ZNF521 sustains the differentiation block in MLL-rearranged acute myeloid leukemia

Zinc finger protein 521 (ZNF521) is a multiple zinc finger transcription factor and a strong candidate as regulator of hematopoietic stem cell homeostasis. Recently, independent gene expression profile studies have evidenced a positive correlation between ZNF521 mRNA overexpression and MLL-rearranged acute myeloid leukemia (AML), leaving open the question on the role of ZNF521 in this subtype of leukemia. In this study, we sought to analyze the effect of ZNF521 depletion on MLL-rearranged AML cell lines and MLL-AF9 xenograft primary cells. Knockdown of ZNF521 with short-hairpin RNA (shRNA) led to decreased leukemia proliferation, reduced colony formation and caused cell cycle arrest in MLL-rearranged AML cell lines. Importantly, we showed that loss of ZNF521 substantially caused differentiation of both MLL-rearranged cell lines and primary cells. Moreover, gene profile analysis in ZNF521-silenced THP-1 cells revealed a loss of MLL-AF9-directed leukemic signature and an increase of the differentiation program. Finally, we determined that both MLL-AF9 and MLL-ENL fusion proteins directly interacted with ZNF521 promoter activating its transcription. In conclusion, our findings identify ZNF521 as a critical effector of MLL fusion proteins in blocking myeloid differentiation and highlight ZNF521 as a potential therapeutic target for this subtype of leukemia.


INTRODUCTION
Acute myeloid leukemias (AMLs) are heterogeneous tumor blood diseases characterized by deregulated cell proliferation, survival and differentiation of hematopoietic/stem progenitor cells [1]. One of the most aggressive subtypes of AML is characterized by the presence of translocation involving the mixed lineage leukemia gene (MLL, or KMT2A). MLL encodes an H3K4 methyltransferase that forms multiprotein chromatin-modifying complexes required in controlling transcriptional program necessary for the development and maintenance of hematopoiesis [2,3]. Translocations that include MLL count more than 60 different fusion partners, which have been identified in AML, acute lymphoid leukemia, and biphenotypic or chemotherapyrelated leukemias [4]. In pediatric and adult AML, the most common translocation juxtaposes the N-terminal portion of the MLL protein to the C-terminal fragment of the AF9 fusion partner in the t(9;11)(p22;q23) generating the oncogenic MLL-AF9 fusion protein [5][6][7]. MLL translocations contribute to leukemogenesis subverting self-renewal program and block of hematopoietic differentiation [5,8]. Transformation by MLL-AF9 induced specifically aberrant expression of several transcriptional target genes involved in stem cell selfrenewal, maintenance and repression of differentiationassociated genes [5,[9][10]. Among these MLL targets

Research Paper
Oncotarget 26130 www.impactjournals.com/oncotarget genes, such as HOXA9 and MEIS1, it is well established the crucial role played in MLL-induced leukemia [11,12], however many other of the genes strongly deregulated by MLL fusion proteins remain poorly characterized.
Zinc finger protein 521 (ZNF521) is a C2H2-type zinc finger transcription factor containing an aminoterminal motif that binds to the nucleosome remodeling and histone deacetylase (NuRD) complex, which is associated with transcriptional repression and conserved among other zinc finger proteins, including Friend of GATA (FOG1 and FOG2), BCL11A, and SALL family members [13]. Initially identified for its expression restricted to human CD34+ progenitor cells [14], ZNF521 has been shown to repress erythroid differentiation by inhibiting GATA-1 activity [15], and to block B-lymphoid differentiation in primary hematopoietic progenitors by antagonizing early B-cell factor 1 (EBF1) [16].
Deregulated expression of ZNF521 mRNA has been observed in medulloblastoma, lymphoblastic lymphoma and acute leukemia [17][18][19]. Recently, knock-in mice models for E2A-HLF and E2A-PBX1 involving fusion genes in B-lineage acute lymphoblastic leukemia (B-ALL) have demonstrated that enhanced expression of Zfp521, the murine counterpart of human ZNF521, cooperates to leukemia transformation, and an upregulation of ZNF521 was found in human B-ALL samples bearing E2A-HLF or E2A-PBX1 fusion oncogenes. Therefore, an altered expression of ZNF521 may be an important cofactor contributing to hematopoietic cell transformation. Recently, high expression of ZNF521 has been observed in pediatric AML, particularly in those cases carrying MLL gene rearrangements [20,21]; however the role of ZNF521 in MLL-rearranged AML is currently unknown. In this study, we examined the role of ZNF521 in MLL-rearranged AML cell lines and primary ex vivo MLL-AF9-expressing cells and showed that depletion of ZNF521 impaired AML progression by inducing myeloid differentiation.

