Resistance to anticancer therapies remains a major obstacle to improving survival rates in cancer; it is often driven by epigenetic alterations. Recently, we discovered a new mechanism of therapy resistance that is not fully driven by epigenetic remodeling but involves a switch in the activity of an epigenetic enzyme, EZH2. Despite its well-established activity as part of PRC2 for mediating gene repression by H3K27me3 deposition, new evidence suggests the importance of other, so-called noncanonical activities mediated by variation of PRC2 composition and posttranslational modifications (PTMs) of EZH2. Interestingly, we showed that noncanonical EZH2 was associated with resistance to retinoic acid (RA) in acute promyelocytic leukemia (APL) and that depletion of pan-EZH2 activities was beneficial in killing relapse-initiating cells. However, the precise noncanonical roles of EZH2 remain unclear, especially in non-APL acute myeloid leukemia (AML). Using public transcriptomic data, we explored these functions in non-APL AML. We showed that patients with high EZH2 levels are enriched in metabolic profiles associated with chemotherapy resistance (HighOXPHOS) and that noncanonical EZH2 is linked to relapse after AraC treatment. Notably, AraC-resistant AML cells responded well to the EZH2 degrader MS177 but not to the enzymatic inhibitor tazemetostat. Although MS177 combined with AraC had limited effect, strong synergy was observed when MS177 was combined with RA, venetoclax, or azacytidine. This suggests that degrading EZH2 sensitizes resistant AML cells to specific therapies. Finally, omics analyses revealed specific PTMs and partner variations in EZH2 that may mediate resistance. These findings highlight the contribution of noncanonical EZH2 to AraC resistance in non-APL AML and support the therapeutic potential of targeting these activities to overcome treatment resistance.

