Overexpression of RUNX3 Represses RUNX1 to Drive Transformation of Myelodysplastic Syndrome

RUNX transcription factors are critical for development and normal tissue homeostasis and have been characterized as an oncogene or tumor suppressor in the pathogenesis of various tumors (1, 2). RUNX family members, including RUNX1, RUNX2, and RUNX3, are highly conserved in the runt domain, which binds to the consensus DNA sequence and is involved in dimerization with the common cofactor CBFβ. Loss-of-function mutations and the deletion of RUNX1 are frequently observed in hematologic malignancies, including myelodysplastic syndrome (MDS), in patients (3, 4), and the hematopoietic cell-specific deletion of Runx1 impairs the differentiation of megakaryocytes and lymphocytes, which leads to the development of myeloproliferative neoplasm (MPN)-like disease in mice despite a long latency (5, 6), indicating the tumor-suppressive function of RUNX1.

Previous studies demonstrated that the repressed expression of RUNX3 due to promoter DNA hypermethylation promoted the development of solid tumors and was associated with poor survival outcomes (e.g., colon, renal, and lung cancers; ref. 7). RUNX3 is expressed in hematopoietic stem and progenitor cells (HSPC); however, its expression levels decline with aging in humans and mice (8), and this may contribute to the emergence of the aging phenotype in hematopoiesis characterized by anemia and enhanced granulopoiesis (9, 10). The deletion of RUNX3 and the promoter hypermethylation and RUNX1-ETO–induced transcriptional suppression of RUNX3 have been reported in acute myeloid leukemia (AML) and chronic myeloid leukemia cells (11–13). The deletion of Runx3 impaired the differentiation of erythrocytes, but retained the production of myeloid cells, such as granulocytes, which exhibited higher sensitivity to a GCSF treatment in vitro, and resulted in the development of the myeloproliferative phenotype in mice (14), partially sharing the phenotype of Runx1-deficient mice. Furthermore, the deletion of both Runx1 and Runx3 in hematopoietic cells facilitated the development of bone marrow (BM) failure and MPN diseases in mice due to alterations in the transcription-dependent and transcription-independent functions of Runx genes (15). Therefore, the loss of Runx3 induces a compensatory function of Runx1, but also promotes the development of MPN upon the deletion of Runx1, indicating a function for RUNX3 as a tumor suppressor.

However, emerging studies have suggested an oncogenic function for RUNX3 in the pathogenesis of hematologic malignancies, including MDS, because genomic amplifications in the 1p36 region including the RUNX3 gene have been observed in patients with MDS (16). The overexpression of RUNX3 has been correlated with poor clinical outcomes in patients with AML harboring the FLT3-ITD mutation (17). FLT3-ITD has been shown to activate the expression of Runx3 in murine Tet2-dificient AML cells, but also confers a chemoresistant property to human AML cells (18, 19). The Epstein–Barr virus (EBV) oncoprotein enhances the expression of RUNX3 by activating the superenhancer of RUNX3, which promotes the proliferation of transformed lymphoblastoid cells (20). On the basis of these findings, we herein confirmed the enhanced expression of RUNX3 mRNA and protein in a subset of patients with MDS by analyzing our cohort and published datasets (21, 22), and attempted to elucidate the mechanisms by which the overexpression of RUNX3 induces the transformation of MDS in vivo.

The TET2 enzyme oxidizes 5-methylcytosine (5mC) and regulates the region-specific removal of 5mC, and loss-of-function mutations in TET2 are one of the most common driver mutations in myeloid malignancies, including MDS (23, 24), as well as in clonal hematopoiesis in healthy aged individuals (25, 26). The loss of Tet2 promotes the self-renewal function of HSCs and results in the development of MPN-like disorders in mice, even in longer observation periods (27, 28). In this study, we generated a RUNX3-expressing Tet2-deficient MDS mouse model showing human MDS features characterized by pancytopenia, dysplastic myeloid cells, and impaired lymphopoiesis. We found that RUNX3-expressing Tet2-deficient HSPCs markedly suppressed the expression of the Runx1 protein and its target genes, in which RUNX3 bound to consensus DNA sequences, and remodeled the binding sites of Runx1. In addition, RUNX3 overexpression enhanced the transcription of Myc target genes, which showed higher sensitivity to the inhibition of the Myc-Max heterodimerization. The present results revealed the mechanisms by which the overexpression of RUNX3 exerts oncogenic effects on the cellular function of and transcriptional program in Tet2-deficient stem cells to initiate the transformation of MDS.

All mice were in the C57BL/6 background. Tet2 conditional knockout (Tet2^flox/flox) mice were described previously (27), and crossed with Rosa26:Cre-ERT2 mice (TaconicArtemis GmbH) for conditional deletion. Two milligrams of tamoxifen (T5648, Sigma-Aldrich) were administered via an intraperitoneal injection for 5 consecutive days twice to completely delete the Tet2 gene. C57BL/6 mice congenic for the Ly5 locus (CD45.1) were purchased from Sankyo-Lab Service. All experiments using these mice were performed in accordance with our institutional guidelines for the use of laboratory animals and approved by the Review Board for Animal Experiments of Kumamoto University (Kumamoto, Japan). All mouse experiments were performed without randomization and blinding.

