HL-60 acute myeloid leukemia (AML FAB M2, Cat# ACC 3) and MV4-11 acute monocytic leukemia (AML FAB M5, Cat# ACC 102) AML cell lines were obtained from the German Collection of Microorganisms and Cell Cultures (DSMZ) and maintained in RPMI 1640 (Cat# R8758, Sigma-Aldrich) supplemented with 10% FBS (Gibco), 100 μg/mL penicillin and 100 μg/mL streptomycin. Bone marrow aspirates of newly diagnosed patients with AML were collected after informed consent and with the approval of the institutional review boards and ethic committees of the University of Jena. Samples were kindly provided by Professor Andreas Hochhaus. Primary cells were cultured in Stemline hematopoietic stem cell expansion medium (Cat# S0189 Sigma-Aldrich).

The following reagents were used in this study: Puromycin (Cat# A1113802, Invitrogen); ATRA (Cat# R2625, Sigma-Aldrich); GSK-343 (Cat# 14094, Cayman Chemical); BIX-01294 (Cat# 13124, Cayman Chemical); UNC1999 (Cat# 14621, Cayman Chemical) and Dimethyl sulfoxide (DMSO, Cat# D8418, Sigma-Aldrich). HKMTI-1-005 was provided by Professor Matthew Fuchter. Antibodies used for immunoblot and immunoprecipitation were as follows: rabbit polyclonal anti-H3K27^me3 (Cat# 07–473, Upstate Biotechnology); rabbit polyclonal anti-H3K9^me2 (Cat# ab115159, Abcam); rabbit polyclonal anti-GAPDH (Cat# ab9485, Abcam); rabbit polyclonal anti-EZH2 (Cat# ab186006, Abcam); mouse monoclonal anti-EZH2 (clone AC22, Cat# 3147, Cell Signaling Technology); mouse monoclonal anti-RARα (clone 9α-9A6, Cat# 39971, Active Motif); rabbit polyclonal anti-RARα (Cat# sc-773, Santa Cruz). EZH2 shEZH2#1 and shEZH2#2 oligos (reverse/complement of mature antisense sequence GAC​TCT​GAA​TGC​AGT​TGC​T and CAT​GTA​GAC​AGG​TGT​ATG​A respectively) were cloned into pGreenPuro lentiviral expression vectors (System Biosciences) and Z-competent bacteria (Z-competent E. coli Transformation Buffer Set, Cat# T3002, Zymo Research) were then transformed with the vectors. Clones were selected and inserts were sequenced (Source BioScience Lifesciences). Plasmid DNA was purified (PureLink HiPure Plasmid Midiprep Kit, Cat# K210004, Invitrogen) and used for transient transfections and the standard production of lentiviruses in HEK-293TN cells (Cat# LV900A-1, System Biosciences). Lentiviral supernatants were prepared, and HL-60 cells infected with viral particles were selected with puromycin and expanded for further experiments.

Inhibitors were titrated to determine the concentration values at which 50% of cell viability occurred (GI₅₀ values) in each cell line. In brief, cells were seeded (∼20,000 cells per well) in 96-well flat bottom plates in a total of 150 µL culture media ± compounds and grown for 72 h in triplicates. Cell viability was determined using the CellTiter-Glo Luminescent Cell Viability Assay (Cat# G7570, Promega).

Cells were lysed in RIPA buffer (Cat# R0278, Sigma) for immunoblot analysis. Protein concentration from whole cell lysate was measured (DC Protein Assay kit II, cat# 5000112, Bio-Rad) and 10 µg–20 µg of protein per well was separated on 10% NextGel (Cat# M256, Amresco) or gradient pre-cast Mini-PROTEAN TGX gels (BioRad). Proteins were transferred on Hybond ECL membrane (GE Healthcare). Membranes were blocked using 5% milk or BSA in PBS-Tween (0.1%), incubated overnight at 4°C with primary antibodies (described above), washed in PBS-Tween and incubated with anti-mouse (Cat# 155–035-174) and anti-rabbit (Cat# 211–032-171) light chain specific antibodies HRP-secondary antibodies (Jackson ImmunoResearch). Bands were visualized using ECL kit (Thermo Scientific) and X-ray films (Kodak) or Gel Doc system (BioRad).

