Orthogonal proteogenomic analysis identifies the druggable PA2G4-MYC axis in 3q26 AML

Despite advancements in next-generation technologies, progresses in treating subtypes of acute myeloid leukemia (AML) driven by aberrantly activated transcription factors (TFs) have been limited. However, the integration of orthogonal omics led to the identification of drugs for "undruggable" TFs. Here, we sought to discover modulators of the EVI1/MECOM gene, the most lethal oncogenic TF, hyper expressed in AML carrying chromosome 3q26 abnormalities. To this end, we employed an integrative strategy that combines phenotypic, gene expression-based and proteomic high-throughput approaches to identify modulators of EVI1. We screened and scored 5294 compounds enriched for Food and Drug Administration (FDA) approved or bioactive drugs at 1 µM on their ability to suppress the proliferation of 3q26 AML models (EVI1High). From EVI1Null transcriptional signatures, we defined an EVI1 "On" and "Off" state and used Connectivity Map to identify inducers of an EVI1 "Off" status. Histone deacetylase inhibitors (HDACis) emerged as the top class of compounds able to repress leukemia proliferation by suppressing EVI1 signature. These molecules induced apoptosis and reduced EVI1 levels in AML cell lines and in primary EVI1High blast cells while the EVI1 expression sensitizes cells to HDACis. We translated these results in an "N-of-1" clinical trial proposing entinostat in association with azacitidine to patients with relapsed/refractory 3q26 AML based on a compassionate use program for this regimen. Pharmacodynamic modulation of EVI1 was assessed at 6 hours (T6) after entinostat administration in patients and in matched patient-derived leukemia xenograft (PDLX). EVI1 expression decreased in blasts 6 hours after entinostat administration and EVI1 depletion paralleled clinical improvements. This effect was not observed in patients treated with cytarabine. To exclude a contribution of azacitidine in our trial, we completed an in vivo efficacy study and we demonstrated a reduction of AML blasts in mice treated with entinostat but not with azacitidine. In addition, longitudinal single-cell RNA sequencing analysis (scRNASeq) of bone marrow AML cells collected before and after azacitidine-entinostat patients’ treatment cycle and revealed an inhibitory effect on Myc targets genes in the leukemic clone. This effect was also observed in cell lines treated with HADCis suggesting that multiple HDACis regulates the Evi1-Myc axis in 3q26 AML. HDACis-mediated or genetic silencing of EVI1 led to a direct decrement of MYC. To dissect the EVI1 co-transcriptional complex and explain the modulation of Myc signaling in EVI1High AML, we performed a rapid immunoprecipitation mass spectrometry of endogenous proteins (RIME) assay. Eleven EVI1 interactors were targets of Myc signaling, including PA2G4, an RNA-binding protein modulated by HDACis. Importantly, PA2G4 silencing or inhibition (WS6) depleted MYC at protein levels recapitulating the effect of HDACis. Chromatin immunoprecipitation assay (ChIP) showed a reduction of EVI1 and MYC occupancy at PA2G4 promoter in 3q26 AML cells lines, after WS6 treatment suggesting that these genes are involved in PA2G4 control in 3q26 leukemia. Consequently, WS6 reduced leukemic growth in vivo by suppressing an EVI1-MYC signature at single-cell level without inducing toxicities suggesting that PA2G4 is a druggable mediator of EVI1 complex. In contrast, PA2G4 overexpression rescued EVI1High AML cell lines from the effects of HDACis. These data support the evidence that PA2G4 is a mediator of the Evi1-Myc axis in AML. In conclusion, we introduce a clinical treatment approach with HDACis to suppress EVI1 in 3q26-AML patients. Additionally, our findings highlight PA2G4 as a potential therapeutic target, urging the development of PA2G4-MYC-EVI1 complex disruptors in this AML subtype.