Ezh1 Targets Bivalent Genes to Maintain Self-Renewing Stem Cells in Ezh2-Insufficient Myelodysplastic Syndrome.

Polycomb repressive complex (PRC) 2 represses transcription through histone H3K27 trimethylation (H3K27me3). We previously reported that the hematopoietic-cell-specific deletion of Ezh2, encoding a PRC2 enzyme, induced myelodysplastic syndrome (MDS) in mice, whereas the concurrent Ezh1 deletion depleted hematopoietic stem and progenitor cells (HSPCs). We herein demonstrated that mice with only one Ezh1 allele (Ezh1+/-Ezh2Δ/Δ) maintained HSPCs. A chromatin immunopreciptation sequence analysis revealed that residual PRC2 preferentially targeted genes with high levels of H3K27me3 and H2AK119 monoubiquitination (H2AK119ub1) in HSPCs (designated as Ezh1 core target genes), which were mostly developmental regulators, and maintained H3K27me3 levels in Ezh1+/-Ezh2Δ/Δ HSPCs. Even upon the complete depletion of Ezh1 and Ezh2, H2AK119ub1 levels were largely retained, and only a minimal number of Ezh1 core targets were de-repressed. These results indicate that genes marked with high levels of H3K27me3 and H2AK119ub1 are the core targets of polycomb complexes in HSPCs as well as MDS stem cells.

Introduction

Polycomb-group proteins are responsible for the regulation of gene expression by catalyzing repressive histone modifications. They form several types of complexes, including Polycomb repressive complex 2 (PRC2) and PRC1, which are responsible for histone H3 lysine 27 mono-, di-, and tri-methylation (H3K27me1/2/3) and H2AK119 mono-ubiquitination (H2AK119ub1), respectively. PRC2 consists of EED, SUZ12, and the PRC2 enzymatic component, enhancer of zeste homolog 2 (EZH2). PRC2-mediated H3K27me3 plays critical roles in the stage-specific repression of developmental regulator genes, which frequently exhibits bivalency with an active histone modification, H3K4me3, during cellular differentiation. The deregulation of this system has been observed in various types of cancers (Conway et al., 2015, Margueron and Reinberg, 2011).

EZH1 is another enzymatic component of PRC2 and has a catalytic SET domain showing high homology to that of EZH2. The H3K27me2/3 activity of EZH1 is markedly weaker than that of EZH2 in vivo (Margueron et al., 2008), and, consistent with this finding, no mutation in EZH1 has been identified, at least in hematological malignancies. The biological importance of EZH1 is considered to be as a backup enzyme for EZH2, which compensates for EZH2 deficiencies in transcriptional repression in embryonic stem (ES) cells, skin stem cells, and hematopoietic cells (Ezhkova et al., 2011, Margueron et al., 2008, Mochizuki-Kashio et al., 2015, Shen et al., 2008). In addition, several groups have revealed that EZH1 forms a non-canonical PRC2 complex that is associated with active transcription (Henriquez et al., 2013, Mousavi et al., 2012, Stojic et al., 2011, Xu et al., 2015). Another intriguing but controversial issue would be the tissue-specific compensation between EZH1 and EZH2.

PRC2-mediated H3K27me3 cooperates with H2AK119ub1 to repress gene expression. H2AK119ub1 is the epigenetic modification catalyzed by canonical and variant (non-canonical) PRC1s, which contain a RING finger E3 ligase, Ring1B or Ring1A, as the enzymatic component. H2AK119ub1 functions down- and upstream of H3K27me3. In the well-established model, PRC2-induced H3K27me3 recruits canonical PRC1, containing CBX as the H3K27me3-binding module. On the other hand, recent studies have reported the existence of variant PRC1s, which lack CBX proteins but bind to a stretch of unmethylated CpG sites and induce H2AK119ub1, independently of PRC2 (Blackledge et al., 2015, Holoch and Margueron, 2017, Kondo et al., 2016).