INTRODUCTION: Mutations in the Enhancer of Zeste Homolog 2(EZH2) occur in ~ 4% of Acute Myelogenous Leukemia (AML) cases, and confer an adverse prognosis. Most of these mutations lead to loss of enzymatic function, causing a genome wide reduction in the deposition of the H3K27me3 repressive histone mark. EZH2 mutations (MUT) that cause low expression of the protein are associated with chemoresistance in AML. We hypothesized that low levels of wild-type (WT) EZH2 protein would produce a similar phenotype and bad prognosis. METHODS: We measured the levels of EZH2 and 433 other proteins using Reverse Phase Protein Arrays in 806 newly diagnosed, fresh, pre-treatment AML samples. Protein expression was normalized to non-G-CSF treated, normal bone marrow (NBM)-derived CD34+ cells. EZH2 mutation status, treatment, and outcome data were known for 529 patients, of which 24 (4.5%) had EZH2 mutation (MUT) and 505 were WT. LogRank tests were used to compared outcomes; Fisher's Exact, Pearson's Chi-squared or Wilcoxon tests for comparing variables; Pearson's correlation for protein correlation (p<0.01 and correlation coefficient >0.3); Wilcoxon tests adjusted by FDR for differential expression (p<0.05 and LFC>0.5); and Cox proportional hazards models (CoxPH) for Uni-(UV) and Multi-variate (MV) analysis. RESULTS: The cohort was divided into tertiles and regrouped into upper 1/3rd, named High (WT N=178, MUT N=5) and lower 2/3rds, termed Low (WT N=327, MUT N=19). Notably, 95% of AML cases had lower EZH2 protein compared to the median of NBM-derived CD34+ cells. Low EZH2 cases were older, had lower WBC count and blast percentage, and a higher frequency of secondary AML, unfavorable cytogenetics, -5/5q-, -7/7q-, and mutations in ASXL1 and RUNX1, but lower rates of inv16, t(8;21), 11q13 and mutations in FLT3, KIT, NPM1 and RAS (P<0.001, <0.001, <0.001, 0.002, 0.003, 0.02, <0.001, 0.001, 0.003, 0.009, <0.001, 0.007, <0.001, 0.04, <0.001, 0.007). EZH2 MUT patients showed an Overall Survival (OS) similar to WT-Low, but inferior to WT-High (5ys OS: MUT-Low=11%, MUT-High=20%, WT-Low=18%, WT-High=34%; P=0.007). Moreover, High and Low protein expression conferred opposite prognostic impact depending on whether patients received Ara-C based combinations (AraC) or Venetoclax plus Hypomethylating agent (VH) in EZH2-WT patients. Notably, the OS of WT-Low was bad, independent of therapy, but superior compared to WT-High treated with VH (5ys OS: AraC WT-High=48%, AraC WT-Low=29%, VH WT-High=0%, VH WT-Low=17%; P<0.001). EZH2 levels did not affect Remission Duration (RD) for WT patients treated with Ara-C (P=NS), but RD was significantly longer in WT-Low treated with VH, compared to WT-High with the same therapy (5ys RD: AraC WT-High=64%, AraC WT-Low=59%, VH WT-High=0%, VH WT-Low=39%; P<0.001). In the CoxPH UV model of OS, EZH2 levels were prognostic, independent of mutation status, along with other clinical, cytogenetic and molecular features (e.g. age, complex karyotype, 2nd AML, mutations in CEBPA, IDH1/2, etc.). In the MV model, EZH2 levels retained significance, as well as age, 2nd AML and some genetic features (e.g. -5/5q-, NPM1 MUT and others). We compared the protein signature of WT- Low patients with EZH2-MUT to verify whether they share a ‘mutant-like’ behavior. The differential expression (DE) analysis, did not yield any proteins, but 9 proteins related to cell cycle transitions (including mitosis), DNA Damage Response, and p53 signaling were positively correlated with both groups, suggesting commonality. In contrast, 48 proteins were DE between WT-Low and WT-High demonstrating functional differences due to loss of protein expression, with VEGF signaling and negative regulation of apoptosis enriched in WT-Low. CONCLUSIONS: Similar to EZH2 MUT cases, patients with Low-WT had poor outcomes, regardless of therapy, and the proteomic signature of EZH2 Low-WT was similar to EZH2 MUT, suggesting a similar pathophysiology in response to low EZH2 protein expression, regardless of cause. As therapies against EZH2 are being developed we postulate that these may be applicable to a much larger AML population, considering both EZH2 MUT (4%) and Low EZH2 WT (62%), greatly increasing their utility. Notably, those with WT-High responded poorly to VH, suggesting that these cases should not receive that regimen. Finally, proteomics could be leveraged to prospectively identify Low EZH2 WT cases that benefit from EZH2-directed therapy.

Hematopoietic stem cells (HSCs) are essential for lifelong blood and immune cell production, but aging impairs their function, reducing self-renewal and regenerative capacity, contributing to hematopoietic system deterioration. Several hallmarks of HSC aging have been identified over the years; however, the molecular mechanisms behind this process are not fully understood. Among these, we identified the transcription factor PLZF (Zbtb16) as a novel regulator of HSC aging. Using a PLZF loss-of-function mouse model and integrative genomic approaches, including RNA-seq and ATAC-seq, we demonstrated that PLZF mitigates aging-associated HSC phenotypes under regenerative stress in middle-aged mice. These include clonal expansion, myeloid bias, diminished reconstitution potential, and altered cell cycle dynamics of HSCs. Transcriptomic and chromatin profiling of PLZF-deficient HSCs revealed dysregulation of key pathways involved in the cell cycle, metabolism, and inflammation—all hallmarks of aging. To further characterize these mechanisms, we performed single-cell RNA-seq on HSCs from young and old wild-type (WT) and PLZF-deficient mice. In young mutants, we observed an expansion of proliferative multipotent progenitor-like populations (MPP2/MPP3), whereas HSCs show affected inflammatory pathways, including interferon and TGF-β signaling compared with WT HSCs. On the other hand, aged mutants exhibited an accumulation of quiescent long-term HSCs compared with their WT counterpart. These findings suggest that PLZF preserves HSC function during aging, preventing their exhaustion and mitigating inflammatory stress. Our results posit PLZF as a promising molecular target to preserve HSC function, and ongoing studies aim to define PLZF activities that counteract age-related hematopoietic decline.

