Selective Requirement of MYB for Oncogenic Hyperactivation of a Translocated Enhancer in Leukemia

Next-generation sequencing has greatly improved our knowledge about the location, distribution, and frequency of recurrent gene mutations in cancer (1, 2). The focus has previously been on the identification and understanding of mutations in protein coding regions. However, many mutations are found in intergenic regions as well (3), which now receive broad attention (4–8). Those studies demonstrate that malignant transformation does not only rely on coding mutations in proto-oncogenes but may also depend on aberrant regulation of oncogene expression. Well-described mechanisms of aberrant gene activation include generation of novel enhancers by nucleotide substitution, focal amplification of enhancers, loss of boundaries between topologically associated domains, or enhancer hijacking by chromosomal rearrangements (9–15).

Chromosomal inversion or translocation between 3q21 and 3q26 [inv(3)(q21;q26) or t(3;3)(q21;q26)] in acute myeloid leukemia (AML) results in the aberrant expression of the proto-oncogene EVI1 located at the MDS1 and EVI1 complex locus (MECOM) at 3q26 (16–19). Our group and others reported that hyperactivation of EVI1 is caused by a GATA2 enhancer translocated from chromosome 3q21 to EVI1 on chromosome 3q26 (12, 20). Upon translocation, this hijacked GATA2 enhancer appears to behave as a superenhancer and is marked by a broad stretch of H3K27 acetylation (12, 20). In this study, we aimed to unravel the mechanism by which the hijacked GATA2 superenhancer leads to EVI1 activation. We generated a model to study EVI1 regulation in inv(3)/t(3;3) AML cells by inserting a GFP reporter 3′ of endogenous EVI1 and introduced an inducible Cas9 construct. To uncover important elements in this hijacked enhancer, we applied CRISPR/Cas9 scanning and identified motifs essential for driving EVI1 transcription. We demonstrated a single regulatory element in the translocated GATA2 enhancer that is critical for the regulation of EVI1 expression, with an essential role for MYB through binding to the translocated enhancer. Treatment of inv(3)/t(3;3) AML cells with peptidomimetic MYB:CBP/p300 inhibitor decreased EVI1 expression and induced leukemia cell differentiation and cell death.

In inv(3)/t(3;3) AMLs, the GATA2 superenhancer is translocated to MECOM, driving expression of EVI1 (12, 20). We investigated whether GATA2 enhancer–driven transcription of EVI1 in inv(3)/t(3;3) is reversible in leukemia cells. In primary inv(3)/t(3;3) AML, immature CD34⁺CD15⁻ cells can be discriminated from more mature CD34⁻CD15⁻ and CD34⁻CD15⁺ cells (Fig. 1A, left; Supplementary Fig. S1A and S1B, left). Whereas EVI1 is highly expressed in CD34⁺CD15⁻ cells, mRNA and protein levels decline in the CD34⁻CD15⁻ fraction and are almost completely lost in CD34⁻CD15⁺ cells in inv(3)/t(3;3) primary AML as well as in MUTZ3 cells, an inv(3) AML model (Fig. 1A and B, right; Fig. 1C; Supplementary Fig. S1A and S1B, right). Because 3q26 rearrangements are present in all fractions, as determined by three-colored FISH (Supplementary Fig. S1C), we conclude that transcription of EVI1 can be reversed in AML cells despite the presence of a 3q26 rearrangement. In vitro culture of sorted MUTZ3 cells revealed that only the EVI1-expressing CD34⁺CD15⁻ cells were competent to proliferate (Supplementary Fig. S1D), in agreement with previous observations showing that EVI1 depletion results in loss of colony formation and induction of differentiation (12). Thus, although AML cells with inv(3)/t(3;3) depend on EVI1, transcription of this gene remains subject to regulation and can be repressed, with major consequences for cell proliferation and differentiation.

Our findings indicate that interference with EVI1 transcription may be an entry point to specifically target inv(3)/t(3;3) AMLs. To study the molecular mechanisms of EVI1 transcriptional activation by the hijacked GATA2 enhancer, we introduced a GFP reporter 3′ of EVI1 at the translocated allele, which is the only allele expressed in MUTZ3 cells. A T2A self-cleavage site was introduced in between EVI1 and GFP separating the two proteins (Fig. 2A; Supplementary Fig. S2A and S2B). Knockdown of EVI1 using two unique EVI1-specific short hairpin RNAs (Fig. 2B) resulted in a reduction of the GFP signal (Fig. 2C). Subsequently, a construct with tight doxycycline-controlled expression of Cas9 was introduced into MUTZ3-EVI1-GFP cells (Supplementary Fig. S2C and S2D) and used to target the translocated GATA2 enhancer and study EVI1 regulation. Deletion of approximately 1,000 bp in the −110-kb (−77 kb in mouse) distal GATA2 enhancer (21, 22) using two specific single-guide RNAs (sgRNA; Supplementary Table S1) resulted in a severe decrease in GFP expression upon doxycycline treatment (Fig. 2D). We sorted the GFP-expressing cells into three fractions and observed that enhancer deletion was most pronounced in the GFP^lo FACS-sorted cells (Fig. 2E, lower band). The GFP^lo fraction also contained the lowest GFP and EVI1 mRNA levels (Fig. 2F and G). Cells from the GFP^lo fraction, which showed reduced EVI1 expression, formed fewer colonies than GFP^hi cells in methylcellulose (Fig. 2H). Only colonies obtained from the GFP^hi fraction consisted of cells able to multiply when placed in liquid culture (Supplementary Fig. S2E). Immunophenotyping of the colonies revealed that GFP^hi fractions predominantly consisted of immature CD34⁺CD15⁻ cells, whereas, in contrast, the GFP^lo fraction contained the highest number of differentiated CD15⁺CD34⁻ cells (Supplementary Fig. S2F). Together, this established a doxycycline-inducible Cas9-expressing inv(3)/t(3;3) AML model (MUTZ3-EVI1-GFP) for studying the transcriptional control of EVI1 via a GFP reporter.

