall survival in three independent cohorts comprising 92 patients with nd non-M3 AML across diverse (cyto-)genetic subgroups. Subsequent treatment with Bcl2 inhibitors venetoclax or navitoclax demonstrated their senescence-preferential killing. By interrogating 977 nd all-subtype AML patients from three additional independent cohorts, a TIS-derived gene classifier unexpectedly identified M3-AML/acute promyelocytic leukemia cases with very high accuracy in a PML-RARA-agnostic fashion. We found the polycomb-repressive complex 2 component Suz12 to regulate this classifier. Non-M3-AML cell lines in TIS differentiated upon all-trans retinoic acid (ATRA), or died in response to the Suz12-targeting histone deacetylase (HDAC) inhibitor panobinostat. CONCLUSIONS TIS predicts superior long-term outcome in nd AML and confers unifying plasticity across molecularly highly diverse AML subtypes towards an M3-like state, which enables consolidative senolysis by Bcl2 family or HDAC inhibition, or senodifferentiation by ATRA as novel plasticity-based treatment concepts. To characterize molecular changes occurring in response to AML chemotherapy, we obtained gene expression profiles (GEP) by RNA sequencing (RNA-seq) of 11 paired samples of ex vivo-DNR-exposed vs. UT non-M3 AML blasts.