CEBPA repression by MECOM blocks differentiation to drive aggressive leukemias

Abstract

Acute myeloid leukemias (AMLs) have an overall poor prognosis with many high-risk cases co-opting stem cell gene regulatory programs, but the mechanisms through which these programs are propogated remain poorly understood. The increased expression of the stem cell transcription factor, MECOM, underlies a key driver mechanism in largely incurable AMLs. However, how MECOM results in such aggressive AML phenotypes remains unknown. To address existing experimental limitations, we engineered and applied targeted protein degradation with functional genomic readouts to demonstrate that MECOM promotes malignant stem cell-like states by directly repressing prodifferentiation gene regulatory programs. Remarkably and unexpectedly, a single node in this network, a MECOM-bound cis-regulatory element located 42 kilobase (kb) downstream of the myeloid differentiation regulator CEBPA is both necessary and sufficient for maintaining MECOM-driven leukemias. Importantly, the targeted activation of this regulatory element promotes differentiation of these aggressive AMLs and reduces leukemia burden in vivo. These findings suggest a broadly applicable approach for functionally dissecting oncogenic gene regulatory networks to inform improved therapeutic strategies.

Graphical abstract

Figure 1.

FKBP12^F36V degron facilitates rapid degradation of endogenous MECOM in AML cells. (A) Schematic illustrating the gene-editing strategy to knock in an FKBP12^F36V degron, 2xHA tag, and eGFP at the C-terminus of the endogenous MECOM locus in human MUTZ-3 AML cells. (B) GFP expression assessed by flow cytometry in CD34⁺ vs CD34^– MUTZ-3 MECOM-FKBP12^F36V cells. (C) Time course western blot analysis of MECOM protein levels in MUTZ-3 cells following treatment with dTAG^V-1 (5-25 nM) or dimethyl sulfoxide (DMSO). (D) Volcano plot showing changes in protein abundance in MUTZ-3 MECOM-FKBP12^F36V cells treated for 2 hours with 500 nM dTAG^V-1 vs DMSO as assessed by mass spectrometry. n = 3 independent replicates. (E) MECOM ChIP-seq of MUTZ-3 MECOM-FKBP12^F36V cells treated with 500 nM dTAG^V-1 vs DMSO (n = 3). Each row represents a single MECOM(HA)-bound peak. Heat map is centered on ChIP-peak summits ±500 bp. (F) Bivariate plot showing CD34 and CD14 expression levels in MUTZ-3 MECOM-FKBP12^F36V cells treated with 500 nM dTAG^V-1 vs DMSO. (G-H) Percentage of CD34⁺ and CD14⁺ cells as observed in panel F. n = 3 independent replicates. Mean and standard error of the mean (SEM) are shown. (I) Viable cell count by trypan blue exclusion of MUTZ-3 MECOM-FKBP12^F36V cells treated with 500 nM dTAG^V-1 vs DMSO. n = 3 independent replicates. Mean and SEM are shown. eGFP, enhanced green fluorescent protein; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; kDA, kilodalton.

Figure 2.