The minimally translocated region of the GATA2 super-enhancer is 18 kb long (12). In MUTZ3 and MOLM1, which are both inv(3) AML models, this highly H3K27 acetylated region (Fig. 3A, yellow) contains four loci of open chromatin determined by Assay for Transposase-Accessible Chromatin using sequencing (ATAC-seq; Fig. 3A, orange), of which two show strong p300 occupancy (Fig. 3A, red). To identify, in an unbiased fashion, which elements of the 18-kb translocated region control EVI1 transcription, we used a CRISPR/Cas9-based enhancer scanning approach (Fig. 3A). We constructed a lentiviral library containing 3,239 sgRNAs covering the 18-kb translocated region (Fig. 3A; Supplementary Table S2) and transduced it into MUTZ3-EVI1-GFP cells at a low multiplicity of infection. After neomycin selection and cell expansion, the cells were treated with doxycycline to induce Cas9 expression, and cells displaying reduced GFP reporter expression (GFP^lo) were selected by flow cytometric sorting at day 5 and day 7. The sgRNAs were amplified from genomic DNA and deep-sequenced to identify the sgRNAs that were enriched in the GFP^lo fraction. The log₂ fold change of three independent experiments was combined as shown in Fig. 3B, which demonstrated a strong correlation between the sgRNAs enriched in GFP^lo cells at day 5 and day 7 (Fig. 3B).

Five sgRNAs targeting EVI1 were the top-scoring hits in the GFP^lo fraction (indicated in blue), whereas sgRNAs targeting the safe harbor AASV1 locus (in red) were not enriched, emphasizing the specificity and sensitivity of the assay (Supplementary Fig. S3A). sgRNAs with a minimum of threefold enrichment in the GFP^lo fraction all clustered in a small region of approximately 700 bp (Fig. 3C). This region is a known p300-interacting region, which belongs to the −110-kb distal GATA2 enhancer (21, 22). This p300-interacting region is occupied by a heptad of transcription factors (SCL, LYL1, LMO2, GATA2, RUNX1, FLI1, and ERG) that regulate gene expression in hematopoietic stem and progenitor cells (HSPC; Supplementary Fig. S3B; refs. 23, 24). Approximately 40 sgRNAs within this region, with at least a twofold enrichment in the GFP^lo fraction, were selected and cloned into a lentiviral construct with iRFP720 for individual testing. The loss of GFP signal at day 7 in the iRFP⁺ fraction (gating strategy; see Supplementary Fig. S3C) highly correlated with the enrichment of those 40 sgRNAs in the GFP^lo fraction as observed in the enhancer scan (Fig. 3D). An efficiently cutting sgRNA that was not enriched in the enhancer scan did not affect GFP signal upon doxycycline exposure (Supplementary Fig. S3D and S3E). Deep amplicon sequencing of the −110-kb enhancer region upon targeting by 36 individual sgRNAs revealed frequent mutations in motifs for MYB-, GATA-, RUNX-, MEIS-, XBP-, and ETS -binding sites, which were among the highest conserved (Fig. 3E; Supplementary Fig. S3F; Supplementary Table S3).

Four sgRNAs (i.e., #3, 8, 11, and 16), generating the highest GFP^neg (EVI1^neg) fraction in the single-guide validation experiments, all targeted the same region containing a potential MYB-binding motif (Fig. 4A). The strong reduction of GFP expression, as tested for three of those guides (Fig. 4B), was accompanied by loss of EVI1 protein (Fig. 4C) and mRNA (Fig. 4D). EVI1 loss was accompanied by differentiation into CD34⁻CD15⁺ cells in the sgRNA8-targeted GFP^lo fraction (Fig. 4E), in line with the findings in primary AML cells (Fig. 1A and B, left; Supplementary Fig. S1A and S1B, left). Strikingly, sgRNA8-directed mutations within the enhancer did not affect GATA2 protein (Fig. 4C) or mRNA levels (Fig. 4D). Western blot analysis on sorted fractions of sgRNA8-treated cells revealed a strong reduction of EVI1 but not of GATA2 in GFP^lo cells (Fig. 4F). Amplicon sequencing within the GFP^lo sorted fraction of sgRNA8-treated cells revealed that almost 97% of the aligned sequences, including the translocated and nontranslocated allele, were mutated (Fig. 4G). In approximately 86% of all aligned sequences, the MYB motif was mutated. In 14%, a 20-bp deletion fully eliminated the predicted MYB DNA-binding motif (Fig. 4H). We carried out pulldown experiments in which equal amounts of MUTZ3 nuclear lysates (Supplementary Fig. S4A) were exposed to beads with immobilized 100-bp enhancer DNA fragments representing wild-type (WT) or MYB-motif mutant enhancer DNA, as defined in Fig. 4H. Western blot analysis confirmed MYB binding to the 100-bp WT enhancer fragment (Fig. 4I). MYB binding to the M1 or M2 mutants was severely reduced, but it was preserved in the M3 mutant, in which the MYB DNA–binding motif was retained (Fig. 4I). We conclude that, in inv(3)/t(3;3) AML, transcription of EVI1 depends on the presence of a MYB DNA-binding motif in the translocated enhancer. Strikingly, this MYB motif appears less relevant for the transcription of GATA2 in the nontranslocated allele.