Acute Promyelocytic Leukaemia (APL) arises from an aberrant chromosomal translocation involving the Retinoic Acid Receptor Alpha (RARA) gene, predominantly with the Promyelocytic Leukaemia (PML) or Promyelocytic Leukaemia Zinc Finger (PLZF) genes. The resulting oncoproteins block the haematopoietic differentiation program promoting aberrant proliferative promyelocytes. Retinoic Acid (RA) therapy is successful in most of the PML::RARA patients, while PLZF::RARA patients frequently become resistant and relapse. Recent studies pointed to various underlying molecular components, but their precise contributions remain to be deciphered. We developed a logical network model integrating signalling, transcriptional, and epigenetic regulatory mechanisms, which captures key features of the APL cell responses to RA depending on the genetic background. The explicit inclusion of the histone methyltransferase EZH2 allowed the assessment of its role in the resistance mechanism, distinguishing between its canonical and non-canonical activities. The model dynamics was thoroughly analysed using tools integrated in the public software suite maintained by the CoLoMoTo consortium (https://colomoto.github.io/). The model serves as a solid basis to assess the roles of novel regulatory mechanisms, as well as to explore novel therapeutical approaches in silico.

Cytogenetic normal AML with NPM1 mutations forms a distinct AML entity, associated with an intermediate prognosis and a heterogeneous response to treatment. We previously described an epigenetic biomarker, defined by the level of H3K27me3 on 70kb of the HIST1 cluster in patient blast DNA. This epigenetic mark separates cytogenetically normal NPM1 mut AML into two groups of patients differing in their survival rate following chemotherapy. To better characterize the influence of the biomarker on disease progression, we performed transcriptomic and histone mark profiling on patient blasts according to the level of H3K27me3 HIST1 . Our integrated analysis revealed that the two groups of patients display differences in terms of transcriptomic, chromatin landscape and cell surface markers, which could explain the clinical difference. Our profiling revealed novel targets and therefore constitutes an (epi)transcriptomic resource for NPM1 AML. It also highlights the power of epigenetic profiling to dissect the heterogeneity of a single AML genetic entity.### Competing Interest StatementThe authors have declared no competing interest.

