NPM1 mutation mediated HOXBLINC lncRNA activation promotes AML leukemogenesis

Nucleophosmin ( NPM1 ) is the most commonly mutated gene in acute myeloid leukemia (AML) resulting in aberrant cytoplasmic translocation of the encoded nucleolar protein (NPM1c + ). NPM1c + maintains a unique leukemic gene expression program, characterized by activation of HOXA / B clusters and MEIS1 oncogene to facilitate leukemogenesis. However, the mechanisms by which NPM1c + controls such gene expression patterns to promote leukemogenesis remain largely unknown. Here, we show that the activation of HOXBLINC, a HOXB locus-associated long non-coding RNA (lncRNA), is a critical downstream mediator of NPM1c + -associated leukemic transcription program and leukemogenesis. HOXBLINC loss attenuates NPM1c + driven leukemogenesis by rectifying the signature of NPM1c + leukemic transcription programs. Furthermore, overexpression of HoxBlinc ( HoxBlinc Tg) in mice enhances HSC self-renewal and expands myelopoiesis, leading to the development of AML-like disease, reminiscent of the phenotypes seen in the Npm1 mutant knock-in ( Npm1 c/+ ) mice. HoxBlinc Tg and Npm1 c/+ HSPCs share significantly overlapped transcriptome and chromatin structure. Mechanistically, HoxBlinc binds to the promoter regions of NPM1c + signature genes to control their activation in HoxBlinc Tg HSPCs, via MLL1 recruitment and promoter H3K4me3 modification. Our study reveals that HOXBLINC lncRNA activation plays an essential oncogenic role in NPM1c + leukemia . HOXBLINC and its partner MLL1 are potential therapeutic targets for NPM1c + AML. remain largely unknown. Our RNA-seq analyses on HOXBLINC ko vs. control NPM1c + OCI-AML3 cells and HoxBlinc Tg vs. NPM1 c/+ or WT LSK cells demonstrate that HOXBLINC regulates NPM1c + signature genes including anterior HOXB genes where HOXBLINC resides and genes located on other chromosomes such as HOXA9 - 10 , MEIS1 and RUNX1 . ChIRP-seq analyses reveal that HoxBlinc binds to the promoters of both resident anterior HoxB cluster genes and distant target genes such as posterior HoxA, Meis1 and Runx1 . These results indicate that the regulation of the NPM1c + signature genes by HoxBlinc is achieved through the direct HoxBlinc binding via cis and trans actions. Indeed, GSEA and GO analyses signify enrichment of genes for NPM1 -mutated signature, HOXA9 oncogenic pathway, Wnt and JAK-STAT signaling HOXBLINC vs. control OCI-AML3 cells NPM1 vs. WT LSK cells. Additional functional analyses demonstrate that normal hematopoiesis requires a tightly controlled HOXBLINC expression, and its misregulation caused by NPM1c + is an oncogenic event in leukemogenesis. Our findings on the direct oncogenic role of HOXBLINC in NPM1 -mutated AML could serve as a blueprint for implicating lncRNAs in AML leukemogenesis. assays, assays cytokine sequential replatings.


Introduction
Nucleophosmin (NPM1) mutations are the most frequently recurring genetic abnormalities in patients with acute myeloid leukemia (AML), occurring in approximately 50% of adult and 20% of childhood AML with normal karyotypes 1,2 . NPM1-mutated AML has been included as a distinct AML entity in the World Health Organization (WHO) classification 3 .
NPM1 encodes a protein that is normally located in the nucleolus and has multiple functions such as biogenesis of ribosomes and maintenance of genomic stability 4 . NPM1 mutations result in cytoplasmic mislocalization of the mutant protein (NPM1c + ), which is critical for its role in leukemogenesis 5,6 . As an AML-initiating lesion, NPM1c + maintains a distinctive transcriptional signature in AML cells, characterized by upregulation of HOXA and HOXB cluster genes and their oncogenic cofactor MEIS1 7,8 . However, the precise mechanisms by which NPM1c + drives the leukemic gene expression programs remain unclear.