ZNF521 is aberrantly overexpressed in pediatric MLL-rearranged AML
Previously, by use of microarray analysis we found a frequent ZNF521 overexpression in pediatric AML with MLL rearrangements [21]. To validate these data and analyze the relationship between ZNF521 expression and distinct MLL-fusion genes, we performed quantitative real-time PCR (qRT-PCR) in an independent cohort of 50 pediatric AML patients (16 MLL-rearranged and 34 non-MLL-rearranged; Supplementary Table 1) and 7 normal bone marrow (BM) controls. We found that ZNF521 was expressed at significantly higher level in AML patients with MLL rearrangements compared to non-rearranged AML and normal controls (P < 0.001, Figure 1A), The analysis of ZNF521 expression between the most frequent MLL rearrangements detected in pediatric AML did not reveal significant difference based on MLL fusion partners (data not shown). In addition, we analyzed the expression of ZNF521 in 6 MLL-rearranged and 6 non-MLLrearranged human leukemic cell lines. Similarly, leukemic cell lines with MLL rearrangements, with the exception of those carrying MLL-AF4 fusion transcripts, showed significantly higher ZNF521 mRNA levels compared to cell lines with other abnormalities (P < 0.05, Figure 1B). Thus, our data indicate that ZNF521 is likely involved in MLL-mediated transformation in AML.

ZNF521 depletion reduces cell viability and causes cell cycle arrest without inducing apoptosis of MLL-rearranged AML cell lines
To determine whether ZNF521 is functionally important in MLL-rearranged AML, we first examined the effects of ZNF521 knockdown on the cell proliferation using a panel of human MLL-rearranged AML cell lines, including, THP-1, NOMO-1 (both expressing MLL-AF9), ML-2 (expressing MLL-AF6) and OCI-AML4 (expressing MLL-ENL). To suppress ZNF521, we used GFP-tagged lentiviral vectors expressing anti-ZNF521 shRNAs (ZNF004 and ZNF710) or a nontargeting shRNA sequence (shScram). After assessing transduction efficiency by flow cytometry (range 30-80%) (Supplementary Figure 1 and data not shown), GFP-positive cells were sorted and maintained under standard cell culture conditions for subsequent analysis. As expected, in all four cell lines downregulation of ZNF521 varied between 60% and 75% compared to ZNF521 mRNA expression in shScram-transduced cells, and this correlated with a decrease in ZNF521 protein amount (Supplementary Figure 2). In addition, ZNF521 knockdown progressively reduced viability of all the transduced cell lines (Figure 2A), and it inhibited colony formation ability of MLL-rearranged cells, measured 2 weeks after transduction ( Figure 2B). In order to get a deeper insight, cell cycle analysis and apoptosis induction were assessed in GFP-positive MLL-rearranged cells. At day 7, we observed an accumulation of cells in G1 phase (17%-77%) in three out of four cell lines (THP-1, NOMO-1 and ML-2) expressing anti-ZNF521 shRNAs. This was most likely due to S phase reduction (from 29% to 65%) rather than G2/M alterations ( Figure 2C). However, annexin V/DAPI assay measured at day 4 and day 7 demonstrated that ZNF521 knockdown did not caused increased apoptosis ( Figure 2D), suggesting that ZNF521 may be involved in proliferation and differentiation of MLL-rearranged cells rather than in cell survival. To substantiate this hypothesis, GFP-sorted transduced THP-1 and ML-2 cells were collected on glass slides by cytospin and stained with antibodies against p21 (CDKN1A) and p27 (CDKN1B) cell cycle inhibitors [22]. www.impactjournals.com/oncotarget At day 7, we observed an increase of both p21 and p27 protein expression in ZNF521 knockdown cells, suggesting a prolonged G1/S transition as the main reason for the aforementioned cell cycle arrest (Supplementary Figure 3). Taken together, these findings indicate that ZNF521 expression is essential in the growth potential of MLL-rearranged AML cell lines.