Chromatin immunoprecipitation sequencing (ChIP-seq) revealed MYB occupancy at the −110-kb GATA2 enhancer in MUTZ3 and in inv(3)/t(3;3) AML patient cells (Fig. 5A and B, green tracks). MYB also occupied the −110-kb GATA2 enhancer in CD34⁺ cells (Supplementary Fig. S4B, green track). Based on a heterozygous single-nucleotide polymorphism (SNP) in the −110-kb GATA2 enhancer in MUTZ3, the translocated allele (EVI1) can be discriminated from the nontranslocated (GATA2) allele (12). We found approximately seven times more MYB occupancy at the translocated allele (Fig. 5A, right), in agreement with the finding that p300 occupancy (Fig. 5A, red track) was also detected predominantly at the translocated enhancer (Fig. 5A, right). Furthermore, H3K27Ac signal (Fig. 5A, right) and open chromatin (ATAC; Supplementary Fig. S4C) were five times more prevalent at the translocated enhancer. No SNPs were present in primary AMLs to discriminate MYB binding to the different alleles. However, based on two SNPs in the 18-kb region (Fig. 5B, left), we observed a strong H3K27Ac allelic skewing of the primary inv(3)/t(3;3) AML, predicted to be biased to the translocated allele (Fig. 5B, right). These data suggest that MYB and p300 interact with the −110-kb enhancer preferentially at the translocated allele. In sgRNA8-treated MUTZ3 cells (+doxycycline), MYB binding to the −110-kb site was significantly decreased compared with control (–doxycycline) cells (Fig. 5C). This loss was GATA2 enhancer–specific because genome-wide MYB chromatin occupancy, which includes the MYB target gene BCL2, did not change in +Dox cells (Supplementary Fig. S4D and S4E). Importantly, the decrease of MYB binding at the −110-kb enhancer upon sgRNA8 treatment was greater within the translocated allele (Fig. 5C, right). Using Cut&Run, we demonstrated that H3K27Ac was severely decreased at the enhancer in GFP^lo-sorted cells (Fig. 5D, blue track) compared with GFP^hi-sorted cells (Fig. 5D, green track) following sgRNA8 treatment. Moreover, SNP analysis revealed that the remaining H3K27Ac at the enhancer in GFP^lo cells occurred predominantly at the nontranslocated allele (GATA2; Fig. 5D, right). These data demonstrate that mutating the MYB-binding motif at the translocated −110-kb enhancer decreases MYB binding, thus inactivating the enhancer and reducing EVI1 transcription.

MYB is expressed in MUTZ3 cells, regardless of their differentiation status (Supplementary Fig. S4F). To study whether MYB is important for EVI1 expression, MYB-specific sgRNAs were introduced into MUTZ3-EVI1-GFP cells. At days 3 and 6 post doxycycline induction, loss of MYB expression was evident, which was accompanied by a decrease of EVI1 protein (Fig. 6A). In contrast, in line with the effects of mutating the MYB-binding motif, knockout of MYB did not decrease GATA2 protein expression (Fig. 6A). This suggests that MYB is not functioning upstream of GATA2 via this motif in inv(3) cells. When we either knocked out MYB or mutated the MYB DNA-binding motif with sgRNAs in K562 cells (Supplementary Fig. S4G), a model without a 3q26 rearrangement, we also did not see an effect on GATA2 protein levels (Supplementary Fig. S4H).

The activity of MYB can be repressed using the peptidomimetic inhibitor MYBMIM, which impairs the assembly of the MYB:CBP/p300 complex (25). In MUTZ3 cells, treatment with 25 μmol/L MYBMIM caused a 50% reduction of viable cells, whereas the inactive MYBMIM analogue TG3 showed no effect (Supplementary Fig. S4I). Treatment of MUTZ3 cells with 20 μmol/L MYBMIM strongly reduced EVI1 protein levels (Fig. 6B) without affecting MYB levels (Fig. 6B). Consistent with the MYB knockout experiment (Fig. 6A), MYBMIM treatment did not alter GATA2 protein levels (Fig. 6B). A 2-day exposure of MUTZ3 cells to MYBMIM reduced the number of colonies in methylcellulose (Fig. 6C). Flow cytometric analysis of MYBMIM-treated colony cells revealed increased maturation (CD34⁻CD15⁺ cells) in comparison with TG3-treated controls (Fig. 6D). We next introduced a FLAG-Evi1 retroviral construct (26) allowing for constitutive murine Evi1 expression in MUTZ3 cells (Supplementary Fig. S4J). Loss of colony formation upon MYBMIM treatment was partly rescued by Evi1 overexpression (Fig. 6E). Similarly, the mild effect of MYBMIM on differentiation of MUTZ3 cells (Fig. 6F, MYBMIM-EV) was reduced (Fig. 6F, MYBMIM-Evi1). This indicates that the effect of MYB interference on MUTZ3 cells is at least partly mediated via EVI1. Moreover, whereas MYBMIM treatment did not reduce MYB protein, it decreased MYB occupancy at the GATA2 enhancer (Fig. 6G). p300 occupancy also decreased but to a lesser extent than MYB (Fig. 6G). MYB binding was reduced at several sites, including the BCL2 enhancer (Supplementary Fig. S4K). MYBMIM, but not TG3, reduced viability of inv(3)/t(3;3) AML patient cells (n = 3; Fig. 6H), and treatment of AML primary cells with MYBMIM reduced EVI1 protein levels without affecting levels of MYB or GATA2 (Fig. 6I). Finally, MYBMIM affected neither GATA2 nor EVI1 levels in normal CD34⁺ cells (Fig. 6J), suggesting that MYB has no effect on the GATA2 enhancer or on EV1 in normal HSPCs. In contrast to MUTZ3 cells, MYBMIM did not reduce the number of CD34⁺ colonies in methylcellulose (Supplementary Fig. S4L). Thus, targeting MYB represents a promising therapeutic possibility in the context of inv(3)/t(3;3) AMLs with EVI1 overexpression.

