The basic helix-loop-helix transcription factor SHARP1 is an oncogenic driver in MLL-AF6 acute myelogenous leukemia

Acute Myeloid Leukemia (AML) with MLL gene rearrangements demonstrate unique gene expression profiles driven by MLL-fusion proteins. Here, we identify the circadian clock transcription factor SHARP1 as a novel oncogenic target in MLL-AF6 AML, which has the worst prognosis among all subtypes of MLL-rearranged AMLs. SHARP1 is expressed solely in MLL-AF6 AML, and its expression is regulated directly by MLL-AF6/DOT1L. Suppression of SHARP1 induces robust apoptosis of human MLL-AF6 AML cells. Genetic deletion in mice delays the development of leukemia and attenuated leukemia-initiating potential, while sparing normal hematopoiesis. Mechanistically, SHARP1 binds to transcriptionally active chromatin across the genome and activates genes critical for cell survival as well as key oncogenic targets of MLL-AF6. Our findings demonstrate the unique oncogenic role for SHARP1 in MLL-AF6 AML.

T he MLL (mixed lineage leukemia) gene is located on chromosome 11q23 and encodes a large histone methyltransferase. MLL constitutes a large protein complex, binding to DNA and positively regulates the clustered homeobox (HOX) genes through histone 3 lysine 4 (H3K4) methyltransferase activity of the SET domain 1,2 and histone acetyltransferase activity of p300/CBP, MOZ, and MOF interacting with the PHD or TA domain [3][4][5] . The translocation of 11q23 is one of the most frequent chromosomal abnormalities in acute leukemia, and this rearrangement fuses the genomic region encoding the N-terminus of MLL to a sequence encoding the Cterminus of one of a number of fusion partner proteins, resulting in loss of chromatin modification potential. MLL-fusion protein (MLL-FP) acquires a unique transcriptional machinery recruiting the transcriptional elongation complex, EAP (elongation assisting protein), that includes p-TEFb (positive transcription elongation factor b), which phosphorylates RNA polymerase 2 and results in sustained transcriptional elongation 6 . The MLL-FP also interacts with DOT1L (disruptor of telomeric silencing 1-like), a specific H3K79 methyltransferase; di-and tri-methylated H3K79 (H3K79me2/3) are epigenetic hallmarks of active transcription by MLL-FPs 7 . Pharmacological inhibition or genetic deletion of DOT1L substantially suppresses MLL-rearranged (MLLr) AML 8,9 , indicating it as a therapeutic target.
More than 70 genes have been characterized as partner genes of MLL in acute leukemia 10 . Although the partner proteins have various functions and cellular localizations, most of the MLL-FPs share a principle machinery in their transcriptional regulation. AF4, AF9, AF10, and ENL are nuclear partner proteins that form a part of the transcriptional elongation complex, and these fusion partners account for more than 80% of all clinical cases of MLLr acute leukemias 10 . On the other hand, MLL-AF6 represents the most common leukemogenic fusion of MLL to a cytoplasmic partner protein. AF6 is not identified in the components of the major transcriptional elongation complex 7,11 . Nevertheless, MLL-AF6 also recruits EAP and DOT1L complexes to target chromatin via an unknown mechanism and activates transcriptional elongation of target genes 7,12 and the unique underlying mechanisms for MLL-AF6-driven leukemogenesis have not been fully elucidated. Here, we identify a basic helix-loop-helix transcription factor SHARP1 as a MLL-AF6 specific target gene and revealed its unique oncogenic role, representing a potential therapeutic target.
