Runt-Related Transcription Factor 3 Promotes Acute Myeloid Leukemia Progression

Acute myeloid leukemia (AML) is an aggressive hematological malignancy with high relapse/refractory rate. Genetic and epigenetic abnormalities are driving factors for leukemogenesis. RUNX1 and RUNX2 from the Runt-related transcription factor (RUNX) family played important roles in AML pathogenesis. However, the relationship between RUNX3 and AML remains unclear. Here, we found that RUNX3 was a super-enhancer-associated gene and highly expressed in AML cells. The Cancer Genome Atlas (TCGA) database showed high expression of RUNX3 correlated with poor prognosis of AML patients. We observed that Runx3 knockdown significantly inhibited leukemia progression by inducing DNA damage to enhance apoptosis in murine AML cells. By chromatin immunoprecipitation sequencing (ChIP-seq) analysis, we discovered that RUNX3 in AML cells mainly bound more genes involved in DNA-damage repair and antiapoptosis pathways compared to that in normal bone marrow cells. Runx3 knockdown obviously inhibited the expression of these genes in AML cells. Overall, we identified RUNX3 as an oncogene overexpressed in AML cells, and Runx3 knockdown suppressed AML progression by inducing DNA damage and apoptosis.


INTRODUCTION
Acute myeloid leukemia (AML) is one of the most common hematologic malignancies that is characterized by clonal expansion of abnormally differentiated myeloid blasts (1,2). High treatment failure rate of AML that is caused by frequent relapses and limited treatments challenges the clinical management of AML (3). Genetic and epigenetic abnormalities, such as NPM1 mutation, DNMT3a mutation, and MLL rearrangement, are determinants of AML pathogenesis and always relate to AML prognosis (4). Consequently, it is imperative to further decipher the genetic and epigenetic characteristics of AML to identify more new molecular targets for AML treatment improvement.
Super-enhancer is a special enhancer identified to enhance the transcription of key oncogenes in various cancer cells, such as prostate cancer cells, T cell acute lymphocytic leukemia cells, and multiple myeloma cells (5)(6)(7). In mouse AML cells, some genes critical in leukemogenesis, including Myc, Meis1, and Runx2, are also super-enhancer-associated genes (8), which indicates that super-enhancers may dedicate to AML pathogenesis.
The Runt-related transcription factor (RUNX) family consists of three members, RUNX1, RUNX2, and RUNX3. The tumorrelated functions of RUNX1 and RUNX2 were well studied. Especially, RUNX1 plays a critical role in AML pathogenesis (9). Meanwhile, RUNX3 was the least investigated (10). As a transcription factor, the heterodimer of RUNX3 and a beta subunit form a complex that binds to the core DNA sequence found in a number of enhancers and promoters and further activate or suppress transcription (11). RUNX3 was previously regarded as a tumor-suppressor gene for its inactivation promotes the progression of gastric cancer, lung cancer, colorectal cancer, and bladder cancer by upregulating oncogenes, such as YBX1 and GLI1, or abrogating ARF-P53 pathway (12)(13)(14)(15)(16). Conversely, RUNX3 is overexpressed and exhibits oncogenic activities in ovarian cancers, basal cell carcinomas, and head and neck cancers (17)(18)(19). As well, RUNX3 overexpression drives the transformation of myelodysplastic syndrome, another myeloid malignancy, by repressing RUNX1 (20) and predicts poor prognosis in childhood AML (21). However, the relationship between RUNX3 and AML pathogenesis remains mysterious.
In this study, we discovered that RUNX3 is a super-enhancerassociated gene and highly expressed in AML cells. Runx3 knockdown in murine AML cells efficiently impeded AML progress. Furthermore, we proved that RUNX3 bound and upregulated the expression of genes involved in DNA repair and antiapoptosis pathways to promote AML progression.

