The study got the approval of the Clinical Ethical Committee of the Second Hospital of Shanxi Medical University. Informed consent was signed by each eligible participant.

Thirty-three bone marrow samples were collected from the Second Hospital of Shanxi Medical University from 2015 to 2018. All the samples were confirmed as MDS (N = 11), MDS-AML (N = 11), or cancer-free individuals (N = 11) by bone marrow puncture and/or biopsy. Among them, MDS-AML referred to the patients who were definitely diagnosed with MDS and then turned into AML. MDS patients and MDS-AML patients were matched according to age and gender. Bone marrow samples were extracted with human peripheral blood lymphocyte separation solution (TBD, Tianjin, China) by density gradient method. Total RNA was isolated from monocytes using RNAiso Plus reagent (Takara, Dalian, China) and reverse transcribed into cDNA using a reverse transcription quantitative polymerase chain reaction kit (Takara), and then stored at −80°C.

NHD13 mice were purchased from Jackson Laboratory and C57BL/6 mice were purchased form Kunming Institute of Zoology, Chinese Academy of Sciences [SYXK (Yunnan) K2015-0003]. Mice were raised in a specific pathogen-free animal facility. Food and water were provided ad libitum. The expressions of EZH2 and EHMT2 in peripheral blood of mice were detected at 4 months old. Then peripheral blood was collected regularly and the blood condition was detected. Blood samples were collected from mice at the age of 14 months (420 days), and then all the mice were euthanized by an intraperitoneal injection of pentobarbital (800 mg/kg) (Kopaladze, 2000; Zatroch et al., 2017).

Human MDS/AML cells (SKM-1 cells) were purchased from BeNa Culture Collection (Beijing, China) and cultured in Roswell Park Memorial Institute-1640 medium containing 10% fetal bovine serum in 95% humidified air with 5% CO₂ at 37°C.

The small interfering RNA (siRNA)-NC-1, si-EZH1, si-NC-2, si-EZH2, si-NC-T2, and si-EHMT2 were designed and synthesized by GenePharma (Shanghai, China), and the sequence is shown in Supplementary Table 1. Overexpression vectors of EZH2 (pcDNA3.1-EZH2) and EHMT2 (pcDNA3.1-EHMT2) and empty vectors were constructed by Zoman Biotechnology Co., Ltd. (Beijing, China). Then, the constructed vectors and siRNAs were transfected into cells using Lipofectamine 2000 (Invitrogen Inc., Carlsbad, CA, United States).

SKM-1 cells were treated with EZH2 inhibitor EPZ-6438 (5 μM, Yeasen Biotech Co., Ltd., Shanghai, China) and EHMT2 inhibitor BIX-01294 (2.5 μM, Sigma-Aldrich, Merck KGaA, Darmstadt, Germany). Briefly, cells in logarithmic growth phase were seeded into 96-well plates (1 × 10⁵ cells/mL) supplemented with culture medium, and treated with 5 μM EPZ-6438 for 4 days (Knutson et al., 2014) and 2.5 μM BIX-01294 for 3 days, respectively (Huang et al., 2017). The treated cells were allocated as EPZ group and BIX group, respectively.

The treated cells were seeded into the 96-well plates (2 × 10³ cells/well), and CCK-8 solution was added at 24, 48, and 72 h. The optical density of each well was measured at 450 nm. The experiment was repeated three times in each group.

SKM-1 cells under different treatments were fixed with 70% ethanol, stained with 300 mL propidium iodide (MultiSciences Biotech Co., Ltd., Hangzhou, Zhejiang, China) in the dark, and detected on the flow cytometer (MoFlo Astrios EQ, Beckman Coulter, Inc., CA, United States) to analyze the cell cycle.

The culture medium of SKM-1 cells under different treatments was removed, and the cells were washed with phosphate-buffered saline (PBS), incubated with EdU solution for 2 h, and then photographed under a fluorescence microscope (Olympus, Tokyo, Japan).

SKM-1 cells under different treatments were seeded into the 12-well plates (1 × 10⁴ cells/well), and incubated at 37°C for 1 week until cell colonies were observed. The colonies were stained with crystal violet and counted.