Comprehensive genome sequencing studies identified change-of-function mutations in EZH2, which increase H3K27me3 levels and reduce H3K27me2 levels, in patients with follicular and diffuse large B-cell lymphomas (Morin et al., 2010). Loss-of-function mutations in EZH2 have also been identified in patients with myelodysplastic syndrome (MDS) (3%–13%), myeloproliferative neoplasms (MPN) (3%–13%), and MDS/MPN overlap disorders (8%–15.6%), which are all clonal myeloid disorders originating from HSCs (Iwama, 2017, Sashida and Iwama, 2017). Since EZH2 is located at chromosome 7q36.1, chromosomal abnormalities, such as −7 and 7q-, result in deletions of EZH2 in hematological malignancies (Honda et al., 2015). We demonstrated using mice models that the hematopoietic-cell-specific deletion of Ezh2 caused a number of hematological malignancies, such as MDS, MDS/MPN, and MPN (Mochizuki-Kashio et al., 2015, Muto et al., 2013, Sashida et al., 2014, Sashida et al., 2016). Collectively, these findings suggest a tumor suppressor role for EZH2 in hematological malignancies. Furthermore, we found that in the absence of Ezh1, the loss of Ezh2 did not induce any hematological malignancies due to the exhaustion of hematopoietic stem cells (HSCs). These findings showed that Ezh1 plays an essential role in Ezh2-insufficient MDS and that at least either Ezh1 or Ezh2 is required for normal HSC maintenance (Mochizuki-Kashio et al., 2015). The requirement of PRC2 for HSC maintenance has been reported in another mouse model in which Eed was deleted in a hematopoietic-cell-specific manner (Xie et al., 2014). Cdkn2a was identified as one of the critical target genes (TG) of PRC2 for HSC function because its deletion partially rescued the exhaustion of Eed-deleted HSCs, implying the existence of other PRC2 targets involved in HSC maintenance. However, the molecular mechanism underlying Ezh1-mediated maintenance of HSCs and MDS stem cells in an Ezh2-insufficient setting has not been addressed.

In the present study, we deleted all genes encoding PRC2 enzymatic components, except for a single allele of Ezh1, in hematopoietic cells in mice (Ezh1Ezh2). Using these mice, we revealed that a single allele of Ezh1 is enough for Ezh2-insufficient MDS development. We profiled Ezh1 core TG, which are critical for the maintenance of HSCs as well as MDS stem cells. Furthermore, H2AK119ub1 mediated by PRC1 has been suggested to contribute to the transcriptional repression of Ezh1 core TG in the setting of a PRC2 insufficiency.

Results

Ezh1Ezh2 Mice Maintain HSC Functions

We previously reported that Ezh2 mice developed heterogeneous hematological malignancies, mostly MDS and MDS/MPN, whereas Ezh1Ezh2 (DKO) mice did not develop any disease due to the exhaustion of HSCs (Mochizuki-Kashio et al., 2015). These findings clearly indicated an important role for Ezh1 in the maintenance of HSCs and tumor-initiating cells in the setting of an Ezh2 insufficiency. To clarify the function of Ezh1 in hematopoiesis and MDS, we generated Ezh1Ezh2 mice to analyze the impact of a one-allele loss of Ezh1 in Ezh2 mice. Bone marrow (BM) cells from Cre-ERT control, Cre-ERT; Ezh1, Cre-ERT; Ezh2, and Cre-ERT; Ezh1Ezh2 mice (CD45.2) were transplanted into lethally irradiated CD45.1 recipient mice. Ezh2 was deleted by intraperitoneal injections of tamoxifen 1 month post-transplantation (Figure 1A). We hereafter refer to recipient mice reconstituted with Cre-ERT control, Cre-ERT; Ezh1, Cre-ERT; Ezh2, and Cre-ERT; Ezh1Ezh2 cells as wild-type (WT), Ezh1, Ezh2, and Ezh1Ezh2 mice, respectively. Genomic PCR and RNA sequencing (RNA-seq) analyses confirmed the efficient deletion of Ezh2 in Ezh2 and Ezh1Ezh2 mice (Figures 1B and 1C). RNA-seq revealed that Ezh1 mRNA levels were reduced by approximately 50% in Ezh1Ezh2 cells (Figure 1B). A western blot analysis confirmed reductions in the global levels of tri- and di-methylation at histone H3 lysine 27 (H3K27me3 and me2) and the methylation to acetylation switch at H3K27 (Pasini et al., 2010) in Ezh2 and Ezh1Ezh2 cells. The loss of one Ezh1 allele had a minimal impact on the global levels of histone modifications at H3K27 (Figure 1D). Intriguingly, the chimerism of Ezh1Ezh2 donor cells, like that of WT, Ezh1, and Ezh2 mice, was almost 100% in peripheral blood (PB) at least for 6 months after Ezh2 deletion (Figure 1E). Ezh1Ezh2 mice showed morphological dysplasia in PB cells (Figure 1F) as Ezh2 mice did (Mochizuki-Kashio et al., 2015), and also showed macrocytic anemia, leukopenia, and increased apoptosis of BM erythroblasts (data not shown). These results indicate that Ezh1Ezh2 mice developed MDS and maintained MDS stem cells as well as HSCs for a long term.