Resistance to anti-cancer therapeutics remains one of major obstacles to improving survival rates in cancer; it arises in a multitude of ways including accumulation of epigenetic alterations. Recently, we discovered a new mechanism of treatment resistance that is not fully driven by epigenetic remodeling but nevertheless implies an activity switch of an epigenetic enzyme, EZH2. Despite its well-established activity as part of PRC2 for mediating gene repression by H3K27me3 deposition, new evidence points the importance of others, so-called non-canonical activities mediated by variation of PRC2 composition and kinase-related post-translational modifications (PTMs) of EZH2. We previously reported that non-canonical EZH2 activity was associated with retinoic acid (RA) resistance in acute promyelocytic leukemia (APL) (Poplineau, Blood, 2022). RA resistant cells (i) expressed genes involved in DNA repair, replication and proliferation processes, that we called ReP signature, (ii) kept the potential to develop leukemia in vivo and (iii) were marked by high EZH2 expression. Targeting pan-EZH2 activities (canonical/non-canonical) was necessary to eliminate RA resistant cells, which underlies a dependency of these cells on an EZH2 non-canonical and non-methyltransferase activity and the necessity to degrade EZH2 to overcome RA resistance in APL. In the present work, we questioned whether EZH2 non-canonical activity may contribute more generally to resistance in non-APL AML. To do so, we questioned non-APL public AML transcriptomic datasets (BeatAML and TCGA). At first, we classified non-APL AML patients with high and low EZH2 expression and revealed that patients with high EZH2 expression were enriched in metabolic signatures reflecting chemotherapy resistance (i-e HighOXPHOS). Next, we computed a new EZH2-related signature (i-e; 61 common genes to BeatAML and TGCA obtained by retaining the 200 most differentially expressed genes in patients with high and low EZH2 expression) and found that this signature was positively correlated with AraC low responder patients. Besides, we showed that our computed non-canonical and non-methyltransferase EZH2 and ReP signatures were associated (i) with HighOXPHOS and leukemic stem cells maintenance signatures in AML patients and (ii) relapse of patients following AraC treatment. Next, we investigated the relevance of targeting EZH2 in AML using an EZH2 PROTAC synthetized in our laboratory and derived from the previously published MS177 EZH2 degrader (Wang, Nat. Cell Biol., 2022). Using several leukemic cell lines, we showed that chemo-resistant cell lines, more especially AraC resistant cells, were highly sensitive to EZH2 degradation (fifty times more sensitive than AraC sensitive cells) but were not sensitive to the catalytic EZH2 inhibitor Tazemetostat. We then studied the potential synergistic effect between the EZH2 degrader and AraC and did not observe any additional beneficial effect with the combo therapy. However, we showed that EZH2 degradation has strong synergistic effect with RA in killing AML cells in vitro, suggesting that EZH2 degradation may sensitize chemotherapy resistant cells to RA. Finally, by coupling omics approaches on AML cell line resistant to AraC, we are currently investigating mechanisms that dictate EZH2 activities (canonical/non-canonical) and characterizing these activities at the functional level. Proteomics screening revealed specific PTMs on EZH2 and variation in its interacting partners that may contribute to AraC resistance in AML. Altogether, these data reinforce (i) the intimate link between EZH2 non-canonical activity and AraC resistance in non-APL AML and (ii) support the need to preclinically investigate the therapeutic value of targeting non-canonical EZH2 activities not only to sensitize non-APL AML to RA but also to overcome chemotherapy resistance in non-APL AML.

Supplementary Figure from Pharmacologic Reduction of Mitochondrial Iron Triggers a Noncanonical BAX/BAK-Dependent Cell Death

Cancer cell heterogeneity is a major driver of therapy resistance. To characterize resistant cells and their vulnerabilities, we studied the PLZF-RARA variant of acute promyelocytic leukemia (APL), resistant to retinoic acid (RA), using single-cell multi-omics. We uncovered transcriptional and chromatin heterogeneity in leukemia cells. We identified a subset of cells resistant to RA with proliferation, DNA replication and repair signatures, that depend on a fine-tuned E2F transcriptional network targeting the epigenetic regulator Enhancer of Zeste Homolog 2 (EZH2). Epigenomic and functional analyses validated the driver role of EZH2 in RA resistance. Targeting pan-EZH2 activities (canonical/non-canonical) was necessary to eliminate leukemia relapse initiating cells, which underlies a dependency of resistant cells on an EZH2 non-canonical activity and the necessity to degrade EZH2 to overcome resistance. These findings provide critical insights into the mechanisms of RA resistance that allow us to eliminate treatment-resistant leukemia cells by targeting EZH2, thus highlighting a potential new targeted therapy approach. Beyond RA resistance and APL context, we obtained preliminary results suggesting that the non-canonical functions of EZH2 are also associated with chemotherapy resistance (e.g. metabolic signatures such as high OxPHOS signatures) in non-APL AML and relapse of patients following AraC treatment (unpublished data). This new finding may probably open new therapeutic opportunity (EZH2 degradation) beyond the APL context. Altogether, given the emerging non-canonical roles of EZH2, our study provides important insights into the pathogenesis of leukemia and for the development of novel therapeutic strategies targeting EZH2 in a variety of hematological malignancies. Cancer cell heterogeneity is a major driver of therapy resistance. To characterize resistant cells and their vulnerabilities, we studied the PLZF-RARA variant of acute promyelocytic leukemia (APL), resistant to retinoic acid (RA), using single-cell multi-omics. We uncovered transcriptional and chromatin heterogeneity in leukemia cells. We identified a subset of cells resistant to RA with proliferation, DNA replication and repair signatures, that depend on a fine-tuned E2F transcriptional network targeting the epigenetic regulator Enhancer of Zeste Homolog 2 (EZH2). Epigenomic and functional analyses validated the driver role of EZH2 in RA resistance. Targeting pan-EZH2 activities (canonical/non-canonical) was necessary to eliminate leukemia relapse initiating cells, which underlies a dependency of resistant cells on an EZH2 non-canonical activity and the necessity to degrade EZH2 to overcome resistance. These findings provide critical insights into the mechanisms of RA resistance that allow us to eliminate treatment-resistant leukemia cells by targeting EZH2, thus highlighting a potential new targeted therapy approach. Beyond RA resistance and APL context, we obtained preliminary results suggesting that the non-canonical functions of EZH2 are also associated with chemotherapy resistance (e.g. metabolic signatures such as high OxPHOS signatures) in non-APL AML and relapse of patients following AraC treatment (unpublished data). This new finding may probably open new therapeutic opportunity (EZH2 degradation) beyond the APL context. Altogether, given the emerging non-canonical roles of EZH2, our study provides important insights into the pathogenesis of leukemia and for the development of novel therapeutic strategies targeting EZH2 in a variety of hematological malignancies.