SKM-1 cells were subjected to ChIP assay referring to previous literature (Dagdemir et al., 2013). Cells were detached with trypsin and counted using Millipore Scepter 2.0 Cell (Thermo Fisher Scientific, Jiangsu, China). And 1 × 10⁶ cells were used for each treatment. Cells were incubated for 8 min in the medium, fixed with formaldehyde and then crosslinked with 1.25 M glycine for 5 min at room temperature. All chromatin preparation and ChIP reaction were carried out at 4°C. The crosslinked cells were washed with PBS-inhibitor (NaBu 20 mM), and the cell membrane was cleaved with the HighCell ChIP kit. Chromatin was prepared in TPX tube with shear buffer S1 and 1 × protease inhibitor, and then broken into fragments of about 500 bp by ultrasound. The size of the fragments was examined on agarose gel, and the cut chromatin was frozen at −80°C. ChIP reaction was performed using the Diagenode kit on the SX-8X IP STAR compact automation system (Diagenode) for all IP procedures. According to the HighCell ChIP kit protocol, IP DNA was purified using DNA Isolation buffer with 2 μg antibody [anti-H3K27me3 (10 μg for 25 μg of chromatin, ab6002, Abcam, Cambridge, MA, United States) or anti-H3K9me1 (5 μg for 10⁶ cells, ab9045, Abcam) or anti-H3K9me2 (4 μg for 25 μg of chromatin, ab1220, Abcam)] and non-immune immunoglobulin G (IgG). Each auto-ChIP sample was performed using the Auto Histone ChIP-seq kit and contained 1 μg input chromatin. The reaction lasted for 2 h. The antibody was coated with protein A-coated magnetic beads, and then incubated for 10 h at 4°C for IP reaction. Afterward, 25 μL system of DNA IP or DNA input (total DNA), 1 × SYBR Green Supermix (Applied Biosystems, Inc., Carlsbad, CA, United States) and TSH2B (pp-1041–500, Diagenode; positive control of methylation) promoter were used for reverse transcription quantitative polymerase chain reaction (RT-qPCR), with the sequence of DLX5 promoter region as primer.

Total RNA was extracted from cells using TRIzol one-step reagent (Invitrogen), and then the concentration and purity of RNA were determined using UV analysis and formaldehyde deformation electrophoresis. The fluorescent qPCR reaction was performed on the instructions of the RT-qPCR kit (Thermo Fisher scientific). Primers (Table 1) were designed and synthesized by Sangon Biotech (Shanghai, China). Amplification curve and dissolution curve were confirmed after reaction. The relative expression of genes was calculated by 2−ΔΔCt method, with glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as the internal reference.

Cells in each group were lysed in radio-immunoprecipitation assay buffer containing protease inhibitors (Sigma-Aldrich). The lysate was centrifuged at 16000 g and 4°C for 20 min to collect the supernatants. The concentration of protein extracted from cells was tested using the Pierce bicinchoninic acid assay kit (Beyotime, Shanghai, China). Then, the protein was separated using sodium dodecyl sulfate-polyacrylamide gel electrophoresis and transferred onto polyvinylidene difluoride membranes (Millipore, Bedford, MA, United States). The membranes were blocked with 5% skim milk for 2 h and cultured with the primary antibodies at 4°C overnight. Thereafter, the membranes were cultured with the secondary antibody for 1 h, and developed and visualized using the enhanced chemiluminescence reagent. The gray value of the target band was analyzed by Image J software (National Institutes of Health, MD, United States). The antibodies used were as follows: H3K27me3 (1:1000, ab6002, Abcam), β-actin (1:1000, ab8227, Abcam), H3K9me1 (1:1000, ab9045, Abcam), H3K9me2 (1:1000, ab1220, Abcam), EZH1 (1:1000, ab189833, Abcam), EZH2 (1: 1000, ab150433, Abcam), and EHMT2 (1:1000, ab185050, Abcam).

SPSS 21.0 (IBM Corp., Armonk, NY, United States) was utilized for data analysis. Shapiro–Wilk test showed that the data in each group were in normal distribution. Data are expressed as mean ± standard deviation. The Mann–Whitney U test or t test was adopted for analysis of comparisons between two groups. The one-way analysis of variance (ANOVA) was applied for comparisons among multi-groups, followed by Tukey’s multiple comparisons test. Kaplan–Meier curve was used to analyze the AML transformation and survival of mice, and Log-rank method was used to test the differences between groups. The p value was obtained from a two-tailed test, and p < 0.05 meant a statistical difference.

