EVI1 is an oncogene inappropriately expressed in the bone marrow (BM) of ∼10% of myelodysplastic syndrome (MDS) patients. This disease is characterized by severe anemia and multilineage myeloid dysplasia that are thought to be a major cause of mortality in MDS patients. We earlier reported on a mouse model that constitutive expression of EVI1 in the BM led to fatal anemia and myeloid dysplasia, as observed in MDS patients, and we subsequently showed that EVI1 interaction with GATA1 blocks proper erythropoiesis. Whereas this interaction could provide the basis for the erythroid defects in EVI1-positive MDS, it does not explain the alteration of myeloid differentiation. Here, we have examined the expression of several genes activated during terminal myelopoiesis in BM cells and identified a group of them that are altered by EVI1. A common feature of these genes is their regulation by the transcription factor PU.1. We report here that EVI1 interacts with PU.1 and represses the PU.1-dependent activation of a myeloid promoter. EVI1 does not seem to inhibit PU.1 binding to DNA, but rather to block its association with the coactivator c-Jun. After mapping the PU.1-EVI1 interaction sites, we show that an EVI1 point mutant, unable to bind PU.1, restores the activation of PU.1-regulated genes and allows a normal differentiation of BM progenitors in vitro. [Cancer Res 2009;69(4):1633–42]
The inappropriate expression of EVI1 (ecotropic viral integration site 1) has been associated with aggressive myelodysplastic syndrome (MDS) that often progresses to acute myeloid leukemia (AML) and, less frequently, with chronic myelogenous leukemia (1–4). MDS patients with activation of EVI1 are characterized by a poor outcome due to multilineage dysplasia, severe anemia, and general cytopenia (5).
EVI1 is a 145-kDa nuclear protein with two zinc finger domains containing respectively seven and three zinc finger motifs. EVI1 is not normally detected in hematopoietic cells, and its ectopic expression in bone marrow (BM) progenitors leads to an alteration of granulopoiesis and a block of erythropoiesis in response to cytokines in vitro (6–8). Several studies have been carried out with murine hematopoietic cell lines; in particular, it was shown that EVI1 blocks the granulocytic differentiation of the 32Dcl.3 cells in response to granulocyte colony-stimulating factor (G-CSF; ref. 6) and that overexpression of Sox4 inhibits cytokine-induced granulocyte maturation by induction of EVI1 expression (9). More recently, we showed that a single zinc finger of EVI1 blocks the G-CSF–stimulated 32Dcl.3 differentiation by interacting with RUNX1 (7). Taken together, these studies indicate that the effect of EVI1 on this cell line is complex and may involve several pathways.
To study the role of EVI1 in a normal environment rather than in a cell line, we earlier expressed this oncogene in the BM of transplanted mice and found that the reconstituted animals developed fatal anemia and cytopenia and had defects in terminal myelopoiesis, as seen in MDS patients (10). As we reported later, the erythroid defects seem to be due to the inactivating binding of EVI1 to GATA1, which impairs the expression of the erythropoietin receptor and, therefore, the response of the cell to erythropoietin (8). However, although the studies with the 32Dcl.3 cell line strongly suggest that EVI1 could affect myelopoiesis in normal progenitors, it still is not clear which pathways are altered by this oncogene.
Here, we have evaluated the potential effect of EVI1 on PU.1, an Ets family member required for the commitment of alternative hematopoietic lineages (11). Indeed, mice lacking PU.1 do not make mature macrophages, functional granulocytes, B cells, and T cells (11–15). It was proposed that the choice of hematopoietic progenitors for a specific lineage is determined by graded levels of PU.1 and that high or medium PU.1 level favors respectively macrophage or granulocyte differentiation, whereas lower concentrations are sufficient for B-cell development (16–20). The regulation of myeloid or lymphoid promoters by PU.1 is modulated by its interaction with other transcription factors, such as GATA-1, C/EBPs, RUNX1, c-Jun, IRFs, and B cell–specific activator protein (21–25), suggesting that the role of PU.1 in the hematopoietic lineage commitment is determined not only by PU.1 concentration but also by interacting partner factors.