FACS analysis of CD11b (MAC-1) expression by AML cell lines and apoptosis analysis of patient samples after drug treatment were performed using an APC-conjugated human CD11b-specific mouse monoclonal antibody at a 1:10 dilution (Cat# 555388, BD Pharmingen) on a BD LSRII FACS machine (Becton Dickinson) with FACS Diva software. For analysis of apoptosis, the Annexin-V FITC apoptosis detection kit II (Cat# 556570, BD Pharmingen) was used according to the manufacturer’s instructions with Propidium Iodide (Cat# R37169, Invitrogen) at 1 μg/mL on Guava EasyCyte flow cytometer with InCyte software (Merck).

For the morphological analysis of myeloid cell differentiation, cytospins were prepared by centrifugation in 150 µL PBS at a speed of 300 rpm for 5 min using Superfrost Plus (VWR) positively charged glass slides. Cytospun slides were stained at room temperature with May-Grünwald-Giemsa (Sigma) and cellular morphology was examined using an Axioscope 2 plus microscope with an AxioCam MRc camera (Zeiss).

Duolink in situ proximity ligation assays (PLA, Sigma Aldrich) were performed on HL-60 AML cells fixed in 4% paraformaldehyde for 20 min, permeabilized with 0.5% Triton X-100, and blocked with 1% BSA for 30 min at room temperature followed by incubation with paired primary antibodies, anti-EZH2 and anti-RARα, overnight at 4°C. PLA detection was performed as recommended by the manufacturer. Images were taken and analyzed using the Zeiss LSM700 confocal microscope and analyzed using Duolink image analysis software.

Experiments were performed at least 3 times, unless otherwise indicated. Error bars represent mean ± standard deviation. Statistical analysis was performed using the unpaired t-test unless otherwise indicated. p-values < 0.05 were considered as statistically significant. Computations were performed using Prism (GraphPad version 9.5.0, San Diego, CA).

GeneChip Human Gene 1.0 ST Array and GeneChip, WT PLUS Reagent Kit, GeneChip WT Terminal Labeling and Controls Kit, GeneChip Hybridization, Wash, and Stain Kit, processed with GeneChip Fluidics Station 450, GeneChip Hybridization Oven 645 and GeneChip Scanner 3,000, were used and standard protocols provided by Affymetrix and Ambion were followed. In brief, total RNA was extracted from 72 h-treated cells and control cells, cDNA, cRNA and then single-stranded cDNA were generated with amplification and purification steps in between, the latter being digested, labelled and hybridized to the oligonucleotide probes on the chips overnight. Staining, washing and scanning on the final day was carried out and the images were processed by the Affymetrix GeneChip Operating Software (GCOS) to produce .CEL files, which were then used for the bioinformatics analysis of the arrays.

Arrays were normalized with Partek Genomics Suite (https://www.partek.com/partek-genomics-suite/) using GC-RMA, after which “batch-remove” function in Partek was applied to normalize for batch-to-batch and date-to-date variation between samples. False discovery rate (FDR) testing across all conditions was applied to generate gene lists with statistically significant (FDR ≤ 0.05) differentially expressed genes between the conditions. The resulting gene lists were further refined by setting fold change (1.3-fold change) and overall expression (top 40th percentile) cut-offs and were analyzed by the following online tools: Database for Annotation, Visualization and Integrated Discovery (DAVID, https://david.ncifcrf.gov) and Gene Set Enrichment Analysis (GSEA, https://www.gsea-msigdb.org). For GSEA, .gct and .cls files were manually created and uploaded in the software. q-values (equivalent to FDR-adjusted p-values) ≤ 0.05 were set as a cut-off and enriched categories were then further analyzed using online tools including Venny (http://bioinfogp.cnb.csic.es/tools/venny/index.html).