The detection of bone marrow samples collected from MDS and MDS-AML patients in the Second Hospital of Shanxi Medical University from 2015 to 2018 showed that EZH2 expression in bone marrow of MDS patients was notably lower than that of healthy non-cancer individuals (p < 0.05; Figure 1A), but EZH2 expression in MDS-AML patients showed a trend of high expression (p < 0.001; Figure 1B). EHMT2 was always highly expressed in MDS and MDS-AML patients (both p < 0.001; Figures 1A,B). The general information of participants is shown in Table 2.

NHD13 mice faithfully reproduced all the key features of MDS, including decreased peripheral blood cells, abnormal bone marrow hyperplasia and increased apoptosis, and conversion to acute leukemia at 4–14 months of age (Lin et al., 2005; Slape et al., 2008). The survival of NHD13 mice (N = 25) were observed within 420 days, and the expressions of EZH2 and EHMT2 in peripheral blood of mice were detected at the age of 4 months. The results showed that EZH2 expression in NHD13 mice was lower than that in healthy C57BL/6 mice (p < 0.05; Figure 2A). According to the median expression of EZH2 in NHD13 mice, the mice with the expression of EZH2 higher than the median were allocated as the EZH2 high expression group (relative EZH2 mRNA expression >0.45, N = 12), and the mice with the expression of EZH2 lower than the median were allocated as the EZH2 low expression group (relative EZH2 mRNA expression ≤0.45, N = 13). Comparing the two groups of mice, it was found that NHD13 mice with high expression of EZH2 transformed from MDS to AML in a short period of time, while the mice with low expression of EZH2 took a relatively long time to transform into AML, even without the presence of AML transformation (Figure 2C). Additionally, the survival rate of NHD13 mice with high expression of EZH2 was significantly lower (p < 0.05; Figure 2E). EHMT2 expression in NHD13 mice was notably increased compared with that in healthy mice (p < 0.05; Figure 2B). Similarly, according to the median expression of EHMT2, mice were designated into EHMT2 high expression group (relative EHMT2 mRNA expression >2.13, N = 12) and EHMT2 low expression group (relative EHMT2 mRNA expression ≤2.13, N = 13). It was found that mice with high expression of EHMT2 developed from MDS to AML more quickly (Figure 2D), and the survival rate of NHD13 mice with high expression of EHMT2 was significantly reduced (p < 0.05; Figure 2F). These results suggested that the expressions of EZH2 and EHMT2 were related to the transformation from MDS to AML.

Our data showed that EZH2 had an opposite trend in MDS and AML. To further investigate the role of EZH2 in MDS and MDS-AML, we transfected SKM-1 cells with si-EZH2 or pcDNA3.1-EZH2. The results of RT-qPCR and Western blot confirmed the transfection efficiency (p < 0.05; Figures 3A,B). SKM-1 cells transfected with si-EZH2 showed significantly reduced proliferation ability and blocked cell cycle, while SKM-1 cells transfected with pcDNA3.1-EZH2 had the opposite trend (all p < 0.05; Figures 3C–F). It was indicated that the deletion of EZH2 inhibited MDS cell proliferation, while the overexpression of EZH2 facilitated cell proliferation.

SKM-1 cells were also transfected with si-EHMT2 or pcDNA3.1-EHMT2. The results of RT-qPCR and Western blot confirmed the transfection efficiency (p < 0.05; Figures 4A,B). Compared with the untransfected cells, SKM-1 cells transfected with si-EHMT2 showed notably reduced proliferation ability and blocked cell cycle, while SKM-1 cells transfected with pcDNA3.1-EHMT2 had enhanced proliferation ability (all p < 0.05; Figures 4C–F). Briefly, EHMT2 promoted SKM-1 cell proliferation.