All samples were obtained at Tokyo Medical University (Tokyo, Japan) after written informed consent compliant with the Declaration of Helsinki had been obtained from patients. Patient anonymity was ensured, and this study was approved by the Institutional Review Committee at Tokyo Medical University (Tokyo, Japan).

Paraffin-embedded BM aspiration clots were submitted to xylene and alcohol for deparaffinization, followed by the blockade of endogenous peroxidase. Samples were incubated with an anti-RUNX3 antibody (R3-5G4, Santa Cruz Biotechnology) at 4°C overnight. Envision Dual Link System-HRP (Dako) was used as the secondary antibody for 45 minutes, followed by visualization with the DAB chromogen (Dako) and counterstaining using hematoxylin.

RUNX3 was constructed from cDNA encoding human RUNX3 and subcloned into the retrovirus vector pMYs IRES-EGFP. Virus supernatant (VSV-G pseudotyped retroviral supernatant) was prepared by transfecting the 293GPG packaging cell line with an empty control or the FLAG-tagged RUNX3 retrovirus vector plasmid using the calcium phosphate transfection method. Virus supernatant was concentrated by centrifugation at 6,000 × g for 16 hours. The final titers of retroviral supernatants were 40,000 to 50,000 infectious units/μL, as assessed by transducing serially diluted viral supernatants into the human Jurkat cell line. Hematopoietic cells were then incubated with the virus supernatant in 10 μg/mL protamine sulfate and 10 ng/mL RetroNectin (Takara), and infected cells were further incubated in serum-free SF-03 medium supplemented with 50 μg/mL mouse stem cell factor (SCF) and 50 μg/mL human thrombopoietin (TPO).

293T and Jurkat cell lines were obtained from RIKEN, and were cultured in DMEM or RPMI1640 containing 10% FBS in a humidified incubator, respectively. 293GPG was provided by Dr. R. Mulligan (29). PCR Mycoplasma Test Kit (Takara) was used to test for Mycoplasma contamination of all cell lines.

Colony assays were performed using Methocult M3234 (Stem Cell Technologies) supplemented with 20 ng/mL mouse SCF, 20 ng/mL mouse IL3, 20 ng/mL human TPO, and 2 units/mL human erythropoietin. The number of colonies was counted on day 10 of culture. In the replating assay, colonies were scored on day 10 and 3 × 10⁴ pooled cells were replated in the same medium.

Flow cytometry and cell sorting were performed by utilizing the following antimurine antibodies (clone and catalog numbers): CD45.2 (104, 109820), CD45.1 (A20, 110730), Gr1 (RB6-8C5, 108404), CD11b/Mac1 (M1/70, 101208), Ter119 (116204), CD127/IL7Rα (A7R34, 121104), B220 (RA3-6B2, 103212), CD4 (L3T4, 100526), CD8α (53-6.7, 100714), CD117/c-Kit (2B8, 105812), Sca-1 (D7, 108114), CD34 (MEC14.7, 11-0341-85), CD71 (R27217, 113808), CD150 (TC15-12F12.2, 115924), and FcγRII-III (93, 101308). These antibodies were purchased from BioLegend, eBioscience, R&D Systems, or TONBO Biosciences. The lineage mixture solution contained biotin-conjugated anti-Gr1, B220, CD4, CD8α, Ter119, and IL7Rα antibodies. Apoptotic cells were stained with an anti-Annexin V–PE antibody (BD 556422) and 7-AAD (BD 559925). To evaluate cell-cycle progression, cells were fixed and permeabilized according to the manufacturer's instructions, and then detected using an anti-Ki67–Alexa Fluor 647 antibody (Biolegend 350510). All flow cytometric analyses and cell sorting were performed on FACSAriaIII or FACSCantoII (BD Biosciences).

Murine HSPCs were purified from pooled BM cells (harvested from 3–5 mice) and were fixed by 0.5% or 1.0% paraformaldehyde at 37°C for 5 minutes and then lysed. Cells were sonicated 15 times at an amplitude of 50% for 10 seconds. Samples were incubated with anti-Runx1 antibody–conjugated (Abcam, ab23980), anti-FLAG antibody-conjugated (Sigma, F1804), or anti-H3K27Ac antibody–conjugated (Active motif, MABI 0309) Dynabeads protein A/G at 4°C overnight. Inputs and immunoprecipitates were incubated at 65°C for 4 hours for reverse cross-linking, and DNA was purified using the MinElute PCR Purification Kit (Qiagen). Chromatin immunoprecipitation sequencing (ChIP-seq) libraries were generated using the ThruPLEX DNA-seq Kit (Rubicon Genomics). Bowtie2 (version 2.2.6; default parameters) was used to map the reads to the reference genome (UCSC/mm9). HOMER (version 4.9) was used for de novo motif discovery in the peaks of superenhancers and enhancers. ChIP-seq data were deposited in the DDBJ database under the accession numbers DRA008820 and DRA009681.