In this paper, we show that the inappropriate expression of EVI1 in BM progenitors leads to the impaired differentiation of myeloid cells both in vitro and in vivo by repression of a subset of PU.1-regulated genes that are normally activated during terminal myelopoiesis. Our data show that two contiguous zinc finger motifs within the EVI1 proximal domain interact with the winged helix-turn-helix motif in the Ets domain of PU.1 and abolish its association with the cofactor c-Jun. This suggests that the EVI1-PU.1 interaction deregulates PU.1 function by blocking the association with coactivating partners rather than by displacing PU.1 from the DNA. The disruption of the two zinc finger motifs by point mutations destroys the interaction with PU.1 and allows the proper differentiation of murine BM progenitors in response to granulocyte macrophage colony-stimulating factor (GM-CSF) and G-CSF in vitro. Our results provide yet an additional mechanism by which EVI1 affects myelopoiesis, further explaining the complexity of this oncogene and the difficulty in identifying the precise pathways that need to be corrected to reestablish normal differentiation.
Materials and Methods
Cell culture and transfection. The adherent cell lines 293T, NIH3T3, and Phoenix (American Type Culture Collection) were maintained in DMEM supplemented with 10% calf serum (Life Technologies) and were transfected with the calcium phosphate precipitation method or with Escort V reagent (Sigma-Aldrich) according to the manufacturer's instructions. Growth and infection of 32Dcl3 cells were performed as described (6).
DNA cloning. The pECE-PU.1 plasmid was a gift of Dr. Ackermann (University of Illinois at Chicago). The cDNA encoding Flag-EVI1, HA-EVI1, HA-EVI1(283-1051), HA-EVI1(1-283), HA-EVI1(1-158), HA-EVI1(159-283), HA-EVI1(Δ4-5), PU.1, and its deletion mutant PU.1(Δ210) were all cloned in the BamHI/XhoI sites of the pCMV/myc/nuc vector (Invitrogen). PU.1 was also cloned in the pBK-CMV vector under T3 promoter for in vitro translation. To generate glutathione S-transferase (GST) fusions, EVI1 was subcloned in frame to GST into the BamHI/SmaI sites of pGEX-2T (Amersham). The EVI1 mutant HA-EVI1(6+7Mut) containing four-point mutations (C190A, C193A, C219A, and C222A) was generated by PCR. HA-EVI1 and HA-EVI1(6+7Mut) were subcloned in the EcoRI/BglII sites of the pMSCV vector (Clontech). All PCR reactions were performed with high-fidelity Pfu-DNA polymerase (Stratagene). All cloning junctions and PCR-generated fragments were verified by DNA sequencing.
Western blot and coimmunoprecipitation. Cells were harvested 48 h after transfection, and the proteins in the lysates were immunoprecipitated by incubation with anti-PU.1 antibodies (Santa Cruz) for 4 h at 4°C, then with protein A-sepharose beads (Invitrogen) for 1 h at 4°C. For Western blot analysis, the beads were washed thrice for 5 min at 4°C, and the proteins were separated by SDS-PAGE.
GST fusion pull-down assay. The expression and purification of GST and GST-EVI1 and their interaction with PU.1 were carried out as described (26). In vitro translated PU.1 was generated by using TNT Coupled Reticulocyte Lysate System (Promega), according to the manufacturer's instructions.
Reporter gene assays. We used PCR to amplify the regions from nucleotide −416 to nucleotide +71 of the human M-CSFR promoter. After DNA sequencing, the fragment was cloned into XhoI/HindIII sites in the luciferase reporter vector pGL3 (Promega). For normalization of transfection efficiency, we used pRL-TK plasmid (Promega) that expresses Renilla luciferase. NIH3T3 cells were transiently cotransfected with 4 μg of M-CSFR-Luc reporter gene and the effector plasmids [4 μg of PU.1 and 8μg of EVI1 or EVI1(6+7Mut)]. All measurements were done in triplicate, and the experiments were repeated thrice.
Quantitative real-time PCR. Total cellular RNA was extracted from 32Dcl.3 and BM cells using TRIzol reagent (Life Technologies BRL) according to the manufacturer's protocol. The cDNA was prepared according to the First Strand cDNA Synthesis kit protocol (Fermentas). Quantitative real-time PCR (RQ-PCR) was performed in a 25-μL reaction containing 2 μL of cDNA, 12.5 μL of the iQ SYBR Green Supermix (Bio-Rad Laboratories), and 300 nmol/L primers. RQ-PCR amplification was performed on the iCycler-iQ detection system, and the data were collected and analyzed using iCycler iQ version 3.0 software (Bio-Rad).