In order to assess the impact of EZH2 expression on myeloid differentiation of human non-APL AML, we first examined the effect of EZH2 knockdown (KD) via lentiviral transduction of two short hairpin sequences (shEZH2#1 and shEZH2#2) targeting separate sequences on EZH2 mRNA in MYC-amplified human HL-60 cells (AML FAB M2) (Collins et al., 1977), a long-established model system in this context (Birnie, 1988). HL-60 cells do not have any reported mutations for EZH1/KTM6B, EZH2/KTM6A, G9A/EHMT2 or GLP/EHMT1 as reported by the Catalogue of Somatic Mutations in Cancer (COSMIC v97, https://cancer.sanger.ac.uk/cosmic) and cBioPortal (https://www.cbioportal.org/) analysis of the Cancer Cell Line Encyclopedia (https://sites.broadinstitute.org/ccle/) (Cerami et al., 2012; Ghandi et al., 2019). According to available data on cBioPortal, mutation of EZH2 is uncommon in AML cell lines, with a single cell line identified as harboring a Y646C predicted gain-of-function mutation (SKM-1) (Palau et al., 2017) and three cell lines harboring predicted loss-of-function mutations: PL-21 (Quentmeier et al., 2003), OCI-AML5 (Wang et al., 1991), and P31/FUJ (Hirose et al., 1982). Treatment with 1 µM ATRA typically differentiates HL-60 cells effectively towards granulocyte-like cells, with more than 90% of cells reflecting band (immature) or segmented (mature) neutrophils (Manda-Handzlik et al., 2018). Upon ATRA treatment, these cells express the myeloid lineage cell-surface CD11b marker but not CD14 that is associated with maturation to monocytes and macrophages. In order to evaluate the impact of compounds on the promotion of myeloid differentiation of HL-60 cells, experiments were performed using a sub-optimal ATRA concentration of 0.1 µM. Consistent with results indicating a role for EZH2 in blocking differentiation in an MLL-AF9 driven mouse model of AML (Tanaka et al., 2012), EZH2 KD induced differentiation in HL-60 cells as measured by flow cytometric analysis of the percentages of cells expressing CD11b (CD11b positive, CD11b+ cells) in either the absence or presence of ATRA (ranging between 2.44-fold and 3-fold, Figure 1A, left panel; Supplementary Figure S1). EZH2 KD was confirmed by immunoblot analysis of EZH2 protein and H3K27^me3 (Figure 1B) and in agreement with the increase in the percentage of CD11b+ cells. After May-Grünwald-Giemsa staining, increased numbers of cells displaying morphological changes consistent with myeloid differentiation (nuclear lobulation and increased cytoplasmic:nuclear ratio) were observed when EZH2 KD cells treated with ATRA were compared with cells treated with ATRA alone (Figure 1C). The increase in expression of CD11b in EZH2 KD cells treated with ATRA was not accompanied by an increase in cells undergoing necrosis or late-stage apoptosis as measured by propidium iodide (PI) uptake (Figure 1A, right panel), but analysis of cell morphology showed the presence of apoptotic blebs on the cell surface that was indicative that the cells were undergoing apoptosis (Figure 1C). It has been demonstrated previously that apoptotic blebs can form prior to a loss of overall membrane permeability (Wickman et al., 2013).