Total RNA was extracted using an ISOGEN (Nippon gene), and cDNA was synthesized using the SMARTer Pico PCR cDNA Synthesis Kit (Clontech). ds-cDNA was fragmented and cDNA libraries were generated using the KAPA HyperPlus Library Preparation Kit (KAPA Biosystems) and FastGene Adaptor Kit (Fastgene). Sequencing was performed using NextSeq500 (Illumina) with a single-read sequencing length of 60 bp, and the CLC genomic Workbench was used to analyze and visualize sequencing data. RNA sequencing data have been deposited in the DDBJ database under the accession numbers DRA008434 and DRA009681.

Total RNA was isolated using the RNeasy Mini kit and then reverse transcribed by the ThermoScript RT-PCR system (Invitrogen) with an oligo-dT primer. qRT-PCR was performed on LightCycler 480 (Roche) using SYBR Premix ExTaq (Takara) or FastStart Universal Probe Master (Roche) with a Universal Probe Library (Roche). Expression levels were normalized to that of Gapdh. Primers for PCR are listed in Supplementary Table S1.

Whole cell lysates were used for Western blotting. Briefly, Lin-Kit+ BM cells were isolated from mice and lysed in 1 × SDS sample buffer (50 mmol/L Tris-HCl, pH 6.83, 1.5% SDS, 10% glycerol). In the co-immunoprecipitation (IP) assay, 293T cells transfected with the indicated plasmids were lysed in modified RIPA buffer (50 mmol/L Tris pH 7.5, 150 mmol/L NaCl, 1% NP40, 0.25% sodium deoxycholate, and 1 mmol/L EDTA). Lysates were incubated with an anti-HA antibody (3F10, Roche, 11867423001) and rotated at 4°C overnight. Lysates were mixed with Dynabeads protein A/G at 4°C for 1 hour, and then washed twice with IP buffer I (50 mmol/L Tris pH 7.5, 150 mmol/L NaCl, 0.5% NP40, 0.25% sodium deoxycholate) followed by IP buffer II (50 mmol/L Tris pH 7.5, 0.1% NP40, 0.05% sodium deoxycholate). The following antibodies were used for Western blotting: FLAG (Sigma, F1804), HA-Tag (Cell Signaling Technology, C29F4), RUNX3 (MBL, R3-5G4), RUNX1 (Sigma, R0406; Abcam, ab23980), MYC (Abcam, Y69), Actin (Santa Cruz Biotechnology, C4), H3 (Abcam, ab1791), and H3K27ac (Active motif, MABI 0309). Uncropped immunoblot images are available in Supplementary Fig. S1.

The promoter or enhancer region of Cebpa, Csf1r, and Hnrnpa0 was amplified by PCR using mouse genomic DNA. PCR primer sequences and amplified genomic regions (mm9) were as follows: Cebpa 5′-CTGTGGTCAACTTCTAGTGGCTTTC-3′, 5′-AGCAGAGCAACCTTAACACTCATTG-3′, and chr7:35912057+35912683; Csf1r 5′-CCCAAGTGGTGTGCCAAGTG-3′, 5′-CACAGAGGGCTGACCACACC-3′, and chr18:61250779+61250955; Hnrnpa0 5′-CCGATGAACAGCTTACAGAGCTGCG-3′, 5′-CAGAGGGTTCCGCACGGGAG-3′, and chr13:58229640+58230427. Amplified products were cloned into the NheI and EcoRV sites of the pGL4.23 vector (Promega). 293T cells were transiently transfected with vector plasmids and cell extracts were prepared 48 hours after transfection. Luciferase activity was assessed by performing the Promega Dual-Luciferase reporter assay system (Promega, E1910).

All statistical tests were performed using Graph Pad Prism version 7 (GraphPad Software). The significance of differences was measured by an unpaired two-tailed Student t test or the Mann–Whitney nonparametric test. A P value of less than 0.05 was considered to be significant. No statistical methods were used to predetermine sample sizes for animal studies.

Sequencing data that support the results of this study have been deposited in DDBJ under the accession numbers DRA008434, DRA008820, and DRA009681 for RNA sequencing and ChIP sequencing.

