Global Identification of EVI1 Target Genes in Acute Myeloid Leukemia

The ecotropic virus integration site 1 (EVI1) transcription factor is associated with human myeloid malignancy of poor prognosis and is overexpressed in 8–10% of adult AML and strikingly up to 27% of pediatric MLL-rearranged leukemias. For the first time, we report comprehensive genomewide EVI1 binding and whole transcriptome gene deregulation in leukemic cells using a combination of ChIP-Seq and RNA-Seq expression profiling. We found disruption of terminal myeloid differentiation and cell cycle regulation to be prominent in EVI-induced leukemogenesis. Specifically, we identified EVI1 directly binds to and downregulates the master myeloid differentiation gene Cebpe and several of its downstream gene targets critical for terminal myeloid differentiation. We also found EVI1 binds to and downregulates Serpinb2 as well as numerous genes involved in the Jak-Stat signaling pathway. Finally, we identified decreased expression of several ATP-dependent P2X purinoreceptors genes involved in apoptosis mechanisms. These findings provide a foundation for future study of potential therapeutic gene targets for EVI1-induced leukemia.


Evidence for the Role of EVI1 in Myeloid Leukemia
The ecotropic virus integration site 1 (EVI1) is an oncogenic transcription factor associated with human myeloid malignancy and several solid epithelial cancers [1,2,3]. Aberrant EVI1 expression occurs in 8-10% of human adult acute myeloid leukemia (AML) and strikingly up to 27% of pediatric mixed lineage leukemia (MLL) rearranged leukemias [4]. EVI1 is one of several protein isoforms encoded by the MECOM locus at human chromosome 3q26 which also yields the MDS1 and MDS-EVI1 protein isoforms [5]. The role of MDS1 and MDS-EVI1 in malignancy is still unclear, while the EVI1 transcription factor, specifically the 135kDa isoform has been reported as a malignant contender [6]. EVI1 overexpression in human AML most frequently occurs with rearrangements at chromosome 3q26 [7,8]. The MLL-AF9 fusion oncoprotein has also been shown to activate the MECOM locus in the setting of AML [9].
Although previous studies have certainly supported the role of EVI1 in myeloid malignancy, establishing an experimental system with consistent disease induction has been challenging. Forced expression of Evi1 in murine lineage-negative bone marrow (BM) cells via retroviral transduction followed by transplantation back into irradiated recipients has yielded conflicting results. Buonamici et al demonstrated Evi1 transduced BM in C57BL6 recipients developed lethal myelodysplastic syndrome (MDS) 8-12 months after bone marrow transplantation (BMT), but none developed AML [10]. In another study, Cuenco et al showed none of the mice that received BM cells transduced with the Evi1 retrovirus developed AML [11]. In contrast to these results, Yoshimi et al showed C57BL6 mice transplanted with Evi1-transduced bone marrow cells all developed AML (characterized by high percentage of blasts on BM smear, positive myeloid markers for leukemic cells, marked splenomegaly) and died within 6-11 months after BMT [12]. Furthermore, a separate study demonstrated Evi1 does not induce AML alone, but requires co-expression with Hoxa9/ Meis1 to drive leukemogenesis [13]. Collectively, the current data does not support a specific experimental approach by which Evi1 overexpression by itself consistently induces leukemogenesis.

EVI1 Binds DNA to Induce Leukemic Transformation
The Evi1 gene spans 65 kb of genomic DNA with 16 exons which generate 3 different isoforms (135kDa [14], 123kDa [15], 103kDa [16,17]). The 135kDa and 123kDa isoforms both contain two zinc finger domains, ZF1 and ZF2 that bind DNA in a sequence specific manner [18,19]. The 103kDa isoform lacks ZF1 domain fingers 6 and 7, and fails to bind DNA via that domain [16]. We previously demonstrated ZF1 binds to the motif GACAAGATA with high affinity and specificity in vitro [19] and showed ZF1, but not ZF2 is critical for malignant activity [20,21]. Zhang et al recently demonstrated ZF1 DNA binding can be inhibited with a pyrrole-imidazole polyamide with high specificity and affinity [20].
Several studies have identified EVI1 downstream target genes associated with putative leukemogenic functions [22,23,24,25,26,27,28,29]. Direct EVI1 binding to the promoter of Gata2, an essential regulator of HSC proliferation [30], was demonstrated by ChIP-qPCR. Gata2 has been reported to be aberrantly expressed in 87% of de novo AML cases [31]; our analysis of RNA expression data from AML patients shows a good correlation between EVI1 and GATA2 expression (Pearson correlation (r) of 0.42-0.52; unpublished data). However a definitive requirement for Gata2 in EVI1-induced leukemogenesis has yet to be shown. A genome wide transcription factor binding study for EVI1 has been reported recently for a human ovarian cancer cell line [28]. The study demonstrated over 25% of EVI1occupied genes were also bound by activator protein 1 (AP1), providing evidence for a synergistic cooperative interaction between EVI1 and AP1, specifically the FOS protein. AP1 controls important cellular processes such as apoptosis, cellular differentiation and proliferation and has been described as a ''nuclear decision-maker'' critical for determining life or death cell fate decisions [32]. Taken together, these studies provide evidence that EVI1 directly binds critical genes associated with malignant transformation.