Hematopoietic stem cell (HSC) aging is a multifactorial event leading to changes in HSC properties and functions, which are intrinsically coordinated and affect the early hematopoiesis. To better understand the mechanisms and factors controlling these changes, we developed an original strategy to construct a Boolean model of HSC differentiation. Based on our previous scRNA-seq data, we exhaustively characterized active transcription modules or regulons along the differentiation trajectory and constructed an influence graph between 15 selected components involved in the dynamics of the process. Then we defined dynamical constraints between observed cellular states along the trajectory and using answer set programming with in silico perturbation analysis, we obtained a Boolean model explaining the early priming of HSCs. Finally, perturbations of the model based on age-related changes revealed important deregulations, such as the overactivation of Egr1 and Junb or the loss of Cebpa activation by Gata2. These new regulatory mechanisms were found to be relevant for the myeloid bias of aged HSC and explain the decreased transcriptional priming of HSCs to all mature cell types except megakaryocytes.

Cancer cell heterogeneity is a major driver of therapy resistance. To characterize resistant cells and their vulnerabilities, we studied the PLZF-RARA variant of acute promyelocytic leukemia, resistant to retinoic acid (RA), using single-cell multiomics. We uncovered tran-scriptional and chromatin heterogeneity in leukemia cells. We identified a subset of cells resistant to RA with proliferation, DNA replication, and repair signatures that depend on a fine-tuned E2F transcriptional network targeting the epigenetic regulator enhancer of zeste homolog 2 (EZH2). Epigenomic and functional analyses validated the driver role of EZH2 in RA resistance. Targeting pan-EZH2 activities (canonical/noncanonical) was neces-sary to eliminate leukemia relapse-initiating cells, which underlies a dependency of resistant cells on an EZH2 noncanonical activity and the necessity to degrade EZH2 to overcome resistance. Our study provides critical insights into the mechanisms of RA resistance that allow us to eliminate treatment-resistant leukemia cells by targeting EZH2, thus highlighting a potential targeted therapy approach. Beyond RA resistance and acute promyelocytic leukemia context, our study also demonstrates the power of single-cell multiomics to identify, characterize, and clear therapy-resistant cells.