In a gene expression analysis of 183 patients with MDS, as reported previously (22), RUNX3 mRNA expression levels in CD34⁺ HSPCs were higher in 7 of 55 patients and in 4 of 80 refractory anemia (RA) and refractory anemia with excess blasts (RAEB) patients than those in the HSPCs of healthy controls (Fig. 1A). We also found that 2 of 55 patients with MDS showed the increased expression of RUNX3 in HSPCs by analyzing another cohort dataset (Fig. 1B; ref. 21). On the basis of the enhanced expression of RUNX3 mRNA in HSPCs, we assessed the expression level of the RUNX3 protein in the BM tissues of patients with MDS in our cohort and found that 4 of 20 patients showed higher expression levels than those in healthy controls (Fig. 1C and D). Detailed clinical information on our cohort was provided in Supplementary Table S2. Some RUNX3 high-expressing patients harbored high-risk -7/7q- anomalies, whereas RUNX3 low-expressing patients did not (Supplementary Fig. S2). Notably, in the other cohort of 122 patients with MDS (30), survival was significantly shorter in RUNX3 high-expressing patients than in RUNX3 low-expressing patients (median survival 702.5 days vs. 1,533 days, P = 0.0287; Fig. 1E). By utilizing the dataset on gene expression in HSPCs (21, 22), we found similar TET2 mRNA expression levels in RUNX3 high- and low-expressing patients in these cohorts; however, RUNX3 high-expressing HSPCs showed significant enrichments in the expression of upregulated and downregulated genes defined in murine RUNX3-Tet2^Δ/Δ MDS HSPCs (as described below; Supplementary Fig. S3). In addition, RUNX3 high-expressing HSPCs showed markedly positive enrichments in the TNFα and TGFβ signaling pathways over RUNX3 low-expressing HSPCs (Fig. 1F), implying that RUNX3 overexpression contributes to activating the expression of genes in these key oncogenic pathways in MDS. Thus, RUNX3 appears to function as an oncogene in the development of MDS in patients showing a worse outcome.

To establish whether RUNX3 promotes the development of MDS in vivo, we generated a hematopoietic cell-specific RUNX3-expressing Tet2-deficient mouse model because the loss-of-function mutation in TET2 is one of the most common genetic mutations observed in clonal hematopoiesis in aged individuals (23, 24), and in patients with MDS showing stronger RUNX3 expression (30). We initially deleted Tet2 by activating Cre recombinase via intraperitoneal injections of tamoxifen in Cre-ERT2 and Tet2^flox/flox;Cre-ERT2 mice. Two months after the tamoxifen injection, hematopoietic stem cells (HSC) purified from mice were infected with an empty control or FLAG-tagged RUNX3 retrovirus vector under liquid culture conditions supplemented with SCF and TPO showing comparable transduction efficacies (Supplementary Fig. S4), and were transplanted into lethally irradiated CD45.1⁺ wild-type (WT) recipient mice together with freshly harvested CD45.1⁺ WT BM cells (Fig. 2A). We hereafter refer to recipient mice reconstituted with control, RUNX3-expressing, Tet2^Δ/Δ, and RUNX3-expressing Tet2^Δ/Δ HSCs as WT, RUNX3-Tet2^wt/wt, Tet2^Δ/Δ, and RUNX3-Tet2^Δ/Δ mice, respectively. We confirmed the overexpression of the RUNX3 protein in c-Kit+ cells (Fig. 2B), and the decreased expression of Tet2 mRNA in Lineage⁻c-Kit⁺Sca-1⁺ (LSK) HSPCs isolated from these mice (Fig. 2C).

Complete blood count (CBC) analyses revealed no significant changes between these genotypes at 2 months posttransplantation (Supplementary Fig. S5). While neither RUNX3-Tet2^wt/wt mice nor Tet2^Δ/Δ mice showed significant changes in CBC at 9 months posttransplantation, platelet counts in RUNX3-Tet2^Δ/Δ mice gradually decreased from 7 months and these mice developed thrombocytopenia at 9 months posttransplantation (Fig. 2D). Moribund RUNX3-Tet2^Δ/Δ mice showed lower frequencies of CD45.2⁺GFP⁺ cells in B and T cells but higher frequency in mature myeloid cells in peripheral blood (PB) than RUNX3-Tet2^Δ/Δ mice at the predisease stage (Fig. 2E). Although survival times were not shorter in RUNX3-Tet2^Δ/Δ mice than in Tet2^Δ/Δ mice (median survival 455 days vs. 351.5 days, P = 0.4862), which develop MDS/MPN showing leukocytosis as reported previously, moribund RUNX3-Tet2^Δ/Δ mice had severe neutropenia (Fig. 2F), anemia, and thrombocytopenia (Fig. 2D), accompanied by dysplastic neutrophils, anisocytosis, and giant platelets in PB and dysplastic megakaryocytes in BM and the spleen (Fig. 2G), which are characteristic features of MDS in patients. While RUNX3-Tet2^Δ/Δ mice showed similar numbers of BM cells to mice with other genotypes (Fig. 2H), spleen weights were variable in RUNX3-Tet2^Δ/Δ MDS mice (Fig. 2I) and the expansion of Lineage⁻Sca-1⁺c-Kit⁺ (LSK) HSPCs was observed in the spleen (Supplementary Fig. S6), indicating extramedullary hematopoiesis in RUNX3-Tet2^Δ/Δ MDS mice. We found that RUNX3-Tet2^wt/wt mice and Tet2^Δ/Δ mice developed MDS and MDS/MPN according to the criteria for murine MDS and the MDS/MPN subcategories (31), whereas RUNX3-Tet2^Δ/Δ stem cells dominantly induced the development of MDS in mice showing dysplasia and cytopenia, but lacking leukocytosis in PB (Fig. 2J). Thus, the overexpression of RUNX3 compromises normal hematopoiesis and drives the development of MDS in the absence of Tet2 in vivo.