Figure 1

Efficient Deletion of Ezh1 and Ezh2 in Hematopoietic Cells

(A) The experimental scheme of BM transplantation (BMT) and hematopoietic-cell-specific deletion of Ezh1 and Ezh2.

(B) Snapshots of RNA-seq signals at Ezh1 and Ezh2 gene loci in Lin−Sca-I+c-Kit+ cells (LSK cells) obtained from WT, Ezh1, Ezh2, and Ezh1Ezh2 recipient mice 3 months (Mo) after the tamoxifen treatment.

(C) Genomic PCR on Lin-c-Kit+ cells (LK cells) isolated as described in (B), using the tail genomic DNA of donor mice as control.

(D) Western blot analysis of global histone modification levels in hematopoietic progenitor cells (HPCs). LK cells isolated as described in (B) were subjected to a western blot analysis using anti-H3K27me3, H3K27me2, H3K27me1, H3K27ac, and histone H3 antibodies.

(E) The chimerism of donor-derived CD45.2+ cells in the PB of recipient mice.

(F) Smear preparation of PB from WT and Ezh1Ezh2 mice 6 months after the deletion of Ezh2 observed after May-Giemsa staining. Scale bar, 10 μm.

To further assess the HSC function of Ezh1Ezh2 HSCs, we first performed competitive repopulating assays (Figure 2A). BM cells from each genotype were transplanted with a half number of CD45.1+ competitor cells. Although the chimerism of Ezh1Ezh2 donor cells was slightly lower than that of Ezh2 donor cells in PB, Ezh1Ezh2 donor cells maintained hematopoiesis for a long term (Figure 2B). In contrast, DKO donor cells were totally outcompeted by competitor cells. We then performed serial transplantation using WT, Ezh2, and Ezh1Ezh2 BM cells without competitor cells (Figure 2C). Of interest, Ezh1Ezh2 BM cells did not engraft in 4 of 10 secondary recipients (Figure 2D) and showed inefficient engraftment in the remaining 6 recipients, although they eventually established high chimerism comparable to that of WT and Ezh2 cells 6 months after the secondary transplantation (Figure 2E). A genomic PCR of LK cells obtained 9 months after the secondary transplantation revealed the complete deletion of Ezh2, ruling out that incompletely excised Ezh2 in HSCs contributed to the recovery of chimerism (Figure 2F). These results indicate that HSC function was attenuated but still maintained in Ezh1Ezh2 mice.

Figure 2

Ezh1Ezh2 Mice Maintain HSC Functions

(A) The experimental scheme of competitive repopulating assays. Two million BM cells (CD45.2) from Cre-ERT, Cre-ERT;Ezh1, Cre-ERT;Ezh2, and Cre-ERT;Ezh1Ezh2 were transplanted into lethally irradiated recipient mice (CD45.1) with 1 × 106 competitor BM cells (CD45.1), and Ezh2 was then deleted by injecting tamoxifen 1 month (Mo) post-transplantation.

(B) The chimerism of donor-derived CD45.2+ cells in the PB of recipient mice is shown as the ratio of chimerism values before treatment with tamoxifen (left panel). The chimerism 4 months after the tamoxifen treatment is summarized in a bar graph (right panel). Data are shown as the mean ± SD (WT, n = 4; Ezh1, n = 5; Ezh2, n = 3; Ezh1Ezh2, n = 4; Ezh1Ezh2, n = 4).

(C) The experimental scheme of serial BM transplantation (BMT). For secondary transplantation, 5 × 106 BM cells from primary recipient mice 3 months after tamoxifen treatment were transplanted into lethally irradiated secondary recipient mice.

(D) A summary of the engraftment rates of donor cells in secondary transplantation.

(E) The chimerism of donor-derived CD45.2+ cells in PB in secondary recipients. Data are shown as the mean ± SD (WT, n = 5; Ezh2, n = 9; Ezh1Ezh2, n = 5).