Depletion of ZNF521 induces myeloid differentiation of MLL-rearranged AML cell lines
Given that ZNF521 can regulate lineage progression of different cell types, including hematopoietic cells [14][15][16], we analyzed whether ZNF521 depletion might influence differentiation in MLL-rearranged leukemic cells. Flow cytometry analysis of CD11b and CD14 myeloid markers was then performed on GFP-positive cells and revealed a change of these markers in 3 out of 4 cell lines transduced with ZNF521 shRNAs ( Figure 3A). The phenotypic changes were also sustained by a more mature macrophage-like morphology observed in all these cell lines upon ZNF521 depletion as compared with transduced control cells ( Figure 3B). Additionally, maturation induced by ZNF521 depletion was also supported by upregulation of C/EBPA and PU.1 mRNA levels, two myeloid differentiation markers ( Figure 3C). Furthermore, a downregulation of ZNF521 expression occurred in response to treatment with all-trans retinoid acid (ATRA) and with Securinine, two differentiation agents administered to THP-1 and NOMO-1 AML cells, respectively (Supplementary Figure 4). In particular, ATRA and Securinine, previously tested on these cell lines by others [23,24], were able to reduce ZNF521 mRNA and protein expression, and stimulate MLL-rearranged cell differentiation, supporting the hypothesis that ZNF521 is required to maintain those cells in an undifferentiated state.

Effects of ZNF521 depletion in patient-derived AML xenograft cells
To extend our findings to primary cells containing MLL-AF9 oncogene, we transduced ZNF521 shRNAs in ex vivo cells obtained from patient-derived xenografts ( Figure 4A). Two out of four patients harboring MLL-AF9 fusion protein (Supplementary Table 3) resulted in successful engraftment into NSG mice. The kinetics of such engraftment, measured by percentage of human CD45+ cells in the peripheral blood varying between 22.3% to 42.2%, ranged from 47 to 67 days and led to expansion of leukemic cells with the same immunophenotype and cytogenetic features of the original patient sample (data not shown). Ex vivo experiments demonstrated that ZNF521 depletion strongly impaired viability and colony formation of mononuclear cells obtained from two primary MLL-AF9 AML patientderived xenografts ( Figure 4B, 4C). Most importantly, an increased expression of myeloid differentiation markers CD11b and CD14 ( Figure 4D) and morphological features of mature monocytes/macrophage blast-like was observed ( Figure 4E). These findings suggest that ZNF521 Oncotarget 26132 www.impactjournals.com/oncotarget overexpression is critical to maintain an immature phenotype consistent with the MLL-rearranged cell lines results.