We then performed a detailed phenotypic analysis of the BM cells of all genotype mice 2 months posttransplantation and moribund RUNX3-Tet2^Δ/Δ MDS mice. Pre-MDS RUNX3-Tet2^Δ/Δ mice showed similar frequencies of CD45.2⁺GFP⁺ cells in LSK HSPCs and GMP cells in BM, while RUNX3-Tet2^Δ/Δ MDS mice subsequently showed larger numbers of these cells in HSPCs and GMPs (Fig. 3A). On the basis of the increased proportion of RUNX3-Tet2^Δ/Δ cells in HSPCs, we performed Ki67 and Annexin V staining by flow cytometry and found that RUNX3-Tet2^Δ/Δ mice showed a slightly higher frequency of Ki67-positive cells (Fig. 3B), but less apoptosis in HSPCs in BM (Fig. 3C). The results obtained also showed that RUNX3-Tet2^Δ/Δ HSPCs maintained larger numbers of in vitro replating colonies than Tet2^Δ/Δ HSPCs showing monocyte colonies in the third plating and also generated mixed-lineage colonies (e.g., GEMM and GM; Fig. 3D). To further assess the proliferative capacity of RUNX3-Tet2^Δ/Δ cells in vivo, we performed the competitive transplantation of CD45.2⁺GFP⁺ HSPCs purified from primary transplanted mice with freshly harvested WT CD45.1⁺ BM cells into lethally irradiated recipient mice (Fig. 3E). WT HSPCs did not repopulate in this experimental setting based on the low frequencies of CD45.2⁺GFP⁺ cells in PB (Fig. 3F) and HSPCs in BM at 6 months posttransplantation (Fig. 3G), while RUNX3-Tet2^Δ/Δ mice showed a higher frequency of CD45.2⁺GFP⁺ cells in myeloid cells, but negligible amounts in lymphoid cells in PB (Fig. 3F), and showed the upregulated expression of the RUNX3 protein in c-Kit+ cells in BM (Fig. 3H). RUNX3-Tet2^Δ/Δ mice showed a similar CD45.2⁺GFP⁺ frequency in HSPCs, CMPs, and GMPs to those in single mutant mice in BM (Fig. 3G), indicating that RUNX3 overexpression and/or a Tet2 deficiency sustained the repopulating capacity of HSPCs and myeloid progenitor cells in vivo. In contrast, RUNX3-Tet2^Δ/Δ mice showed a smaller frequency of CD45.2⁺GFP⁺ in megakaryocyte/erythroid progenitor (MEP) cells than that in Tet2^Δ/Δ mice (Fig. 3G), which may have contributed to thrombocytopenia in primary transplanted mice (Fig. 2D). Thus, although RUNX3-Tet2^Δ/Δ HSPCs maintain their repopulating capacity, their differentiation is impaired, leading to the development of MDS.

To elucidate the mechanisms by which RUNX3-Tet2^Δ/Δ HSPCs develop MDS in vivo, we performed gene expression analyses using RNA sequencing on HSPCs isolated from WT, RUNX3-Tet2^wt/wt, Tet2^Δ/Δ, and RUNX3-Tet2^Δ/Δ mice at 2 months posttransplantation, one Tet2^Δ/Δ MDS/MPN mouse, two Tet2^Δ/Δ MDS mice, and two RUNX3-Tet2^Δ/Δ MDS mice. As expected, a principal component analysis revealed that RUNX3-Tet2^Δ/Δ cells at the pre-MDS and MDS stages clustered closer, but separately from WT, Tet2^Δ/Δ, and Tet2^Δ/Δ MDS cells (Fig. 4A). We observed 68, 258, and 151 upregulated genes and 279, 615, and 415 downregulated genes in RUNX3-Tet2^wt/wt, Tet2^Δ/Δ, and RUNX3-Tet2^Δ/Δ cells, respectively, at the pre-MDS stage from those in WT cells (Fig. 4B; Supplementary Data 1). We also found that RUNX3-Tet2^Δ/Δ MDS HSPCs shared approximately 50% upregulated and downregulated genes in RUNX3-Tet2^Δ/Δ HSPCs at the pre-MDS stage (Fig. 4C), but shared fewer dysregulated genes with Tet2^Δ/Δ MDS and MDS/MPN cells (Supplementary Fig. S7), indicating an altered transcriptional program in HSPCs during the development of Tet2-deficient MDS due to the overexpression of RUNX3. We performed gene ontology (GO) enrichment analyses and found that RUNX3-Tet2^Δ/Δ MDS cells showed positive enrichments in genes involved in TGFβ, TNFα, and inflammatory response pathways, as expected from the result showing that human MDS cells express RUNX3 (Fig. 1E), but negative enrichments in genes in immune responses and leukocyte differentiation (Fig. 4D; Supplementary Data 2). RUNX3-Tet2^Δ/Δ MDS cells showed a positive enrichment in the expression of RUNX1-ETO target genes, but negative enrichments in other acute leukemia–related fusion genes, such as MLL-AF9 and BCR-ABL1 (Fig. 4D), indicating that RUNX3-Tet2^Δ/Δ cells suppress the transcriptional function of Runx1, which has been shown to be repressed by RUNX1-ETO (32, 33).