Multiomic profiling of MECOM-depleted cells reveals a predominantly repressive role at target sites. (A) Schematic representation of experimental protocol for multiomic characterization of dTAG^V-1–treated MUTZ-3 MECOM-FKBP12^F36V cells. The CD34⁺, GFP⁺ MECOM-expressing population was preenriched via MACS before treatment with 500 nM dTAG^V-1 or DMSO. Cells were then harvested and processed for bulk RNA-seq, ATAC-seq, and PRO-seq to profile transcriptional and epigenetic changes. (B-C) Volcano plots representing changes in nascent gene expression assessed via PRO-seq in MUTZ-3 MECOM-FKBP12^F36V cells treated with dTAG^V-1 vs DMSO for 1 and 4 hours. n = 3 independent replicates (supplemental Table 4). (D) Volcano plot representing changes in gene expression assessed via bulk RNA-seq in MUTZ-3 MECOM-FKBP12^F36V cells treated with dTAG^V-1 vs DMSO for 6 hours. n = 3 independent replicates (supplemental Table 3). (E) Volcano plot representing changes in chromatin accessibility as assessed by ATAC-seq in MUTZ-3 MECOM-FKBP12^F36V cells treated with dTAG^V-1 vs DMSO for 6 hours. n = 3 independent replicates. Red data points represent chromatin peaks that are also bound by MECOM as assessed by MECOM-HA ChIP-seq. There are 837 of these sites that are schematically highlighted in the top right corner of the plot (supplemental Table 5). (F-G) Assessment of eRNA transcription levels at 837 MECOM-bound differentially accessible peaks measured from PRO-seq data. (F) Average PRO-seq read density across all MECOM-regulated cisREs with ±2000 bp on each side of the peak summit in dTAG^V-1–treated vs DMSO-treated samples. (G) Box plot showing average PRO-seq read density in aggregate for each MECOM-regulated cisRE ±500 bp on each side of the peak summit in dTAG^V-1–treated vs DMSO-treated samples. Two-sided Student t test was used for comparisons. n = 3 independent replicates. (H) Unbiased motif enrichment analysis of ATAC-seq differentially accessible peaks between dTAG^V-1–treated and DMSO-treated samples. (I) Venn diagram comparing gene expression and chromatin accessibility changes across sequencing modalities. Bulk RNA-seq DEGs from 6 hours and 24 hours dTAG^V-1 treatment, PRO-seq DEGS from 4 hours dTAG^V-1 treatment, and genes in proximity (within 1 MB) to at least 1 MECOM-bound, differentially accessible ATAC-seq peak were overlapped to yield a consensus MECOM gene network consisting of 122 genes. Cutoffs for bulk RNA-seq and PRO-seq were P < .05 (supplemental Tables 6 and 7). Peak-to-gene proximity was determined using the GREAT. (J) Schematic depiction of MECOM’s interaction with transcriptional corepressor CtBP2 via MECOM’s PLDLS motif. This protein-protein interaction can be inhibited by a genetically encoded 4x-PLDLS peptide inhibitor (top) or if MECOM’s PLDLS motif were mutated to PLASS (bottom). (K) H3K27ac and CtBP2 ChIP-seq analysis. Heat map (left) displays CtBP2 ChIP-seq signal at MECOM-regulated cisREs in MUTZ-3 cells expressing a 4x-PLDLS peptide inhibitor of the MECOM-CtBP2 interaction compared with cells expressing 4x-PLASS control. Heat map (right) showing H3K27ac ChIP-seq signal at MECOM-regulated cisREs in MUTZ-3 MECOM-FKBP12^F36V cells treated with 500 nM dTAG^V-1 or DMSO for 6 hours. (L-M) Experimental overview for lentiviral MECOM add-back rescue experiment. (L) MUTZ-3 MECOM-FKBP12^F36V cells were transduced with lentiviruses constitutively expressing either WT MECOM (EVI1 isoform) or MECOM PLDLS>PLASS along with a TagRFP transduction reporter at high MOI. (M) CD34 expression assessed by flow cytometry as a function of treatment duration (500 nM DMSO vs dTAG^V-1) (bottom). Histogram of CD34 expression at day 15 (top). Samples were transduced 48 hours before treatment. n = 3 independent technical replicates. Mean and standard deviation are shown, but many are hidden due to low variation between replicates. DEGs, differentially expressed genes; DAPs, differentially accessible peaks; eRNA, enhancer RNA; GFP, green fluorescent protein; MACS, magnetic-activated cell sorting; MB, megabase; MOI, multiplicity of infection; ns, not significant; nt, nucleotide; TF, transcription factor; WT, wild-type.

Figure 2.