Basic helix-loop-helix (bHLH) transcription factor SHARP1 (also known as BHLHE41 or DEC2) was the highest and the most significantly upregulated MLL-AF6 target gene (average log 2 fold change 4.650, -log 10 p value 13.32) (Fig. 1b and Table 1). Although SHARP1 was identified as a common retroviral integration site in the genomes of AKXD murine myeloid tumors 19 , suggesting a potential role in leukemogenesis, there have not been further studies on its role in leukemogenesis. Importantly, SHARP1 was decreased in most cases of other subtypes of AML as well as normal bone marrow (NBM) CD34 + cells (Fig. 1c). Moreover, to test these findings, unsupervised hierarchical gene-expression clustering of leukemic blasts of adult AML patients from two independent cohorts was performed. Three cases, in a cohort of 285 AML cases that were studied using gene expression profiling, showed high SHARP1 expression levels (Fig. 1d). These three cases were in a cluster that was highly enriched for AMLs with a MLL-rearrangement (MLLr-AML) 20 and all three carried a t (6;11). Gene expression profiling of a second cohort of AMLs (n = 268) revealed two more cases with high SHARP1 expression, which also carried a t(6;11), and were clustered within a group of patients with MLLr-AML as well (Fig. 1e). In these two cohorts, all of the MLL-AF6 AML cases showed high SHARP1 expression. These findings prompted us to investigate whether SHARP1 plays an important role in the pathogenesis of MLL-AF6 AML.
MLL-AF6 directly upregulates SHARP1 by DOT1L. In human AML cell lines, consistent with our findings in the gene expression profiles from the multiple AML cohorts, SHARP1 mRNA was expressed highly in ML-2, CTS and SHI-1 cells, all of which harbor t(6;11)(q27;q23), whereas it was undetectable in MOLM-14, MV4-11 and Kasumi-1, which harbor t(9;11)(p22;q23), t(4;11)(q21;q23), and t(8;21)(q22;q22), respectively (Fig. 2a). MLL-FP complex contains MEN1 (Multiple Endocrine Neoplasia syndrome type 1, also called MENIN), which binds to the Nterminus of MLL, linking it to LEDGF (Lens Epithelium-Derived Growth Factor). The association of MEN1 or LEDGF with MLL is required for chromatin localization of the complex and transcription of their target genes, which are crucial for MLLrleukemias development 18,21 . A histone methyltransferase, DOT1L is a subunit of MLL-FP complexes and solely responsible for both H3K79 di-and tri-methylation (H3K79me2/3). ChIP-seq analysis of ML-2 cells demonstrated that posterior HOXA genes (HOXA7-10) were bound by MLL N /MEN1/LEDGF and enriched with H3K79me2/3, which is a hallmark of DOT1L recruitment to active chromatin, whereas the region of anterior HOXA genes (HOXA1-6) were neither bound by MLL N /MEN1/LEDGF nor enriched with H3K79me2/3 (Fig. 2b). In the SHARP1 gene locus, MLL N /MEN1/LEDGF localized across the transcribed region concomitantly with high enrichment of H3K79me2/3 (Fig. 2b). These findings were verified by ChIP-quantitative PCR (qPCR) of the promoter regions of the SHARP1 gene using antibodies against MLL N and H3K79me2 and ChIP-qPCR of HOXA9 promoter was used as a positive control ( Supplementary Fig. 2a). To confirm these findings in another MLL-AF6 AML cell line, we performed an independent ChIP-seq analysis of SHI-1 cells which expresses both MLL and MLL-AF6, demonstrating that MLL N binds to SHARP1 gene loci, as well as posterior HOXA genes locus (Fig. 2c). To ascertain the unique MLL-AF6 binding, we analyzed MLL N and H3K79me2 ChIP-seq data of THP-1 (MLL-AF9) and MV4-11 (MLL-AF4) cells and found that neither MLL N binding nor H3K79me2 enrichment was observed at SHARP1 loci ( Supplementary Fig. 2b). Collectively, our results indicate that SHARP1 is a unique transcriptional target of MLL-AF6 and its expression is not suppressed at the post-transcriptional level in the other MLLr-AML subtypes.  To examine whether MLL-AF6 regulates SHARP1 expression, we performed MLL-AF6 knockdown using two independent lentiviral shRNA targeting MLL N (shMLL #1 and #2) in ML-2 cells. Reduction in MLL-AF6 resulted in suppressed SHARP1 mRNA expression (Fig. 2d). Pharmacological inhibition of DOT1L results in robust and selective ablation of H3K79 methylation, leading to suppressed transcription of MLL-FP target genes, such as HOXA gene cluster and MEIS1 9 . To investigate whether SHARP1 expression relies on DOT1L activity, ML-2 and MOLM-14 cells were treated with the selective aminonucleoside inhibitor EPZ5676 22 . Consistent with the previous study, H3K79me2 was reduced in both cell lines (Fig. 2e). Importantly, EPZ5676 treatment dramatically reduced SHARP1 expression at both mRNA and protein levels in ML-2 cells, whereas no significant change was observed in MOLM-14 cells at mRNA level ( Fig. 2f, g). Collectively, these results demonstrate that MLL-AF6 and MEN1/LEDGF directly bind to the SHARP1 gene locus to positively regulate its expression through DOT1L activity.