We attempted to clarify whether the overexpression of RUNX3 inhibits the expression and/or function of Runx1. While RUNX3-expressing HSPCs showed similar expression levels of Runx1 mRNA (Fig. 4E), RUNX3-Tet2^wt/wt, and RUNX3-Tet2^Δ/Δ c-Kit⁺ progenitor cells both had markedly decreased expression levels of the Runx1 protein (Fig. 4F). Because the stability of the RUNX1 protein is enhanced by binding to CBFβ (34), we transiently transfected RUNX1, RUNX3, and CBFβ vectors in human cells and performed an IP-Western blotting experiment to assess binding between RUNX1, RUNX3, and CBFβ. We found that RUNX3 inhibited the binding of RUNX1 to CBFβ (Supplementary Fig. S8), which supports the overexpression of the RUNX3 protein decreasing Runx1 protein levels due to their competitive binding to CBFβ. Consistent with the results obtained in the GO analysis, the gene set enrichment analysis (GSEA) revealed that RUNX3-expressing cells had significantly decreased or increased expression levels of canonical Runx1 target genes (Fig. 4G), which were upregulated or downregulated in Runx1-knockout cells defined by a microarray (GSE40155; Supplementary Data 3). RUNX3-Tet2^Δ/Δ MDS HSPCs showed significantly lower or higher expression levels of Runx1 target genes than Tet2^Δ/Δ MDS HSPCs (Fig. 4G). Among Runx1 target genes, we performed qRT-PCR using these HSPCs and confirmed that RUNX3-Tet2^Δ/Δ MDS cells showed the significantly weaker expression of Runx1 target tumor suppressor genes, such as Cebpa and Csf1r (Fig. 4H), the expression levels of which were slightly decreased in these HSPCs in RNA sequencing. Therefore, RUNX3 overexpression reduces the expression of the Runx1 protein in cells and dysregulates the transcription of Runx1 target genes.

On the basis of the dysregulated expression of Runx1 target genes, to clarify whether RUNX3 impeded the DNA binding of and transcriptional function of RUNX1, we first performed Runx1–ChIP-seq on WT, RUNX3-Tet2^wt/wt, and RUNX3-Tet2^Δ/Δ cells and FLAG-tagged RUNX3–ChIP-seq on RUNX3-Tet2^wt/wt and RUNX3-Tet2^Δ/Δ cells after expanding HSPCs in an in vitro culture. To assess the level of active histone modifications in Runx-binding sites, we also performed H3K27ac–ChIP-seq on HSPCs isolated from mice of all genotypes. As expected, RUNX3- and Runx1-binding peaks were enriched at the promoter to transcription start site and intron regions in these cells (Fig. 5A; Supplementary Fig. S9). Motif enrichment analyses of RUNX3– and Runx1–ChIP-seq revealed robust enrichments in consensus Runx-binding sequences, supporting the accuracy of these ChIP-seq data (Table 1 and Supplementary Data 4). While RUNX3-binding peaks mostly overlapped between RUNX3-Tet2^wt/wt and RUNX3-Tet2^Δ/Δ cells, we found that Runx1-binding peaks showed a large variation between WT, RUNX3-Tet2^wt/wt, and RUNX3-Tet2^Δ/Δ cells (Fig. 5B). The Runx1-binding sites identified in WT cells overlapped with those in RUNX3-Tet2^Δ/Δ cells, which were partly shared by RUNX3-binding sites in RUNX3-Tet2^Δ/Δ cells (Fig. 5C), indicating that the overexpression of RUNX3 and loss of Tet2 remodeled the binding regions of Runx1. Among these Runx1 target genes, by visualizing ChIP-seq data, we found that RUNX3-Tet2^Δ/Δ cells showed RUNX3-binding peaks in the promoter regions of the Cebpa, Csf1r, and Hnrnpa0 genes (Fig. 5D; Supplementary Fig. S10).

While we did not detect any changes in H3K27ac expression levels in c-Kit⁺ progenitor cells among these different genotypes (Fig. 5E), Runx1 target genes showed lower levels of the H3K27ac modification in RUNX3-expressing cells than in control cells (Fig. 5F and G), supporting the inhibition by RUNX3 of the transcription of Runx1 target genes (Fig. 4H). To determine whether RUNX3 competed with RUNX1 to activate the promoter activity of their target genes, we performed a luciferase assay in which activity was regulated by the promoter/enhancer region of either Cebpa, Csf1r or Hnrnpa0 containing RUNX-binding motif sequences (Fig. 5H). In this experimental setting, the transfection of RUNX3 expression suppressed luciferase activity driven by RUNX1 expression in a dose-dependent manner (Fig. 5I). Therefore, RUNX3 overexpression remodels the binding regions of Runx1 with the loss of Tet2, but also impedes the transcriptional function of Runx1 accompanied by the reduction of H3K27ac modification in the region-dependent manner.