Biologic Effects of EVI1
AML cells harbor dysfunction of one or more of the following decision processes: cellular differentiation, programmed cell death (apoptosis) and cellular growth control. In regards to differentiation, EVI1-induced leukemic cells have been associated with defects in terminal myeloid differentiation, specifically disruption of granulocytic and erythroid commitment [33,34]. Morishita et al first reported Evi1 overexpression in 32Dc13 myeloid cells inhibits terminal differentiation to granulocytes in response to granulocytecolony stimulating factor (G-CSF) [34]. However it was later shown that native 32Dc13 cells harbor a proviral insertion at Evi1 and overexpress both mRNA and protein [35]. In addition, this assay is difficult to interpret, since the EVI1-overexpressing cells undergo cell death upon treatment with G-CSF. Another study showed Evi1 overexpression in BM progenitors result in impaired myeloid terminal differentiation associated with a subset of genes regulated by PU.1 binding [36]. More recently, Evi1 has been shown to be preferentially expressed in HSCs and required for the maintenance of hematopoiesis [37,38]. However, there is still a paucity of data connecting EVI1 binding to specific gene targets and how it influences definitive hematopoietic cell lineage decisions.
In addition to blocked differentiation, Evi1 leukemic cells also demonstrate resistance to apoptosis which has been associated with ineffectiveness of chemotherapy regimens, high relapse rates and poor prognosis [39]. The survival advantage conferred by Evi1 in myeloid leukemic cells has been well studied [40,41,42]. Kurokawa et al showed EVI1 directly interacts with and inhibits c-Jun N-terminal kinase (JNK) to protect cells from JNK-activated stress-induced cell death [42]. EVI1 ZF1 also binds and activates the BCL-XL promoter in the colon carcinoma HT-29 cell line overexpressing EVI1, resulting in inhibition of apoptosis [40]. However, a role for the deregulation of JNK and BCL-XL in leukemogenesis has not been directly addressed. We have also shown that Evi1 knockdown in DA-1 leukemic cells induces apoptotic features such as DNA fragmentation, reduction in mitochondrial membrane potential and cleavage of procaspases 3 and 9 (delCampo et al, in revision). Previous studies demonstrate a single amino acid mutation (R205N) in ZF1 prevents EVI1 binding to DNA [23]. Preliminary data shows DA-1 leukemic cells overexpressing the R205N mutant EVI1 exhibit significantly increased apoptosis, supporting the notion that ZF1 DNA binding is critical in suppressing apoptosis (del Campo et al, in revision).
Collectively, there appears to be good evidence for EVI1-induced anti-apoptosis mechanisms, but additional studies are needed to confirm these findings and to flesh out the precise mechanism.
Finally, inappropriate Evi1 expression has been associated with aberrant cell cycle regulation resulting in excessive proliferation [43,44,45]. Abnormal cellular proliferation mediated by the TGFb pathway has frequently been cited in Evi1 expressing cells. EVI1 has been reported to interact with and repress SMAD3 function, resulting in loss of TGFb-induced antiproliferative effects [43]. However, the relevance of this to AML is not clear. Evi1 has also been shown to accelerate the cell cycle of Rat-1 fibroblasts [45], murine 32Dcl3 myeloid cells [44] and murine embryonic stem cells [46]. However according to other reports, the cell cycle and proliferative activity of HEL cells is not influenced by EVI1 overexpression [47]. These conflicting data seem to indicate that EVI1-regulated proliferative effects in AML have yet to be elucidated.
Various other biologic functions regulated by EVI1 downstream gene targets have also been identified by ChIP assay and confirmed by PCR experiments. These functions include disruption of normal hematopoiesis [48], growth arrest in response to stressful stimuli [27], calreticulin function [24], and microRNA gene silencing [26].
Despite these many findings, a cohesive mechanism by which Evi1 induces leukemogenesis remains elusive.
Here, we report for the first time ChIP-Seq combined with RNA-Seq expression profiling in Evi1-overexpressed leukemic cells. We found that deregulation of genes involving apoptosis, differentiation and proliferative mechanisms likely all contribute to the development of Evi1 leukemogenesis. Specifically, we identified EVI1 directly binds to and downregulates a master myeloid differentiation regulator gene, Cebpe, in both Evi1 overexpressed leukemic cell lines. We found a high number of downstream gene targets of Cebpe were also downregulated in EVI1 leukemic cells. We also identified EVI1 binds to and deregulates Serpinb2 as well as numerous genes involved in the Jak-Stat signaling pathway to drive cellular differentiation. Finally, we found several ATPdependent P2X purinoreceptors involved in apoptosis mechanisms, particularly P2rx7, to be significantly downregulated.