Multiomic profiling of MECOM-depleted cells reveals a predominantly repressive role at target sites. (A) Schematic representation of experimental protocol for multiomic characterization of dTAG^V-1–treated MUTZ-3 MECOM-FKBP12^F36V cells. The CD34⁺, GFP⁺ MECOM-expressing population was preenriched via MACS before treatment with 500 nM dTAG^V-1 or DMSO. Cells were then harvested and processed for bulk RNA-seq, ATAC-seq, and PRO-seq to profile transcriptional and epigenetic changes. (B-C) Volcano plots representing changes in nascent gene expression assessed via PRO-seq in MUTZ-3 MECOM-FKBP12^F36V cells treated with dTAG^V-1 vs DMSO for 1 and 4 hours. n = 3 independent replicates (supplemental Table 4). (D) Volcano plot representing changes in gene expression assessed via bulk RNA-seq in MUTZ-3 MECOM-FKBP12^F36V cells treated with dTAG^V-1 vs DMSO for 6 hours. n = 3 independent replicates (supplemental Table 3). (E) Volcano plot representing changes in chromatin accessibility as assessed by ATAC-seq in MUTZ-3 MECOM-FKBP12^F36V cells treated with dTAG^V-1 vs DMSO for 6 hours. n = 3 independent replicates. Red data points represent chromatin peaks that are also bound by MECOM as assessed by MECOM-HA ChIP-seq. There are 837 of these sites that are schematically highlighted in the top right corner of the plot (supplemental Table 5). (F-G) Assessment of eRNA transcription levels at 837 MECOM-bound differentially accessible peaks measured from PRO-seq data. (F) Average PRO-seq read density across all MECOM-regulated cisREs with ±2000 bp on each side of the peak summit in dTAG^V-1–treated vs DMSO-treated samples. (G) Box plot showing average PRO-seq read density in aggregate for each MECOM-regulated cisRE ±500 bp on each side of the peak summit in dTAG^V-1–treated vs DMSO-treated samples. Two-sided Student t test was used for comparisons. n = 3 independent replicates. (H) Unbiased motif enrichment analysis of ATAC-seq differentially accessible peaks between dTAG^V-1–treated and DMSO-treated samples. (I) Venn diagram comparing gene expression and chromatin accessibility changes across sequencing modalities. Bulk RNA-seq DEGs from 6 hours and 24 hours dTAG^V-1 treatment, PRO-seq DEGS from 4 hours dTAG^V-1 treatment, and genes in proximity (within 1 MB) to at least 1 MECOM-bound, differentially accessible ATAC-seq peak were overlapped to yield a consensus MECOM gene network consisting of 122 genes. Cutoffs for bulk RNA-seq and PRO-seq were P < .05 (supplemental Tables 6 and 7). Peak-to-gene proximity was determined using the GREAT. (J) Schematic depiction of MECOM’s interaction with transcriptional corepressor CtBP2 via MECOM’s PLDLS motif. This protein-protein interaction can be inhibited by a genetically encoded 4x-PLDLS peptide inhibitor (top) or if MECOM’s PLDLS motif were mutated to PLASS (bottom). (K) H3K27ac and CtBP2 ChIP-seq analysis. Heat map (left) displays CtBP2 ChIP-seq signal at MECOM-regulated cisREs in MUTZ-3 cells expressing a 4x-PLDLS peptide inhibitor of the MECOM-CtBP2 interaction compared with cells expressing 4x-PLASS control. Heat map (right) showing H3K27ac ChIP-seq signal at MECOM-regulated cisREs in MUTZ-3 MECOM-FKBP12^F36V cells treated with 500 nM dTAG^V-1 or DMSO for 6 hours. (L-M) Experimental overview for lentiviral MECOM add-back rescue experiment. (L) MUTZ-3 MECOM-FKBP12^F36V cells were transduced with lentiviruses constitutively expressing either WT MECOM (EVI1 isoform) or MECOM PLDLS>PLASS along with a TagRFP transduction reporter at high MOI. (M) CD34 expression assessed by flow cytometry as a function of treatment duration (500 nM DMSO vs dTAG^V-1) (bottom). Histogram of CD34 expression at day 15 (top). Samples were transduced 48 hours before treatment. n = 3 independent technical replicates. Mean and standard deviation are shown, but many are hidden due to low variation between replicates. DEGs, differentially expressed genes; DAPs, differentially accessible peaks; eRNA, enhancer RNA; GFP, green fluorescent protein; MACS, magnetic-activated cell sorting; MB, megabase; MOI, multiplicity of infection; ns, not significant; nt, nucleotide; TF, transcription factor; WT, wild-type.

Figure 3.