As we further analyzed gene expression profiles in HSPCs in an unsupervised manner, we found that the overexpression of RUNX3 and/or loss of Tet2 resulted in positive enrichments in the expression of the MYC hallmark and TGFβ signaling pathways (Fig. 6A). Given the importance of these biological pathways in the pathogenesis of MDS, we found that RUNX3-Tet2^Δ/Δ MDS cells showed the stronger expression of genes in the MYC hallmark and TGFβ signaling pathways than RUNX3-Tet2^Δ/Δ pre-MDS cells (Fig. 6B). RUNX3-Tet2^Δ/Δ MDS HSPCs did not show a significant change in the expression of Myc hallmark genes or Myc-RUNX3-common target genes (as defined below) from Tet2^Δ/Δ MDS HSPCs (Fig. 6A), implying that the loss of Tet2 exerted a dominant effect on the transcription of Myc target genes. These genotype cells showed similar mRNA and protein expression levels of Myc (Fig. 6C and D), in which blood-specific enhancers harbored similar levels of the H3K27ac modification (Supplementary Fig. S11), implying that the overexpression of RUNX3 and loss of Tet2 promote the transcriptional function of Myc rather than increasing the expression of Myc. A motif enrichment analysis of RUNX3–ChIP-seq revealed significant enrichments in E-box and Myc/Max-binding motif sequences in RUNX3-Tet2^Δ/Δ cells (Table 1 and Supplementary Data 4); therefore, we found that RUNX3-binding sites were shared by Myc-binding sites, which were identified by Myc–ChIP-seq in RUNX3-Tet2^Δ/Δ cells (Fig. 6E). Mutually, Myc–ChIP-seq revealed a significant enrichment in the Runx-binding motif sequence in RUNX3-Tet2^Δ/Δ cells (Table 1). In fact, Myc-RUNX3 common target genes showed a positive enrichment in RUNX3-Tet2^Δ/Δ HSPCs rather than in either of the single mutant HSPCs (Fig. 6A).

To determine whether Myc was critical for the cell growth capacity of RUNX3-expressing and/or Tet2-deficient HSPCs, we treated these HSPCs with the Myc inhibitor, 10058-F4, which blocks its heterodimerization with Max under in vitro culture conditions. Although Tet2^Δ/Δ cells showed the increased expression of Myc hallmark genes (Fig. 6A), they showed a similar cell growth capacity after the treatment with 10058-F4 (Fig. 6F), suggesting that the loss of Tet2 increased the expression of Myc target genes in a Myc-Max–independent manner. The number of RUNX3-expressing cells was significantly decreased by the treatment with 10058-F4 (Fig. 6F), indicating the transcriptional function of Myc was critical for the cell growth property of RUNX3-Tet2^Δ/Δ cells. Although further studies are needed to assess the in vivo effects of Myc inhibition, the overexpression of RUNX3 promoted the transcriptional function of Myc and sensitivity to the inhibition of Myc-Max heterodimerization, at least under this in vitro condition.

The expression levels of RUNX transcription factors are fine tuned by signaling pathways, transcription factors, and epigenetic modifiers at the promoter and enhancer regions of RUNX in a cell type–specific manner in development and adult tissue homeostasis, while the dysregulation of these processes initiates tumorigenesis (1). RUNX3 is known to function as a tumor suppressor in various tumors including MPN (7), and we demonstrated the enhanced expression and oncogenic function of RUNX3 in the pathogenesis of MDS in patients and mice. Although, it currently remains unclear how RUNX3 is activated in MDS HSPCs in patients, in this study, the ectopic expression of RUNX3 clearly facilitated the initiation of Tet2-deficient MDS in mice following alterations in the transcriptional program involved in Runx1-mediated hematopoiesis and cancer-related biological pathways (e.g., MYC and TGFβ; a model is shown in Fig. 6G), which were partly shared in RUNX3 high-expressing MDS HSPCs in patients, thereby supporting our mouse model recapitulating the pathogenesis of human MDS.

Runx1 is expressed in HSPCs and its loss impairs normal hematopoiesis, leading to the development of MPN-like disease in vivo, a similar phenotype to that observed in Runx3-deficient mice (5, 14), suggesting a functional redundancy between RUNX1 and RUNX3 for regulating the expression of their common target genes. A Runx1 deficiency is compensated for Runx3 in terms of the proliferative capacity of HSPCs being maintained and the deletion of both genes leading to BM failure disease in mice (15). In contrast, direct transcriptional inhibition between RUNX genes has been implicated in pathogenesis because RUNX3 directly binds to the promoter of RUNX1 to repress the transcription of RUNX1 and promote the development of EBV-mediated B-cell malignancies, which markedly enhance the expression of RUNX3 due to its activated superenhancer (20, 35, 36). In this study, RUNX3-expressing MDS cells suppressed the transcription of canonical Runx1 target genes following the remodeling of Runx1-binding sites via RUNX3 accompanied by a reduced level of the H3K27ac modification. Among these common peaks, we found that RUNX3-Tet2^Δ/Δ MDS cells markedly suppressed the expression of Runx1 target genes, such as Cebpa and Csf1r, which regulate differentiation and play tumor-suppressive roles in myeloid transformation including MDS (37–39), indicating that the region-specific competition of RUNX proteins and recruitment of repressor complexes drive the suppression of target genes. Although the functional roles of these repressed genes in RUNX3-Tet2^Δ/Δ MDS cells currently remain unclear, RUNX3-Tet2^Δ/Δ HSPCs impeded the differentiation and production of mature myeloid cells and platelets and MDS rather than MDS/MPN subsequently developed in mice. These results indicate that the overexpression of RUNX3 drives MDS clones and/or suppresses MPN clones in Tet2-deficient stem cells; however, further studies are needed to elucidate the mechanisms underlying competition between MDS clones and MPN clones in vivo.