Gene expression changes after ZNF521 depletion in THP-1 cells
To investigate the gene expression pattern in MLL-AF9 AML cells expressing high levels of ZNF521, we performed microarray analysis of shZNF521or shScram-transduced THP-1 cells. Since that the differentiation was overt after 7 days of transduction as reported above, we performed gene expression profiling at day 4 after transduction. A total of 158 genes showed a significant change of expression (>1.5-fold change, FDR< 0.05), 58 were upregulated while 100 were downregulated ( Figure 5A and Supplementary Table 4). Gene Set Enrichment Analysis (GSEA) confirmed that ZNF521 depletion affected cell cycle progression and cell fate differentiation related genes [25] ( Figure 5B, 5C). These results showed also positive enrichment of genes downregulated in CD133+ HSCs when compared with the CD133-cell [26], and negative enrichment of embryonic stem cells (ESC) associated genes [27] (Figure 5D, 5E). The enrichment of stemness-related genes found by our analysis is in line with proposed role of ZNF521 in the regulation of hematopoietic stem cell homeostasis [28]. Furthermore, GSEA revealed a negative enrichment with genes that are upregulated in MLL-rearranged pediatric AML compared with non-MLL-rearranged AML [29]  shRNAs (ZNF004 or ZNF710) or non-targeting scramble control (shScram). GFP+ cells were sorted 4 days after transduction and placed in appropriate medium. Graphs show percentage of GFP+ cells measured at day 4, day 7 and day 10, normalized to the percentage of shScram cells. Data are represented as mean ± SD of at least three independent experiments. *P < 0.05, **P < 0.001, ***P < 0.0001, t-test. (B) Colony formation of GFP+ cells transduced with ZNF521 shRNAs or shScram. Error bars represent mean ± S.D. of three independent experiments. **P < 0.001, ***P < 0.0001, t-test. (C) Cell cycle distribution at day 7 of ZNF521 knockdown cells and control shScram of gated GFP+ cells. Data are represented as mean ± SD of three independent experiments. **P < 0.001, ***P < 0.0001, t-test.  Figure 5F, 5G). Interestingly, the ZNF521 depletion gene set revealed positive enrichment with genes that are upregulated in hematopoietic precursors conditionally expressing HOXA9 and MEIS1, including HOXA9 target genes upregulated in hematopoietic stem cells [8,30] (Figure 5H, 5I). In addition, genes up-regulated or downregulated upon knockdown of HOXA9 [11] were also similarly regulated in ZNF521-transduced THP-1 cells ( Figure 5J, 5K). Considering individual genes, we found deregulated genes with known relevance in MLLfusion-mediated AML (TET1, CDK6 and Musashi2) [31][32][33] and in myeloid progenitors differentiation (CD14 and MEF2A) [34] (Supplementary Figure 5). Taken together, these results indicate that ZNF521 expression negatively modulates genes involved in myeloid differentiation, and is required to maintain expression programs associated with MLL-induced transformation.

ZNF521 gene promoter is activates by MLL fusion proteins
Finally, to investigate the molecular mechanism that upregulates ZNF521 in MLL-rearranged AML, we performed luciferase reporter and ChIP assays using the Flag-tagged MLL-AF9 expression plasmid. To this end, we generated a series of constructs in which 5.0 kb of the Oncotarget 26134 www.impactjournals.com/oncotarget genomic region upstream of the ZNF521 transcription start site (TSS) was subdivided in 4 fragments (ZNF521P1, ZNF521P2, ZNF521P3 and ZNF521P4) and inserted into a pGL4-basic reporter plasmid ( Figure 6A). Luciferase assays in 293T cells showed that MLL-AF9 strongly activated the promoter region that lay between -1.3 to -3 kb (ZNF521P3) of the TSS ( Figure 6A). To further confirm the region of ZNF521 activated by MLL-AF9, we generated 3 constructs (ZNF521P3.1, ZNF521P3.2 and ZNF521P3.3) spanning the ZNF521P3 fragment ( Figure 6B). We found that the pGL4-ZNF521P3.3 construct showed the highest luciferase activity ( Figure 6B), indicating that the MLL-AF9 responsive elements likely reside between −1.0 and −1.6 kb upstream of the ZNF521 TSS. Furthermore, to determine whether ZNF521 activation was MLL fusiondependent, we performed ZNF521-driven luciferase reporter assay in another MLL fusion gene (MLL-ENL) and in two non-MLL-associated fusion genes such as AML1-ETO and PML-RARAα. We observed that both AML1-ETO and PML-RARAα yielded only a minimal luciferase activity compared with MLL-ENL that showed even a higher promoter binding affinity than MLL-AF9 (> 2.5 fold) ( Figure 6C). Besides, wild-type (WT) MLL did not affect luciferase activity under the same settings, providing evidence that only MLL-fusion proteins likely activate ZNF521 expression ( Figure 6C). Consistent with these results, ChIP analyses showed that both MLL-AF9 and MLL-ENL bind to ZNF521 promoter region in transfected 293T cells ( Figure 6D). In order to validate MLL-AF9 binding to the ZNF521 promoter in AML cells, we performed ChIP with lysate from NOMO-1 and HL60 cell lines that endogenously expressing MLL-AF9 and WT MLL, respectively. Since that MLL-AF9 lacks the MLL-C portion of WT MLL, an anti-MLL N-terminal (MLLN) and an anti-MLL C-terminal (MLLC) antibodies were used for this experiment. ChIP assays showed that MLLN bound specifically to the ZNF521 promoter region in NOMO-1 but Oncotarget 26135 www.impactjournals.com/oncotarget not in HL60 ( Figure 6E; upper panel). By contrast, there was not apparently association with MLLC and ZNF521 in both NOMO-1 and HL60 cells ( Figure 6E; lower panel). Together, these findings demonstrated that ZNF521 promoter is specifically bound by MLL-AF9, and provide further evidence that MLL fusion oncoproteins may drive aberrant expression of ZNF521, which may in turn lead to a block in differentiation.