Dysregulation of HOXA/B genes is a dominant mechanism of leukemic transformation and hematopoietic stem/progenitor cell (HSPC) deregulation 9 . A wide variety of molecular determinants including transcriptional factors, epigenetic regulators (e.g. polycomb and trithorax proteins), microRNAs, chromatin structure, and long non-coding RNAs (lncRNAs) are known to control HOX gene expression. However, their relationship with each other to fine-tune the HOXA/B gene expression pattern in AML remains to be elucidated. Recently two HOXA/B loci associated lncRNAs, HOTTIP and HOXBLINC, were shown to regulate transcription of HOXA/B genes through influencing epigenetic landscape [10][11][12] . HOXBLINC has been reported to play a critical role in hematopoietic specification during development through its cis-acting function to coordinate anterior HOXB gene expression via recruitment of the SETD1A/MLL1 histone H3K4 methyltransferase complexes 10 .
Located in the anterior HOXB locus and serving as a regulator of HOXB gene transcription, the role of HOXBLINC in HSPC biology and leukemogenesis remains unknown. In this study, we demonstrate that HOXBLINC is upregulated restrictively in NPM1c + AML. Npm1 mutant knock-in (Npm1 c/+ ) and HoxBlincTg HSPCs share significantly overlapped chromatin signatures and gene expression profiles in their upregulated genes as compared to WT HSPCs, including the NPM1c + signature HoxA/B cluster genes and homeobox oncogene Meis1.
Transgenic overexpression of HoxBlinc lncRNA in hematopoietic cells led to the development of an AML-like disease by triggering HSC self-renewal and expanding myelopoiesis, similar to the phenotypes displayed by Npm1 c/+ mice, while inhibition of HOXBLINC in NPM1c + AML cells mitigates leukemogenesis. Importantly, HoxBlinc overexpression in HSPCs increases its binding to NPM1c + signature genes and drives the leukemic specific transcription program in HSPCs by recruiting the MLL1 complex to reorganize local chromatin signatures. Together, our studies provide compelling evidence for the potent oncogenic role of HOXBLINC in NPM1c + -mediated leukemogenesis. HOXBLINC lncRNA and MLL1 could serve as potential therapeutic targets for the treatment of NPM1c + AML.

Results
HOXBLINC is specifically upregulated in NPM1c + AML. Up-regulation of HOX genes, especially HOXA and HOXB cluster genes are not only a characteristic but also a dominant mechanism for the pathogenesis of AML 9 . HoxBlinc lncRNA has been shown to be required to activate anterior HoxB gene transcription during development 10 . To determine whether HOXBLINC is aberrantly expressed along with the HOXB genes in AML, we performed RT-qPCR on bone marrow mononuclear cells (BMMNCs) from a cohort of AML patients (NPM1c + , n=25; and NPM1-wt, n=40; see patient information in Table S1) as compared to both BMMNCs (n=16) and CD34 + cells (n=11) from normal individuals. Interestingly, a dramatic upregulation of HOXBLINC was observed specifically in NPM1c + AML patients ( Figure 1A) as compared to NPM1-wt patients and normal CD34 + cells. When the RNA-seq data from TCGA-LAML datasets consisting of a cohort of 181 AML patients was analyzed for HOXBLINC expression, significantly higher HOXBLINC expression was observed in NPM1c + , but not MLL-rearranged (MLLr + ) AML patients as compared to NPM1c -MLLrpatients ( Figure 1B). The expression of HOXBLINC was positively correlated with the expression levels of NPM1c + signature genes including anterior HOXB genes, HOXA9 and MEIS1, but not HOXB13 in this AML cohort ( Figure S1A). Interestingly, AML patients with high HOXBLINC expression (the top thirty percentile of patients) had a significantly shortened survival as compared to patients with low HOXBLINC expression (the bottom thirty percentile, Figure S1B). Consistently, HOXBLINC was highly expressed in NPM1c + OCI-AML3 and IMS-M2 AML cells, but not NPM1-wt AML cells such as MLLr + MOLM-13, MV4-11, THP-1, NOMO-1 and OCI-AML2 cells, as well as BCR-ABL1 + K562 and JAK2V617F + SET-2 cells ( Figure S1C). These data collectively indicate that HOXBLINC is upregulated specifically in NPM1c + AML patients.