Although multiple examples of hijacked enhancers causing uncontrolled expression of proto-oncogenes have been reported in various types of cancer (8, 10, 12, 27, 28), insight into their altered biological function remains limited. Elucidating these functions could provide opportunities for tailored interference and tools for therapeutic exploitation. Our unbiased CRISPR/Cas9 scan of the translocated 18-kb region in inv(3)/t(3;3) AMLs revealed a single region of approximately 1 kb essential for EVI1 activation and leukemogenesis. This distal GATA2 enhancer contained several conserved transcription factor DNA-binding motifs, including an element preferentially occupied by MYB at the translocated allele (Fig. 7, top). Strikingly, mutating this MYB-binding motif in the enhancer at both alleles strongly decreased the expression of EVI1 but not of GATA2. GATA2 was also not affected in another leukemia line or in normal HSPCs. Together, these findings support a unique role for MYB in driving EVI1 expression via the translocated enhancer and suggest a potential vulnerability in inv(3)/t(3;3) AMLs. Indeed, peptidomimetic inhibition of MYB:CBP/p300 assembly in inv(3)/t(3;3) AML cells reduced EVI1 but not GATA2 protein levels, causing myeloid differentiation and cell death (Fig. 7, bottom). This strengthens the hypothesis that interfering with EVI1 expression via MYB may constitute a new entry point for targeting these AMLs. The fact that targeting MYB specifically compromises EVI1 expression compared with GATA2 points to the possibility of selectively targeting leukemia cells while sparing GATA2 in normal HSPCs (Fig. 6J), in which GATA2 is a vital regulator.

Although MYB encodes a transcription factor essential for normal hematopoiesis (29), there is also overwhelming evidence that it plays a critical role in malignant transformation. MYB was first discovered as an oncogene (v-myb) within the avian myeloblastosis virus (AMV) genome that generated myeloid leukemias in chickens (30, 31). Its critical involvement in superenhancer activity was previously shown in human T-cell acute lymphoblastic leukemia (9, 32). Mutations in noncoding regions near TAL1 or LMO2 create de novo binding sites for MYB, leading to the formation of new MYB-bound superenhancers that drive uncontrolled transcription of those target genes. Furthermore, MYB binds to a translocated superenhancer driving MYB expression in adenoid cystic carcinoma, creating a positive feedback loop sustaining its own expression (28). MYB is also frequently overexpressed in human myeloid leukemias (33, 34), and AML cells can be addicted to high levels of MYB and thus be more vulnerable to MYB inhibition than normal hematopoietic progenitor cells (35). However, the mechanisms whereby MYB drives transformation to AML are not fully understood. To our knowledge, our results in this study represent the first example of a mechanism by which MYB drives oncogene activation in AML (Fig. 7).

MYB occupies the translocated GATA2 enhancer at a level considerably higher than the nontranslocated enhancer. This may reflect increased chromatin accessibility as determined by H3K27Ac ChIP-seq and ATAC-seq. The mechanisms driving this open chromatin pattern at the translocated locus remain a focus of future studies. However, translocation of the enhancer to a new location places it in proximity to distinct promoters and regulatory elements, which may ultimately affect chromatin accessibility and MYB binding. In support of this hypothesis, mutating the MYB DNA-binding site or interference with MYB function causes reduced expression of EVI1 but not GATA2.

The coactivators CBP and p300 are major mediators of MYB transcriptional activity (36, 37). Therefore, specifically targeting the MYB:CBP/p300 interaction has been the focus of most small molecules seeking to inhibit MYB activity (25, 38–41). Experiments using the peptidomimetic inhibitor MYBMIM, which blocks the formation of MYB:CBP/p300 complex, showed a severe loss of EVI1 activity. As reported by Ramaswamy and colleagues (25), we also observed that MYBMIM caused loss of MYB binding to the enhancer, with largely preserved total cellular levels of MYB. Concurrently, we observed that MYBMIM treatment did not inhibit p300 occupancy at the enhancer to the same extent as MYB occupancy. This partially retained p300 binding could be explained by the presence of other transcription factors bound at the GATA2 enhancer that also recruit CBP/p300 (Supplementary Fig. S3B). MYBMIM reduced MYB binding at multiple sites, which may be relevant in other leukemias in which MYB is essential (33–35). Therefore, it is not surprising that other AML cell lines (25, 42) respond to MYBMIM as well. Although initial results with MYBMIM peptide treatment of inv(3)/t(3;3) AML cells are a promising proof of concept, MYBMIM peptide is very unstable in vivo (A. Kentsis; personal communication). Thus, development of small molecules with improved bioavailability that interfere with MYB:CBP/p300 complex will be required to investigate the relevance of MYB inhibition in vivo.

Our CRISPR/Cas9 scan identified one p300-interacting region containing a MYB DNA-binding motif to be important for EVI1 expression. Although mutations in the MYB DNA-binding motif had the biggest impact on EVI1 expression, other mutations also reduced EVI1 levels. These included mutations in consensus DNA-binding sites for GATA, RUNX, MEIS, XBP, and ETS factors. Interestingly, some of these factors have been demonstrated to occupy the −110-kb enhancer in CD34⁺ cells, including RUNX1, ERG, and GATA2 (24). MYB binding and activity at the −110-kb GATA2 enhancer most likely occur in conjunction with p300 as well as transcription factors such as RUNX1 and ERG. This is in accordance with other studies showing colocalization and potential cooperation between these factors and MYB (25, 43, 44). Therefore, combinatorial targeting of MYB and other transcription factors may synergistically affect EVI1 expression. This knowledge provides a rationale to develop new compounds to treat inv(3)/t(3;3) AML, which can be tested in our newly developed model.

Our findings provide important insight into the mechanisms of oncogenic enhancer-driven gene activation in AML. The selective MYB motif requirement for enhancer function at the translocated but not the normal allele constitutes a novel paradigm in which chromosomal aberrations reveal critical motifs that are nonfunctional at their endogenous locus. In principle, this paradigm may be extrapolated to other enhancer-driven cancers and even nonmalignant pathologies.