Although originally described as transcriptional activator, SPI1/PU.1, a major player in haematopoiesis whose alterations are associated with haematological malignancies, has the ability to repress transcription. Here, we investigated the mechanisms underlying gene repression in the erythroid lineage, in which SPI1 exerts an oncogenic function by blocking differentiation. We show that SPI1 represses genes by binding active enhancers that are located in intergenic or gene body regions. HDAC1 acts as a cooperative mediator of SPI1-induced transcriptional repression by deacetylating SPI1-bound enhancers in a subset of genes, including those involved in erythroid differentiation. Enhancer deacetylation impacts on promoter acetylation, chromatin accessibility and RNA pol II occupancy. In addition to the activities of HDAC1, polycomb repressive complex 2 (PRC2) reinforces gene repression by depositing H3K27me3 at promoter sequences when SPI1 is located at enhancer sequences. Moreover, our study identified a synergistic relationship between PRC2 and HDAC1 complexes in mediating the transcriptional repression activity of SPI1, ultimately inducing synergistic adverse effects on leukaemic cell survival. Our results highlight the importance of the mechanism underlying transcriptional repression in leukemic cells, involving complex functional connections between SPI1 and the epigenetic regulators PRC2 and HDAC1.

Abstract Cancer cell metabolism is increasingly recognized as providing an exciting therapeutic opportunity. However, a drug that directly couples targeting of a metabolic dependency with the induction of cell death in cancer cells has largely remained elusive. Here we report that the drug-like small-molecule ironomycin reduces the mitochondrial iron load, resulting in the potent disruption of mitochondrial metabolism. Ironomycin promotes the recruitment and activation of BAX/BAK, but the resulting mitochondrial outer membrane permeabilization (MOMP) does not lead to potent activation of the apoptotic caspases, nor is the ensuing cell death prevented by inhibiting the previously established pathways of programmed cell death. Consistent with the fact that ironomycin and BH3 mimetics induce MOMP through independent nonredundant pathways, we find that ironomycin exhibits marked in vitro and in vivo synergy with venetoclax and overcomes venetoclax resistance in primary patient samples. Significance: Ironomycin couples targeting of cellular metabolism with cell death by reducing mitochondrial iron, resulting in the alteration of mitochondrial metabolism and the activation of BAX/BAK. Ironomycin induces MOMP through a different mechanism to BH3 mimetics, and consequently combination therapy has marked synergy in cancers such as acute myeloid leukemia. This article is highlighted in the In This Issue feature, p. 587

Single-cell transcriptomic technologies enable the uncovering and characterization of cellular heterogeneity and pave the way for studies aiming at understanding the origin and consequences of it. The hematopoietic system is in essence a very well adapted model system to benefit from this technological advance because it is characterized by different cellular states. Each cellular state, and its interconnection, may be defined by a specific location in the global transcriptional landscape sustained by a complex regulatory network. This transcriptomic signature is not fixed and evolved over time to give rise to less efficient hematopoietic stem cells (HSC), leading to a well-documented hematopoietic aging. Here, we review the advance of single-cell transcriptomic approaches for the understanding of HSC heterogeneity to grasp HSC deregulations upon aging. We also discuss the new bioinformatics tools developed for the analysis of the resulting large and complex datasets. Finally, since hematopoiesis is driven by fine-tuned and complex networks that must be interconnected to each other, we highlight how mathematical modeling is beneficial for doing such interconnection between multilayered information and to predict how HSC behave while aging.