Loss of HOXBLINC perturbs NPM1c + -mediated transcription programs and
leukemogenesis. To confirm HOXBLINC activation is a downstream event of NPM1c + and determine the role of HOXBLINC in NPM1c + -mediated transcription regulation and leukemogenesis, we performed RNA-seq analysis on LSK cells isolated from the Npm1 mutant knock-in (Npm1 c/+ ) mice 13 . As compared to WT LSK cells, Npm1 c/+ LSK cells had 871 downregulated genes and 980 upregulated genes, including HoxBlinc and the NPM1c + signature genes HoxB2-5, HoxA7, [9][10][11]Meis1 and Runx1 ( Figure 1C, S1D), some of which are confirmed by qPCR ( Figure S1E). Gene Ontology (GO) and gene set enrichment (GSEA) analyses revealed that the upregulated genes in Npm1 c/+ vs. WT LSK cells are enriched with cell fate commitment, cell cycle, myeloid cell proliferation, stem cell maintenance, Wnt and Jak-STAT signaling pathways, pathways in cancer, as well as AML NPM1-mutated and HOXA9 oncogenic pathway ( Figure S1F,G).
We next examined the effect of HOXBLINC loss on NPM1c + -mediated transcription regulation and leukemogenesis using NPM1c + OCI-AML3 cells. We created CRISPR-dCas9-KRAB mediated HOXBLINC epigenetic silencing clones (HOXBLINCi) by targeting the KRAB repressive domain to the HOXBLINC promoter ( Figure S2A). We then compared genome-wide transcriptome changes between control and HOXBLINCi OCI-AML3 cells by performing RNAseq analysis. Consistently, NPM1c + OCI-AML3 cells exhibited high expression of HOXBLINC lncRNA and the common NPM1c + AML signature genes ( Figure 1D). Interestingly, inhibition of HOXBLINC in OCI-AML3 cells significantly impaired the transcription of many NPM1c + signature genes such as HOXB2-5, HOXA9-11, RUNX1 and MEIS1 ( Figure 1D). GSEA and GO analyses revealed that loss of HOXBLINC affects the pathways and genes involved in AML with NPM1-mutated, HOXA9 pathway, pathways in cancer, cell cycle, cell fate commitment, myeloid cell differentiation, and Wnt and JAK-STAT signaling pathways ( Figures 1E&F, S2B), similar to those observed in the upregulated genes of Npm1 c/+ vs. WT LSK cells ( Figure S1F,G).
When we transplanted a Dox-inducible HOXBLINC KD OCI-AML3 clone into NOD-scid IL2Rγ null (NSG) mice (2x10 5 cells/mouse) followed by Dox induction or vehicle treatment, the Dox-treated recipient mice had a significantly prolonged survival as compared to the recipients without Dox treatment ( Figure 1H). At 30 days after transplantation, Dox-treated mice had significantly lower chimerism of the hCD45 + cell population in the BM, spleen and peripheral blood (PB) of the recipients compared to the untreated animals ( Figure S2E). These results indicate that HOXBLINC KD suppresses OCI-AML3 leukemic cell proliferation both in vitro and in vivo, likely through the normalization of NPM1-mutation induced abnormal gene expression patterns ( Figure 1D). In addition, we silenced HOXBLINC expression in primary AML cells with or without NPM1c + mutation (#1315: NPM1c + ;FLT3wt, #921: or NPM1c + ;FLT3mu (#921) had significantly prolonged survival as compared to mice transplanted with control cells ( Figure 1I). FACS analysis revealed that HOXBLINCi dramatically decreased the hCD45 + cell chimerism in BM, spleen and PB of recipients ( Figure   S2F). In contrast, HOXBLINCi neither prolonged the survival nor decreased hCD45 + cell chimerism in mice transplanted with NPM1wt;MLLr + (#LPP4) AML cells ( Figures 1I, S2F).