Cell line sequence data generated in this study have been deposited at the EMBL-EBI ArrayExpress database (ArrayExpress, RRID:SCR_002964) under accession numbers E-MTAB-9939 (RNA sequencing); E-MTAB-9949 (ATAC-seq), E-MTAB-9946 (Cut&Run sequencing), E-MTAB-9945 (amplicon sequencing), E-MTAB-9948 (CRISPR enhancer scan), and E-MTAB-9959 (ChIP-seq). ChIP-seq and ATAC-seq data derived from donors or patients have been deposited at the European Genome-phenome Archive (RRID:SCR_004944) under the accession number EGAS00001004839. This study did not generate any unique codes. All software tools used in this study are freely or commercially available.

The MUTZ3 cell lines (DSMZ cat. ACC-295, RRID:CVCL_1433) were cultured in αMEM (HyClone) with 20% FCS and 20% conditioned 5637 medium. The 293T cells (DSMZ catalog no. ACC-635, RRID:CVCL_0063) were cultured in DMEM (Gibco) with 10% FCS. K562 cells (DSMZ catalog no ACC-10, RRID:CVCL_0004) were cultured in RPMI (Gibco) with 10% FCS. All cell lines were supplemented with 50 U/mL penicillin and 50 μg/mL streptomycin. Viable frozen AML cells and viable (frozen) bone marrow or cord blood CD34⁺ cells were thawed and suspended in Iscove's Modified Dulbecco's Medium supplemented with 20% BIT medium (StemCell Technologies), 1× β-mercaptoethanol (1,000×; Life Technologies), 6 μg/mL LDL (Sigma-Aldrich), human IL6, IL3, G-CSF, GM-CSF at 20 ng/mL, and FLT3 and SCF at 50 ng/mL (PeproTech). Cell lines were obtained from DSMZ and regularly confirmed to be Mycoplasma free by the MycoAlert Mycoplasma Detection Kit (LT07–318; Lonza) according to the manufacturer's instructions.

The repair template was generated using Gibson Assembly (NEB). Both homology arms were PCR amplified from MUTZ3 genomic DNA using Q5 polymerase (NEB). The first homology arm consists of a part of the intron and last exon of EVI1 minus the STOP codon. The second homology arm consists of part of the 3′–untranslated region with the PAM sequence of sgRNA omitted. The T2A-eGFP was PCR amplified from dCAS9-VP64_2A_GFP (RRID:Addgene_61422). All fragments were cloned using Gibson assembly into the PUC19 (Invitrogen) backbone. sgRNA sequence AGCCACGTATGACGTTATCA was cloned into pX330-U6-Chimeric_BB-CBh-hSpCas9. Cells were nucleofected with pX330 vector (RRID:Addgene_42230) containing the sgRNA and Cas9 and the repair template using the Nucleofector 4D (Lonza) with Kit SF and program DN-100. GFP⁺ cells were sorted using a FACS AriaIII (BD Biosciences). In a second sorting round, GFP⁺ cells were single cell sorted and tested for proper integration. Clone 1A5 was transduced with lenti pCW-Cas9 (RRID:Addgene_50661), puromycin selected (1 μg/mL), and subsequently single cell sorted based on GFP positivity and tested for inducible Cas9 expression. Clone 3E7 was used for the screen, which we called MUTZ3-EVI1-GFP.

Samples of the selected patients presenting with AML were collected from the Erasmus MC Hematology Department biobank (Rotterdam, the Netherlands). The karyotype of patients with AML used in this study was as follows: AML-1, 45,XX, inv(3)(q2?1q26),–7; AML-2, 45,XY,inv(3)(q22q26),–7; and AML-3, 45,XX,t(3;3)(q21;q26),–7. Leukemic blast cells were purified from bone marrow or blood by standard diagnostic procedures. All patients provided written informed consent in accordance with the Declaration of Helsinki. The Medical Ethical Committee of the Erasmus MC has approved usage of the patient rest material for this study.

Cells were lysed in lysis buffer (20 mmol/L Tris-HCl, 138 mmol/L NaCl, 10 mmol/L EDTA, 50 mmol/L NaF, 1% Triton, 10% glycerol, 2 mmol/L NA-vanadate) containing Complete Protease Inhibitors (CPI; Roche #4693159001). Protein levels were detected using antibodies against EVI1 (Cell Signaling Technology, cat. 2265, RRID:AB_561424), MYB (Millipore, catalog no. 05–175, RRID:AB_2148022), FLAG (Sigma-Aldrich, catalog no. F3165, RRID:AB_259529), β-Actin (Sigma-Aldrich, catalog no. A5441, RRID:AB_476744), GAPDH (Santa Cruz Biotechnology, catalog no. sc-25778, RRID:AB_10167668), CAS9 (BioLegend, catalog no. 844301, RRID:AB_2565570), or GATA2 (kind gift of E.H. Bresnick, Department of Cell and Regenerative Biology, Madison, Wisconsin). Proteins were visualized using the Odyssey infrared imaging system (LI-COR Biosciences).

Cell sorting was performed using the FACS Aria flow cytometer (BD Biosciences, RRID:SCR_013311) into a 96-well plate format or into batch culture. Flow cytometric analysis on MUTZ3 cells was done with GFP/RFP or antibody stainings for CD34-PE-CY7 (BD Biosciences, catalog no. 348811, RRID:AB_2868855) and CD15-APC (Sony, #2215035) or CD15-BV510 (BioLegend, catalog no. 323028, RRID:AB_2563400). Intracellular stainings with EVI1 (Cell Signaling Technology, catalog no. 2256, RRID:AB_561017) or Rabbit (DA1E) mAb IgG XP Isotype Control (Cell Signaling Technology, catalog no. 3900, RRID:AB_1550038) were performed using Foxp3/Transcription Factor Staining Buffer Set (00–5523–00; eBioscience). Cells were measured on a BD Canto or BD LSR II flow cytometer (BD Biosciences), and data were analyzed using FlowJo software (FlowJo, RRID:SCR_008520).