We next tested GSK-343 in HL-60 (Figure 2; Supplementary Figures S2, S3) and MV4-11 cells (Lange et al., 1987), which were established from a patient with biphenotypic B-myelomonocytic leukemia (AML FAB M5) (Supplementary Figures S2, S4, S5). MV4-11 cells harbor a t(4;11)(q21;q23) MLL/AF4 (KMT2A/AFF1) chromosome translocation and a FLT3 internal tandem duplication (Quentmeier et al., 2003). MV4-11 cells do not have any reported mutations for EZH1/KTM6B, EZH2/KTM6A or G9A/EHMT2, but do harbor a missense mutation in G9A/EHMT1 resulting in a conservative leucine to valine substitution at amino acid position 1059 (L1059V) (Cerami et al., 2012; Ghandi et al., 2019). When treated with ATRA, MV4-11 cells undergo myelomonocytic differentiation, expressing both CD11b and CD14 markers (Gocek et al., 2012; Ma et al., 2016). As with HL-60 cells, experiments using MV4-11 were performed using a sub-optimal ATRA concentration of 0.1 µM. We found that GSK-343 moderately decreased the percentage of metabolically active, viable, cells as measured by ATP quantitation with CellTiter-Glo for both HL-60 (Figure 2A, left panel) and MV4-11 cells (Supplementary Figure S4A, left panel) with GI₅₀ values of 13.22 µM (Supplementary Figure S2A) and 10.12 µM (Supplementary Figure S2C), respectively. We did not observe the presence of apoptotic blebs on the cell surface of HL-60 cells (Figure 2C, left panel), but blebbing was evident on MV4-11 in either the absence or presence of ATRA (Supplementary Figure S4C, left panel). For HL-60 cells, while GSK-343 treatment did not induce myeloid differentiation in the absence of ATRA, it caused a small but significant increase in the percentage of CD11b+ cells in the presence of ATRA (Figure 2A, right panel and Supplementary Figure S3), although this was not accompanied by a clear change in cell morphology associated with myeloid differentiation (Figure 2C, left panel). GSK-343 treatment failed to induce myeloid differentiation in MV4-11 cells as evidenced by a lack of increase in the proportion of CD11b+ cells or changes in cell morphology (Supplementary Figures S4A, right panel; (Supplementary Figures S4C, left panel and Supplementary Figure S5). We also tested the pyridone-containing and SAM-competitive EZH1 and EZH2 inhibitor UNC1999 (Konze et al., 2013) for its ability to induce cell death and myeloid differentiation in HL-60 cells and the results obtained were similar to those for GSK-343 (Supplementary Figure S6). UNC1999 decreased cell viability in HL-60 cells in a dose-dependent manner with a GI₅₀ value of 7.22 µM (Supplementary Figure S6A,B). UNC1999 did not induce myeloid differentiation in the absence of ATRA but, consistent with GSK-343, did cause a small but significant increase in the percentage of CD11b+ cells in the presence of ATRA (Supplementary Figure S6C).

Inhibition of EZH2 with HKMTI-1-005 decreased cell viability for both HL-60 (Figure 2B, left panel) and MV4-11 cells (Supplementary Figure S4B, left panel) in a dose-dependent manner at lower concentrations than GSK-343, with GI₅₀ values of 2.43 µM (Supplementary Figure S2B) and 2.45 µM (Supplementary Figure S2C), respectively. Apoptotic blebbing was evident on both HL-60 (Figure 2C, right panel) and MV4-11 (Supplementary Figure S4C, right panel) cells treated with HKMTI-1-005 in the presence of ATRA. In contrast with GSK-343, HKMTI-1-005 treatment increased the percentage of CD11b+ HL-60 cells both in the absence (3.2-fold) and presence (2.1-fold) of ATRA (Figure 2B, right panel and Supplementary Figure S3), and promoted a cell morphology associated with myeloid differentiation Figure 2C, right panel). Similarly, HKMTI-1-005 treatment increased the percentage of CD11b+ MV4-11 cells 4.2-fold and 2.9-fold in the absence and presence of ATRA, respectively, (Supplementary Figure S4B, right panel and Supplementary Figure S5), also promoting a cell morphology associated with myeloid differentiation (Supplementary Figure S4C, right panel).