RUNX3 Is a Super-Enhancer-Associated Gene Only Highly Expressed in Acute Myeloid Leukemia Cells Instead of in Normal Blood Cells
To identify super-enhancer-associated genes that are unique in AML cells, we analyzed H3K27ac chromatin immunoprecipitation sequencing (ChIP-Seq) data of three types of normal blood cells, including neutrophils (NEs), monocytes (MOs), and hematopoietic stem cell progenitor cells (HSPCs), and AML cells (AML1#-3#). We found 3,436 super-enhancer-associated genes in normal blood cells. Meanwhile, 528 super-enhancer-associated genes were consistently identified in AML cells ( Figure 1A). Furthermore, two independent RNA sequencing (RNA-seq) datasets (GSE128910 and GSE138702) were analyzed and revealed that 485 genes were identified as being differentially expressed (adjusted p < 0.05) with ≥1.5-fold differential expression between the groups were consistently overexpression. Accordingly, we found three abnormally highly expressed genes (RUNX3, TMEM50B, and TGOLN2) those were superenhancer-associated genes only in AML cells ( Figure 1A and Supplementary Figure S1). Furthermore, we analyzed The Cancer Genome Atlas (TCGA) database and found that only RUNX3 expression was positively associated with poor prognosis in AML (RUNX3, p = 0.02; TMEM50B, p = 0.14; TGOLN2, p = 0.81; Figure 1B). Moreover, we observed that the expression level of RUNX3 was remarkably higher in bone marrow cells from AML patients than that from healthy volunteers (15.39-fold increase; Figure 1C and Table 1). Consistently, both mRNA and protein levels of RUNX3 in MLL-AF9-induced murine AML cells were significantly elevated compared to those in normal murine bone marrow cells (7.72-fold increase of Runx3 mRNA expression; Figures 1D-F).
Taken together, we demonstrate that RUNX3 is a superenhancer-associated gene only highly expressed in AML cells instead of in normal blood cells and probably exerts pro-tumor function on AML cells.

Runx3 Knockdown Inhibits Acute Myeloid Leukemia Progression In Vivo
To further explore the potential pro-tumor role of Runx3 in AML, equal numbers of control (Vector) or Runx3 knockdown (Runx3 KD) murine AML cells were transplanted into syngeneic wild-type (WT) recipients ( Figure 2A). To evaluate the effects of Runx3 reduction, we first sorted green fluorescent protein (GFP) + leukemia cells from Vector and Runx3 KD AML mice and confirmed that shRNA specific for Runx3 led to decreased RUNX3 expression by qRT-PCR and Western blot analysis (80.3% reduction of Runx3 mRNA expression; Figures 2B, C and Supplementary Figure S2). We found that Runx3 KD significantly ablated AML cells in the peripheral blood (PB) [79.5% reduction on day 28 (d28), 46.9% reduction on d45; Figures 2D, E] and reduced disease burden in the bone marrow (37.4% reduction on leukemic cell frequency, 47.2% reduction on leukemic cell number; Figure 2F). Furthermore, the spleen and liver of AML mice were significantly enlarged, and Runx3 knockdown significantly alleviated these symptoms (44.9% and 35.7% reduction of spleen weight and liver weight, respectively; Figure 2G). Consistently, histological analysis showed that AML mice in Runx3 KD group had fewer leukemia cell infiltration in the peripheral blood, spleen, and liver ( Figure 2H). Importantly, Runx3 knockdown significantly prolonged the survival of AML mice [Median overall survival (MOS) 81 days in Runx3 knockdown group compared to 56 days in vector control group; Figure 2I].
These results demonstrate that Runx3 knockdown suppresses the progression of MLL-AF9-induced AML in mice, which supports our hypothesis that Runx3 is oncogenic in AML.

Runx3 Knockdown Induces DNA Damage and Apoptosis in Acute Myeloid Leukemia Cells In Vivo
We next investigated how Runx3 knockdown suppressed the development of AML. Through flow cytometric analysis of control (Vector) or Runx3 KD AML cells, we found a slight increase in the percentage of G2/S/M-phase cells and G0-phase cells, accompanied by a minor decrease in the percentage of G1phase cells after Runx3 knockdown (1.14-fold and 1.34-fold increase in G2/S/M-phase cells and G0-phase cells, 12.8% reduction in G1-phase cells, respectively; Figure 3A and Supplementary Figure S3). Reduction of Runx3 in leukemic cells therefore appeared to induce G0/G1 arrest, which was consistent with a reduction of leukemia burden. Furthermore, Runx3 knockdown obviously increased DNA damage in AML cells (1.81-fold increase in g-H2AX + cells, 3.93-fold increase in g-H2AX foci per cell; Figures 3B, C). Annexin V and 7-AAD double staining showed that upon Runx3 knockdown, the percentage of total apoptotic cells in AML cells was remarkably increased to 13.4% compared with 8.6% in control group (1.56fold increase) ( Figure 3D).
Altogether, these results illustrate that Runx3 knockdown induces DNA damage and apoptosis in leukemia cells in vivo.