Direct MECOM gene network is repressed in primary leukemia cells. (A) UMAP of 40 866 cells derived from 11 patients with leukemias driven by MLLrs sequenced using scRNA-seq. UMAPs were colored from top to bottom by patient, whether MECOM is expressed, and MECOM expression counts per cell. (B) Same UMAPs as panel A and colored by the expression signatures of normal HSCs and normal monocytes derived from Lambo et al and MECOM-regulated genes identified to be activated after depletion of MECOM (Figure 2). (C) Quantification of the 3 signatures from panel B compared between MECOM-positive leukemias (n = 5) and MECOM-negative leukemias (n = 6). Comparisons were performed by randomly taking the average over 10 iterations of 1000 randomly sampled cells from both samples expressing MECOM and samples not expressing MECOM to avoid uninformative P values close to 0. Significance was calculated using 2-sided Wilcoxon signed-rank tests corrected for multiple testing using BH. (D) Differential expression of all analyzed genes (n = 28 113) between leukemias expressing MECOM and leukemias that did not express MECOM. Differential expression was performed using MAST using 10 iterations of 1000 randomly selected MECOM-positive cells and 1000 randomly selected MECOM-negative cells to prevent uninformative P values. BH corrected P values and log fold changes shown are the average of 10 iterations. BH, Benjamini-Hochberg; UMAP, uniform manifold approximation and projection.

Figure 4.

Direct MECOM chromatin network is repressed in primary leukemia cells. (A) UMAP of 64 682 cells generated using scATAC-seq from remissions of patients with pediatric AML (n = 20). Cells were colored by predicted cell type derived using label transfer of matching scRNA-seq data. Labels were derived from Lambo et al. (B) UMAP showing the same cohort as panel A and colored by lineage scores. Lineage scores were calculated as the total insertions in 5000 accessible sites in each lineage, normalized to accessible sites in the other 2 lineages (accessible sites derived from Lambo et al³⁶). (C) Cells colored by chromatin accessibility at MECOM-bound loci that were identified to increase in accessibility after the depletion of MECOM. Scores were calculated by ATAC-seq reads at MECOM-bound loci divided by ATAC-seq reads in the TSS and corrected for Tn5 bias. Scores were scaled to the 99th quantile to reduce the effect of outliers. (D) Spearman correlation between lineage scores and MECOM cisRE scores. Each dot represents 1 cell, with cells colored by density. (E) UMAP showing a trajectory inferred using Monocle from inferred HSCs to inferred monocytes. (F) UMAP showing the scaled expression of MECOM in counts from linked scRNA data across cells from remissions. (G) Heat maps showing the scaled expression of MECOM-regulated genes (n = 122) and cisREs (n = 837) along the monocyte trajectory (pseudotime). Each column represents an aggregated minibulk from cells across the inferred pseudotime (100 bins in total). Normalized gene expression scores are derived from linked scRNA samples, with ATAC-seq signal normalized by TSS insertions and Tn5 bias. Both gene expression and ATAC-seq signal were scaled across all cells in the pseudotime. cDC, conventional dendritic cell; CLP, common lymphoid progenitor; GMP, granulocyte-monocyte progenitor; NK, natural killer; pDC, plasmacytoid dendritic cells.

Figure 5.