Whole Transcriptome Analysis using RNA-Seq
To identify genes differentially expressed between Evi1 overexpressed (DA-1, NFS-60) and in shRNA Evi1 knockdown cells, RNA was extracted to generate transcriptome-wide expression profiles. Genes with expression levels significantly greater or reduced relative to the control shRNAs (shLuc and shScr) cell lines have been termed upregulated and downregulated, respectively. High throughput parallel sequencing using the Illumina Genome Analyzer IIx revealed 806 significantly deregulated (p,0.05) genes in DA-1 cells (481 upregulated, 325 downregulated in the Evi1 overexpressed cells compared to the Evi1 shRNA knockdown) and 782 deregulated genes in the NFS-60 cell line (437 upregulated, 345 downregulated, Dataset S1).
A total of 35 genes were significantly upregulated and 42 genes were significantly downregulated in both cell lines ( Table 2). We identified a 2-fold downregulation of Cebpe, a master regulator of terminal myeloid differentiation, in both the murine EVI1 leukemic cell lines. However significance was only reached in the NFS-60 cell line due to the low number of RNA-Seq reads in the DA-1 cell line for the Cebpe gene (Dataset 1). A U937 human leukemic cell line with Evi1 overexpression via retroviral infection also confirmed significant downregulation of Cebpe by PCR ( Figure 1a). We also found a high number (N = 6) of significantly downregulated direct gene targets of C/EBP-e in DA-1 leukemic cells (Dataset S2, Figure 2a and 2b). In NFS-60 leukemic cells, 3 C/EBP-e direct gene targets were also significantly downregulated (Dataset S3, Figure 2c). These results demonstrate EVI1 leukemic cells not only exhibit downregulation of Cebpe expression, but also suppression of downstream target genes of the master differentiation regulator.
Expression levels of numerous genes associated with the regulation of Jak-Stat signaling (a principal pathway by which cytokines and growth factors control differentiation, proliferation and apoptosis) were found to be aberrant in both EVI1 leukemic cell lines ( Figure 3). Socs1 (supressor of cytokine signaling 1), an inhibitor of STAT transcription factors, was significantly downregulated by 5.7-fold in DA-1 EVI1 leukemic cells, p = 0.01 (Dataset S2), and by 4.4-fold in NFS-60 EVI1 leukemic cells, p = 0.02 (Dataset S3). In NFS-60 leukemic cells, Stat1 and Stat5 expression levels were also significantly upregulated (p = 0.02 and p = 0.01 respectively, Dataset S3). Phosphorylation of STAT1 in Evi1 overexpressed cells was tested in two separate human hematopoietic cell lines with verified Evi1 overexpression (Kasumi 3 and U937+Evi1). Marked increased total STAT1 protein was present in Kasumi 3 cells (an established human leukemic cell line with Evi1 overexpression) at baseline compared to the control. There was also an increase in phosphorylated STAT1 in Kasumi 3 cells ( Figure 4). U937 wildtype and U937+Evi1 overexpressed cells did not show a marked difference in total STAT1 or phosphorylated STAT1 protein levels (data not shown).
Several genes associated with cell cycle regulation, specifically those in the serine protease inhibitor (Serpin) family, were significantly downregulated in both EVI1 leukemic cell lines. These included Serpinb2 and Serpinf1. There was a striking 11.4fold decrease in Serpinb2 expression in DA-1 EVI1 leukemic cells (Dataset S2), and an 11.5-fold decrease in NFS-60 leukemic cells (Dataset S3). Using conventional and q PCR, we were also able to demonstrate marked Serpinb2 downregulation in the two human hematopoietic cell lines with Evi1 overexpression, Kasumi 3 and U937+Evi1 ( Figure 5). Serpinf1 (also known as pigment epitheliumderived factor, PEDF) was also significantly reduced (DA-1 cells 3.4-fold downregulation, p = 0.02, and NFS-60 cells 2.1-fold downregulation, p = 0.01, Datasets S2 and S3).

ChIP-Seq for EVI1 DNA Binding Sites
To globally identify direct gene targets of EVI1, we performed ChIP-Seq experiment. DNA bound to EVI1 from the DA-1 murine leukemic cell line was precipitated using both anti C-and N-terminal EVI1 mouse antisera [23]. The generated sequencing reads were mapped to the mouse genome (mm9) by using the bowtie program [53]. This resulted in around 5 million uniquely mapped reads. To identify EVI1 binding peaks, we applied Model-based Analysis of ChIP-Seq (MACS) program [54], which was designed to analyze data generated by short read sequencers such as from the SOLiD platform to first estimate peak size and location, using SAM files as an input. We identified 16,745