As expected, treatment with 10 µM GSK-343 almost completely abolished H3K27^me3 levels in both HL-60 (Figure 2D) and MV4-11 (Supplementary Figure S4D) cells. However, treatment with the GI₅₀ concentration of HKMTI-1-005 (2.5 µM) diminished, rather than abolished, H3K27^me3 levels in the presence of ATRA in both HL-60 (Figure 2D) and MV4-11 (Supplementary Figure S4D) cells. A decrease but not abolition of H3K27^me3 levels with 2.5 µM HKMTI-1-005 in the AML cell lines examined here is in agreement with previous data for MDA-MB-231 breast cancer cells, where treatment with an approximate IC₅₀ concentration of HKMTI-1-005 diminished H3K27^me3 by around half (Curry et al., 2015). Consistent with its G9A/GLP inhibitory activity, 2.5 µM HKMTI-1-005 treatment also reduced levels of H3K9^me2 and H3K9^me3 both in the absence and presence of ATRA (Figure 2D). To evaluate the effect of G9A/GLP inhibition alone on cell viability and myeloid differentiation, we also treated HL-60 cells with BIX-01294. In agreement with previous results (Savickiene et al., 2014), we found that G9A/GLP inhibition with BIX-01294 at the GI₅₀ concentration of 2 µM (higher concentrations were extremely cytotoxic, Figures 3A, B) also promoted a moderate increase in the percentage of CD11b+ HL-60 cells (Figure 3C), which was accompanied by a reduction in H3K9^me2 (Figure 3D). These results are consistent with a study using another substrate-competitive G9A/GLP inhibitor, A-366, which was found to increase CD11b expression and decrease cell viability with concomitant diminution in H3K9^me2 levels (Pappano et al., 2015). Interestingly, we found that co-treatment with BIX-01294 and GSK-343 did not increase the proportion of CD11b+ cells in the absence of ATRA and led to a small decrease in the proportion of CD11b+ cells in the presence of ATRA (Figure 3C).

Lastly, we tested GSK-343 and HKMTI-1-005 against four samples of primary AML cells (Figure 4). We found that treatment with the HKMTI-1-005 concentration used for the cell lines for 48 h −/+ ATRA effectively diminished the proportion of live primary AML cells, with a concomitant increase in early apoptotic cells. There was no increase in the proportion of dead cells at the 48-h time point, and two samples, Patient #2 and Patient #3, showed an increase in the percentage of late apoptotic cells. Treatment with 10 µM GSK-343 −/+ ATRA or ATRA alone, by contrast, did not affect the proportion of live, early or late apoptotic primary AML cells. We failed to see induction of CD11b expression for any of the four AML patient samples tested here (data not shown). It should be noted that we and others have observed in ATRA-based therapy a high level of variation in the induction of cell-surface CD11b, a lack of correlation between decreases in cell viability and CD11b induction, and also that morphological features associated with granulocytic differentiation do not necessarily correlate with the presence CD11b (Schenk et al., 2012; Klobuch et al., 2018; Kahl et al., 2019). The limited number of samples studied here, as well as the 48-h treatment (which was used due to non-specific decrease in cell viability at 72 h), mean that further investigation is required as for this study we had limited access to primary samples.

We next investigated the consequences for transcriptional programs in HL-60 cells as a result of inhibiting the activities of EZH2 by EZH2 KD, or treatment with GSK-343 or HKMTI-1-005, either alone or in combination with ATRA. Following 72 h treatment, we performed gene expression arrays and transcriptomic analysis. Initial post-normalization principal component analysis (PCA) revealed that, consistent with its role as a ligand for RARα (a key mediator of myeloid differentiation) (Collins, 2002), ATRA treatment caused the greatest change to the transcriptomic program in HL-60 cells. However, while the transcriptional signatures for HL-60 cells following single agent GSK-343 and HKMTI-1-005 treatment were more similar, HKMTI-1-005/ATRA co-treatment resulted in a transcriptional signature that resembled that for ATRA-treated EZH2 KD cells as compared with GSK-343/ATRA co-treatment (Figure 5A). In accordance with the minimal differentiation response in HL-60 cells treated with GSK-343, the transcriptional signature of these cells was more similar to that of untreated control as evidenced by the smaller variation along the z-axis (PC2) between these two samples compared to that between HKMTI-1-005 treated cells and untreated control, especially in the presence of ATRA (Supplementary Figure S7A). Thus, PCA analysis indicated that SAM-competitive inhibition of EZH2 failed to trigger a transcriptional program associated with myeloid differentiation, regardless of the ability of GSK-343 to effectively deplete global H3K27^me3 levels as measured by immunoblot analysis. Gene ontology (GO) analysis of upregulated genes (FDR < 0.05, FC = ±1.3) provided further support for the finding that EZH2 KD and EZH2 targeting via HKMTI-1-005 induced similar transcriptomic changes that are likely to underpin the previously observed phenotypes of myeloid differentiation, while GSK-343 failed to direct the fate of the cells towards a more mature state. Unsurprisingly, the presence of ATRA was insufficient to activate certain expression signatures related to myeloid differentiation (Supplementary Figure S7B), and we found that in the absence of ATRA EZH2 KD and inhibition with HKMTI-1-005, but not SAM-competitive inhibition with GSK-343, upregulated the same ATRA-driven transcriptional programs that are characteristic for mature myeloid cells including those involved in inflammatory response, response to wounding, and defense response (Figure 5B).