Oligo GEArray. The mouse Hematopoiesis Oligo-GEArray was purchased from SABiosciences. The analysis was performed with the RNA extracted from BM cells infected with the empty vector, EVI1, or EVI1(6+7Mut). This kit uses a linear RNA amplification and labeling procedure to synthesize labeled antisense cRNA. The GEArray membrane was hybridized overnight at 60°C with the cRNA. After extensive washings, the chemiluminescent detection kit was applied to the membrane for 1 min. Multiple images were collected with an AutoChemi System (UVP). The intensity of each spot was quantified and normalized to three internal standards.
Chromatin immunoprecipitation. Chromatin immunoprecipitation (ChIP) analysis was performed with 32Dcl.3 cells infected with the empty vector or expressing EVI1. For analysis of PU.1 occupancy on the M-CSFR promoter, the proteins were cross-linked to the DNA with 1% formaldehyde. An anti-PU.1 antibody (Santa Cruz) was used with protein G-sepharose to adsorb and immunoprecipitate the protein-DNA complexes. The beads were washed and eluted, and the cross-linking of protein/DNA complexes was reversed by heating at 65°C for 8 h. DNA was extracted with phenol/chloroform and precipitated. The DNA fragments were analyzed by PCR.
Colony formation assay. Lineage negative (Lin−) murine hematopoietic progenitors were isolated from the BM of C57BL/6J mice, as previously described (27). Retroviral particles encoding full-length or mutant EVI1 were prepared by transfection of the Phoenix packaging cells with the plasmids pMSCV-EVI1, pMSCV-EVI1(6+7Mut), or with the empty retroviral vector. The day after the culture medium containing the retrovirus was collected and used for spinoculation of Lin− BM cells in the presence of polybrene (4 μg/mL). On day 2, the spinoculation process was repeated, and on day 3, the cells were counted and plated in methylcellulose with cytokines (SCF, 100 ng/mL; IL-3, IL-6, and GM-CSF, 10 ng/mL) and neomycin (1 mg/mL) for selection. After 7 days, the cells were replated with GM-CSF or G-CSF. Seven days later, the colonies were counted and the morphology of the cells was analyzed after Wright-Giemsa staining.
EVI1 impairs myeloid differentiation in vivo. The peripheral blood (PB) and the BM smears obtained from our EVI1-positive mice (10) were reevaluated more carefully. We determine that 75% of the neutrophils were hypersegmented and that 20% of the total monocytes were dysplastic. These aberrant features were already present in the PB of the mice 6 months after BM transplantation. The maturative defects were observed only in myeloid cells (Fig. 1A,, I and II), whereas the lymphoid cells in the PB of the EVI1-positive mice seemed normal (Fig. 1A,, III). Nucleated cells in the PB of the control mice looked normal as expected (Fig. 1A,, IV–VI). These defects were also observed in BM smears, which revealed a high number of hypersegmented neutrophils (75% of the total neutrophils) in the EVI1-positive mice at time of death (Fig. 1B,, left), but not in the controls (Fig. 1B,, right). The aberrant traits described for the EVI1 mice strongly resemble the abnormal elements found in the PB of the EVI1-positive MDS patients (Fig. 1C , I and II), suggesting that in vivo EVI1 not only blocks erythroid differentiation but it can also deregulate the maturation of myeloid cells.
EVI1 impairs myeloid differentiation in vitro. To determine whether these defects can be reproduced in vitro, Lin− cells isolated from murine BM were infected with a recombinant EVI1-expressing retrovirus or with the empty retroviral vector, as described (8, 27). After selection with G418, equal numbers of cells were plated in semisolid medium in the presence of GM-CSF or G-CSF. After 7 days in culture with either cytokine, the cells were collected for morphology analysis. Cytospin preparations of EVI1 cells stimulated with either GM-CSF or G-CSF were very similar. Differential cell count showed that most of the cells (85%) were promyelocytes or earlier precursors and very few mature neutrophil or monocytes were observed (less then 15%). In addition, more than half of the entire cell population had dysplastic aspects, including nuclear cytoplasmatic maturative asynchronization and aberrant mitosis. These abnormal promyelocytes were characterized by an unusual presence of very large aberrant granules (Fig. 1D,, left). These abnormal features were not present in the control cells that had the normal appearance of all the maturative stages of the myeloid lineage, including mature neutrophils and monocytes (Fig. 1D , right).