DISCUSSION
We present data showing that pediatric AML patients carrying MLL translocations have a significantly upregulation of ZNF521 expression independently of the fusion partner involved in the translocation with MLL. The overexpression of ZNF521 is a robust transcriptional feature of MLL-rearranged AML, consistent across independent adult and pediatric microarray datasets [19][20][21]. From these data, we started our study with the aim to decrypt the ZNF521 function as transcription factor in MLL-rearranged AML, and understand if it might deserve attention as potential therapeutic target.
A major hallmark of leukemia and a consequence of MLL fusion proteins expression is a block in hematopoietic differentiation [35]. On this way, our data show that the most relevant effect of ZNF521 depletion  Oncotarget 26137 www.impactjournals.com/oncotarget was to enhance myeloid differentiation of leukemia cells as evidenced by changes in cell morphology, immunophenotype and increase of a myeloid-specific gene expression in MLL-rearranged cell lines and primary cells. The requirement of ZNF521 in the maintenance of an undifferentiated status associated with MLL-rearranged AML was also supported by the fact that ZNF521 expression drastically decreased upon treatment with specific differentiation-induced agents, such as ATRA. The observed growth defect, cell cycle arrest and reduced colony formation upon ZNF521 depletion were secondary to cells entering into a differentiation program. Thus, MLL fusion proteins might promote leukemogenesis not only by HOXA9 and MEIS1 upregulation, but also by keeping the ZNF521 overexpressed, which in turn contributes to a block of differentiation or to the maintenance of an undifferentiated state of leukemia cells.
Consistent with this finding, others have reported that loss of ZNF521 enhanced erythroid differentiation and increased B-lineage maturation in cell lines and primary hematopoietic progenitor cell, respectively [15,16]. Moreover, it is well established that Zfp521, the mouse counterpart of human ZNF521, in other cellular contexts including, embryonic stem cells (ESCs), neural cells, osteoblasts and chondrocytes mainly function to control cell differentiation of primitive or mature cells by modulating the activity of specific transcription factors [36][37][38][39]. Consistent with these findings, our GSEAs in THP-1 cells depleted for ZNF521 showed enrichment of hematopoietic stem cells (HSCs)-and ESC-associated downregulated genes and sets of genes associated with differentiation program [25,26]. Based on these findings, we assumed that ZNF521 has not only a role in promoting self-renewal and maintenance of HSC but it also acts in MLL rearranged AML. Furthermore, our results support the direct activation of ZNF521 by MLL fusion proteins increasing the importance of this transcription factor in the transformation of the leukemia cells. In fact, we showed enrichment of set of genes related to MLL fusion-dependent transformation signatures as well as to HOXA9-mediated gene expression program [11,30,40]. Thus, the events documented after ZNF521 depletion, which in part resemble what has been previously observed in MLL-rearranged cells after loss of HOXA9, gave a further support that ZNF521 plays a critical role in MLL-fusion-mediated leukemia. Interestingly, the expression of either HOXA9, a canonical downstream target for MLL-rearranged leukemia [10,41] or ZNF521 have been shown to be restricted in CD34+ progenitor cells [28,30].
Nevertheless, in our gene-expression analysis, loss of ZNF521 does not affect HOXA9 expression, implying that both are MLL-dependent but might act in a nonmutually exclusive and additive manner.
Supporting the idea that ZNF521 is particularly required for MLL-mediated leukemia, our data of luciferase reporter and ChIP assays revealed that ZNF521 is a direct target of both MLL-AF9 and MLL-ENL fusion proteins. We defined a genomic region of 555 bp in 5′ ZNF521 promoter that is thought to be crucial for ZNF521 activation by MLL fusion proteins. This finding is consistent with prior observations that showed how the modulation of MLL-AF9 levels resulted in concordant changes in ZNF521 expression in different human in vitro models [42,43]. Surprisingly, the inspection of ChIP-seq data from Bernt et al. [44] did not show peak in the vicinity of the Zfp521 gene in an MLL-AF9 mouse leukemia model. Of note, this is also observed for other well-known targets of MLL fusion proteins such as EVI1 and PLZF [45,46] About ZNF521, this can be explained by the different approaches used, and the fact that in mouse BM Zfp521 is primarily expressed in the HSC fraction and significantly reduced in granulocyte-monocyte-progenitor cells (GMPs) (http://servers.binf.ku.dk/bloodspot/?gene= ZFP521&dataset=nl_mouse_data), in which the analysis has been done. Future ChIP-seq experiments on human transformed HSC will likely shed further light on ZNF521-MLL-AF9 target gene specificity.
In summary, this study unravels the antidifferentiation function of ZNF521 in MLL-rearranged cells and showed the mechanism by which ZNF521 participates in MLL-fusion mediated transformation. This data also indicate that ZNF521 is highly expressed in the majority of MLL-rearranged AML pediatric patients, and thus ZNF521 could be a potential molecular target for this subtype of aggressive leukemia.