To investigate whether Sharp1 deletion affects the initiation of other subtypes of MLLr-AML, the MLL-AF9 fusion gene 15 was retrovirally transduced into LSK cells from Sharp1 +/+ or Sharp1 −/− mice and subsequently 200,000 cells were transplanted into sublethally irradiated (650 rads) CD45.1 + congenic mice ( Supplementary Fig. 4c). Recipients from both groups succumbed to leukemia with a similar median survival in primary transplantations (median survival; 74 vs 70 days, p = 0.302). Secondary transplantation was performed by injecting 200,000 leukemic whole BM cells into sublethally irradiated (650 rads) congenic mice, which did not exhibit any survival difference (median survival; 23.5 vs 23 days, p = 0.848) ( Supplementary  Fig. 4d), demonstrating that Sharp1 deletion does not affect development or propagation of MLL-AF9 AML. Consistent with these findings, Sharp1 mRNA was elevated in MLL-AF6 AML cells compared to BM Granulocyte-Macrophage Progenitor (GMP) and granulocytes, which are the phenotypic normal    hematopoietic counterparts for AML cells 15 , while the expression in MLL-AF9 AML was comparable to normal BM GMP and higher than granulocytes ( Supplementary Fig. 4e), suggesting that upregulated Sharp1 may confer oncogenic properties to murine MLL-AF6 AML, but not to MLL-AF9 AML.
To investigate the reconstitution ability of Sharp1 −/− hematopoietic stem and progenitor cells (HSPCs), we performed competitive transplantation assays by injecting 500,000 BM cells from Sharp1 +/+ or Sharp1 −/− mice into lethally irradiated (900 rads) CD45.1 + CD45.2 + congenic mice along with equal number of BM cells from CD45.1 + congenic mice. In PB chimerism analysis, no differences were observed in percentage of donor cells in myeloid, B, and T cell lineages between recipients of Sharp1 +/+ and Sharp1 −/− BM cells over a period of 16 weeks after the transplantation (Fig. 5d, Supplementary Figs. 5c and 5d). Collectively, these results demonstrate that Sharp1 deletion does not affect steady-state hematopoiesis, as well as the ability of HSPCs to differentiate into multi-lineage cells and reconstitute hematopoiesis. SHARP1 binds to actively transcribed genes. Given these findings and the known functions of SHARP1 as a bHLH transcription factor, we hypothesized that SHARP1 binds to target genes and regulates their expression, which are important for the development and maintenance of AML. To delineate the direct transcriptional target genes, we performed ChIP-seq using antibodies against SHARP1 in ML-2 cells and identified 7,443 SHARP1-bound genes. Consistent with the known binding to Ebox with high affinity as a homodimer 25,26 , CACGTG was the most enriched motif in the binding regions across the genome, and it was increased near the binding peaks ( Fig. 6a and Supplementary Fig. 6a). A large proportion of SHARP1 occupancy was located at the proximal promoter (−1 kb, +100 bp from TSS, 36%), intronic (28%), and intergenic regions (27%) (Fig. 6a). SHARP1 was considered to function as a transcriptional repressor, either by direct or indirect binding to DNA, and interacts with DNA-bound transcription factors, such as C/EBP or MyoD, and recruits G9a, HDAC1, and SIRT1 to binding sites, resulting in alteration of histone modifications [27][28][29] . To delineate chromatin accessibility within SHARP1 binding sites, we overlaid them with the regions enriched with active enhancer and promoter marks (H3K4me3 and H3K27ac) 30 and the repressive mark (H3K27me3) 31 . Remarkably, SHARP1 binding sites were enriched in H3K4me3 and H3K27ac marks within the gene loci (Fig. 6a), defined as the promoter region (−2 kb) and gene body, of highly expressed transcripts (H3K4me3; 6459 genes, H3K27ac; 5840 genes), whereas the binding sites enriched in the H3K27me3 sites are located within the gene loci of poorly transcribed genes (1055 genes) (Fig. 6b). The genes that were known to be bound by SHARP1 protein (CLOCK, PER1, SHARP2, and MLH1) 32,33 demonstrated a SHARP1 peak in their promoters (Fig. 6c).