Mapk signaling 6
Pathways in cancer 6 DAVID analysis was performed for significantly upregulated and downregulated genes in both the Evi1 overexpressing DA-1 and NFS-60 leukemic cell lines. doi:10.1371/journal.pone.0067134.t001 significant peaks by using the cutoff of 1.00e-05 for the p-value. We then mapped those peaks on genome-wide scale relative to RefSeq mouse genes ( Figure 6a). 7.1% of peaks were within 1kb of the transcription start site (TSS). A de novo motif discovery algorithm, MEME [54], was performed on the top 1000 ranked EVI1 ChIP-Seq peaks. MEME identified an AGGAAG ETS-like motif (E-value = 2.1e-193). We then refined this motif by running TPD [55]   previously reported EVI1 ChIP-Seq study in ovarian cancer cells which reported 5097 EVI1 significant binding peaks with an ETSlike motif, and over 2000 direct gene targets bound by EVI1 through the ETS-like motif [28].
To provide biological meaning to the significant EVI1 peaks, the Stanford GREAT Analysis Tool was used to assign peaks to nearby annotated genes [56]. EVI1 peaks were significantly associated with 8565 annotated genes (Dataset S5). Of the 35 significantly upregulated and 42 downregulated genes shared by both EVI1 leukemic cell lines, 86% exhibited significant EVI1 DNA binding and deregulation of transcription (Dataset S6).  Cebpe, Socs1 and Ube1l were all noted to have significant EVI1 binding. Seven significant EVI1 binding sites were found for Cebpe (Figure 7a), 5 with the AGGAAG ETS-like motif and 2 of which were in the promoter region (-2185, 22585 relative to the TSS).
Significant EVI1 binding sites were identified for several of the C/ EBP family of genes (Cebpa Cebpb, Cebpd, Cebpe and Cebpg). However .2-fold changes in gene expression were only present for Cebpe in Eight significant EVI1 DNA binding sites were identified for Socs1 (Figure 7b), 7 with the ETS-like motif and 3 of which were within the promoter region (-307, 2657 and 21007 relative to the TSS). Significant EVI1 binding sites were also identified for Socs3 (2 significant binding sites at 25610, 210360 relative to TSS), Socs4, Socs5, and Socs7, but with the exception of Socs3 in NFS-60 cells, the expression levels for these genes were not significantly different in EVI1 leukemic cells. For Osm, 7 significant EVI1 binding sites were found, 6 which were within the promoter region ( Figure 7c). Four of the promoter region Osm binding sites had the ETS-like binding motif. Two significant EVI1 DNA binding sites were identified for Ube1l (Figure 7d), both of which were within the promoter region (-397 and 23447 relative to the transcription start site, TSS) and had the ETS-like motif.
Six significant EVI1 DNA binding sites were found for Serpinb2 (Figure 7e), all of which had the ETS-like motif, 2 of which were within the promoter region (-1745, 23945 relative to the TSS). Two significant EVI1 DNA binding sites were identified for Serpinf1 (-6601 and +7384 relative to TSS), both of which had the ETS-like motif.
Regarding genes regulating cellular death, ChIP-Seq revealed 7 EVI1 binding sites for the P2rx7 gene (Figure 7f), all of which had the ETS-like motif, 3 within the promoter region (+451, +901, +1701 relative to the TSS).
We validated ChIP-Seq peaks for selected genes near or in promoter regions by standard and quantitative PCR analysis using EVI1 antisera and no antibody immunoprecipitated chromatin ( Figure 8).

Shared DNA Binding Sites with Other Transcription Factors
An unusually large number of EVI1 binding sites were identified within 1.5kb of annotated genes, indicating binding within promoter regions and raising the possibility of interactions with other transcription factors ( Figure 9). To determine if other transcription factors might bind within the 61.5 kb regions centered about the annotated EVI1 DNA binding sites, we performed an analysis using the MATCH program and TRANS-FAC database [57]. In DA-1 leukemic cells, 79 transcription factors were found to share binding within the promoter regions of EVI1 target genes (p,0.05). In NFS-60 leukemic cells, 67 had shared binding (p,0.05). Sixty two of the same transcription factors were identified to be present in both EVI1 leukemic cell lines. Of these ELK1, an ETS-like transcription factor was found to significantly share binding with EVI1 promoter regions (p = 0.003).

Integrated Functional Pathway Analysis
To determine the important biological pathways involved with genome-wide EVI1 transcription factor binding in Evi1 overexpressed leukemic cells, DAVID analysis was performed for the 8565 annotated genes significantly associated with EVI1 peaks. The most significant KEGG pathway based on global EVI1 binding were Pathways in cancer (p = 2.5E 215 ), followed by Jak-Stat signaling (p = 2.3E 211 ), Mapk signaling (p = 9.1E 29 ) and Chemokine signaling (p = 1.6E 27 ). Direct EVI1 target genes were also significantly enriched for KEGG pathways specific for Acute Myeloid Leukemia (p = 5.4E 26 ), Apoptosis (p = 3.9E 24 ), Hematopoietic Cell Lineage (p = 3.5E 25 ) and p53 signaling (p = 2.1E 23 ).
DAVID analysis was also performed for the 4,585 annotated genes associated with an EVI1 ChIP-Seq peak with an ETS-like binding motif within its promoter region. Jak-Stat signaling was the most significantly enriched KEGG pathway associated with the annotated genes harboring an AGGAAG ETS-like motif (p,.6E 27 , Figure 10). EVI1 bound to the promoter regions of 78% of the major genes involved in the Jak-Stat pathway.
Gene set enrichment analysis (GSEA) using curated gene sets from published genomic studies (GSEA Modules C2-C5) was performed to identify distinct molecular signatures for the global EVI1 gene targets. Only genes with significant EVI1 binding sites and deregulation of mRNA transcription were used as input data for the analysis. GSEA revealed these genes were significantly associated with signatures only involving cancer or cancer-oriented genes (C4 Module) (p,0.001, FDR ,0.1).