RUNX3 Binds to Cell Cycle-Related Genes in Both Normal Bone Marrow Cells and Acute Myeloid Leukemia Cells but Specifically to DNA Repair and Antiapoptosis-Related Genes Only in Acute Myeloid Leukemia Cells
To determine the molecular mechanism of the oncogenic activity of RUNX3 in AML cells, we analyzed the genomic distribution of RUNX3 in bone marrow cells from normal mice and AML mice by RUNX3 chromatin immunoprecipitation followed by nextgeneration sequencing (ChIP-seq). The bone marrow cells from primary AML mice were collected 35 days after transplantation at which point the percentage of AML cells (GFP + cells) was 97.5% ( Figure 4A and Supplementary Figures S4A, B). The number of RUNX3 peaks and corresponding genes in the leukemia group was significantly higher than those in the normal cells ( Figure 4B). RUNX3 peaks were enriched at introns, promoters, intergenic sites, and exons of genes in normal cells and leukemia cells ( Figure 4C). By analyzing RUNX3bound genes in normal bone marrow cells and AML cells, we found that 4,667 genes were able to be bound by RUNX3 in both normal bone marrow cells and AML cells. There were 5,845 RUNX3-bound genes that could be found in AML cells but not in normal bone marrow cells ( Figure 4D).
To further explore the difference between RUNX3-bound genes in normal bone marrow cells and AML cells, Gene Ontology (GO) enrichment analysis was performed. We found that most of the 4,667 RUNX3-bound genes in both types of cells were cell cycle-related genes, but the RUNX3 peak values were higher in AML cells (Supplementary Figures S4C, D). More importantly, most of the 5,845 genes specifically bound by RUNX3 in AML cells were related to DNA repair and the negative regulation of apoptosis, such as Chek1, Ddb1, Rad51c, Rpa2, Bcl-2, and Mcl-1 ( Figures 4E, F). Surprisingly, many classical AML-related oncogenes that have been reported were found to be bound by RUNX3 in AML cells, such as Myc, Cd93  Figure S4E). Furthermore, we discovered that RUNX3 tended to bind with these classical genes related to DNA repair, antiapoptosis, and leukemogenesis around their promoter areas ( Figure 4F and Supplementary Figure S4E). Interestingly, RUNX3 bound some classical DNA-repair genes at their enhancer areas, while it bound no antiapoptotic genes at their enhancer areas (Supplementary Figure S4F).
These results indicate that RUNX3 probably directly regulates genes related to cell cycle, DNA repair, and apoptosis in AML cells.

DISCUSSION
Genetic and epigenetic abnormalities drive leukemogenesis and determine the prognosis of AML (4). They are complex and dynamically evolving (26). There is still much work to do to uncover the full genetic and epigenetic landscape of AML. In our study, we found that RUNX3 was an obvious highly expressed gene in AML cells. According to TCGA-LAML database, the high expression of RUNX3 was positively related to poor prognosis of AML patients. These discoveries suggest a potential role of RUNX3 in AML progression.
As reported, diverse solid tumors, including gastric, colorectal, lung, and bladder cancers, exhibit low expression of RUNX3 (27). Low expression of RUNX3 is caused by gene deletion and epigenetic alteration. Epigenetic alteration is the most common one (28). The RUNX3 gene is regulated by two promoters, P1 and P2. At the P2 promoter, there is a large CpG island that is often hypermethylated in tumor cells to silence RUNX3 (29). However, the P2 promoter in AML cells is unmethylated, and demethylating agents fail to increase RUNX3 expression level in AML cells (21). Moreover, we discovered that RUNX3 in AML cells was regulated by a superenhancer that had strengthened transcriptional regulating ability. These evidences explain why differs from that in other solid tumor cells, RUNX3 expression in AML cells is elevated.
The function of RUNX3 in cancers is controversial. Primarily, RUNX3 was reported as a tumor-suppressor gene in multiple cancers (12)(13)(14). Then, in ovarian cancers and head and neck cancers, RUNX3 showed oncogenic activity (17,19). Consistent with the human information from TCGA database, our results showed that Runx3 knockdown in MLL-AF9-induced AML cells retarded AML progression. This indicates that RUNX3 probably plays a pro-tumor role in AML. RUNX3 regulates transcription by binding enhancers and promoters (11). So, we performed ChIP-seq in both murine normal bone marrow cells and AML cells to investigate the mechanism of the oncogenic activity of RUNX3 in AML. We further discovered that compared to that in normal bone marrow cells, RUNX3 in AML cells tended to bind genes enriched in DNA repair (Chek1, Ddb1, Rad51c, and Rpa2), antiapoptosis (Bcl2, Bcl2l10, Bcl2l12, and Mcl1), and leukemogenesis (Myc, Cd93, Kit, Ikzf2, Fto, and Sox4) pathways. Myc is a classical oncogene in various cancers. The activation of Myc by RUNX3 had already been demonstrated to be the main cause of the oncogenic function of RUNX3 (30,31). Also, Ddb1, Chek1, and Rad51c are essential genes involved in DNA-damage repair of AML cells (32,33). Bcl2 and Mcl1 are critical antiapoptosis genes and successfully used as treatment targets for AML (34). We demonstrated that in murine AML cells, RUNX3 mainly bound these classical genes at their promoter sites. With combined analysis of public datasets of H3K27ac ChIP-seq, we discovered that RUNX3 bound some key DNA-repair factors at their enhancer areas. However, we failed to find RUNX3 that bound any antiapoptotic factors at their enhancer areas. Our results further proved that Runx3 knockdown significantly downregulated the expression of these DNA-repair and antiapoptotic genes in murine AML cells. This suggests that RUNX3 directly upregulates the expression of DNA-repair genes by controlling both their promoters and enhancers while upregulating the expression of antiapoptotic genes only by controlling their promoters. Decreased expression of DNA-repair genes resulted in increased DNA damage, which ultimately collaborated with the influence of reduced antiapoptotic factors to induce more apoptosis of AML cells in vivo. Also, our results showed that RUNX3 knockdown slightly disturbed the normal cell cycle of AML cells in vivo. Altogether, we elucidate that RUNX3 promotes AML progression not only by activating Myc transcription but also by directly regulating oncogene network covering DNA repair and apoptosis. Further studies are warranted to determine the detailed mechanism of how RUNX3 regulates the oncogene network. Collectively, our study identified RUNX3 as an oncogene in AML, which conferred a new treatment target for AML therapy.