Nuclear lysates for pulldown experiments were prepared as described (45). Oligonucleotides for affinity purification were ordered as custom-synthesized oligos from Integrated DNA Technologies (IDT; see Supplementary Table S4). DNA pulldown was performed as described by Karemaker and Vermeulen (45) with minor changes. Essentially, per DNA pulldown, 500 pmol of annealed oligos was diluted to 600 μL in DNA-binding buffer (DBB; 1 mol/L NaCl, 10 mmol/L Tris pH 8.0, 1 mmol/L EDTA, 0.05% NP40) and incubated with washed beads [10 μL Streptavidin Sepharose High performance bead slurry (GE Healthcare #17511301), washed once with PBS + 0.1% NP-40 and once with DBB] for 30 minutes at 4°C while rotating. After washing once with 1 mL DBB and twice with 1 mL protein incubation buffer [PIB; 150 mmol/L NaCl, 50 mmol/L Tris pH 8.0, 0.25% NP40, 1 mmol/L DTT with CPI (Roche #4693159001)], the immobilized oligos on beads were combined with 500 μg nuclear extracts in a total volume of 600 μL PIB with 10 μg competitor DNA [5 μg poly-dldC (Sigma-Aldrich #81349_500ug) and 5 μg poly-dAdt (Sigma-Aldrich #P0883_50UN)] and incubated for 90 minutes at 4°C while rotating. Beads were washed three times with 1 mL PIB and twice with 1 mL PBS. To elute proteins from the oligo probes, beads were resuspended in 20 μL 1× Western blot protein sample buffer and incubated at 95°C for 15 minutes while shaking. The beads were spun down, and the eluate was loaded on a protein gel. A 40-μg nuclear extract sample was prepared directly from the nuclear lysate as input sample for Western blot.

MUTZ3, primary AMLs, or CD34⁺ cells were cultured in medium as described above, plus MYBMIM or control peptide TG3 at indicated concentrations. For measuring viability of MUTZ3 or primary AMLs, cells were seeded in an opaque-colored 96-well plate at 15,000 cells/well in a total volume of 100 μL medium containing MYBMIM or control peptide TG3 at indicated concentrations (20 μmol/L MYBMIM or control peptide TG3 for primary AMLs). Cell viability was assessed 72 hours after treatment using CellTiter-Glo cell viability assay according to the manufacturer's protocol (Promega). Luminescence was measured on the Victor X3 plate reader (PerkinElmer). Rescue experiments in MUTZ3 cells were performed by retroviral overexpression of murine pMY-FLAG-Evi1-IRES-GFP or an empty vector (EV) control. The pMY vectors were kind gifts of T. Sato (26) in which FLAG was inserted 5′ of Evi1. Evi1- or EV-overexpressing cells were cultured in the presence of 20 μmol/L MYBMIM for 48 hours. For colony cultures following peptide treatment, 2,000 MUTZ3 cells or 500 CD34⁺ cells were plated in MethoCult (StemCell Technologies) with 100 U/mL penicillin/streptomycin. For protein lysates and ChIP experiments, cells were cultured containing 20 μmol/L MYBMIM or control peptide TG3 and harvested after 48 hours of peptide treatment.

FISH was performed and reported according to standard protocols based on the International System of Human Cytogenetics Nomenclature (2016; ref. 46). MECOM FISH was performed according to the manufacturer's protocol using the MECOM t(3;3); inv(3)(3q26) triple-color probe (Cytocell, LPH-036).

The sgRNAs (Supplementary Table S1) were cloned into pLentiV2_U6-IT-mPgk-iRFP720 (J. Zuber) using BsmBI restriction sites or px330 using BbsI or were in vitro transcribed using the T7 promoter. Lentiviruses were prepared by transfecting 293T cells with lentiviral packaging constructs pSPAX2/pMdelta2.G and sgRNA cloned into pLentiV2_U6-IT-mPgk-iRFP720. Transfections were performed using Fugene 6 (Promega) according to the manufacturer's protocol. For in vitro transcribed sgRNA oligos containing the T7 promoter, target sequence and the Tail annealing sequence were annealed, filled in, and transcribed using the Hi-scribe T7 kit (NEB). Turbo DNAse (Invitrogen) was added and sgRNAs were cleaned up using the RNA clean&concentrator kit (Zymo). Concentration of sgRNAs was estimated using Qubit (Invitrogen). RNP complexes were formed incubating sgRNA and Cas9 (IDT) for 20 to 30 minutes at room temperature before nucleofection using the Neon (Thermo Fisher) with buffer R with settings 1,500 V, 20 ms, 1 pulse for MUZT3 or 1,350 V, 10 ms, 4 pulses for K562. Genomic DNA was extracted at an indicated timepoint after transfection using Quick Extract buffer (Epicenter) PureLink Genomic DNA Mini Kit (Invitrogen) and checked for targeting by PCR using Q5 polymerase (NEB) or amplicon sequencing.