EVI1 down-regulates a set of genes required for myeloid differentiation. We used RQ-PCR to determine whether the expression of a group of genes specifically activated during myeloid differentiation is affected by EVI1. Because we and other investigators have shown that 32Dcl.3 cells have a block to G-CSF–induced granulocytic differentiation when EVI1 is expressed (6, 7), we stimulated control and EVI1-expressing 32Dcl.3 cells with G-CSF for 4 days only. After this time, RNA was extracted, and the expression of several genes was quantified relative to the endogenous expression of β-actin. We tested neutrophil cytochrome b/gp91phox (Cybb), myeloperoxydase (Mpo), lactoferrin (Ltf), cathepsin G (Ctsg), M-CSFR (Csf1r), lysozyme (Lyzl6), neutrophil gelatinase/lipocalin2 (Lcn2), and cathelicidin antimicrobial peptide (Camp). All these genes are activated during myeloid differentiation (6, 28–30). The results (Fig. 2A) indicate that the transcription of Cybb, Mpo, Ltf, Ctsg, Csf1r, and Lyzl6 were significantly reduced in EVI1-positive 32Dcl.3 cells. To confirm the effects of EVI1 in primary cells, RQ-PCR assays were repeated with RNA extracted from BM progenitors infected with EVI1− or the empty retrovirus and cultured in the presence of GM-CSF or G-CSF. As we noted for 32Dcl.3 cells, in response to G-CSF, EVI1 down-regulated Cybb, Mpo, Ltf, Ctsg, Csf1r, and Lyzl6, but it also reduced the expression level of Lcn2 (Fig. 2B). We did not find a significant difference in gene expression when the cells were cultured with either GM-CSF or G-CSF (data not shown). Although RQ-PCR was chosen because of its detection sensitivity, sequence specificity, and reproducible quantification compared with DNA microarrays (31), we also performed a larger gene expression analysis. We hybridized the RNA extracted from the vector-positive or the EVI1-positive BM progenitors to an array specifically designed for profiling the expression of 113 genes related to the development of all blood cell lineages from progenitor cells. We found that EVI1 deregulated the expression of a subset of genes that control either erythroid or myeloid differentiation. In particular, Spp1, Nos2, and Cbfb were strongly down-regulated in the BM progenitors expressing EVI1 (Fig. 2C). Cbfb encodes CBFβ, the non-DNA-binding partner of CBFA2/AML1/RUNX1 in the CBF complex required for coactivation of Csfr1 and Mpo (32, 33). The osteopontin (Spp1) and the iNOS (Nos2) genes are highly expressed in monocytes and associated with several inflammatory states (34, 35).
EVI1 interacts with PU.1. We thought that the repression of such myeloid lineage–specific genes could be the result of the deregulation of a master myeloid transcription factor rather than the effect of EVI1 directly on each one of these genes. We focused on PU.1 because this factor is critical to myeloid lineage commitment and because Cybb and Csf1r, repressed by EVI1, are also undetectable in PU.1−/− mice (14, 36). First, we determined whether PU.1 levels were reduced in EVI1-positive cells. The results obtained by RQ-PCR in 32Dcl.3 and BM progenitor cells showed that there was no significant difference in PU.1 level when EVI1 was expressed (Fig. 2D,, top). These data were confirmed by Western blotting (Fig. 2D,, bottom). To determine whether EVI1 interacts with endogenous PU.1, we used coimmunoprecipitation assays in naive and EVI1-positive 32Dcl.3 cells. Proteins from the cells were immunoprecipitated with a polyclonal anti-PU.1 antibody, and EVI1 was detected by Western blotting only in 32Dcl.3-EVI1 cells. The results (Fig. 3A,, lane 4) show that EVI1 interacts with endogenous PU.1. As negative control, cellular proteins were also immunoprecipitated with normal immunoglobulins (Fig. 3A , lanes 1 and 2).