Animals
The C57BL/6 mice and CD45.1 mice (6-8 weeks old, weighing 18-22 g) were all raised in the specific pathogen-free (SPF)-level animal breeding facility of the Experimental Animal Center of

Definition of Enhancers and Super-Enhancers
We downloaded H3K27ac ChIP-seq data from a public database (NE SRR1915572, MO SRR787551, HSPC SRR2094192, AML1# SRR3503794, AML2# SRR3503797, AML2# SRR3503801, mouse GSE117443). Enhancers were stitched, and super-enhancers were identified using ROSE (https://bitbucket.org/young_ computation/rose). Briefly, constituent enhancers were stitched together if they are within a certain distance and ranked by their input-subtracted signal of H3K27ac. And then, we separated super-enhancers from typical enhancers by identifying an inflection point of H3K27ac signal; the slope here was 1. We run ROSE with a stitching distance of 12,500 bp and allowed enhancers within 12,500 bp to be stitched together. In addition, we used a transcription start site (TSS) exclusion zone of 5,000 bp. Finally, Rose GeneMapper tool was used to annotate the genes within the 50-kb range of the super-enhancers.

Survival Analysis of the Genes in The Cancer Genome Atlas Dataset
LAML data from TCGA were used to perform validation with the Gene Expression Profiling Interactive Analysis (GEPIA) database (http://gepia.cancer-pku.cn) (35). Furthermore, Kaplan-Meier curves were generated from the GEPIA database. The overall survival (OS) was estimated using the log-rank test, and p-value <0.05 was considered to denote statistically significant data.

Patient Specimens
The AML patients' specimens used in this study were derived from the routine clinical management in the Third Affiliated Hospital, Sun Yat-sen University. The procedure was approved by the ethics committee of the Third Affiliated Hospital, Sun Yatsen University in accordance with the international guidelines and the ethical standards outlined in the Declaration of Helsinki. Mononuclear cells were isolated from the patient bone marrow with Ficoll-Hypaque and then processed to extract mRNA.

Quantitative RT-PCR
Total mRNA was extracted from sorted GFP + cells using MagZol ™ Reagent (R4801-03, Magen) according to the manufacturer's instructions. mRNA purity and quantity were determined with NanoDrop (Thermo Scientific) before qPCR analysis. For qRT-PCR, equal amounts of mRNA samples were reverse transcribed into cDNA using the TransScript All-in-One First-Strand cDNA Synthesis SuperMix for qPCR (One-Step gDNA Removal) Kit (AT341, Transgen). Quantitative realtime PCR was performed on Bio-Rad CFX96 Touch ™ Real-Time PCR Detection system using SYBR Green I Master Mix reagent (11203ES03, YEASEN).