Thus, HOXBLINC perturbation decreased tumor burden and attenuated leukemic progression in vivo most likely specific for NPM1c + AML patients.
Transgenic expression of HoxBlinc in hematopoiesis leads to lethal AML-like disease in mice. It has been shown that activation of a humanized NPM1c + knock-in allele in mouse HSCs (NPM1 c/+ ) causes Hox gene overexpression, enhanced self-renewal and expanded myelopoiesis, as well as development of delayed-onset AML 14 . Since NPM1c + activates HOXBLINC which is critical for NPM1c + -mediated transcription program and leukemogenesis, it is important to determine whether HOXBLINC activation is sufficient to cause abnormal hematopoieiss and myeloid malignancies similar to the NPM1 c/+ mice. We first examined the HoxBlinc expression pattern along the HSC differentiation hierarchy. HoxBlinc expression was high in long-term (LT) and short-term (ST) HSCs, decreased in progenitor cells (MPP, CMP and GMP) and was further decreased in the mature lineage cell populations except the B220 + B cells (Figure 2A). The expression pattern of HoxBlinc in hematopoiesis suggests that this lncRNA might play an important role in regulating HSPC function. To investigate the impact of HoxBlinc activation on normal hematopoiesis and leukemogenesis in vivo, we generated a HoxBlinc transgenic (Tg) mouse model in which full-length mouse HoxBlinc cDNA was inserted into mouse genome under the control of Vav1 promoter and enhancer (HS321/45-vav vector) to ensure the expression of transgene specifically in hematopoiesis ( Figure 2B). Two founder HoxBlincTg mice were obtained. The transgene was inserted into the intron of the Bin1 gene on chromosome 18q for Tg Line #1 ( Figure 2C). The expression levels of HoxBlinc RNA in BM cells were ~18and 3-folds greater than the endogenous HoxBlinc expression in Tg Line #1 and #2, respectively ( Figure 2D).
Monitoring of a cohort of HoxBlincTg (Line #1) mice showed that within 1 year of age, 67% of HoxBlincTg mice (10 of 15) died or were sacrificed because of a moribund condition, whereas none of the WT mice (n=12) died ( Figure 2E). Moribund HoxBlincTg mice exhibited weight loss, hepatosplenomegaly, enlarged lymph nodes as well as pale footpads and femurs as compared to WT ( Figures 2F, S3A). Peripheral blood examination revealed marked leukocytosis due to elevated immature myeloid cells and neutrophils, thrombocytopenia and severe anemia in these moribund HoxBlincTg mice ( Figure S3B). Morphologically, May-Grünwald-Giemsa stained PB smears showed significantly increased blasts ( Figure 2G We next examined the effect of HoxBlinc overexpression on the self-renewal and repopulation capacity of HSCs using in vitro replating and paired-daughter cell assays and in vivo competitive transplantation. A significantly higher replating potential was observed in each of the four successive rounds of replating in HoxBlincTg LSK cells than WT cells ( Figure 3D).
Both symmetric and asymmetric cell divisions are required for the preservation of normal HSC pool and continuous production of sufficient blood cells. Paired-daughter cell assays using WT and HoxBlincTg primitive CD34 -LSK cells showed a higher proportion of HoxBlincTg CD34 -LSK cells with symmetric self-renewal capacity, while the cells that underwent symmetric differentiation or asymmetric self-renewal were reduced as compared to WT ( Figure 3E).
Competitive transplantation assays showed that the donor cell (CD45.2 + ) chimerism in the PB of recipients transplanted with HoxBlincTg BM cells steadily increased, reaching ~80% 7 months after transplantation, while the CD45.2 + cell population in the PB of mice receiving WT BM cells remained ~50% ( Figure 3F). Furthermore, HoxBlincTg BM cells contributed to greater proportions of Gr-1 + /Mac1 + cells in the PB than WT cells in recipient mice (data not shown).