The interaction between EVI1 and PU.1 is direct. We used GST fusion pull-down assay to confirm the association between EVI1 and PU.1 and to determine whether this interaction was direct. Recombinant GST and GST-EVI1 expressed in Escherichia coli were purified using glutathione-sepharose beads, as described (6), and incubated with in vitro translated PU.1 labeled with [35S]methionine. After several washings, the sepharose beads were subjected to SDS-PAGE and the separated proteins were analyzed by autoradiography. The results show an association of PU.1 with GST-EVI1 (Fig. 3B,, lane 3) but not with GST (Fig. 3B,, lane 2). The expression of PU.1 is shown in Fig. 3B (lane 1).
The sixth and the seventh zinc finger motifs of EVI1 bind to PU.1. To determine which domain of EVI1 was responsible for the association with PU.1, we generated four HA-EVI1 deletion mutants. The first mutant, HA-EVI1(283-1051), consists essentially of the entire protein, excluding the proximal zinc finger domain; the second mutant, HA-EVI1(1-283), contains only the proximal zinc finger domain; the third mutant, HA-EVI1(1-158), encodes the first four zinc finger motifs; the fourth mutant, HA-EVI1(159-283), encodes zinc finger motifs 5 to 7. These four deletion mutants, as well as the full-length HA-EVI1, were transiently expressed in 293T cells and tested in coimmunoprecipitation assays to evaluate their capacity to bind PU.1. As shown in Fig. 3C, PU.1 interacts with HA-EVI1 (lane 9), HA-EVI1(1-283) (lane 8), and HA-EVI1(159-283) (lane 6), but not with the deletion mutants HA-EVI1(283-1051) (lane 10) nor with HA-EVI1(1-158) (lane 7), suggesting that the fifth, sixth, and seventh zinc finger motifs of the proximal domain are responsible for PU.1 recognition. Figure 3C (right) shows that the proteins were expressed at comparable levels. To identify the zinc finger motif(s) that mediate EVI1-PU.1 interaction, we introduced Cys-to-Ala mutations in the fifth, sixth, or seventh motifs of EVI1 and tested these mutants in coimmunoprecipitation assays with PU.1. We found that the disruption of the fifth zinc finger motif did not affect EVI1 binding to PU.1; however, mutations of the sixth or the seventh motif significantly decreased the intensity of the coimmunoprecipitation band. These results suggested that perhaps EVI1 recognizes PU.1 with these two motifs and prompted us to disrupt both the sixth and the seventh zinc fingers in HA-EVI1. As shown in Fig. 3D, HA-EVI1(6Mut) and HA-EVI1(7Mut) showed reduced binding to PU.1 (Fig. 3D,, lanes 6 and 7) compared with HA-EVI1 (Fig. 3D,, lane 5). However, the double mutant HA-EVI1(6+7Mut) completely failed to bind to PU.1 (Fig. 3D,, lane 8). Fig. 3D (right) shows that the proteins were expressed at comparable levels.
The COOH terminal region of PU.1 is essential for EVI1 recognition. PU.1 contains three functional domains: the activation domain (amino acids 1–120), the PEST domain (amino acids 120–170), and the Ets domain containing the DNA-binding motif (amino acids 170–254; Fig. 4A,, top). The winged helix-turn-helix motif in the Ets domain of PU.1 (amino acids 210–254) is the region interacting with c-Jun, C/EBP-α, and RUNX1 (22, 32, 37–39). Given the importance of this COOH terminus, we generated the deletion mutant HA-PU.1(Δ210) lacking the distal part of the Ets domain (Fig. 4A,, bottom) and tested this mutant in coimmunoprecipitation assays. The results confirm that EVI1 strongly associates with PU.1 (Fig. 4B,, lane 4) and show that this interaction fails to occur with the deletion mutant PU.1(Δ210) (Fig. 4B,, lane 5), suggesting that the COOH terminus of the Ets domain contains the major interacting region. Figure 4B (right) reports the expression of the proteins.