SHARP1 does not influence circadian clock genes expression.
SHARP1 is one of the regulators of the mammalian molecular clock 33 . Circadian clock genes generally play a critical role in cancer cells with tumor suppressive or oncogenic properties in a context-dependent manner 35,36 . In leukemias, PER2 (period circadian clock 2) was identified as a downstream target of C/EBPα and had its genitive impact in promoting AML initiation 37 . A recent study demonstrated that perturbation of the core circadian protein heterodimer, CLOCK/BMAL1, induced myeloid differentiation of AML cells and depleted LSC, highlighting the importance of clock genes in AML 38 . SHARP1 is regulated by the CLOCK/BMAL1 and represses their transcriptional activity by competing for DNA binding or direct interaction with BMAL1 33 . Thus, SHARP1 functions as a negative regulator for PER1, SHARP2, and SHARP1 itself in a feedback loop. Consistently, SHARP1 was bound to the promoter of the circadian clock genes (CLOCK, PER1, and SHARP2) in ML-2 cells (Fig. 6c, d). We asked whether upregulated SHARP1 induce the aberrant expression of the clock genes, which have a potential to affect AML activity. We investigated the expression of ten circadian clock genes (SHARP2, BMAL1, CLOCK, CRY1, CSNK1E, PER1, PER2, PER3, CUL1, and NR1D) in AML patients, comparing MLL-AF6 to other MLLr or non-MLLr AML, none of which exhibited aberrant expression (Supplementary Fig. 6d) (Fig. 7a). To delineate the correlation of genome-wide occupancy between SHARP1 and MLL-AF6, we overlaid SHARP1 and MLL N bound regions obtained from ChIP-seq analysis. Notably, 78 out of 92 MLL-AF6 target genes were SHARP1-bound (Fig. 7b, Supplementary  Table 3), and 14 of the co-target genes were downregulated upon SHARP1 knockdown (Fig. 7c), whereas none of these genes were upregulated (≥1.5-fold). The gene set includes previously identified oncogenic targets of MLL-FPs, such as MEF2C 15 , CDK6 39 , and RUNX2 8 , which demonstrated co-localization of MLL, MEN1, and SHARP1 within the promoter loci, accompanied with enrichment of H3K79me2/3 (Fig. 7d). These results suggest that SHARP1 cooperates with the MLL-AF6 protein complex, and expression of some MLL-AF6 target genes depend on SHARP1.
To determine whether SHARP1 forms a complex with MLL-AF6, we carried co-immunoprecipitation (co-IP) experiments in nuclear extracts from MLL-AF6 cell lines ML-2 and SHI-1. AF6  Median value is plotted in the box. c Representative SHARP1 binding peaks in the known target gene loci (circadian clock genes and MLH1). d Pathway analysis for the genes in the SHARP1 and H3K4me3 co-bounded regions within the promoter and gene body or MLL N co-IPs failed to detect SHARP1, which may have been obscured by heavy chain bands. To circumvent this issue, we performed co-IP in 293 T cells that were transfected with MLL-AF6 and SHARP1, and demonstrated a robust interaction between the two proteins (Fig. 7e). Intriguingly, we also observed interaction between SHARP1 and MLL-AF9 ( Supplementary  Fig. 7a), indicating that SHARP1 interacts with the portion of MLL that is present in both MLL fusions. Using a series of MLL deletion mutants 14 , we identified a region, amino acids 541-1251 (541-1251aa) of MLL, which was responsible for interaction with SHARP1. We did not observe an interaction with MLL (1-540aa), while the interactions with other MLL mutants were comparable to that of MLL (1-1251aa) (Fig. 7f, g)   only might provide a unique mechanism in regulation of the MLL-AF6 target genes.