(F) Genomic PCR of LK cells obtained from WT, Ezh2, and Ezh1Ezh2 recipient mice 9 months after secondary transplantation.

**p < 0.01; and ***p < 0.001.

Ezh1 Preferentially Targets Developmental Regulator Genes in Ezh2-Insufficient HSPCs

Irrespective of the global loss of H3K27me3, Ezh1Ezh2 mice maintained HSC function (Figure 2), whereas DKO mice showed the depletion of HSCs without developing any disease (Figure 2) (Mochizuki-Kashio et al., 2015). These results indicate that the residual H3K27me3 marks mediated by only one allele of Ezh1 in Ezh1Ezh2 mice maintained repression of critical PRC2 TG in HSCs as well as MDS stem cells. To characterize the genome-wide distribution of residual H3K27me3, chromatin immuniprecipitation (ChIP) sequencing (ChIP-seq)/DNA sequencing was performed on hematopoietic stem and progenitor cells (HSPCs) purified from WT, Ezh1, Ezh2, and Ezh1Ezh2 mice. The enrichment of H3K27me3 ChIP signals over input signals at the promoter region (transcription start site [TSS] ± 2.0 kb) of each RefSeq gene is shown in scatterplots (Figure 3A). We defined 5,701 genes with H3K27me3 ChIP signals >2-fold over input signals at the promoter region (TSS ± 2.0 kb) in WT cells as PRC2 TG (Figure 3B and Table S1). The Ezh1 deletion had a limited effect on H3K27me3 levels, whereas the levels were markedly lower in Ezh2 and Ezh1Ezh2 cells than in WT cells, as expected from western blot data (Figure 1D). Only 970 genes were marked with residual H3K27me3 in Ezh1Ezh2 cells, and we defined them as “Ezh1 core target genes” (Ezh1 core TG) (Figure 3B and Table S1). H3K27me3 levels at the promoter regions of Ezh1 core TG were significantly higher than those of PRC2 TG (Figure 3C) and were significantly lower in Ezh1 and Ezh1Ezh2 cells (Figure 3C). However, Ezh1 core TG retained high levels of H3K27me3 mark, even in Ezh1 cells (Figure 3C), suggesting that they are regulated not only by Ezh1 but also by Ezh2. The majority of H3K27me3 genes in Ezh1Ezh2 cells (Ezh1 core TG) were also included in those in WT, Ezh1, and Ezh2 cells (Figure 3D). A heatmap clearly showed that the levels of H3K27me3 around the TSS were profoundly lower in Ezh2 and Ezh1Ezh2 cells than in WT and Ezh1 cells (Figure 3E). These results suggest that in the setting of a PRC2 insufficiency, residual PRC2 is distributed to a restricted number of TG to maintain HSC functions. A gene ontology (GO) analysis showed that Ezh1 core TG were remarkably enriched in genes involved in development, cell differentiation, and transcription (Figure 4A). A total of 37.7% of the Ezh1 core TG were transcription factors or DNA-binding proteins (Figure 4B). The expression of Ezh1 core TG was more tightly repressed in the LSK HSPCs of all genotypes than in PRC2 TG (Figure 4C). Collectively, these results suggest that residual PRC2 preferentially targets developmental regulator genes in Ezh2-insufficient HSPCs.

Figure 3

Profiling of PRC2 Target Genes in Ezh1Ezh2 HSPCs

(A) Scatterplots showing the relationship of the fold enrichment values of H3K27me3 ChIP signals against the input signals (ChIP/input) at TSS ± 2.0 kb of RefSeq genes (listed in RefSeq ID) between WT and Ezh1, Ezh2, or Ezh1Ezh2 LK cells 3 months after the tamoxifen treatment.

(B) Bar graph showing the number of H3K27me3 genes that showed ≥ 2-fold enrichment in the level of H3K27me3.

(C) Box-and-whisker plots showing the H3K27me3 levels of PRC2 target genes (TG) and Ezh1 core TG in the indicated mice. Boxes represent 25–75 percentile ranges. Vertical lines represent 10–90 percentile ranges. Horizontal bars represent medians. Mean values are indicated by red dots.

(D) Venn diagram showing the overlap between H3K27me3 genes in Ezh1Ezh2 LSK cells (Ezh1 core TG) and those in WT, Ezh1, or Ezh2 LSK cells.

(E) Heatmap showing the levels of H3K27me3 at the range of TSS ± 10.0 kb.