Discussion
The ecotropic virus integration site 1 (EVI1) is an oncogenic transcription factor associated with a wide range of human malignancies including AML. EVI1 is an independent biomarker that confers poor prognosis in AML. We report here the first genome-wide study of EVI1 DNA binding sites in leukemic cells. We confirmed EVI binding to and deregulation of a select number of previously-reported EVI1 downstream gene targets (Table 3), but not others (Table 4). We also identified novel EVI target genes involved in terminal myeloid differentiation, cell cycle regulation and apoptosis previously unreported in EVI1-induced leukomogenesis. Furthermore, we found the majority of significant EVI1 binding sites contained an ETS-like motif.

EVI1 Binds and Deregulates a Major Terminal Myeloid Differentiation Gene
C/EBP-e (CCAAT/enhancer binding protein-e) is a well established regulator of myeloid lineage differentiation and is critical for the terminal differentiation of granulocytes [58,59,60]. Seven significant EVI1 binding sites, 2 of which were within the promoter region, were identified for Cebpe. This was associated with a 2-fold downregulation of Cebpe in both the Evi1 overexpressed leukemic cell lines. Unlike other C/EBP family proteins, Cebpe expression is restricted to hematopoietic cells, and its activation is associated with terminal differentiation of neutrophils and eosinophils [58]. Koeffler et al demonstrated Cebpe knockout mice exhibit neutrophils blocked at the myelocytes and metamyelocytes stage. Clonogenic assays revealed a significant decrease in the number of myeloid colonies, and a significant increase in Lin-Sca1+c-Kit+ colonies [59]. The Yale group showed neutrophils with Cebpe knockout have bilobed nuclei, lack secondary granules and mRNA for secondary granule proteins, and exhibit aberrant chemotaxis [61].
As a master regulator of terminal myeloid differentiation, C/ EBP-e binds and activates several downstream gene targets to produce mature granulocytes. To generate a mature neutrophil, a series of committed steps occur from the pluripotent hematopoietic stem cell, which differentiates into the myeloblast, promyelocyte, myelocyte, and finally the band stage. The presence of secondary   16,745 significant peaks, defined as a difference in number of reads between the control rabbit IgG and EVI1 antiserum antibody yielding a p-value ,0.001, were identified based on a Poisson distribution. a) Of the 16,745 generated significant peaks (p,0.001), 45.5% were within introns, 35.0% within distal intergenic region, 7.1% were within the proximal (within 1kb) of the TSS. b) A 500bp DNA sequence extracted around each significant peak was matched to de novo consensus sequence discovery programs. The AGGAAG ETS-like motif was identified and refined in 88% of the significant EVI1 ChIP-Seq binding sites. doi:10.1371/journal.pone.0067134.g006 granules marks the transition from the promyelocyte to the fully committed myelocyte stage [62]. Secondary granule protein genes such as lactoferrin, transcobalamin I, neutrophil collagenase, and neutrophil gelatinase are direct targets of C/EBP-e [63,64]. We identified several downregulated C/EBP-e downstream gene targets in EVI1 leukemic cells. In both Evi1 overexpressed leukemic cell lines, expression of neutrophil collagenase (Mmp8) and neutrophil gelatinase-associated lipocalin (Lcn2) were significantly reduced. In the DA-1 leukemic cells, 2 major genes involved in eosinophil maturation (Epx and Prg2), were also significantly downregulated. We identified at least 6 different downstream C/ EBP-e direct target genes to be downregulated in EVI1-induced leukemic cells. These results suggest it is unlikely that EVI1 directly regulates critical genes involved in myeloid differentiation individually, but binds to and downregulates a master regulator. To our knowledge this is the first report of Cebpe deregulation in EVI1-induced leukemia.