Strikingly, the BM CD45.2 + Lincells of HoxBlincTg recipients are comprised of significantly higher LK and LSK cells than that of CD45.1 + Lincells in HoxBlincTg recipients and Lincells in WT recipients ( Figure S5B). Consistently, a higher proportion of immature myeloid cells were observed in the BM of HoxBlincTg recipients ( Figure S5D Table S3). In addition, significant portions of the genes with promoter accessibility gain by either NPM1c + or HoxBlincTg were also upregulated ( Figure S6E). These genes included the NPM1c-signature genes HoxB2-5, HoxA9-10, Meis1, and Runx1, as well as other target genes such as Stat1 and Cdr2 (Figures S1D; 4F,G & S6F,G,) that also play critical roles in HSC regulation and leukemogenesis. As a control, no significant changes in chromatin accessibility were observed in the HoxD and Lypla1 loci, and the expression of HoxD or Lypla1 genes was not altered by HoxBlinc upregulation caused by either NPM1c + or HoxBlincTg ( Figure S6G & data not shown). These results suggest that overexpression of HoxBlinc lncRNA specifically activates NPM1c + signature genes via enhancing enhancer/promoter chromatin accessibility in HSPCs.
HoxBlinc directly binds to target genes and mediates chromatin interactions to drive gene regulatory networks in HSPCs. CTCF boundaries facilitate enhancer/promoter interactions within confined topologically associated domains (TADs). We recently reported that a CTCF boundary in the posterior HOXA locus establishes and maintains an active TAD to drive posterior HOXA gene expression 16 . To examine whether HoxBlinc overexpression affects CTCF defined anterior TAD domain and enhancer/promoter regulatory networks in the anterior HoxB locus, circular chromosome conformation capture using high throughput sequencing (4C-seq) was performed using the several HoxB locus CTCF binding sites (CBSs) as viewpoints in HoxBlincTg vs. WT Linc-Kit + cells ( Figure 5A). When the CBS located between HoxB4 and B5 (CBS4/5, which overlaps HoxBlinc gene) was used as a viewpoint, CBS4/5 interacted with the +43Kb CBS (+43CBS, Figure 5B). CBS4/5 also contacted each of the anterior HoxB genes ( Figure 5B), suggesting that either CBS4/5 or more likely HoxBlinc communicates with anterior HoxB gene promoters. When +43CBS was used as a viewpoint, the interaction of CBS4/5 and +43CBS was confirmed and +43CBS communicated with anterior HoxB genes too ( Figure 5B).
Interestingly, HoxBlinc overexpression intensified each of these long-range interactions within the anterior HoxB locus mediated by CBS4/5 and/or +43CBS ( Figure 5B). In contrast, +73Kb CBS (+73CBS) and HoxB13 CBS (CBS13) did not interact with the anterior HoxB genes, although +73CBS interacted with CBS5/6 and CBS8/9, which however was not affected by Hox genes, AML, HSC proliferation, Wnt signaling, and cell cycle ( Figure S7E). Integration of ChIRP-seq, RNA-seq, and ATAC-seq datasets from WT and HoxBlincTg HSPCs revealed that around 74% of the genes with increased HoxBlinc binding exhibited a ≥2 folds increase in gene expression levels ( Figure 5F) and 44.7% of them showed increased promoter chromatin accessibility ( Figure 5G). These data revealed that HoxBlinc acts as an epigenetic regulator to control target gene expression through remodeling promoter chromatin accessibility.
Transcription motif analysis showed that the top HoxBlinc bound motifs in HoxBlincTg HSPCs are transcription factors important for hematopoiesis, such as CTCF and PU.1 ( Figure S7F, Table   S4). These data demonstrated that HoxBlinc directly binds to hematopoietic specific target genes, mainly the NPM1c + signature genes and mediates long range chromatin interactions to drive gene regulatory networks in HSPCs.

