Oncogene EVI1 Drives Acute Myeloid Leukemia Via a Targetable EVI1-CtBP1/2 Interaction and By Repressing Myeloid Regulator CEBPA

Overexpression of EVI1 (located at 3q26), as a result of 3q26 rearrangements, causes the development of one of the most aggressive, chemotherapy-resistant forms of AML. We hypothesize that insight into the function of EVI1 in AML will lead to the identification of targetable vulnerabilities for those patients. EVI1 encodes a DNA-binding zinc-finger protein that has a critical role in transcriptional regulation and maintenance of quiescent hematopoietic stem cells. Here, we aim to understand the mechanism by which EVI1 overexpression in bone marrow progenitors drives AML development. To identify its target genes, we made EVI1 rapidly degradable by endogenous fusion with an auxin-inducible degron (AID) tag in inv(3q21;3q26) rearranged MUTZ3 leukemia cells. Degradation of EVI1 strongly increased myeloid differentiation of MUTZ3 cells. Using metabolic labeling of newly produced RNA (SLAM-seq), we found 6 genes to be upregulated after 1.5 hrs of degradation. Out of those genes, CEBPA was also upregulated upon shRNA knockdown of EVI1 or after introducing inactivating mutations in the hijacked GATA2 enhancer that drives EVI1 expression. Those findings suggest persistent upregulation of CEBPA when EVI1 levels are reduced. In line with this, CEBPA expression is 1.83 fold (p=0.00024) lower in AML patients with 3q26 rearrangements (n=18) compared to non-3q26 rearranged cases (n=97). Interestingly, overexpression of CEBPA in MUTZ3 cells bypassed EVI1 driven repression and caused myeloid differentiation as well. We next addressed the question how EVI1 represses CEBPA transcription. The CEBPA locus contains 14 enhancers in a 100kb region around the gene body, among which the myeloid specific +42kb enhancer, essential for neutrophil development. EVI1 binds this +42kb enhancer of CEBPA as determined by chromatin immunoprecipitation sequencing (ChIP-seq) in 3q26-rearranged AML and MUTZ3 cells. After degradation of EVI1, H3K27Ac levels increased at this enhancer, indicating that EVI1 reduces activity of the +42kb CEBPA enhancer. Since EVI1 lacks any enzymatic domains that could modulate chromatin, we sought to further understand how EVI1 represses activity of the +42kb CEBPA enhancer. EVI1 protein immunoprecipitation followed by mass spectrometry revealed interaction with CtBP1/2, factors known to interact with repressors such as Polycomb proteins or HDACs. ChIP-seq showed a strong overlap between CtBP2 and EVI1 binding sites in MUTZ3 cells. Moreover, EVI1 elimination caused loss of CtBP2 binding to the chromatin including to the +42kb CEBPA enhancer. EVI1 harbors a PLDLS domain essential for binding to CtBP1/2. Mutation of this domain abolished CtBP2 binding and reduced transformation of mouse bone marrow progenitors by EVI1, as determined by in vitro replating assays. We reasoned we could exploit the high affinity of EVI1 for CtBP through this single domain by overexpressing a repeat of a small section of EVI1 containing only the CtBP-binding PLDLS-site. Overexpression of "4X-PLDLS competitor peptide" was able to sequester CtBP and thereby prevent it's binding to EVI1. The competitor peptide was able to abolish in vitro replating capacity and re-enabled differentiation of mouse bone marrow progenitors transformed with EVI1. In line with this, transduction of the competitor peptide into MUTZ3 cells led to upregulation of CEBPA as well as growth inhibition. Furthermore, in vivo leukemia outgrowth in a xenotransplant model of 3q26 rearranged AML cell lines MUTZ3 and SB1690CB was fully abolished by the competitor peptide, but not by a mutant control peptide. In conclusion, EVI1 is a repressor of active chromatin regions, including CEBPA via the +42kb myeloid specific enhancer. Repression by EVI1 is dependent on its interaction with CtBP1/2, proteins that recruit repressors of transcription via an essential . A PLDLS domain in EVI1 is essential for this interaction. The competitor peptide experiments point to exploiting this interaction with repressive molecules as a potential therapeutic target in AML with overexpression of EVI1.

No relevant conflicts of interest to declare.