Discussion
Our study reveals a unique mechanism in the leukemogenicity in MLL-AF6 AML. We identified direct MLL-AF6 target genes that are overexpressed in MLL-AF6 AML patients compared to other subtypes of MLLr-AMLs and focused on the bHLH transcription factor, SHARP1, the highest and most significantly upregulated gene. We demonstrated that SHARP1 plays an oncogenic role to maintain clonogenic growth and leukemia-initiating potential, regulating the expression of genes crucial for leukemia cell survival including MLL-AF6 target genes (Fig. 8).
MLLr-leukemias present with related gene expression profiles as a result of a common MLL-FPs driven activation of transcriptional elongation machinery 20,42 . Recent genome-wide ChIP-seq analyses identified various sets of direct target genes of MLL-FP, which consistently included the master regulatory factors (HOXA7, HOXA9, HOXA10, and MEIS1) required for the development of MLLr-leukemias. In this study, we identified 101 genes bound by MLL-AF6, and this number is comparable to the previous studies of other MLL-FPs, in which 165 genes were identified in MLL-AF4 43 , 139 genes in MLL-AF9 8 , and 178 genes in MLL-ENL 44 . In contrast to the recent ChIP-seq analysis which used ML-2 cells 45 , using a different MLL N antibody, we were able to identify SHARP1 as a direct target of MLL-AF6, and validate this finding by ChIP-qPCR. Furthermore, we found that MEN1 and LEDGF, two major MLL-FP subunits, co-bound to SHARP1 gene loci, corroborating our findings.
MLL-FPs recognize their target gene loci through the CXXC domain of MLL and the PWWP domain of LEDGF 14,21,45,46 . The CXXC domain specifically binds to unmethylated CpG DNA, which is enriched in active promoters, whereas the PWWP domain recognizes H3K36me2/3, which generally associates with transcriptionally active regions. Although MLL-FPs functions as an epigenetic reader through these common subunits of the MLL-FP complex, the majority of MLL-AF6 targets are not included in the gene sets of MLL-ENL, MLL-AF9, or MLL-AF4 targets 8,43,44 , raising the possibility that unknown mechanisms may be involved in this process. It is conceivable that distinct cellular functions and localizations of translocation partners may determine the unique target genes. AF6 (MLLT4), also known as afadin, is the most common MLL cytoplasmic partner protein and has a dual residency protein in the plasma membrane and the nucleus 47 . AF6 may be involved in the transcription of unique MLL-AF6 target genes by recruiting transcription factors or co-activators within the nucleus 48 .
Given the high expression level of SHARP1 comparable to those of pivotal oncogenic target genes, HOXA9 and MEIS1, in MLL-AF6 AML patients, we hypothesized that SHARP1 plays an oncogenic role in MLL-AF6 AML cells. SHARP1 may exert contextual tumor suppressive or oncogenic functions, depending on the type of cancer. A recent study demonstrated that SHARP1 is highly expressed in renal cell carcinoma cells and its overexpression accelerated tumor progression in xenograft models 49 , whereas in triple negative breast cancers, SHARP1 mediates the anti-metastatic function of p63 by degrading HIF-1α, and its overexpression is associated with a favorable prognosis 50 . In physiological conditions, SHARP1 is expressed in various tissues, though the expression level is generally low and is upregulated by external stimuli such as cytokines, infection, and hypoxia 26 , indicating its potent role as a positive regulator for cell survival under stress conditions. A recent study demonstrated that SHARP1 expression can be induced by DNA-damaging agents, and that SHARP1 inhibits activation of the p53 pathway including pro-apoptotic genes 51 , providing protection from cytotoxic effects. In agreement with the anti-apoptotic role of SHARP1, SHARP1 knockdown resulted in robust apoptosis in human MLL-AF6 AML cells, accompanied by the upregulation of p53 pathway and apoptosis associated genes ( Supplementary  Fig. 6e). However, a recent study by Coenen et al. demonstrated that shRNA-mediated SHARP1 knockdown did not have any effect in SHI-1 cells 52 , in contrast to our findings. This discrepancy might be explained by the difference in knockdown efficiencies with the use of different shRNAs against SHARP1. In fact, one of the shRNAs was common between their study and ours, and has led to reduced growth and increased apoptosis in SHI-1 cells, even though the differences were only significant in our study. Also, cell lines may acquire mutations that alter original characteristics after long periods of culture, which could explain differences in knockdown between these two studies. Based on our findings in the three MLL-AF6 and the two other MLLr-AML cell lines, as well as genetic deletion in murine AML models, we concluded that SHARP1 plays an oncogenic role in MLL-AF6 AML cells.