To design a high-resolution sgRNA library for the enhancer scan, we considered all possible sgRNA target sites containing a canonical Cas9 PAM site (NGG) on both strands of the minimal 18-kb translocated region. sgRNAs containing a G in positions 1 to 3 of the 20-nt target site were trimmed at this position to favor 20-, 19- or 18-mers (in this order of priority) containing a natural G at the 5′ end as previously described (47). For all other sgRNAs, a G was added to the 5′ end (resulting in a 21-mer). Subsequently, all sgRNAs showing (i) a high number of target sites in the human genome (>5 with no mismatch or >20 with 1 mismatch), (ii) a BsmBI site (interfering with cloning), or (iii) a polyA signal (interfering with packaging) were filtered out. In addition, we added a number of negative controls (82 sgRNAs targeting the AAVS1 region) as well as positive controls (5 sgRNAs targeting EVI1 as well as 313 sgRNAs covering 5 kb of the breakpoint in MUTZ3 cells). The final library of 3,239 sgRNAs (Supplementary Table S2) was synthesized with overhangs for PCR amplification and cloning as one oligo pool (Twist Bioscience) and cloned into the lentiviral vector sgETN (J. Zuber) as previously described (47). The pool of 3,239 sgETN-sgRNAs was transduced in triplicate into MUTZ3-EVI1-GFP. For each replicate, a total of 120 million cells were infected with 3% to 4% transduction efficiency to ensure that each sgRNA is represented predominantly as a single lentiviral integration in >1,000 cells. After neomycin drug selection (1 mg/mL) for 7 days, T0 samples were obtained (5 million cells per replicate), and cells were subsequently cultured in the presence of 1 μg/mL doxycycline. Culture medium was exchanged every 2 days. After 5 days (T5) and 7 days (T7), about 1 million sgRNA-expressing (GFP^lo) cells were sorted for each replicate using a FACS AriaII (BD Biosciences). Genomic DNA from T0, T5, and T7 samples was isolated by two rounds of phenol extraction using PhaseLock tubes (5PRIME), followed by isopropanol precipitation. Deep sequencing libraries were generated by PCR amplification of sgRNA guide strands using primers that tag the product with standard Illumina adapters and a 4-bp sample barcode in a two-step PCR protocol. For each sorted sample, all DNA was used as template in multiple parallel 50-μL PCR reactions, each containing 250 to 500 ng template, 1× AmpliTaq Gold buffer, 0.2 mmol/L of each dNTP, 2 mmol/L MgCl₂, 0.3 μmol/L of each primer, and 1 U AmpliTaq Gold (Invitrogen), which were run using the following cycling parameters: 95°C for 10 minutes; 28 cycles of 95°C for 30 seconds, 52°C for 45 seconds, and 72°C for 30 seconds; and 72°C for 7 minutes. PCR products (367 bp) were combined for each sample and Ampure purified. For the T0 samples and a DNA-pool sample, the amount of input DNA necessary to get a 1,000× coverage was used as input in the PCRs. For the second PCR, 10 ng of input was used per PCR using the following cycling parameters: 95°C for 10 minutes; eight cycles of 95°C for 30 seconds, 57°C for 45 seconds, and 72°C for 30 seconds; and 72°C for 7 minutes. PCR products (448 bp) were combined for each sample and Ampure purified. Libraries were sequenced equimolarly on an Illumina HiSeq 2500 (Illumina) by the Next Generation Sequencing Facility at Vienna BioCenter Core Facilities, member of the Vienna BioCenter, Austria. Multiple experiments (different time points and sorted fractions) were sequenced simultaneously, each identified by a unique barcode. Sequencing data were processed by converting unaligned BAM files into FASTA using bam2fastx. Experiment-specific barcodes (positions 7–10) were extracted together with the sgRNA sequence (positions 31–) into a new FASTA file, which was subsequently reverse-complemented with seqtk seq. Next, the barcodes were used to demultiplex the FASTA file into experiment-specific files with ngs-tools split-by-barcode, using parameters –s 4 –d 1 (i.e., barcode size 4 and maximum 1 mismatch). For each of these files, we counted the number of identical sgRNA sequences with fastx_collapser and assigned them to their known identifiers. These counts were employed for downstream data analysis. To provide a sufficient baseline for detecting sgRNA enrichment in experimental samples, we aimed to acquire >1,000 reads per sgRNA in the sequenced sgRNA pool to compensate for variation in sgRNA representation inherent in the pooled plasmid preparation or introduced by PCR biases. Reads were normalized to the total number of library-specific reads per lane for each condition. To ensure a proper sgRNA representation in the initial plasmid pool, we used a cutoff of more than 10% average reads/sgRNA sequenced in the plasmid pool (resulting in passing of 3,050 out of 3,239 sgRNAs). Enrichment analyses were performed using MAGeCK (48).

H3K27Ac and p300 ChIP-seq data from the inv(3) cell line MOLM1 as well as p300 ChIP-seq data from MUTZ3 were previously generated by our group and are available at ArrayExpress E-MTAB-2224 (12). H3K27Ac (Abcam, catalog no. ab4729, RRID:AB_2118291) ChIPs were performed according to the standard ChIP protocol from Upstate. ChIP with antibodies direct against MYB (Millipore, catalog no. 05-175, RRID:AB_2148022) or p300 (Diagenode, #C15200211) was performed by first cross-linking for 45 minutes with DSG before formaldehyde crosslinking. ChIP samples were processed according to the Illumina TruSeq ChIP Sample Preparation Protocol (Illumina) or Diagenode Library V3 preparation protocol (Diagenode) and either sequenced single-end (1 × 50 bp) on the HiSeq 2500 platform (Illumina) or paired-end (2 × 100 bp) on the Novaseq 6000 platform (Illumina). Briefly, reads were aligned to the human reference genome build hg19 with bowtie (49) for single-end runs and bowtie2 (50) for paired-end runs, and bigwig files were generated for visualization with bedtools genomecov (51) and UCSC bedGraphToBigWig (52). Peaks were determined using the MACS2 program with default parameters (53). The tracks were normalized per million reads and visualized as genome browser profiles using the Fluff package (54).