Patient samples and cell lines
All of the pediatric AML patient samples were obtained at the time of diagnosis from the University-Hospital of Padua and stratified according to the AIEOP AML 2002/01 protocol AML 2002/01 [47]. Patient characteristics are listed in Supplementary Table 1. Seven BM samples from healthy donors were obtained as control. All human myeloid cell lines (THP-1, NOMO-1, OCI-AML4, ML2, HL60, K562, Kasumi-1, NB4, U-937, SEM and REH) were obtained from DSMZ (Braunschweig, Germany) and 293T cells were obtained from ATCC (Manassas, VA, USA). All cell lines were maintained under standard conditions suggested by the manufacturer.

Quantitative real time PCR
Total RNA was extracted with Trizol reagent (Invitrogen) and reverse transcribed into cDNA using the Superscript III First-Strand Synthesis System (Life Technology). The mRNA levels of ZNF521, PU.1, CEBPa, HOXA9 and MEIS1 were measured by quantitative RT-PCR (qRT-PCR) with SYBR green on an AB 7900HT real time system (Applied Biosystem) using the comparative C t method and the GAPDH gene expression Oncotarget 26138 www.impactjournals.com/oncotarget as internal control [48]. The primer sequences for quantitative qRT-PCR are listed in Supplementary Table 2.

Lentiviral shRNA vector, transduction and FACS-sorting
For knockdown studies, two shRNAs against ZNF521 and a control scrambled shRNA (shScram) were used (Mission pLKO.1-puro-CMV-TurboGFP system, Cat Number TCRN0000229710 and TCRN0000229004, Sigma-Aldrich) (see Supplementary Table 2 for shRNA sequences). Lentiviral cell transduction was performed as described previously [49]. After culture in fresh medium, GFP-positive cells were sorted 96 hours after infection using a MoFlo XDP cell sorter (Beckman Coulter) and used for experiments. Alternatively, cells were gated for GFP expression and subjected to flow cytometry analyses. ZNF521 knockdown efficiency was measured by qRT-PCR and western blot analyses.