In contrast to a number of evidence for a transcriptional repressive role 25,27,28,53,54 , SHARP1 activates JunB and Gata3 expression to induce naïve T cells to Th2 T cells, and genetic deletion of Sharp1 leads to reduced histone H3 acetylation at the JunB conserved non-coding sequence and Gata3 promoter 55 . This indicates that Sharp1 regulates chromatin modification at these two loci and functions as a transcriptional activator. We demonstrated that SHARP1 binds to E-box motifs in active chromatin that are marked by H3K4me3 and H3K27ac, which suggests an interaction between SHARP1 and transcriptional activators or co-activators, recruiting chromatin modifiers to the binding sites. This process can be influenced by differences in post-translational modifications of SHARP1 protein, which allows association with distinct protein complexes in different cell contexts. For instance, SHARP1 sumoylation regulates its association with G9a 56 , which determines the SHARP1-dependent functions. We further demonstrated that physical interaction of SHARP1 and MLL-AF6 and co-localization on the promoters of the majority of the MLL-AF6 target genes, some of whose expressions are sensitive to SHARP1 levels. It is also possible that SHARP1 regulates these genes through interaction with constituents of these complexes. Delineating the proteins interacting with SHARP1 in MLL-AF6 AML cells will provide further insights into the process of gene regulation by SHARP1 in leukemia.
We identified a subset of (a) SHARP1 targets and (b) co-targets of MLL-AF6 and SHARP-1 that are critical for leukemogenicity using an integrative analysis of RNA-seq and ChIP-seq datasets in ML-2 cells. However, these genes are not overexpressed in MLL-AF6 AML patients compared to the other subtypes of MLLr-AML. SHARP1 ChIP-seq analysis highlighted that various motifs of potential co-factors are enriched in the promoter of SHARP1 targets ( Supplementary Fig. 7B), suggesting a more complex regulatory mechanism involving other transcription factors. It is plausible that those genes are activated by other transcription factors in different AML subtypes that generally do not express SHARP1. It will be of future interest to investigate how overexpressed SHARP1 influences the recruitment of transcriptional regulatory factors to chromatin, providing a unique mechanism for gene regulation in ML-AF6 AML.
Notably, SHARP1 played more profound role in maintaining the leukemia-initiating potential of L-GMP than whole BM cells in MLL-AF6 AML. This suggests that the pathways or genes that SHARP1 regulates may differ with the differentiation stage of leukemia cells, and may play a more significant role in LSC activity. MLL-AF6 AML patients present with a dismal clinical prognosis due to resistance to initial chemotherapy and high rate of relapse 57 , which may be caused by residual chemotherapy-resistant quiescent LSC 58 . Our finding that Sharp1 is dispensable for normal HSPC function suggests that SHARP1 could be a promising therapeutic target of MLL-AF6 AML LSC. Additionally, we showed that SHARP1 expression is sensitive to treatment with DOT1L inhibitor, indicating that inhibition of DOT1L could be a promising therapeutic approach to eliminate LSC and improve the prognosis in these patients by preventing relapse after chemotherapy.
Cell growth and Apoptosis assay. ML-2, CTS cells, and SHI-1 cells were transduced with lentiviral shRNA, shGFP, shSHARP1#1, or shSHARP1#2, selected with puromycin (0.5 μg/mL), and 4 × 10 4 cells seeded per well in 96-well plates and counted every 2 days by hemocytometer. Trypan blue was used to exclude dead cells. Apoptosis assay were performed according to the manufacturer's instructions using the FITC Apoptosis Assay Kit (BD Biosciences) and analyzed by LSRII flow cytometer.
Mice. Mice were housed at in a sterile barrier facility within the Comparative Medicine facility at the National University of Singapore. All mice experiments performed in this study were approved by Institutional Animal Care and Use Committee (IACUC). Sharp1 −/− mice were described previously 23 . CD45.1 + congenic mice (B6.SJL) and NSG mice were purchased from Jackson Labs.