Acute Myeloid Leukemia Mouse Model
The 293T cells were transfected with retroviral plasmids MSCV-MLL-AF9-IRES-GFP containing MLL-AF9 and GFP cDNA sequences. Bone marrow cells from C57 mice treated with 5fluorouracil (5-FU) for 5 days were infected with retrovirus twice with 24-h interval. The 400K infected cells were mixed with 100K protective cells to intravenously inject into WT recipient mice irradiated with a 9-Gy lethal dose. The number of animals used per experiment is shown in the figure legends.

g-H2AX Immunofluorescence Staining
The cells were transferred to a glass slide and allowed to stand for 1 h to make the cells adhere to the glass slide. After fixation with 4% paraformaldehyde (PFA) for 15 min, cells were permeabilized with 0.5% Triton X-100 at room temperature for 30 min, blocked with 10% goat serum solution at room temperature for 1 h, washed, and incubated with g-H2AX primary antibody (Biolegend, Cat 613404) overnight. After that, the secondary antibody was added dropwise and incubated at room temperature for 1 h, and the high-speed confocal imaging system (Dragonfly CR-DFLY-202 2540) was used for imaging. The g-H2AX foci in 30 cells were counted in each group.

Flow Cytometry
Take 20-30 ml of peripheral blood through the tail vein of the mouse and add to the anticoagulation tube. Take the bone marrow cells from the femur and tibia of the sacrificed mice.

Chromatin Immunoprecipitation
Bone marrow cells were harvested from MLL-AF9-induced AML mice 35 days after transplantation, and bone marrow cells from normal syngeneic mice with the same age served as controls, five mice for each group. Here, 1% formaldehyde in PBS was used to crosslink the cells for 10 min, followed by quenching with 125 mM glycine on ice. Cells were collected and flash frozen in liquid nitrogen, then stored at -80°C for use. Frozen crosslinked cells were thawed on ice and then resuspended in lysis buffer I (50 mM HEPES-KOH pH 7.5, 140 mM NaCl, 1 mM EDTA, 10% glycerol, 0.5% NP-40, 0.25% Triton X-100, protease inhibitors). After rotating for 10 min at 4°C, the cells were collected and resuspended in lysis buffer II (10 mM Tris-HCl pH 8.0, 200 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, protease inhibitors). After rotating for 10 min at 4°C, the cells were collected and resuspended in sonication buffer (20 mM

ChIP-Seq Data Analysis
We aligned the ChIP-Seq data to the mm9 reference genome by bowtie2 with default parameter, followed by removing the multiple aligned reads, PCR duplications with samtools. To eliminate the impact of "Problematic genomic regions", we downloaded the ENCODE blacklist (Consortium, 2012) and discarded the reads aligning this region through bedtools. Finally, we used macs2 to calling peaks with control, setting a q value cutoff of 0.05.

Gene Ontology Analysis
To find the GO terms enriched in RUNX3-bound genes, The clusterProfiler (36) package in R was utilized for the identification and visualization of enriched pathways among differentially expressed genes identified as described above. The functions "enrichGO" were used to identify overrepresented pathways based on the GO databases. Significance in the enrichment analysis was based on p.adjust <0.05. For Figures 4E, G, we reported 10 significant GO Biological Process terms and their associated q values.

Statistics
Data are expressed as means ± SEM. For all experiments, except the determination of survival, data were analyzed by Student's ttests, and differences were considered statistically significant if p < 0.05. The survival of the two groups was analyzed using a logrank test, and differences were considered statistically significant if p < 0.05. * p < 0.05, ** p < 0.01, *** p < 0.001.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: NCBI BioProject PRJNA741044.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by the ethics committee of the Third Affiliated Hospital, Sun Yat-sen University (SYSU). The patients/ participants provided their written informed consent to participate in this study.

AUTHOR CONTRIBUTIONS
WZ, QM, BL and ZS designed and performed most of the experiments and analyzed the data. BL and LL contributed to animal experiments and patient sample assay.WZ and MZ wrote the paper. LL, DL, and MZ supervised the project. All authors contributed to the article and approved the submitted version.

FUNDING
We would like to thank the National Natural Science Foundation of China (NSFC 81870127), National Natural Science Foundation of China (NSFC 81700149), Sanming Project of Medicine in Shenzhen (SZSM201911004), and Research support foundation for post doctors in the seventh affiliated hospital, Sun Yat-sen University (ZSQYRSFPD0017) for generous support.