We next examined all the differentially expressed genes used for the GO analysis (Figure 5C) to attempt to address the differences and similarities between the three ways of targeting EZH2 and gain more insight into the enhanced differentiation or lack of such (Supplementary Tables S1, S2). We found that genes exclusively upregulated by GSK-343 are connected to oncogenic/stem cell programs (e.g., WNT signaling), and included genes involved in transcriptional control, of which GATA2 (Nandakumar et al., 2015), LEF1 (Petropoulos et al., 2008), SMYD2 (Sakamoto et al., 2014) and MLLT3 (AF9) have been strongly implicated in leukemogenesis. On the other hand, EZH2 KD and HKMTI-1-005 treatment, both in the presence and absence of ATRA, confirmed the myeloid maturation of the cells through upregulation of the expression of genes encoding cell surface receptors that are functionally important including CD36 and ITGAM (CD11b). Among the genes that were downregulated, we found that GSK-343 treatment not only failed to trigger myeloid maturation, but also downregulated the expression of receptors that are implicated in myeloid function, including SLAMF7, GPR65 and FCER2 (CD23) (Lattin et al., 2007; Mossalayi et al., 2009; Beyer et al., 2012). Lastly, we carried out gene set enrichment analysis (GSEA) in order to understand the transcriptional changes that characterize the differentiation phenotype observed following EZH2 KD and HKMTI-1-005 treatment. We found that targeting of EZH2, via gene knockdown or substrate-competitive inhibition, affected the “stemness” transcriptional signatures of the cells. We observed downregulation of hematopoietic and leukemic stem cell attributes, including the gene sets “Jaatinen hematopoietic stem cell down” (M6905), and “Gal leukemic stem cell down” (M4502), and enhancement of transcriptional programs that are involved in myeloid cell development (Figure 6), including the gene set “Brown myeloid development up” (M1430). In contrast, GSK-343 induced the opposite effect. There was upregulation of hematopoietic stem cell signatures and a concomitant failure to activate downregulation of leukemic stem cell attributes and upregulation of myeloid differentiation programs (negative enrichment scores). This strong transcriptional basis confirmed the inability of GSK-343 inhibition of EZH2 to trigger differentiation.

Our results indicated that inhibition of canonical EZH2 activity (i.e., methylation of H3K27) via pyridone-containing, SAM-competitive inhibitors such as GSK-343 and UNC1999 was not effective in terms of promoting ATRA-based myeloid differentiation. We therefore sought to identify whether treatment with GSK-343 or HKMTI-1-005 could differentially impact non-canonical EZH2 activity in the context of RARα-mediated myeloid differentiation given that EZH2 had previously been shown to interact with RARα (albeit with less affinity as compared with PML-RARα) (Villa et al., 2007). Using the proximity ligation assay (PLA) (Fredriksson et al., 2002), we first performed a time course experiment with ATRA and found that upon treatment there was a rapid increase in the number of signals per cell, indicative of increased RARα/EZH2 interactions, which peaked between 4 h and 8 h post-treatment (Figure 7A). We next assessed the impact of GSK-343 or HKMTI-1-005 treatment on increased RARα/EZH2 interactions at the 4-h timepoint. The addition of GSK-343 did not significantly increase the number of signals per cell as measured by PLA in the absence of ATRA and led to a small but non-significant decrease in the number of signals per cell following ATRA treatment (Figure 7B). The addition of HKMTI-1-005, by contrast, significantly increased the number of signals per cell as measured by PLA both in the absence and presence of ATRA (5.5-fold and 2.1-fold, respectively) (Figure 7B).