Figure 4

Characterization of Ezh1 Core Target Genes

(A) Gene ontology (GO) analysis of Ezh1 core TG using DAVID Bioinformatics Resources.

(B) Pie graph showing the breakdown of Ezh1 core TG.

(C) Box-and-whisker plots showing the expression levels of PRC2 TG and Ezh1 core TG in LSK cells 3 months (Mo) after the tamoxifen treatment. Boxes represent 25–75 percentile ranges. Vertical lines represent 10–90 percentile ranges. Horizontal bars represent medians. Mean values are indicated by red dots.

Compared with Ezh2 LSK cells, Ezh1Ezh2 LSK cells lost H3K27me3 at the promoters of 1,035 genes (designated as Ezh2/Ezh1Ezh2 loss genes) (Table S1). A GO analysis showed that these genes were also significantly enriched for developmental regulator genes, despite the much lower levels of enrichment compared with Ezh1core TG (Figure S1A, compared with Figure 4A). Initial levels of H3K27me3 around TSS were higher than those of PRC2 TG, but lower than those of Ezh1 core TG (Figure S1B). Their expression levels were moderately, albeit not significantly, increased in Ezh1Ezh2 cells compared with Ezh2 cells (Figure S1C).

De-repression of Ezh1 Core TG Restricts the Proliferative Capacity of HSCs

To examine the impact of the complete loss of PRC2 on the expression of Ezh1 core TG, we performed RNA sequence analysis on DKO LSK cells. DKO LSK cells were purified from DKO mice 1 week after the deletion of Ezh2 due to the immediate depletion of HSCs. The complete deletion of Ezh1 and Ezh2 in DKO LSK cells, which was confirmed by RNA sequence data (Figure S2A), induced the de-repression of PRC2 TG, including Ezh1 core TG (Figure 5A). To identify the PRC2 TG critical for HSC maintenance, we focused on 42 Ezh1 core TG de-repressed in DKO cells relative to Ezh1Ezh2 cells (Figure 5B). Among them, Cdkn2a, Cdkn2b, Pitx1 (Liu and Lobie, 2007), Egr2 (Unoki and Nakamura, 2003), Egr3 (Cheng et al., 2015), Tbx15 (Yuan et al., 2011), and HtrA1 (Supanji et al., 2013) have been reported to negatively regulate cell proliferation or induce cell death. The de-repression of Cdkn2a, Pitx1, Egr2, Egr3, Tbx15, and HtrA in DKO cells was confirmed by a quantitative RT-PCR analysis (Figure 5C). Manual ChIP assays also confirmed that H3K27me3 levels were significantly reduced at the promoter of Cdkn2a, Pitx1, and Egr3 in DKO cells but kept at high levels in Ezh1Ezh2 cells (Figure S2B). Cdkn2a and Cdkn2b are representative targets of PRC1 and PRC2 that are critical for the maintenance of the proliferative capacity of HSCs (Hidalgo et al., 2012, Oguro et al., 2006, Park et al., 2003, Xie et al., 2014). To investigate the effects of the de-repression of other Ezh1 core TG in HSCs and MDS stem cells, we transduced LSK cells with Pitx1, Egr2, Egr3, or Tbx15 lentiviruses. We purified transduced cells expressing Venus as a marker protein of transduction, and then monitored their growth. A quantitative RT-PCR analysis confirmed a significant overexpression of Pitx1, Egr2, and Egr3 (Figure 5D). The overexpression of all the tested genes, particularly Egr2, Egr3, and Tbx15, significantly attenuated the growth of LSK cells under a stem cell culture condition supplemented with Stem cell factor (SCF) and Thrombopoietin (TPO), suggesting that de-repression of each tested gene impairs HSC function (Figure 5E). De-repression of these Ezh1 core TG might account for the exhaustion of HSCs in DKO mice.

Figure 5

Identification of Responsible Genes Involved in HSC Depletion in PRC2-Null Mice

(A) Box-and-whisker plots showing the expression levels of PRC2 TG and Ezh1 core TG in LSK cells 1 week (wk) after the tamoxifen treatment. Boxes represent 25–75 percentile ranges. Vertical lines represent 10–90 percentile ranges. Horizontal bars represent medians. Mean values are indicated by red dots.

(B) Scatterplots showing the relationship between H3K27me3 levels in Ezh1Ezh2 LSK cells and fold expression levels of PRC2 TG in DKO relative to Ezh1Ezh2 LSK cells.