Discussion
In a humanized NPM1c + knock-in mouse model, NPM1c + enhanced HSC self-renewal and expanded myelopoiesis leading to 1/3 rd of the animals developed late-onset AML 14 . Brunetti et al recently reported that specific reduction of NPM1c + lessens key features of the leukemic program 17 . Thus, NPM1 mutation is an AML-driving lesion and maintains leukemia through a gain-of-function by the NPM1c + . NPM1c + -mediated leukemogenesis has been shown to depend on the unique gene expression signatures such as HOXA/B and MEIS1 activation 18 . However, how this aberrant gene expression program is driven and maintained is largely unknown. In this study, we show that NPM1c + deregulates it signature genes, perturbs hematopoiesis and promotes leukemogenesis via the activation of a critical lncRNA, HOXBLINC. HOXBLINC overexpression is strongly association with NPM1 mutations in AML and NPM1c + expression leads to HoxBlinc activation in HSPCs. Although HoxBlinc is involved in normal hematopoietic development, 13 HoxBlinc overexpression plays an essential and sufficient oncogenic role in NPM1c + -mediated signature gene expression, HSPC deregulation and leukemogenesis.
Therefore, our studies identify HoxBlinc activation as a novel and critical downstream mediator for NPM1c + .
In addition to mutations and/or aberrant expression in protein-coding genes, misregulation of lncRNAs perturbs cellular physiology in multiple ways and plays important roles in the development and progression of various cancers 19 . However, how lncRNAs affect the initiation and progression of malignant myelopoiesis remains to be determined. Accordingly, lncRNAs have been profiled in various myeloid leukemias in order to identify potential oncogenic lncRNAs [20][21][22] . Recent studies have shown that LncHSC-2 and Hottip lncRNAs contribute to the control of critical signaling pathways in HSC regulation 12,23 . However, a direct link between lncRNAs and oncogenesis remains elusive in malignant myelopoiesis. Furthermore, the detailed molecular mechanisms underlying lncRNA dysregulation-mediated myeloid malignancy development remain largely unknown. Our RNA-seq analyses on HOXBLINC ko vs.
control NPM1c + OCI-AML3 cells and HoxBlincTg vs. NPM1 c/+ or WT LSK cells demonstrate that HOXBLINC regulates NPM1c + signature genes including anterior HOXB genes where HOXBLINC resides and genes located on other chromosomes such as HOXA9-10, MEIS1 and RUNX1. ChIRP-seq analyses reveal that HoxBlinc binds to the promoters of both resident anterior HoxB cluster genes and distant target genes such as posterior HoxA, Meis1 and Runx1.
These results indicate that the regulation of the NPM1c + signature genes by HoxBlinc is achieved through the direct HoxBlinc binding via cis and trans actions. Indeed, GSEA and GO analyses signify enrichment of genes for NPM1-mutated signature, HOXA9 oncogenic pathway, Wnt and JAK-STAT signaling pathways in HOXBLINC ko vs. control OCI-AML3 cells and HoxBlincTg or NPM1 c/+ vs. WT LSK cells. Additional functional analyses demonstrate that normal hematopoiesis requires a tightly controlled HOXBLINC expression, and its misregulation caused by NPM1c + is an oncogenic event in leukemogenesis. Our findings on the direct oncogenic role of HOXBLINC in NPM1-mutated AML could serve as a blueprint for implicating lncRNAs in AML leukemogenesis.
LncRNAs show a high versatility in their mechanism-of-action, influencing many cellular processes such as spatial conformation of chromosomes, chromatin modifications and RNA transcription 24 . HoxBlinc recruits MLL1 and SETD1a to anterior HoxB loci and controls their gene expression by regulating chromatin states during development 10  In summary, we show that HOXBLINC overexpression is a critical event to drive leukemogenesis by establishing aberrant NPM1c + signature gene expression program via controlling the MLL1 recruitment, chromatin domains and promoter accessibility in cis and trans actions. Our studies, therefore not only provide novel molecular insights into the biology of HSC and NPM1-mutated AML, but also create a unique opportunity for the identification of novel drug targets for NPM1c + AML.