Deregulation of Jak-Stat Signaling in EVI1 Leukemia
Global biological function analysis using all significant EVI1 binding gene targets revealed the Pathways in cancer and Jak-Stat signaling pathways were most aberrant. Given a surprising 88% of the EVI1 binding sites contained an ETS-like AGGAAG binding motif, we repeated the analysis using only EVI1 gene targets with the motif. This revealed the Jak-Stat signaling was the most significantly enriched KEGG pathway. We found EVI1 significantly binds to the promoter region of a remarkable 50 gene targets involved in the Jak-Stat signaling pathway ( Figure 10). Of these 50 genes, expression levels of 10 were significantly aberrant.
Jak-Stat signaling is one of the principal mechanism by which extracellular signals, specifically cytokines and growth factors, are translated into intracellular responses [65]. Various ligands such as erythropoietin, growth hormones, interferons and interleukins bind their cognate receptors which are associated with JAK tyrosine kinases (JAK1, JAK2, JAK 3 and TYK2) [66,67]. Upon ligand binding, JAKs are transphosphorylated and subsequently phosphorylate latent STAT transcription factors in the cytoplasm. Phosphorylated STATs enter the nucleus and activate or repress gene targets critical for cellular differentiation, proliferation and death [68]. STAT transcription factors are regulated through various inhibitory factors, including the suppressor of cytokine signaling (SOCS) proteins [69].
Excessive Jak-Stat signaling activation results in numerous inflammatory diseases and hematopoietic disorders such as essential thrombocythemia, polycythemia vera, myelofibrosis and leukemias [66,67]. JAK2 mutations which induce auto-activation of STAT proteins have been well documented in AML [67]. Constitutive activation of STAT 1, 3 and 5 in proliferating human AML blasts have also been reported [70]. We identified Socs1, which encodes for an inhibitor of STAT transcription factors, was significantly downregulated by 5.7-fold in DA-1 EVI1 leukemic cells, and by 4.4-fold in NFS-60 EVI1 leukemic cells. We identified 8 significant EVI1 DNA binding sites for Socs1, 3 of which were within the promoter region. Two significant EVI1 binding sites were also identified for Socs3, but not for Socs2. Interestingly, we also found EVI1 significantly binds to and overactivates Stat1 and Stat5 genes in one of the Evi1 overexpressed murine cell lines (NFS-60). We therefore examined if phosphorylated STAT1 protein was increased in two separate human hematopoietic cell lines with verified Evi1 overexpression (Kasumi 3 and U937+Evi1). We found an increased level of endogenous STAT1 protein phosphorylation in Kasumi 3 Evi1 overexpressed leukemic cells. However, we also noted a marked elevation of total STAT1 protein in these cells, which was consistent with our mRNA findings. Given the baseline level of total STAT1 was much higher in Evi1 overexpressed leukemic cells, it is unclear at this point if EVI1 directly overactivates Jak-Stat signaling via STAT activa-  tion. Although there is a clear interaction between EVI1 and the Jak-Stat pathway, further studies are necessary to elucidate potential mechanisms.
Osm (oncostatin M), which encodes for a cytokine in the interleukin 6 family, was also significantly downregulated in our EVI1 leukemic cells. The role of OSM in malignancy remains unclear. Yoshimura et al demonstrated Osm is a downstream target of the Jak-Stat pathway [71], transcriptionally induced by cytokines that specifically activate STAT5. OSM has been reported to act as a growth factor in myeloid neoplasms and has also been shown to inhibit proliferation of numerous malignant cell lines, including murine M1 myeloid leukemic cells [72]. OSM also induces differentiation of M1 monocytic leukemia cells and suppresses embryonic stem cell function [73]. We identified 7 significant EVI1 binding sites for Osm, 6 which were within the promoter region. EVI1 binding was associated with a significant decrease in transcription in both DA-1 and NFS-60 leukemic cells (13-fold and 1.8-fold decrease, respectively). This suggests downregulation of Osm may have an important role in failure of myeloid differentiation in EVI1-induced leukemogenesis.
We also identified a significant increase in Ube1l expression in both EVI1 leukemic cell lines (DA-1 cells 2.5-fold and NFS-60 cells 2-fold upregulation). UBE1L is an E1 ubiquitin-like enzyme that is activated at the transcriptional level by type I interferons. UBE1L is required for the conjugation and function of interferon stimulating gene 15 protein (ISG15) [50], a critical modifier of Jak-Stat pathway proteins [51]. Isg15 is one of the strongest genes induced by type I interferons in response to cellular stress and infection. Upregulation of ISG15 activity has been associated with several cancers [74,75]. UBE1L E1 enzyme charges ISG15 by forming a thiolester intermediate suitable for transfer to the UBCH8 E2 enzyme [76]. Cong et al demonstrated multipotent hematopoietic progenitor cells from Ube1L deficient mice exhibit a G2/M phase block and delay in cellular proliferation, without an effect on survival or differentiation functions [77]. We identified 2 significant EVI1 DNA binding sites for Ube1l, both of which were within the promoter region, and associated with a significant increase in Ube1l expression in both EVI1 leukemic cell lines. These results suggest EVI1 leukemic cells may harbor sensitivity to cellular stress or inflammatory responses, resulting in uncontrolled cellular proliferation mediated by aberrant UBE1L-ISG15 activation.

Serpinb2 Downregulation in EVI1 Leukemia
Serpinb2, which encodes for a serine protease inhibitor, was significantly bound by EVI1 and downregulated by .10-fold in both Evi1 overexpressed leukemic cell lines. Serpinb2 encodes for plasminogen activator inhibitor (PAI-2), a coagulation factor that inhibits tissue plasminogen activator and urokinase. PAI-2 exists in a secreted, extracellular glycosylated form and an unsecreted intracellular form. PAI-2 is present in monocytes and exists predominantly in the cell cytosol as a 47 kDa nonglycosylated intracellular form [78]. However the intracellular role of PAI-2 is still being established [79].
Alternatively, the decrease in Serpinb2 expression found in EVI1 leukemic cells may be a marker of reduced differentiation in immature myeloid cells. PAI-2 gene activation has been associated with monocyte differentiation in U-937 monocyte-like cells [85]. Suppressed Serpinb2 expression may be a reflection of EVI1induced inhibition of myeloid differentiation.
The PAI-2 promoter is tightly regulated under the control of an upstream silencer element (located between 21977 and 21675, termed PAUSE-1) and a repressor element (between 2219 and 21100) [86]. We identified a very prominent EVI1 binding site which lies directly within the Serpinb2 silencer element (position 21745), suggesting EVI1 can potentially disrupt or alter normal binding and function of PAUSE-1 transcription factors. A 67kDa PAUSE-1 BP complex has been shown to bind the silencer element. However, cooperative DNA-binding partners have yet to be identified and may be an area for future study. Additionally, AP1-like elements, AP1a (position 2103 to 297) and AP1b (position 2114 to 2108) have been identified to bind to regulatory elements of Serpinb2 and induce transcriptional regulation [86]. We have shown EVI1 binds Serpinb2 to reduce its expression. Bard et al previously demonstrated AP1 physically interacts with EVI1 and frequently shares promoter binding to putative target genes [28]. Collectively, these results suggest the EVI/AP1 may bind Serpinb2 as a complex to reduce expression and increase cellular proliferation in leukemic cells.