(C) A quantitative RT-PCR analysis in WT, Ezh1Ezh2, and DKO LSK cells 1 week after the tamoxifen treatment. mRNA levels were normalized to Hprt1 expression, and relative expression levels are shown as the mean ± SD of triplicate analyses. Cells with expression levels arbitrarily set to 1 are indicated as “1.” ND, not detected.

(D and E) Overexpression of selected Ezh1 core TG in WT LSK cells using a lentivirus. Sorted Venus-positive cells were cultured in the presence of SCF and TPO. (D) A quantitative RT-PCR analysis on cells cultured for 10 days. mRNA levels were normalized to Hprt1 expression, and expression levels relative to the control are shown as the mean ± SD of triplicate analyses. (E) Growth of cells overexpressing the indicated genes. Cell numbers are shown as the mean ± SD of triplicate cultures.

*p < 0.05; **p < 0.01; ***p < 0.001.

H3K4me3 Was Associated with Ezh1 Core TG

Bivalent genes marked with a repressive mark H3K27me3 and an active mark H3K4me3 are the so-called developmental regulator genes that are involved in the regulation of development and differentiation (Bernstein et al., 2006, Zaidi et al., 2017). Given that Ezh1 core TG are rich in developmental regulator genes (Figure 4A), we profiled H3K4me3 in WT and PRC2-insufficient LSK HSPCs by ChIP-seq analysis (Figure 6A). Genes with both H3K27me3 and H3K4me3 ChIP signals >2-fold over input signals at the promoter region (TSS ± 2.0 kb) in WT cells were defined as bivalent genes in LSK cells. We found that 882 of 970 Ezh1 core TG were overlapped with the bivalent genes (Figure 6B) and were significantly enriched for developmental regulator genes, as evident from the data of GO analysis (Figure 6C). A heatmap and the following quantitative analysis showed that the levels of H3K4me3 at the promoters of Ezh1 core TG were remarkably higher than those at other PRC2 TG in both WT and Ezh1Ezh2 cells (Figures 6D and 6E). Levels of H3K4me3 at the promoters of Ezh1 core TG were slightly reduced in Ezh1Ezh2 cells but stayed at higher levels than those at other PRC2 TG (Figures 6A, 6D, and 6E). Taken together, these data suggest that residual PRC2 preferentially targets bivalent developmental regulator genes in its insufficiency.

Figure 6

Profiling of H3K4me3 in Ezh1Ezh2 HSPCs

(A) Scatterplots showing the relationship of the fold enrichment values (ChIP/input) of H3K27me3 and H3K4me3 at TSS ± 2.0 kb of RefSeq genes. WT, Ezh1, Ezh2, and Ezh1Ezh2 LSK cells were obtained from mice 3 months after the tamoxifen treatment. Black and gray dots indicate PRC2 TG and others, respectively.

(B) Venn diagram showing the overlap between Ezh1 core TG and bivalent genes in WT LSK cells.

(C) GO analysis data of Ezh1 core TG overlapping with bivalent genes in (B) using DAVID Bioinformatics Resources.

(D) Heatmap showing H3K27me3 and H3K4me3 levels at the range of TSS ± 10.0 kb.

(E) Box-and-whisker plots showing H3K4me3 levels. Boxes represent 25–75 percentile ranges. Vertical lines represent 10–90 percentile ranges. Horizontal bars represent medians. Mean values are indicated by red dots. ***p < 0.001.