Functional CRISPR screening identifies CEBPA cisRE as a key regulator of myeloid differentiation in high-risk leukemia. (A) Schematic overview of the CRISPR screens used to functionally interrogate MECOM-regulated cisREs. An sgRNA oligo library was designed against MECOM-regulated elements (up to 5 sgRNAs per element, depending on the availability of high-quality sgRNA-targeting sites) and packaged into a lentiviral vector. Two different populations of MUTZ-3 MECOM-FKBP12^F36V cells were then transduced with this sgRNA library virus at an MOI of ∼0.33, in which one population expressed dCas9-KRAB (CRISPRi screen) and another expressed dCas9-VPR (CRISPRa screen). Cells in the CRISPRi screen were treated with 500 nM dTAG^V-1 for the duration of the screen. After 14 days of in vitro culture, cells from the CRISPRi and CRISPRa screens were sorted for phenotypically rescued CD34⁺ cells (up-assay) and differentiated CD34^– cells (down-assay), respectively. Genomically integrated sgRNAs were sequenced to assess relative sgRNA abundance. Both screens were performed with n = 3 independent replicates. (B-C) Volcano plots depicting sgRNA enrichment/depletion from sorted populations compared to plasmid library DNA (supplemental Table 8). The sgRNA library included sgRNAs targeting the TSSs of VHL, ELOB, and ELOC (5 sgRNAs per gene), which form the E3 ubiquitin-ligase complex recruited by dTAG^V-1. (D) Genome browser tracks at the CEBPA locus encompassing the +42 kb cisRE. ATAC-seq tracks from MECOM-FKBP12^F36V cell line models and MECOM ChIP-seq demonstrate increased chromatin accessibility upon dTAG^V-1 treatment. Three top-scoring CEBPA cisRE-targeting sgRNAs were selected for single sgRNA validation experiments. (E) MUTZ-3 dCas9-KRAB cells were infected with sgRNA-expressing lentiviruses targeting either the CEBPA cisRE or a nontargeting sequence. At 48 hours after transduction, cells were treated with 500 nM dTAG^V-1 vs DMSO. Histogram shows CD34 expression at day 9 (top). Percentage of CD34⁺ cells at day 9 (bottom). n = 3 independent replicates. Mean and SEM are shown. (F) qRT-PCR of CEBPA expression in dTAG^V-1–treated cells 3 days posttreatment. Fold change represents ΔΔCt values compared to the sgNT condition. n = 3 independent replicates. Mean and SEM are shown. Two-sided Student t test was used for comparisons. ∗∗∗∗P < .0001. (G) MUTZ-3 dCas9-VPR cells were infected with sgRNA-expressing lentiviruses targeting either the CEBPA cisRE or a nontargeting sequence. Histogram shows CD34 expression at day 9 (top). Percentage of CD34⁺ cells at day 9 (bottom). n = 3 independent replicates. Mean and SEM are shown. (H) qRT-PCR of CEBPA expression in all conditions 3 days posttransduction. Fold change represents ΔΔCt values compared to the sgNT condition. n = 3 independent replicates. Mean and SEM are shown. Two-sided Student t test was used for comparisons. ∗∗∗∗P < .0001. FACS, fluorescence-activated cell sorted; FSC-A, forward scatter area; MOI, multiplicity of infection; qRT-PCR, quantitative reverse transcription polymerase chain reaction; sgNT, nontargeting sgRNA.

Figure 6.

CEBPA cisRE is necessary for differentiation of MECOM-driven AML cells. (A) Primary MECOM⁺ AML cells were harvested from patients at diagnosis and cryopreserved (supplemental Table 9). Cells were thawed for short-term ex vivo culture and electroporated with CRISPR-Cas9 RNPs to induce genetic perturbations at the MECOM vs AAVS1 locus ± CEBPA +42 kb cisRE. (B) Efficiency of gene editing in 3 biologically distinct primary AMLs at the AAVS1, MECOM, and CEBPA (cisRE) loci. Editing estimated using Sanger sequencing of amplicons followed by sequence trace decomposition analysis with Inference of CRISPR Edits (ICE) tool. For CEBPA cisRE, only deletions resulting from dual guide cleavage were counted. n = 3 technical replicates. Mean and SEM are shown. (C-D) qRT-PCR of CEBPA and MECOM expression in all conditions 3 days postelectroporation. Fold change represents ΔΔCt values compared to the sgNT condition. n = 3 technical replicates, and mean and SEM are shown. n = 3 independent replicates, and mean and SEM are shown. Two-sided Student t test was used for comparison. ∗P < .05. (E-G) Immunophenotypic analysis of a primary leukemia sample (patient 1, supplemental Table 9) 6 days postelectroporation. (E) Bivariate plot showing CD34 and CD117 expression assessed by flow cytometry. Black box denotes CD34⁺/CD117⁺ subset. (F) Percentage of CD34⁺/CD117⁺ cells. (G) CD34 expression measured by MFI. n = 3 independent technical replicates. Mean and SEM are shown. A Mann-Whitney test was used for comparisons. ∗P < .05. (H-I) Immunophenotypic analysis of a primary leukemia sample (patient 2, supplemental Table 9) 8 days postelectroporation. (H) Histogram showing CD34 expression assessed by flow cytometry. (I) Percentage of CD34⁺ cells. n = 3 independent technical replicates. Mean and SEM are shown. A Mann-Whitney test was used for comparisons. ∗P < .05. (J-K) Immunophenotypic analysis of primary leukemia (patient 3, supplemental Table 9) 8 days postelectroporation. (J) Histogram showing CD34 expression assessed by flow cytometry. (K) Percentage of CD34⁺ cells. n = 3 independent replicates. Mean and SEM are shown. Two-sided Student t test was used for comparisons. ∗P < .05. MFI, mean fluorescence intensity; sgNT, nontargeting sgRNA.