Disruption of Apoptosis Mediated by Downregulation of ATP-Dependent Purinoceptors
We identified significant downregulation of several genes that encode for ligand gated P2 purinoreceptors, specifically P2rx3, Prx4, and P2rx7 in EVI1 leukemic cells. P2rx7 was of particular interest, given its well established role in regulating apoptosis in macrophages [87,88,89]. P2RX7 is a cell surface ATP receptor involved in rapid cell death via calcium influx, and is primarily expressed in macrophages and neutrophils [87]. The ionotropic ligand gated channel is activated by graded doses of ATP which induces reversible permeabilization of the plasma membrane. After channel opening, calcium influx and rapid depolarization [90,91] leads to a signaling cascade that have been linked to superoxide-mediated mechanisms [89,92]. Suh et al demonstrated that P2RX7 activation is coupled to the generation of superoxides in human neutrophils [89]. However, the mechanism by which the superoxide production cascade occurs remains unclear.
Previous studies have also shown P2RX7 activation results in release of interferon-1b, accumulation of transcription factors that mediate apoptosis, specifically NFAT and NFKb [93,94], and macrophage cell death [87]. P2RX7 activation has also been associated with increased caspase-1 and caspase-3 activity [88]. Caspase inhibitors have shown to inhibit P2RX7-induced NFKb activity [92]. Humphreys et al demonstrated P2RX7 stimulation with ATP rapidly elevates caspase-3 protease activity associated with DNA fragmentation, and is also strongly linked to upregulation of the c-Jun N-terminal kinase pathway [95]. Failure of apoptosis due to P2 purinoreceptor dysfunction has been implicated in previous studies [96,97]. We report here that EVI1 binds to 3 sites within the P2rx7 gene promoter region with significant reduction of P2rx7 transcription leukemic cells. Our study provides evidence for a potential new mechanism of apoptotic deregulation in the development of AML via ion channel regulation.

EVI1 Significantly Binds to an ETS-like Binding Motif
We identified 14,672 ChIP-Seq peaks (88%) with an AGGAAG ETS-like motif. Over 4,500 peaks with this motif were within promoter regions of an annotated gene. Our results are consistent with the only other reported EVI1 ChIP-Seq study, which was performed in human ovarian cancer cells. Their study demonstrated over 5,000 significant EVI1 peaks contained an ETS-like binding motif [28] [28]. The ETS family includes 28 transcription factors in the mouse and has been reported to be important in tissue development and cancer progression [98].
Shared transcription factor analysis revealed the ETS-like transcription factor ELK1, significantly occupied binding sites with EVI1 promoter regions. ELK1 is one of the most studied ETS-like transcription factors [99] and has been implicated in several malignancies, including bladder, breast, esophageal can-  cers and glioblastoma [99,100,101,102,103]. Interestingly, a recent ELK1 ChIP-Seq study demonstrated ELK1 binds to redundant DNA regions in cooperation with another ETS-like transcription factor, GABPA [104]. However, regions that are occupied by ELK1 but not GAPBA were defined as unique regions associated with gene expression of critical cellular functions. Putative ELK1 competition with GABPA, and potentially other ETS proteins, presents an interesting area for additional study.
In summary, these findings represent the first global genomewide study of EVI1 DNA binding associated with whole transcriptome expression analysis. Our results reveal several important genes with an ETS-like binding motif, is involved in terminal myeloid differentiation, cell cycle regulation and apoptosis [ Figure 11]. The Jak-stat pathway and response to inflammatory and stress conditions were notably aberrant. We have previously shown that small molecule inhibitors against EVI1 gene targets can be designed to successfully block its binding [20]. This study provides a list of critical genes that can be targeted for future anti-leukemic therapies. We demonstrate that several gene targets work in concert to drive leukemogenesis. This suggest a cocktail of inhibitors targeting a select number of DNA sites, rather than a drug targeting an isolated gene, may be a more promising approach for developing a cure for EVI1-induced leukemogenesis.