Plasmids constructs, transient transfection and luciferase assay
pMSCV-neo-Flag-MLL-AF9, MSCV-PML-RARA-IRES-GFP, MSCV-AML1-ETO-GFP, pCMVMLL-3xFlag and pCMVMLL-ENL-3xFlag have been previously described [42,[50][51]. Flag-tagged proteins were previously verified by Western blot with anti-Flag M2 antibody (Sigma), as well as the GFP-tagged proteins by expression of green fluorescence protein (GFP) in vitro. The procedure to generate the various constructs of the ZNF521 promoter into pGL4.28 plasmid (Promega) is described in Supplementary Materials and Methods. For luciferase assay, 293T cells were cotransfected with 0.5 mg of the reporter plasmid, 1 mg of expression plasmid or empty vector and 0.5 mg of Renilla luciferase reporter vector (Promega) as internal control for normalization of transfection efficiency, for a total of 2 mg of combined plasmids per well. The cells were then harvested at 48 hours after transfection using a Dual-Luciferase reporter assay system (Promega) and the Victor3 TM 1420 Multilabel Counter (PerkinElmer). Data are presented as the mean ratio for triplicate experiments.

Chromatin immunoprecipitation (ChIP) and PCR detection
ChIP assay was performed using the Imprint Chromatin Immunoprecipitation kit (Sigma), according to the manufacturer's protocol with minor modifications. Briefly, 293T cells (3.5 × 10 6 cells) were transfected with 10 mg of Flag-MLL-AF9 or Flag-MLL-ENL expression plasmids. 48 hours post transfection, were cross-linked with 1% formaldehyde (Sigma) for 15 minute at room temperature. Subsequently, the lysed cells were isolated and sonicated on ice to shear DNA into fragments of 200 bp to 1 kb. Then, the chromatin complexes were incubated into pre-treated Stripwells (Sigma) with anti-Flag M2 monoclonal antibody (Sigma), or normal mouse IgG (Sigma) as indicated. The input DNA was isolated from sonicated lysates before immunoprecipitation as a positive control. Purified DNA was then resuspended in TE buffer (10 mM Tris-HCL and 1 mM EDTA, pH 8.0) for PCR. ChIP assay from 2 × 10 6 of MLL-AF9-expressing NOMO-1 cells or HL60 cells was performed as above reported using a N-terminal MLL monoclonal antibody (Santa Cruz Biotechnology) or a C-terminal MLL polyclonal antibody (Sigma) or a mouse IgG (Sigma) as indicated. Purified ChIP DNA was amplified by regular PCR. Primers amplifying the ZNF521 promoter region and the HOXA9 promoter region used for the ChIP PCR are listed in Supplementary Table 2.

Microarray analysis
Total RNA from sorted THP-1 cells transduced with shRNAs was isolated using Trizol as above reported and processed for microarray analysis using the Affymetrix GeneChip 3′IVT express Kit (Affimetrix) after RNA quality control using Agilent 2100 Bioanalyzer (Agilent). Gene expression profile was performed using a Human Genome U133 2.0 Plus chip (Affymetrix), as previously described [52]. The data were RMA-normalized using R software (http://www.r-project.org/) with BioConductor package (www.bioconductor.org). Shrinkage t test was used to identify differentially expressed genes between shScram and shRNA ZNF521 THP-1 cells selected with a local FDR < 0.05 (FDR). Hierarchical clustering analyses were performed using Euclidian distance and Ward's methods. Gene set enrichment analysis (GSEA) was performed using GSEA version 2.0 software (Broad Institute; http://www.broadinstitute.org/gsea) with genes ranked by difference of class and statistical significance by 1000 gene set permutations. Gene set permutation was used to enable direct comparisons between shScram and shRNA ZNF521 results (< 7 replicates). Median of probes was used to collapse multiple probe sets to a single value per gene for each sample. Gene sets with a FDR < 0.05 were declared to be statistically significant. The microarray gene expression data have been submitted in NCBI's Gene Expression Omnibus (GEO, http://www. ncbi.nlm.nih.gov/geo/) under accession GSE79110