PRC1-Mediated H2AK119ub1 Was Involved in Repression of Ezh1 Core TG

PRC1 largely shares TG with PRC2 because PRC1-mediated H2AK119ub1 functions up- and downstream of PRC2-mediated H3K27me3 (Blackledge et al., 2014, Blackledge et al., 2015, Cooper et al., 2016, Kalb et al., 2014). We hypothesized that H2AK119ub1 also plays a key role in the repression of PRC2 TG in PRC2-insufficient HSPCs. To test this hypothesis, the state of the H2AK119ub1 modification in Ezh1Ezh2 LSK HSPCs was examined. For analysis on DKO HSPCs, c-Kit BM cells that were collected 1 week after tamoxifen treatment were used. A western blot analysis showed that the global level of H2AK119ub1 was not significantly changed by the loss of PRC2, even in DKO cells (Figure 7A). A ChIP-seq analysis of H2AK119ub1 revealed a strong correlation of H2AK119ub1 levels with H3K27me3 levels at the promoters of PRC2 TG, particularly of the Ezh1 core TG in Ezh1Ezh2 cells (Figure 7B). The genes, the promoters of which were marked with H3K27me3 and H2AK119ub1, largely overlapped (Figure 7C). A heatmap and boxplot showed that the deletion of Ezh1 and/or Ezh2 had limited effects on the level of H2AK119ub1 at the promoters of PRC2 TG (Figures 7D and 7E), which is consistent with the global levels of H2AK119ub1 detected in western blots (Figure 7A). Manual ChIP assays confirmed that H2AK119ub1 levels of selected Ezh1 core TG were maintained at the basal levels in DKO as well as Ezh1Ezh2 HSPCs (Figure S2C). These results suggest a role for PRC1-mediated H2AK119ub1 in the repression of PRC2 TG in the setting of a PRC2 insufficiency.

Figure 7

Involvement of H2AK119ub1 in Repression of Ezh1 Core TG

(A) Western blot analysis of global histone modification levels in hematopoietic progenitor cells (HPCs). LK cells obtained from mice 3 months (Mo) or c-Kit+ cells obtained 1 week (wk) after the tamoxifen treatment were subjected to a western blot analysis using anti-H2AK119ub1 and histone H2A antibodies. H2AK119ub1 amounts relative to total H2A are indicated.

(B) Scatterplots showing the relationship of the fold enrichment values (ChIP/input) of H3K27me3 and H2AK119ub1 at TSS ± 2.0 kb of RefSeq genes. WT, Ezh1, Ezh2, and Ezh1Ezh2 LSK cells were obtained from mice 3 months after the tamoxifen treatment. Black and gray dots indicate PRC2 TG and others, respectively.

(C) Venn diagram showing the overlap between genes marked with H2AK119ub1 and those with H3K27me3 at their promoters in WT (upper) and Ezh1Ezh2 (lower) LSK cells.

(D) Heatmap showing H3K27me3 and H2AK119ub1 levels at the range of TSS ±10.0 kb of PRC2 TG in LSK cells 3 months or c-Kit cells 1 week after the tamoxifen treatment.

(E) Box-and-whisker plots showing H3K27me3 and H2AK119ub1 levels of PRC2 TG. Boxes represent 25–75 percentile ranges. Vertical lines represent 10–90 percentile ranges. Horizontal bars represent medians. Mean values are indicated by red dots.

Discussion

Only Ezh2 and its paralog Ezh1 function as H3K27 methyltransferases. No other enzymes have been shown to exhibit similar catalytic activities. We previously reported that the hematopoietic-cell-specific deletion of Ezh2 did not compromise the self-renewal capacity of HSCs and caused heterogeneous hematological malignancies, including MDS, in mice. In contrast, Ezh1Ezh2 mice failed to maintain HSCs as well as MDS stem cells (Mochizuki-Kashio et al., 2015). Here, our ChIP-seq analysis provided a new insight into the epigenetic mechanisms of maintenance of HSCs and MDS stem cells in the setting of an Ezh2 insufficiency.

Another group also reported the requirement of PRC2 for HSC maintenance using Eed-deleted mice, which are similar to Ezh1Ezh2 (DKO) mice in terms of the complete loss of PRC2 activity (Xie et al., 2014). These findings suggest that Ezh1 plays an important role in the maintenance of HSCs and development of MDS in the setting of an Ezh2 insufficiency. In the present study, we demonstrated that only one allele of Ezh1 (Ezh1Ezh2) was sufficient to maintain HSCs as well as MDS stem cells (Figure 2), despite the markedly weaker catalytic activity of Ezh1 than Ezh2 (Margueron et al., 2008). This result indicates that the minimal levels of H3K27me3 mediated by Ezh1 are sufficient for HSC maintenance.