Materials and Methods
shRNA Suppression of Evi11 and RNA-Sequencing Evi1 expressing AML cell lines (DA-1 [105]and NFS-60 [106]) were transduced with retrovirus expressing EVI1 targeting shRNAs (sh11 and sh54) and OFF-targeting shRNAs (shScr and shLuc). After 48 hrs cells were selected for puromycin-resistance for 48 hrs and purified on Ficoll-Paque PLUS gradient (GE healthcare Bio-Sciences AB; Uppsula Sweden). Ficoll-buoyant cells were harvested and Western blot analyses of protein extracts from DA-1 and NFS-60 cells transduced with control (shScr) or anti-Evi1 (sh11) was completed using anti-EVI1 serum. Total RNA was prepared with Qiagen RNeasy plus kit with QIAshredders

Evi1 Overexpression in Human Hematopoietic Cell Lines
Kasumi 3 cells, an established human AML cell line with 3q27 translocation and Evi1 overexpression were purchased from ATCCH and expanded in RPMI-1640 media with 10% FBS. Kasumi 1 cells, a leukemic cell line with 8q21 translocation and without Evi1 overexpression was used as a control to demonstrate downregulation of significant genes. U937 histiocytic lymphoma cells purchased from ATCCH were retrovirally transduced with an Evi1 construct to overexpress EVI1. T293 cells were transfected using the Gag Pol and VSV viral packaging system. Forty eight hours post tranfection the viral supernatant was used to infect U937 wildtype cells. A Western blot was performed to confirm overexpression of EVI1 ( Figure 12). U937 cells with Evi1 overexpression were sorted using flow cytometry using the green fluorescent protein selection marker.

Western Blot in Human Leukemic Cells
Immunoblot analysis using the EVI1 C-terminal antibody was performed. Approximately 3 million cells were lysed in ice cold RIPA buffer with protease and phosphatase inhibitors. A Bradford assay was used to calculate protein concentrations for each sample (Kasumi 1, Kasumi 3, U937 wildtype. U937+ Evi1). Samples were diluted in 5X SDS buffer for equal concentration gel loading. Beta actin was used as a gel loading control ( Figure 4). Protein transfer to a nitrocellulose membrane was completed as previously described. Primary antibodies included anti-phosph-STAT1 (Cell SignalingH, 1:1000 dilution), anti-STAT1 (BD TransductionH, 1:1000 dilution) and anti-beta-actin (Cell SignalingH, 1:1000 dilution). Conjugated HRP sheep-anti mouse and donkey-anti rabbit secondary antibodies were used (GE HealthcareH). The BioradH ChemiDoc TM XRS+ System and Image Lab TM software were used to detect chemiluminescence.

ChIP-Seq and PCR
ChIP was completed using a lyophilized Staph A Cell (PansorbinH) Assay Protocol as previously described [107]. Antisera specific for the N-and C-terminus of the murine EVI1 protein (GenBank TM accession number M21829) were generated using purified protein as previously described [23]. Immunoprecipitation was performed using EVI1 antisera, with no antibody and rabbit IgG as negative controls. Confirmation of genomic DNA for selected EVI1 binding target genes was completed with quantitative PCR reactions using the appropriate positive and   negative control primers (Table 5). Input DNA libraries were prepared and sequenced using the SOLiD TM System (Applied Biosystems) platform [53,54]. PCR reaction primers for cDNA expression in human Kasumi1, Kasumi3, U937 wildtype and U937+Evi1 are listed in Table 6.

Computational and Functional Analysis
The ChIP-Seq read were mapped to a reference genome by bowtie program, allowing two mismatches. Multiple aligned reads were removed from the analysis. The aligned sequence reads were displayed as a track onto the mouse reference genome using the University of California at Santa Cruz (UCSC) genome browser (http://genome.ucsc.edu/index.html) for visual inspection. To determine where the EVI1 bound to the genome, we applied the MACS program to look for areas where there were significantly more enriched reads mapped in the ChIP sample. The Stanford Genomic Regions Enrichment of Annotations Tool (GREAT v1.2.6) was used to assign significant peaks (non-coding genomic regions) to annotated genes in cis [56]. A 500bp DNA sequence was extracted around each peak and run against MEME and TPD programs to identify potential EVI1 consensus binding motifs. The MATCH program in the TRANSFAC database was used to perform the search for the enrichment of other transcription factors binding sites (TFBS) in the 6250 bp regions centered on the annotated EVI1 binding sites. We set up a filtering step utilizing the cross-species conservation information. The conservation score for each TFBS was evaluated to be the average phastCons score within the TFBS region. A cut-off of 0.5 for the conservation score was applied here and Fisher's exact test was used to detected statistical significance.
Integrated functional pathway analysis using DAVID and GSEA were completed for gene lists generated from significantly up and downregulated transcripts with EVI1 DNA binding sites. For GSEA, a hypergeometric distribution was used to determine enrichment score (FDR ,0.10) when compared to curated GSEA gene sets C2-C5. C2-C5 collection of gene sets included data from over 340 PubMed articles and online databases from Biocarta, Gene array, BioScience Corp, KEGG, Reactome, Sigma-Aldrich Pathways, Signal Transduction Knowledge Environment and Signaling Gateway [108].

Supporting Information
Dataset S1 RNA-Seq data in Evi1 overexpressed myeloid leukemic cells (DA-1, NFS-60) and in shRNA Evi1 knockdown cells. To identify genes differentially expressed between Evi1 overexpressed myeloid leukemic cells (DA-1, NFS-60) and in shRNA Evi1 knockdown cells, RNA was extracted to generate transcriptome-wide expression profiles. High throughput parallel sequencing revealed 806 significantly deregulated