Another interesting result was reduced Runx1 protein levels in spite of similar mRNA levels in RUNX3 overexpressing cells. The cofactor CBFβ enhances Runx1 stability by preventing its ubiquitin-mediated degradation (34). The stability of the Runx1 protein is tightly regulated by posttranslational modifications in Runx1 through various modification enzymes (40) because WT MLL increases, whereas the MLL-fusion oncoprotein reduces Runx1 protein levels to promote the development of leukemia, which was inhibited by the ectopic expression of Runx1 (41). The present results showed that the RUNX3 protein competed with the RUNX1 protein for binding to CBFβ, whereas the RUNX1 protein did not inhibit RUNX3 protein binding to CBFβ in an ex vivo setting, supporting the oncogenic property of RUNX3, the overexpression of which reduces RUNX1 protein levels to promote transformation in vivo.

MYC is critical for the development of various tumors through its regulation of cancer-related biological pathways (e.g., cell cycle, apoptosis, and translation). MYC and RUNX family genes have been shown to collaborate to initiate and propagate tumorigenesis, for example, MYC and RUNX2 collaborate to promote the development of blastic plasmacytoid dendritic cell neoplasms as well as T-cell lymphoma, at least in part, due to the suppression of MYC-dependent apoptosis (42, 43). RUNX3 has also been shown to function as an oncogene in natural killer/T-cell lymphoma and is transcriptionally activated by MYC (44). In this study, we demonstrated that RUNX3-Tet2^Δ/Δ cells enhanced the expression of Myc target genes to activate cancer-related biological pathways, which were partly bound by RUNX3, as identified in ChIP-seq and motif enrichment analyses. Although the mechanisms by which the loss of Tet2 or overexpression of RUNX3 activates the transcriptional function of Myc currently remain unclear, RUNX3 directly binds to Myc target genes and facilitates the transcriptional function of Myc, resulting in the development of MDS. Indeed, we found that RUNX3-expressing cells showed greater vulnerability to the functional inhibition of Myc by 10058-F4 than control cells, at least, in an in vitro setting. We also applied this result to human cells and demonstrated that human AML cells strongly expressing RUNX3 with activated histone modifications at the enhancer showed higher sensitivity to 10058-F4 and JQ1, a BRD4 inhibitor that inhibits the function of superenhancers (45, 46), following the moderate reduction of RUNX3 and MYC expression. Further studies are needed to examine the combinatorial inhibition of MYC and activated enhancers of target genes, including the region of the RUNX3 gene in MDS cells in vivo.

In this study, we generated a murine model for Tet2-deficient MDS expressing RUNX3 and demonstrated that the overexpression of RUNX3 exerts oncogenic effects on the cellular function of and the transcriptional program in MDS stem cells. Several MDS patient-derived xenograft murine models have recently been reported using humanized cytokine knock-in mice and supplementing patient-derived mesenchymal stroma cells with human MDS stem cells (47–49). These xenograft models for MDS remain challenging, but may provide a more detailed understanding of the pathogenesis of MDS. An integrated study on genetically engineered mice and these xenograft models is warranted in the future for the development of novel therapeutics that target MDS stem cells.

No potential conflicts of interest were disclosed.

Conception and design: T. Yokomizo-Nakano, G. Sashida

Development of methodology: T. Yokomizo-Nakano, G. Sashida

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T. Yokomizo-Nakano, S. Kubota, J. Bai, A. Hamashima, M. Morii, Y. Sun, S. Katagiri, M. Iimori, D. Tanaka, K. Ohyashiki, A. Iwama, G. Sashida

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T. Yokomizo-Nakano, S. Kubota, A. Kanai, M. Oshima, G. Sashida

Writing, review, and/or revision of the manuscript: T. Yokomizo-Nakano, G. Sashida

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y. Harada, H. Harada, M. Osato, G. Sashida

Study supervision: M. Osato, G. Sashida

The authors thank the members of the Sashida Laboratory for their discussions during the preparation of this manuscript and Mr. Shinji Kudoh and Dr. Takaaki Ito for their technical help. This work was supported in part by a grant from the Uehara Memorial Foundation (to G. Sashida), the Princess Takamatsu Cancer Research Fund (to G. Sashida), the Japanese Society of Hematology (to G. Sashida), the Naito Foundation (to T. Yokomizo-Nakano), the Takeda Science Foundation (to T. Yokomizo-Nakano), and Grants-in-Aid for Scientific Research (16KT0113, 16K19579, 18H02842, 19K22640, and 19K08842) from the Ministry of Education, Culture, Sports, Science, and Technology (MEXT) of Japan.

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