We previously analyzed 15 000 transcriptomes of mouse hematopoietic stem and progenitor cells (HSPCs) from young and aged mice and characterized the early differentiation of the hematopoietic stem cells (HSCs) according to age, thanks to cell clustering and pseudotime analysis 1. In this study, we propose an original strategy to build a Boolean gene network explaining HSC priming and homeostasis based on our previous single cell data analysis and the actual knowledge of these biological processes (graphical abstract). We first made an exhaustive analysis of the transcriptional network on selected HSPC states in the differentiation trajectory of HSCs by identifying regulons, modules formed by a transcription factor (TFs) and its targets, from the scRNA-seq data., From this global view of transcriptional regulation in early hematopoiesis, we chose to focus on 15 components, 13 selected TFs (Tal1, Fli1, Gata2, Gata1, Zfpm1, Egr1, Junb, Ikzf1, Myc, Cebpa, Bclaf1, Klf1, Spi1) and two complexes regulating the ability of HSC to cycle (CDK4/6 - Cyclines D and CIP/KIP). We then defined the relations in the differentiation dynamics we want to model ((non) reachability, attractors) between the HSPC states that are partial observations of binarized activity levels of the 15 components. Besides, we defined an influence graph of possibly involved TF interactions in the dynamic using regulon analysis on our single cell data and interactions from the literature. Next, using Answer Set Programming (ASP) and considering these inputs, we obtained a Boolean model as a final solution of a Boolean satisfiability problem. Finally, we perturbed the model according to aging differences underlined from our regulon analysis. This led us to propose new regulatory mechanisms at the origin of the differentiation bias of aged HSCs, explaining the decrease in the transcriptional priming of HSCs toward all mature cell types except megakaryocytes. Graphical abstract From single cell-RNA seq data and current knowledge in early hematopoiesis (literature and biological database investigation), 3 inputs are obtained to define the network synthesis as a Boolean Satisfiability Problem depending on observations of states in the differentiation process: 1 influence graph of the possible component interactions, 2 discretized component activity levels in the considered states (blue: 0, inactive, white: *, unknown/free, red: 1, active). 3 dynamic relations ((non) reachability, attractors) between the considered states. Then, these inputs were encoded as constraints in Answer Set Programing (ASP) thanks to the Bonesis tool. After the solving, a final solution of a Boolean model of early hematopoiesis is obtained. This model is altered according to the characteristics of aging observed in our scRNA-seq data, in order to identify the main molecular actors and mechanisms of aging.

ABSTRACTCancer relapse is caused by a subset of malignant cells that are resistant to treatment. To characterize resistant cells and their vulnerabilities, we studied the retinoic acid (RA)-resistant PLZF-RARA acute promyelocytic leukemia (APL) using single-cell multi-omics. We uncovered transcriptional and chromatin heterogeneity in leukemia cells and identified a subset of cells resistant to RA that depend on a fine-tuned transcriptional network targeting the epigenetic regulator Enhancer of Zeste Homolog 2 (EZH2). Epigenomic and functional analyses validated EZH2 selective dependency of PLZF-RARA leukemia and its driver role in RA resistance. Targeting pan-EZH2 activities (canonical/non-canonical) was necessary to eliminate leukemia relapse initiating cells, which underlies a dependency of resistant cells on an EZH2 non-canonical activity and the necessity to degrade EZH2 to overcome resistance.Our study provides critical insights into the mechanisms of RA resistance that allow us to eliminate treatment-resistant leukemia cells by targeting EZH2, thus highlighting a potential targeted therapy approach.HIGHLIGHTS- sc-RNAseq identifies PLZF-RARA leukemia heterogeneity and retinoic acid resistant cells- sc-ATACseq refines leukemic cell identity and resolves retinoic acid resistant networks- EZH2 is a selective dependency of PLZF-RARA leukemia and drives retinoic acid resistance- Targeting pan-EZH2 activities (canonical/non-canonical) is necessary to overcome leukemia onset

Cancer cell heterogeneity is a major driver of therapy resistance. To characterize resistant cells and their vulnerabilities, we studied the PLZF-RARA variant of acute promyelocytic leukemia (APL), resistant to retinoic acid (RA), using single-cell multi-omics. We uncovered transcriptional and chromatin heterogeneity in leukemia cells. We identified a subset of cells resistant to RA with proliferation, DNA replication and repair signatures, that depend on a fine-tuned E2F transcriptional network targeting the epigenetic regulator Enhancer of Zeste Homolog 2 (EZH2). Epigenomic and functional analyses validated the driver role of EZH2 in RA resistance. Targeting pan-EZH2 activities (canonical/non-canonical) was necessary to eliminate leukemia relapse initiating cells, which underlies a dependency of resistant cells on an EZH2 non-canonical activity and the necessity to degrade EZH2 to overcome resistance.Our study provides critical insights into the mechanisms of RA resistance that allow us to eliminate treatment-resistant leukemia cells by targeting EZH2, thus highlighting a potential targeted therapy approach. Beyond RA resistance and APL context, our study also demonstrates the power of single-cell multi-omics to identify, characterize and clear therapy-resistant cells.