A double EVI1 point mutant restores M-CSFR promoter activation by PU.1. EVI1 interacts with several transcription corepressors. Previously, we reported that HDAC1 recognizes the proximal zinc finger domain of EVI1 (40), which is the same region that mediates the interaction with PU.1. To determine whether EVI1 inhibits PU.1 function by corepressor recruitment to PU.1 target gene promoters, we first identified the zinc finger motifs that are necessary to interact with HDAC1. We generated EVI1 mutants lacking alternative zinc finger structures and tested the mutants in coimmunoprecipitation assays with HDAC1 (data not shown). We found that the disruption of the fourth and fifth zinc fingers abolishes EVI1 binding to HDAC1 (Fig. 4C,, lane 3). This mutant EVI1(Δ4-5) was still able to bind PU.1 (data not shown). We also determined that EVI1 and the double mutant unable to bind PU.1, EVI1(6+7Mut), strongly interact with HDAC1 (Fig. 4C,, lanes 1 and 2). Taken together, these results suggest that PU.1 and HDAC1 recognize two separate regions of EVI1, and we concluded that recruitment of corepressor to a PU.1-dependent promoter is not the major mechanism of repression by EVI1. To further confirm these results and to determine whether the interaction between EVI1 and PU.1 contributes to the repression of PU.1-dependent genes, we used a reporter plasmid that contains one PU.1-binding site in the M-CSFR promoter linked to the luciferase gene (7, 41). This reporter plasmid and a plasmid expressing PU.1 were used in transient transfections of NIH3T3 cells. The results showed ∼10-fold induction of the reporter gene when PU.1 was added (Fig. 4D,, column 5). When EVI1 or EVI1(Δ4-5), the mutant unable to bind HDAC1, were cotransfected with PU.1, the activation of the promoter was almost completely repressed, suggesting that the interaction of EVI1 with PU.1 inhibits the transactivation of the reporter promoter independently of HDAC1 (Fig. 4D,, columns 6 and 8). When the double mutant EVI1(6+7Mut) was expressed, there was minimal interference with the activity of PU.1 (Fig. 4D,, column 6). Transfection of the cells with EVI1, EVI1(Δ4-5), or EVI1(6+7Mut) alone did not show any significant repression of the M-CSFR promoter (Fig. 4D , columns 2–4). Altogether, these results suggest that EVI1 represses PU.1-activated genes regardless of the HDAC1 presence in the complex and that HDAC1 does not have a major role in the repression of PU.1 function by EVI1.
EVI1 does not affect PU.1-DNA binding. ChIP is a very reliable method to evaluate DNA occupancy in vivo; hence, we used this technique to determine whether EVI1 affects PU.1 binding to DNA. For the assay, we used 32Dcl.3 cells stably transfected with the empty retrovirus or with the EVI1-experssing or the EVI1(Δ4-5)-expressing retrovirus. The ChIP experiments (Fig. 5A) show that anti-PU.1 antibody precipitates endogenous PU.1-chromatin complex in 32Dcl.3-vector cells (Fig. 5A,, lane 4), as well as in 32Dcl.3-EVI1 and 32Dcl.3-EVI1(Δ4-5) cells (Fig. 5A,, lanes 5 and 6). Unrelated purified immunoglobulins were used as a control for the immunoprecipitation (Fig. 5A,, lanes 1–3). To confirm that the amount of the immunoprecipitated chromatin-DNA complex was comparable in each assay, an equal amount of nonimmunoprecipitated DNA was used as control (Fig. 5A , lanes 7–9). Altogether, these results indicate that neither EVI1 nor EVI1(Δ4-5) affect the occupancy of PU.1 on the M-CSFR promoter in vivo, suggesting that the impairment of PU.1 by EVI1 occurs through a different mechanism.
EVI1 blocks c-Jun binding to PU.1. It was shown that c-Jun interacts with the COOH terminus of the Ets domain and enhances the ability of PU.1 to transactivate myeloid promoters, such as M-CSFR (42, 43). PU.1 does not transactivate the M-CSFR promoter in F9 cells, which do not express endogenous c-Jun, unless c-Jun is cotransfected into these cells (39). Our results show that EVI1 binds the region of PU.1 recognized by c-Jun, and we thought that the repression of PU.1 function by EVI1 could occur through the disruption of the activating PU.1/c-Jun complex. Therefore, we used 32Dcl.3 cells that express low levels of endogenous PU.1 and c-Jun to determine whether EVI1 blocks PU.1–c-Jun interaction. As shown in Fig. 5B, in 32Dcl.3-vector cells, c-Jun was detected by Western blotting after PU.1 coimmunoprecipitation (Fig. 5B,, lane 3). However, we did not detect c-Jun in 32Dcl.3-EVI1 cells (Fig. 5B,, lane 4). Figure 5B shows that c-Jun is expressed at a comparable level in 32Dcl.3-vector and 32Dcl.3-EVI1 cells. As negative control, 32Dcl.3 cellular proteins were also immunoprecipitated with normal immunoglobulins (Fig. 5B,, lanes 1 and 2). To confirm these results, we transiently transfected 293T cells with PU.1, EVI1, or a combination of PU.1 and EVI1 to evaluate the capacity of EVI1 to displace the endogenous c-Jun from PU.1 in coimmunoprecipitation assays. When PU.1 was expressed alone, c-Jun could be immunoprecipitated (lane 1). However, when PU.1 was coexpressed with EVI1, the PU.1–c-Jun interaction was disrupted (lane 3). Figure 5C (top) shows protein expression in the cells. These findings suggest that EVI1 could deregulate PU.1 activity by displacement of its coactivator c-Jun.