The enzymatic activity of Ezh1 was previously shown to be markedly weaker than that of Ezh2 in vitro and in vivo (Margueron et al., 2008). Correspondingly, the deletion of Ezh1 alone (Ezh1) affected H3K27me3 levels in HSPCs much less than that of Ezh2. However, in the absence of Ezh2, Ezh1 heterozygosity had a significant impact on H3K27me3 levels at the promoters of PRC2 TG (Figures 3A–3C and 3E). Only 970 genes retained H3K27me3 at their promoters in Ezh1Ezh2 HSPCs (Figure 3B). These 970 genes, defined as “Ezh1 core TG,” were strongly enriched in genes coding for transcription factors and developmental regulators (Figures 4A and 4B). The forced expression of selected Ezh1 core TG that were de-repressed in DKO HSPCs, including Pitx1, Egr2, Egr3, and Tbx15, suppressed the growth of LSK cells in vitro (Figure 5E). Diffuse intrinsic pontine glioma cells with the mutation of H3K27M exhibited attenuated PRC2 function with the global loss of H3K27me3, whereas the H3K27me3 mark was retained by a limited number of genes, including the tumor suppressor gene CDKN2A. The inhibition of residual PRC2 activity abolished tumor cell growth following the de-repression of CDKN2A (Mohammad et al., 2017). Collectively, these findings indicate that residual PRC2 preferentially targets critical genes for the maintenance of stem cells in PRC2-insufficient tumor cells. In contrast, 1,035 Ezh2/Ezh1Ezh2 loss genes, which lost H3K27me3 at the promoters in Ezh1Ezh2 LSK cells compared with Ezh2 LSK cells, also included many developmental regulator genes, and their expression levels were moderately increased in Ezh1Ezh2 cells compared with Ezh2 cells (Figures S1A–S1C). Given that the de-repression of developmental regulator genes of Ezh1 core targets affect the function of HSCs, de-repression of Ezh2/Ezh1Ezh2 loss genes also could account for the defective function of Ezh1Ezh2 HSCs/MDS stem cells as observed in Figure 2.

In addition to H3K27me3, we profiled H3K4me3 in PRC2-insufficient HSPCs (Figure 6A), since the bivalency of these histone modifications are frequently observed at the promoters of developmental regulator genes (Bernstein et al., 2006, Ku et al., 2008, Zaidi et al., 2017). Consistent with the results of GO analysis (Figure 4A), Ezh1 core TG were mostly bivalent genes in LSK cells (Figure 6B). Of note, we detected remarkably higher levels of H3K4me3 at Ezh1 core TG than other PRC2 TG (Figures 6D and 6E). The high levels of H3K4me3 could serve as a marker for selective targeting of residual PRC2 to Ezh1 core TG.

Many of the developmental regulator genes, such as Hox genes, which are well-known targets of PRC2, were included in Ezh1 core TG, but continued to be repressed in PRC2-insufficient LSK cells, even in DKO mice (Figures 4C, 5A, and 5B). They appeared to be repressed by mechanisms other than PRC2-mediated H3K27me3 in HSPCs. H2AK119ub1 cooperates with H3K27me3 to form a repressive chromatin region called the polycomb domain. In the traditional model, PRC1, which catalyzes H2AK119ub1, is recruited to PRC2-mediated H3K27me3. However, recent studies reported an alternative mechanism, in which PRC1 deposits H2AK119ub1 independently of PRC2. This new type of PRC1 complex has been classified as variant PRC1, whereas PRC1 in the traditional model is called canonical PRC1 (Blackledge et al., 2015). H2AK119ub1 levels were previously shown to be largely retained in PRC2-deficient mouse ES cells (Tavares et al., 2012). Consistent with this finding, we demonstrated that DKO HSPCs exhibited almost the complete retention of H2AK119ub1 at the global level as well as at the promoters of PRC2 TG, suggesting that variant PRC1s, but not canonical PRC1, predominantly contribute to the deposition of H2AK119ub1 in HSPCs as well as in mouse ES cells (Figures 7A, 7D, and 7E). Since variant PRC1-mediated H2AK119ub1 recruits PRC2, leading to the deposition of H3K27me3 (Blackledge et al., 2014), variant PRC1-mediated H2AK119ub1 may be one of the mechanisms underlying the preferential targeting of PRC2 to Ezh1 core TG, as well as H3K4me3. In fact, we found that this mark strongly co-localized with H3K27me3 at Ezh1 core TG in Ezh1Ezh2 HSPCs (Figure 7B). Intriguingly, inhibition of Ring1A, an enzymatic component of PRC1, has been shown to have an antitumor effect on an MDS cell line and primary CD34 cells from patients with MDS (Palau et al., 2017). Therefore, the functional inhibition of PRC1, especially variant PRC1, would represent an interesting approach to eradicate MDS stem cells by inducing the de-repression of Ezh1 core TG in Ezh2-insufficient MDS.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.