H3K27Ac (Abcam, catalog no. ab4729, RRID:AB_2118291) Cut&Run libraries for the MUTZ3 bulk and sorted fragments were generated with an input of 200,000 cells. The protocol described by Henikoff and colleagues was used to generate these tracks (55), using a 0.04% Digitonin buffer and with the addition of cOmplete, EDTA-free Protease Inhibitor Cocktail (Roche) and 1 mol/L sodium butyrate (Sigma-Aldrich) to all the buffers. Isolation was done according to the standard phenol chloroform protocol. Cut&Run samples were processed according to the protocol described by the Fazzio group (56) and sequenced paired-end (2 × 100 bp) on the Novaseq 6000 platform (Illumina). Reads were aligned similarly to ChIP-seq.

Open chromatin regions were mapped by the ATAC-seq method as described (57) with a modification in the lysis buffer (0.30 mol/L sucrose, 10 mmol/L Tris pH 7.5, 60 mmol/L KCl, 15 mmol/L NaCl, 5 mmol/L MgCl₂, 0.1 mmol/L EGTA, 0.1% NP40, 0.15 mmol/L spermine, 0.5 mmol/L spermidine, 2 mmol/L 6AA) to reduce mitochondrial DNA contamination. ATAC-seq samples were sequenced paired-end (2 × 50 bp) on the HiSeq 2500 platform (Illumina) and aligned against the human genome (hg19) with bowtie2, allowing for a maximum 2,000-bp insert size. Mitochondrial reads and fragments with mapping quality below 10 were removed.

RNA was isolated using Trizol or the Qiagen Allprep DNA/RNA kit and protocol (Qiagen, #80204). cDNA synthesis was done using the SuperScript II Reverse Transcriptase kit (Invitrogen). Quantitative real-time PCR was performed by using primers (Supplementary Table S4) as described previously (15) on the 7500 Fast Real-time PCR System (Applied Biosystems). For RNA sequencing, sample libraries were prepped using 500 ng input RNA according to the KAPA RNA HyperPrep Kit with RiboErase (HMR; Roche) using Unique Dual Index adapters (Integrated DNA Technologies, Inc.). Amplified sample libraries were paired-end sequenced (2 × 100 bp) on the Novaseq 6000 platform (Illumina) and aligned against the human genome (hg19) using STAR version 2.5.4b. Salmon (58) was used to quantify expression of individual transcripts, which were subsequently aggregated to estimate gene-level abundances with tximport (59). Human gene annotation derived from RefSeq (60) was downloaded from UCSC (ref. 61; RefGene) as a GTF file. Transcript-level abundances were normalized to transcripts per million for visualization.

For amplicon sequencing, we used a PCR-based NGS library preparation method in combination with the TruSeq Custom Amplicon index kit (Illumina). The first PCR for target selection (Supplementary Table S4) was performed using Q5 polymerase (NEB), the second nested PCR, to add the index-adapters, with KAPA HiFi HotStart Ready mix (Kapa Biosystems). Libraries were sequenced paired-end (2 × 250 bp) on the MiSeq platform (Illumina). Reads were trimmed with trimgalore (Trim Galore, RRID:SCR_011847) to remove low-quality bases and adapters and subsequently aligned to the human reference genome build hg19 with BBMap (BBmap, RRID:SCR_016965) allowing for 1,000-bp indels. Mutations introduced by genome editing were analyzed and visualized using CRISPResso2 (62). Mutated sequences consisting of up to 5% of sequenced reads were next analyzed for differential binding with CIS-BP (63).

A. Kentsis reports personal fees from Novartis and personal fees from Rgenta during the conduct of the study; in addition, A. Kentsis has a patent for agents and methods for treating CREB binding protein-dependent cancers pending. J. Zuber reports other support from Boehringer Ingelheim & Co KG during the conduct of the study; personal fees and other support from Quantro Therapeutics GmbH outside the submitted work. No disclosures were reported by the other authors.

L. Smeenk: Conceptualization, data curation, software, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. S. Ottema: Investigation, writing–review and editing. R. Mulet-Lazaro: Data curation, software, formal analysis, writing–review and editing. A. Ebert: Investigation.M. Havermans: Investigation. A. Arricibita Varea: Investigation. M. Fellner: Investigation. D. Pastoors: Investigation. S. van Herk: Investigation. C. Erpelinck-Verschueren: Investigation. T. Grob: Investigation. R.M. Hoogenboezem: Data curation. F.G. Kavelaars: Resources. D.R. Matson: Resources. E.H. Bresnick: ResourcesE.M. Bindels: Investigation. A. Kentsis: Resources. J. Zuber: Conceptualization, resources. R. Delwel: Conceptualization, resources, data curation, software, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing.

We thank our colleague Michael Vermeulen and the Bioptics Facility at IMP for flow cytometric sorting. We are thankful to Tobias Neumann for help with the design of the enhancer scanning strategy as well as to the Zuber group at the IMP for their help with the enhancer scanning CRISPR/Cas9 experiments. We thank the Vermeulen group at the Radboud Institute for Molecular Life Sciences for assistance in the DNA-pulldown experiments. Furthermore, we acknowledge Berna Beverloo and the Department of Clinical Genetics for the FISH analysis and colleagues from the bone marrow transplantation group and the molecular diagnostics laboratory of the Department of Hematology for storage of samples and molecular analysis of the leukemia cells. This work was funded by a fellowship from the Daniel den Hoed, Erasmus MC Foundation (to L. Smeenk), the Koningin Wilhelmina Fonds grant from the Dutch Cancer Society (to R. Delwel), the NIH grant R01 DK68634 (to E.H. Bresnick), Carbone Cancer Center P30 CA014520 (to E.H. Bresnick), the NIH T32 HL07899 (to D.R. Matson), and FWF-SFB grant F4710 of the Austrian Science Fund (to J. Zuber). Research at the IMP was generously supported by Boehringer Ingelheim and the Austrian Research Promotion Agency (headquarter grant FFG-852936).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.