Retrovirus transduction and generation of leukemia. The MLL-AF6 construct, consisting of 35-347 amino acids of the AF6 portion 12  Quantitative PCR. RNA was extracted using RNeasy kit (QIAGEN), reverse transcribed using the QuantiTech Reverse Transcription kit (QIAGEN), and quantitatively assessed using an ABI7500 (Applied Biosystem). For each sample, transcript levels of tested genes were normalized to β-actin using the delta CT method. The highest expression was arbitrarily set to 1 and expressions in the other samples were normalized to this value. All experiments were performed with three replicates. PCR was performed on cDNA using following primers: Chromatin immunoprecipitation assays. Chromatin immunoprecipitation (ChIP) (for both qPCR and sequencing) were performed as previously described 61 , with the following modifications: ML-2 cells were fixed using a 1% formaldehyde (FA) fixation protocol for 10 min for histone marks and SHARP1. For the other proteins, the cells were fixed using a 1% FA fixation protocol for 10 min, while a 45 min, 2 mM disuccinimidyl glutarate (DSG) and a 30 min 1% FA double fixation protocol was used. The antibodies used included SHARP1 (a mixture of H-72 Santa-Cruz, 12688-1-AP Proteintech, and ab175544 Abcam), H3K79me3 (Diagenode, pAb-068-050), MLL1 (Bethyl, A300-086A), H3K4me3 (Active Motif, 39159), H3K27me3 (Millipore, 07-499), H3K27ac (Millipore, 07-360), LEDGF (Bethyl, A300-848A), and MEN1 (Bethyl, A300-105A). Fixed chromatin samples were fragmented using a Bioruptor sonicator (Diagenode) for 30 min at high power in a constantly circulating 4°C water bath to an average size of 200-500 bps. Antibody:chromatin complexes were collected with a mixture of protein A and Protein G Dynabeads (Life Technologies) collected with a magnet, and washed 2 × with a solution of 50 mM HEPES-KOH, pH 7.6, 500 mM LiCl, 1 mM EDTA, 1% NP-40, and 0.7% Na-Deoxycholate. After a TE wash, samples were eluted, RNase and Proteinase K treated, and purified using a QIAGEN PCR purification kit. The primer sequences were HOXA9 promoter: 5′-TGGCTGCTTTTTTATGGCTTCA ATTATTG-3′/5′-CCGCGTGCGAGTGCG-3′, GAPDH promoter: 5′-CCCCTCC TAGGCCTTTGC-3′/5′-GCTGAGAGGCGGGAAAGTT-3′, and SHARP1 promoter: 5′-TGGGTAAACTTGAGTCCCAAAGGAAATT-3′/5′-TGCAAGTTG CTTCTTCTCGAGGC-3′. ChIP samples were quantified relative to inputs as described 61 . Briefly, the amount of genomic DNA co-precipitated with antibody is calculated as a percentage of total input using the following formula ΔCT = CT (input) − CT (chromatin IP), % total = 2 ΔCT × 5.0%. 50 μl aliquot taken from each of 1 ml of sonicated, diluted chromatin before antibody incubation serves as the input, thus the signal from the input samples represents 5% of the total chromatin used in each ChIP. CT values were determined by choosing threshold values in the linear range of each PCR reaction. The % input values of MLL N and H3K79me2 enrichment on HOXA9 (positive control) and SHARP1 loci were normalized against that of the GAPDH locus (negative control) in the respective samples for comparison.
RNA/ChIP-sequencing. RNA-seq libraries were prepared using Illumina Tru-Seq Stranded Total RNA with Ribo-Zero Gold kit protocol, according to the manufacturer's instructions (Illumina, San Diego, California, USA). Libraries were validated with an Agilent Bioanalyzer (Agilent Technologies, Palo Alto, CA), diluted, and applied to an Illumina flow cell using the Illumina Cluster Station. ChIP-seq libraries were prepared using Next ChIP-Seq library prep reagent set (New England Biolabs), and multiplexed (New England Biolabs). Each library was sequenced on an Illumina Hiseq2000 sequencer.