Supplemental Information
Supplemental information includes 9 figures and 8 tables.

Competing Interest
The authors have declared that no competing interests exist.                  (Table S6), the positive band is 589bp. Transgenic founder mice were crossed with WT C57BL/6 mice. HoxBlinc negative siblings of the HoxBlincTg mice were used as controls throughout the study. Two HoxBlincTg lines were used for this study. The mHoxBlinc-set1 and mHoxBlinc-set2 were used as real-time PCR primers to recognize both endogenous and exogenous HoxBlinc lncRNA ( Figure 2D, Table S6). The levels of transgenic expression of HoxBlinc were also confirmed by RNA-seq analysis ( Figure 6A).

Morphological and histological analyses of the hematopoietic organs. PB was collected by tail vein
bleeding and was subjected to an automated blood count (Hemavet System 950FS). PB smears were subjected to May-Grünwald-Giemsa staining for morphological and lineage differential analysis.
Morphological evaluation of BM and spleen cells were performed on cytospins followed by May-Grünwald-Giemsa staining. For histopathological analyses, femurs were fixed in 10% Neutral Buffered Formalin (10% NBF) and demineralized in a solution of 10% EDTA for 1-2 weeks. The specimens and other soft tissues (spleens, lymph nodes and livers) were fixed in 10% Neutral Buffered Formalin and then dehydrated using ethanol and cleared in xylenes. The specimens were then embedded in melted paraffin and allowed to harden. Thin sections (5μm) were cut and floated onto microscope slides. For routine assessment, slides were stained with hematoxylin and eosin (H&E) staining. For MPO and hCD45 immunohistochemical staining, the tissue was rehydrated followed by heat-induced epitope retrieval, peroxidase and serum blocking. Samples were then incubated with MPO (R&D, #MAB3174) or hCD45 antibody (BD, #555485) overnight at 4°C followed by staining with the biotinylated second antibody.
Slides were visualized under a Nikon TE2000-S microscope. Images were taken by a QImaging camera and QCapture-Pro software (Fryer Company Inc.). Chemicals were obtained from Sigma (St. Louis, MO) unless otherwise indicated.
Flow cytometry analysis, cell sorting, and colony assay. Total white blood cells were obtained after lysis of red blood cells with red blood cell lysis buffer (QIAGEN 1045722). Single-cell suspensions from BM, spleen and PB were stained with panels of fluorochrome-conjugated antibodies (listed in Table S7).
The analyses were performed using BD LSRII or LSR Fortessa flow cytometer. All data were analyzed by FlowJo.V10 software. Purified LSK cells were used for the colony and replating assays. Briefly, BM cells from 6-8 weeks old mice were pre-enriched with lineage depletion beads (MiltenyiBiotec, Bergisch Gladbach, Germany) and then stained with c-Kit, lineage, and Sca-1 antibodies (listed in Table S7) and then sorted by BD FACSAriaⅡ. The purity of selected LSK cells were routinely over 98%. The purified LSK were incubated in RPMI1640 containing 30% FBS, 2% BSA, and a combination of cytokines (mG-CSF, 10ng/mL; mIL-3, 5ng/mL; mEPO, 4U/mL; hTPO, 100ng/mL; and mSCF, 100ng/mL). Colonies were scored at 8-10 days. For replating assays, CFU assays were performed with LSK cells in methylcellulose medium supplemented with the same cytokine cocktails. Colonies were passaged every 7 days for 4 sequential replatings. Antibodies used are listed in Table S7.  Table S7) and sorted by BD FACSAriaⅡ. The purity of selected LK cells was routinely over 98%. The LK cells were incubated in RPMI1640 containing 30% FBS, 2% BSA, and a combination of cytokines (mG-CSF, 10ng/mL; mIL-3, 5ng/mL; mEPO, 4U/mL; hTPO, 100ng/mL; and mSCF, 100ng/mL). At weekly intervals, cultures were mixed by pipetting and half of the culture media were removed, which was then replaced by the newly prepared medium with the same combinations of cytokines. Cells in the collected media were counted and used for flow cytometric analysis. Total CFUs generated at each time point in the suspension culture were evaluated by culturing a fraction of the expanded cells in the colony assay as described above.