EVI1(6+7Mut) reestablishes normal myeloid differentiation. To determine whether EVI1(6+7Mut) could bypass the EVI1-induced differentiation block observed in 32Dcl.3 cells in response to G-CSF (6), we generated a 32Dcl.3-EVI1(6+7Mut) cell line. 32Dcl.3 cells expressing EVI1, EVI1(6+7Mut) or the empty vector were cultured in presence of G-CSF. After 3 days, the cells were collected, stained with antibodies that recognize the surface lineage-specific antigens Gr-1 and CD11b, and analyzed by flow cytometry. As expected, half of the 32Dcl.3-vector cells were positive for both markers after this time (Fig. 6A,, left), and because 32Dcl.3-EVI1 do not differentiate, only a minority of the cells expressed both markers (Fig. 6A,, center). In contrast, although the 32Dcl.3-EVI1(6+7Mut) cells were not as advanced in differentiation as the 32Dcl.3-vector cells, they showed an intermediate phenotype, suggesting that disruption of the two zinc finger motifs can partially overcome the EVI1-induced differentiation block (Fig. 6A,, right). To further characterize the effect of this mutant, we quantified the expression of myeloid genes down-regulated by EVI1 when EVI1(6+7Mut), rather than EVI1, was expressed. We therefore repeated the RQ-PCR assays using the RNA extracted from BM progenitors infected with the empty vector EVI1 or EVI1(6+7Mut). The results (Fig. 6B,, left) indicate that the activation of Cybb, Mpo, Ctsg, Csf1r, and Lyzl6 in EVI1(6+7Mut)-positive BM cells was comparable with that of the vector. However, this mutant was not able to rescue Ltf and Lcn2 expression. To evaluate the effect of the mutant EVI1(6+7Mut) on a larger group of genes, we hybridized the RNA extracted from the vector-infected, EVI1-infected, or EVI1(6+7Mut)-infected BM progenitors to the same array described in Fig. 2 and compared the results. We found that the Spp1, Nos2, and Cbfb that were strongly down-regulated by EVI1 were not repressed in the EVI1(6+7Mut)-positive BM progenitors (Fig. 6B,, right). It was shown that Spp1 levels are increased after PU.1 expression (44) and that Nos2 is up-regulated by IRF1 and IRF8 (35), two partner factors that need PU.1 to form the complex that activates also Cybb (45), suggesting a role of PU.1 in the regulation of the myeloid genes that are repressed by EVI1. Finally, to determine the extent at which EVI1(6+7Mut) affects myelopoiesis in vivo, we stably expressed EVI1, EVI1(Δ4-5), EVI1(6+7Mut), or the empty vector in murine Lin− BM cells by retrovirus infection. After a week in culture in presence of GM-CSF or G-CSF, we counted the colonies and recovered the cells to evaluate their potential to differentiate. As we previously reported (8, 27), EVI1 confers a significant increase in clonogenicity in response to GM-CSF, and again, we observed a larger number of colonies generated by expression of EVI1 or EVI1(Δ4-5) compared with the control cells (Fig. 6C,, columns 1–3). However, the number of colonies generated by EVI1(6+7Mut)-positive cells was intermediate (Fig. 6C,, column 4). The morphology was analyzed after Wright-Giemsa staining. The vector cells seemed normal with all maturative stages of the myeloid lineage. In contrast, a large number of immature elements was observed in the EVI1-positive and EVI1(Δ4-5)-positive cell populations (Fig. 6C,, right). However, the BM cells expressing EVI1(6+7Mut) were in an advanced state of differentiation, characterized by large cytoplasm and compact nuclei, and none of the aberrant features induced by EVI1 were observed. Furthermore, as in the vector cells, fully differentiated neutrophils and mature macrophages were observed in EVI1(6+7Mut) cells (Fig. 6C , right). Altogether, these data suggest that the mutant EVI1(6+7Mut) reestablishes the normal function of PU.1, allowing the proper regulation of a subset of myeloid genes leading to the terminal differentiation of the BM progenitors.