Figure 7.

Transient activation of CEBPA cisRE is sufficient to differentiate high-risk, stem cell–like AML cells. (A) qRT-PCR of CEBPA expression 3 days postelectroporation. Fold change represents ΔΔCt values compared to the sgNT condition. n = 3 independent replicates. Mean and SEM are shown. Two-sided Student t test was used for comparison. ∗∗P < .01. (B-D) Immunophenotypic analysis of a primary leukemia sample (patient 1, supplemental Table 9) 4 days postelectroporation. (B) Bivariate plot showing CD34 and CD117 expression assessed by flow cytometry. Black box denotes CD34⁺/CD117⁺ subset. (C) CD34 expression measured by MFI. (D) Percentage of CD34⁺/CD117⁺ cells. n = 3 independent replicates. Mean and SEM are shown. Two-sided Student t test was used for comparisons. ∗∗∗∗P < .0001. (E-F) Immunophenotypic analysis of a primary leukemia sample (patient 1, supplemental Table 9) 12 days postelectroporation. (E) Histogram showing CD11b expression assessed by flow cytometry. (F) Percentage of CD11b⁺ cells. n = 3 independent replicates. Mean and SEM are shown. Two-sided Student t test was used for comparison. ∗∗ P < .01. (G) Viable cell counts by trypan blue exclusion in primary leukemia sample (patient 1, supplemental Table 9) 8 days postelectroporation. n = 3 independent replicates. Mean and SEM are shown. Two-sided Student t test was used for comparison. ∗∗ P < .01. (H) qRT-PCR data of a panel of established HSC genes and LSC17 genes 18 days postelectroporation demonstrating the robust differentiation of a primary leukemia sample (patient 1, supplemental Table 9) following transient activation of CEBPA cisRE. n = 3 independent replicates. Mean and SEM are shown. Two-sided Student t test was used for comparison. ∗ P < .05. (I) Schematic of the experiment to assess the in vivo impact of CEBPA cisRE activation of a xenotransplanted primary leukemia sample (patient 1, supplemental Table 9). Cells were electroporated with mRNA encoding dCas9-VPR and 2 chemically synthesized sgRNAs targeting the CEBPA cisRE or a sgNT. Cells recovered in ex vivo culture for 2 days postelectroporation and then injected via the tail vein. All animals were euthanized 56 days after transplant for analysis of leukemia burden in spleens and BM. (J) Quantification of human cell chimerism (hCD45⁺) in the BM of mice transplanted with 1 × 10⁵ to 1 × 10⁶ cells and spleens of mice transplanted with 1 × 10⁶ cells. n = 4 to 10 xenotransplant recipients as shown. Mean and SEM are shown. Two-sided Student t test was used for comparison. ∗∗∗∗ P < .0001, ∗ P < .05. (K-L) Immunophenotypic analysis of the BM of mice transplanted with 1 × 10⁶ cells. Cells were labeled with a cocktail of antibodies including mouse CD45 and human CD45, CD34, CD117, and CD11b. (K) Bivariate plots depicting the gating strategy for quantification of engrafted leukemia stem/progenitor cells (CD34⁺/CD117⁺) and mature cells (CD11b⁺). Black boxes denote human cell subset (top), CD34⁺/CD117⁺ subset (middle), and CD11b⁺ subset (bottom). (L) Percentage of CD34⁺/CD117⁺ cells (top) and CD11b⁺ cells (bottom). n = 4 to 10 xenotransplant recipients as shown. Mean and SEM are shown. Two-sided Student t test was used for comparison. (M) Hematoxylin and eosin (H&E) staining of bone marrow of mice transplanted with 1 × 10⁶ cells. BM, bone marrow; MFI, mean fluorescence intensity, mRNA, messenger RNA; ns, not significant; sgNT, nontargeting sgRNA.