Paired-daughter cell assay. To examine the frequency of HSCs to undergo self-renewal and differentiation, we performed paired-daughter cell assays 31 Table S6 and key reagents are listed in Table S7. RNA library was prepared with the IlluminaTruSeq strand-specific mRNA sample preparation system. Paired-end RNA-seq was performed by the UF ICBR core facility according to standard protocols. All of the sequencing reads were processed and aligned to the mouse genome assembly (mm9) or human genome assembly (hg19) using TopHat (version 2.0) and Expression level increased or decreased genes were marked with red or blue, respectively. Gene Ontology analysis with the Database for Annotation, Visualization and Integrated Discovery (DAVID) tool (https://david.ncifcrf.gov/, Version 6.8) 36 . Gene set enrichment analysis (GSEA) 37  Peak calling was performed using peak calling algorithm MACS2 with parameters ("-g mm -p 1e-9nolambda -f BAMPE -nomodel -shiftsize=100 --extsize 200") 44 . bedGraphToBigWig program was employed to generate the bigWig file of fragment or read coverages, including control and experimental datasets (https://www.encodeproject.org/software/bedgraphtobigwig/). All sequencing tracks were viewed using the Integrated Genomic Viewer (IGV/2.4.19) 46 . Peaks annotation was carried out with the command "annotatePeaks.pl" from HOMER package (version 4.8) 47 and GREAT 57 . DEseq2 (Benjamini-Hochberg adjusted p< 0.05; FoldChange≥2) were also performed to find the differential binding sites between two peak files, including control and treatment groups with C+G normalized and "reads in peaks" normalized data 58 . The de novo motif analysis was performed by the "findmotifsgenome.pl" from the HOMER package 47 . For each genomic feature (peaks or chromVAR annotation), we calculated the chromatin accessibility median deviation z-score (for chromVAR features) or fragment counts (for peaks) in control and treatment groups with chromVAR package in R language 59,60 . Pearson's correlation coefficient and Pearson's χ2-test were carried out to calculate overall similarity between the replicates of ATAC-seq global open chromatin signatures. All genomics datasets were deposited in the NCBI GEO under accession number (GSE115096).

Xenotransplantation of human OCI-AML3 or primary leukemic cells. Adult NOD.Cg-
Prkdc scid Il2rg tm1Wjl /SzJ (NSG) mice (6-8 weeks old) were pretreated with 280cGy total body irradiation, then 5x10 5 viable AML cells (in 300μl of PBS) were transplanted into each NSG mice by tail-vein injection. After transplantation, the recipients were administered with doxycycline (for Dox inducible HoxBlinc KD) in the drinking water (Sigma D-9891, 1 mg/ml, 1% sucrose, newly prepared every other day) until being sacrificed. Daily monitoring of the mice for symptoms (ruffled coat, hunched back, weakness and reduced motility) and survival time. Human CD45 chimerism in BM, spleen and PB cells were analyzed by flow cytometry as described above.
Quantification and statistical analysis. Differences between experimental groups were determined by the Student's t-test or analysis of variance (ANOVA) followed by Newman-Keuls multiple comparison tests as appropriate. P<0.05 is considered significant. For in vivo experiments, the sample size chosen was based on the generalized linear model with Bonferroni multiple comparison adjustments; with the proposed sample size of at least five mice/group/genotype. Animals were randomly assigned to each study. For all in vitro experiments, at least three independent experiments with more than three biological replicates for each condition/genotype were performed to ensure adequate statistical power. The Log-rank test was used to analyze differences between the survival curves.