EZH2 is an epigenetic regulator that regulates gene transcription by promoting H3K27me3 methylation level (Hosono, 2019). SKM-1 cells were transfected with pcDNA3.1-EZH2 or si-EZH2, and the transfection efficiency is shown in Figures 3A,B. Western blot revealed that H3K27me3 level was increased notably in SKM-1 cells after EZH2 overexpression, but did not decrease after EZH2 silencing (p > 0.05; Figure 5A). The previous literature has demonstrated that EZH1 has overlapping mechanism with EZH2 in MDS (Ling et al., 2019). EZH1 expression was increased significantly when EZH2 was knocked down (p < 0.05; Figure 5B). Therefore, we speculated that the reason why H3K27me3 did not decrease when EZH2 was knocked down may be due to the functional compensation of EZH1. Subsequently, we intervened the expressions of EZH1 and EZH2 in SKM-1 cells simultaneously (Figure 5B shows the transfection efficiency of si-EZH1), and found that H3K27me3 level in SKM-1 cells was notably decreased (p < 0.05; Figure 5A).

Similar to EZH2, EHMT2 inhibits the transcription of tumor suppressor genes by promoting the methylation and dimethylation of H3K9me1/3K9me2 (Kondengaden et al., 2016). We detected the levels of H3K9me1 and H3K9me2 in SKM-1 cells. H3K9me1 and H3K9me2 were decreased significantly in SKM-1 cells transfected with si-EHMT2 (p < 0.001) (Figure 6).

The high methylation and low expression of DLX5 are frequent in AML and MDS, and are related to the transformation from MDS to AML (Zhang et al., 2020). We detected the transcription level of DLX5 in SKM-1 cells of each group and found that transfection of si-EZH2 alone or si-EHMT2 alone could promote the DLX5 mRNA level (Figure 7A). DLX5 mRNA level in NHD13 mice was lower than that in healthy C57BL/6 mice (Figure 7B). DLX5 mRNA level was increased to the highest level when EZH2 and EHMT2 were at low expressions at the same time, and decreased to the lowest level when EZH2 and EHMT2 were at high expressions (Figure 7C) (all p < 0.001). Therein, we speculated that the transcription of DLX5 was co-inhibited by EZH2 and EHMT2.

To confirm the specific mechanism of EZH2 and EHMT2 regulating DLX5 transcription, we detected the binding levels of DLX5 promoter region with H3K27me3, H3K9me1, and H3K9me2 in SKM-1 cells. The binding level of H3K27me3 and H3K9me2 to DLX5 promoter was significantly decreased in SKM-1 cells transfected with si-EZH2 or si-EHMT2, while the binding rate of H3K9me1 to DLX5 promoter was not affected by si-EHMT2 (p < 0.05) (Figure 7D). These results indicated that EZH2 and EHMT2 synergistically catalyze H3K27me3 and H3K9me2 to inhibit DLX5 transcription in SKM-1 cells.

Then EZH1-treated SKM-1 cells were simultaneously transfected with overexpressing EZH2 and EHMT2 or silencing EZH2 and EHMT2 (Figure 7E), which made H3K27me3 and H3K9me2 increase or decrease simultaneously (Figure 7F). The mRNA level of DLX5 was decreased in the cells overexpressing EZH2 and EHMT2, but increased in the cells silencing EZH2 and EHMT2 (all p < 0.001) (Figure 7G). Taken together, EZH2 and EHMT2 synergistically inhibited DLX5 gene transcription in SKM-1 cells.

To verify that the transcription of DLX5 was co-inhibited by H3K27me3 and H3K9me2, we used SKM-1 cells transfected with si-EZH1 and overexpressing EZH2 and EHMT2 as the control group, and treated the cells with 5 μM EZH2 inhibitor EPZ-6438 for 4 days and/or 2.5 μM EHMT2 inhibitor BIX-01294 for 3 days, respectively. It was found that DLX5 mRNA expression was increased whether H3K27me3 or H3K9me2 was inhibited alone, but the increase of DLX5 mRNA expression was the highest when H3K27me3 and H3K9me2 were inhibited at the same time (Figures 8A,B). To avoid the off-target effect of the inhibitors, we also used the EZH2 inhibitor PF-0672630 or EHMT2 inhibitor BRD4770 to treat SKM-1 cells, and compared the levels of H3K27me3 or H3K9me2. The results showed that there was no difference in the effect between inhibitors at the same target (Supplementary Figure 1). The increase of DLX5 expression inhibited the proliferation of SKM-1 cells (Figure 8C). All in all, EZH2 and EHMT2 synergistically inhibited the transcription of DLX5 in MDS cells and then promoted the transformation from MDS to AML.