ChIP-Seq data analysis. In-house ChIP-seq data for SHARP1, MEN1, LEDGF, H3K79me3, H3K4me3, H3K27me3, and H3K27ac in ML-2 and MLL N in SHI-1 were used in this study. ChIP-seq data for MLL N and H3K79me2 in ML-2 were obtained from GSE83671, and those in MV4-11 and THP-1 were from GSE79899. Sequencing quality check for every dataset was performed. Mapping of the reads was performed using BWA to human genome version UCSC Hg19. MACS2 was used for peak calling between the special ChIP samples and the corresponding inputs with filtering parameters of cutoff p value at 0.05, minimal peak fold enrichment of 5, and minimal peak height of 10 reads. Peak integration cross different ChIP-seq datasets was done by utilizing the Bedtools. The Database for Annotation, Visualization and Integrated Discovery (DAVID) v6.7 was used to detect potential significantly altered pathways (https://david.ncifcrf.gov/). SHARP1 DNA binding motif enrichment was computed using the peak-motifs module from the RSAT suite 62 using oligomer length ranging from 6 to 8 nucleotides and the "merge lengths for assembly" option. ChIP-seq heatmaps were generated using ChAsE vers. 1.0.11 63 .
RNA-Seq data analysis. RNA-seq was performed for duplicates of shGFP, shSHARP1#1 and shSHARP1#2. After mapping of RNA-seq data to human genome version UCSC Hg19, normalization among the samples was performed using the total mappable counts. For identification of enriched gene sets or pathways, we utilized the GSEA (Gene Set Enrichment Analysis) software with the tool for preranked gene list (http://www.broad.mit.edu/gsea/) and the database MSigDB. v5.0.
Co-immunoprecipitation. HEK293T cells were transfected with the following plasmids: empty vector (control), HA-tagged SHARP1 (human), MLL-AF6, MLL-AF9 15 , and/or MLL deletion constructs 14 . The cells were lysed in coimmunoprecipitation (co-IP) lysis buffer (10 mM Tris-Cl pH 7.5, 150 mM NaCl, 0.5 mM EDTA and 0.5% NP-40), supplemented with complete EDTA-free protease inhibitor cocktail (Roche) for 30 min on ice. Lysates were cleared at 15,000× g for 15 min at 4°C. Co-immunoprecipitation was performed in co-IP lysis buffer with 2 μg of anti-HA (Santa Cruz, sc-7392) or anti-FLAG antibodies (Sigma, F1804) for 3 h or overnight at 4°C. Following which, three washes of immunoprecipitated proteins were performed with wash buffer (10 mM Tris-Cl pH 7.5, 150 mM NaCl, and 0.5 mM EDTA). The bound proteins were eluted with 2 × laemmli buffer for 5 min at room temperature with constant shaking. Samples were resolved on 7% SDS-PAGE and subjected to Western blot analysis.
Immunoblotting, immunoprecipitation, and antibodies. For whole cell lysis, 1 × 10 7 cells were lysed with radioimmunoprecipitation (RIPA) buffer. Protein concentrations were quantitated with Biorad Protein Assay (Bio-Rad). Proteins were separated on 10% SDS-PAGE gels. Immunoblots were incubated with primary antibody overnight at 4°C, followed by a secondary horseradish peroxidase (HRP)conjugated antibody (1:2000 dilution) at room temperature for 1 h. Details on the antibodies used and dilutions are described in Supplementary Table 5. Full blots are shown in Supplementary Fig. 8.
Flow cytometry. Single-cell suspensions were analyzed by flow cytometry. Antibodies conjugated with phycoerythrin (PE), PE-CY7, fluorescein isothiocyanate (FITC), allophycocyanin (APC), APC-CY7, and Pacific Blue obtained from BD Pharmingen, BioLegend, and eBioscience were used. Details on the antibodies used and dilutions are described in Supplementary Table 4. Stained cells were analyzed with an LSRII flow cytometer and sorted using a FACS Aria (BD Biosciences). Flow Jo 7.5 (Tree Star) was used for data acquisition and analysis.
Statistical information. The statistical significances were assessed by two-sided Student's unpaired t-test using the GraphPad PRISM 5 unless otherwise specified.
Data availability. The data supporting the findings of this study are available in the article or supplementary information files. Any other relevant data are available from the authors upon request. The ChIP-seq and RNA-seq datasets have been deposited in the Gene Expression Omnibus (GEO) repository with the accession code GSE 95511.