Since its identification as a preferential retroviral integration site leading to myeloid tumors in mice, the inappropriate expression of EVI1 has been associated with myeloid, erythroid, and megakaryocytic defects in murine systems (6–8) and with severe human diseases that affect both erythroid and myeloid lineages (46, 47). Indeed, the inappropriate expression of EVI1 in human hematopoietic disorders is one of the worst prognostic factors for patients. In general, EVI1 seems to deregulate the response to cytokines that control lineage commitment and differentiation. Whereas the precise pathways altered by this nuclear oncoprotein are not yet clearly defined, the combined work of several investigators shows that EVI1 interacts with many transcription factors that regulate hematopoiesis, including GATA1 (8), C/EBP-α (37), and RUNX1 (7), providing clues on the complexity of EVI1 inappropriate functions and on the difficulty of identifying the affected pathways.
Among the regulatory hematopoietic factors, PU.1 is perhaps one of the most critical because it controls both lymphopoiesis and myelopoiesis (11, 15, 18), depending on its level of expression, with a lower level favoring B lymphoid commitment and differentiation, and with a higher level favoring the maturation of macrophages and granulocytes (19, 48). The myeloid regulation by PU.1 seems to occur at the terminal differentiation stage rather than at the commitment stage of the precursor cell (49). The studies that we have presented here suggest that EVI1 can disrupt the role of PU.1 on terminal myelopoiesis, leading to the down-regulation of a subgroup of key target genes required for maturation from promyelocyte to neutrophil (30). This is in agreement with the high number of promyelocytes observed in EVI1-positive cells differentiated in vitro. These combined reports are in agreement with our findings that the EVI1-positive mice had apparently normal B cells and only the myeloid lineages were altered.
Our work suggests that EVI1 inactivates PU.1 function by inappropriately interacting with the COOH terminal region of the Ets domain. This conclusion is supported by the results obtained with an EVI1 point mutant unable to bind PU.1 that restores the PU.1-dependent transactivation of the M-CSFR promoter and rescues the myeloid differentiation of the BM progenitors in vitro. The ChIP studies we present here indicate that PU.1 repression by EVI1 probably occurs through a mechanism other than the DNA-binding inhibition described for other transcription factors (7, 8, 37). This conclusion would be in agreement with the weak transactivating potential of PU.1 that requires interactions with cofactors to become fully functional (50). Although this is true also for GATA1, RUNX1, and C/EBP-α, which in absence of cofactors are not strong activators, it was shown that EVI1 interacts with their DNA-binding domain leading to the displacement of the complex from the DNA (7, 8, 37). EVI1 interaction with PU.1 involves the region recognized by c-Jun (42, 43), and the results of our coimmunoprecipitation assays suggest that EVI1 competes with c-Jun, a positive regulator of myeloid differentiation (51) that acts as a PU.1 coactivator of the M-CSFR promoter (39) in a mechanism similar to that described for the GATA1-induced repression of PU.1 during erythroid differentiation (52, 53). Other investigators reported on a differentiation block of myeloid cell lines by binding EVI1 to RUNX1 or C/EBP-α (7, 37), and it is conceivable that the EVI1-induced defects in myelopoiesis could result from the combination of multiple transcription factor deregulation. Nevertheless, the capability of an EVI1 mutant unable to bind PU.1 to allow neutrophil and macrophage differentiation in vitro reinforces the importance of PU.1 in normal myeloid maturation and suggest that the critical EVI1 interaction sites disrupted in the mutant could be exploited for the design of therapeutic agents as a treatment of EVI1-associated disorders.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Grant support: NIH grants R01 HL79580, R01 HL082935 and CA096448 (G. Nucifora).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.