|Home | About | Journals | Submit | Contact Us | Français|
EVI1 is an aggressive nuclear oncoprotein deregulated by recurring chromosomal abnormalities in myelodysplastic syndrome (MDS). The expression of the corresponding gene is a very poor prognostic marker for MDS patients and is associated with severe defects of the erythroid lineage. We have recently shown that the constitutive expression of EVI1 in murine bone marrow results in a fatal disease with features characteristic of MDS, including anemia, dyserythropoiesis, and dysmegakaryopoiesis. These lineages are regulated by the DNA-binding transcription factor GATA1. EVI1 has two zinc finger domains containing seven motifs at the N terminus and three motifs at the C terminus. Supported by results of assays utilizing synthetic DNA promoters, it was earlier proposed that erythroid-lineage repression by EVI1 is based on the ability of this protein to compete with GATA1 for DNA-binding sites, resulting in repression of gene activation by GATA1. Here, however, we show that EVI1 is unable to bind to classic GATA1 sites. To understand the mechanism utilized by EVI1 to repress erythropoiesis, we used a combination of biochemical assays, mutation analyses, and in vitro bone marrow differentiation. The results indicate that EVI1 interacts directly with the GATA1 protein rather than the DNA sequence. We further show that this protein-protein interaction blocks efficient recognition or binding to DNA by GATA1. Point mutations that disrupt the geometry of two zinc fingers of EVI1 abolish the protein-protein interaction, leading to normal erythroid differentiation of normal murine bone marrow in vitro.
EVI1 (ecotropic viral integration site 1) is an aggressive nuclear oncoprotein deregulated by recurring chromosomal abnormalities associated with acute myeloid leukemia or myelodysplastic syndrome (MDS) (1, 25; for a recent review on EVI1, please see reference 14). EVI1 contains 10 zinc finger motifs, grouped in a proximal domain (7 motifs) and a distal domain (3 motifs). In MDS patients, EVI1 expression is a very poor prognostic marker and is associated with erythroid defects and severe erythropoietin (Epo)-unresponsive anemia (18). Since EVI1 was first identified, there has been a considerable effort by many investigators to understand the mechanism by which the protein deregulates erythropoiesis. Kreider et al. (8) showed first that the inappropriate expression of this gene in murine bone marrow (BM) severely impairs erythroid differentiation in vitro. Furthermore, based on electrophoretic mobility shift assay (EMSA) and reporter gene studies with a synthetic DNA promoter containing several GATA repeats, these authors suggested that EVI1 interferes with GATA1 regulation by competitive binding to GATA-DNA sites at GATA1-regulated promoters. Later, three murine EVI1-transgenic mouse lines were generated by Louz et al. (11), and a general, though not fatal, impairment of erythropoiesis was observed in these lines by the investigators. More recently, our group showed that mice transplanted with syngeneic BM that expresses EVI1 by retrovirus infection invariably develop a fatal anemia and a disease with similarity to MDS (2). Furthermore, we showed that GATA1-regulated genes, such as those encoding EpoR and c-Mpl, are strongly downregulated in vivo in the BM of the EVI1-positive animals (2). Altogether, these data leave no doubt that EVI1 plays a dominant repressive role in the deregulation of GATA1-dependent pathways. However, despite being identified almost 20 years ago, not much is know about the normal function of EVI1. Recently, Yatsula et al. (30) identified several genes that are potentially targeted by EVI1, and Yuasa et al. (31) showed that EVI1 directly binds to the GATA2 promoter activating the protein.
GATA1 was identified as a protein that binds to the consensus motif (A/T)GATA(A/G), which is found in the regulatory elements of many erythroid-specific genes. Targeted disruption of GATA1 leads to impairment of erythropoiesis and embryonic death in mice (6, 17). GATA1 contains two zinc fingers, the C-terminal finger (CF), with a strong affinity for the DNA consensus, and the N-terminal finger (NF), which increases the DNA-binding association and mediates the interaction with GATA1 cofactors (5, 12, 29).
In this report, we have readdressed the basis of GATA1 inhibition by EVI1. With a combination of biochemical and cellular assays, we determined that the mechanism of GATA1 repression by EVI1 involves binding of the two proteins rather than competition for the DNA sites. Our data show that two zinc fingers within the EVI1-proximal domain directly interact with the CF of GATA1 and abolish the DNA affinity of the transcription factor. Disruption of the two-zinc-finger structure by point mutations (Cys to Ala and His to Ala) destroys the interaction between the proteins and allows DNA binding and reporter gene activation by GATA1. More importantly, this EVI1 double mutant does not block the erythroid differentiation of murine BM cells observed when EVI1 is present. Our results provide a possible mechanism by which EVI1 exerts a dominant repression over GATA1 functions and point to a protein-protein interaction interface that could be exploited for drug development.
The maintenance and transfection of the 293T, COS7, 32Dcl3, AML14.3D10, and NIH 3T3 cell lines have been previously reported (3, 4). The Phoenix retroviral packaging cells (ATCC) were maintained and transfected as described previously (20). Normal BM cells, isolated from C57BL mice, were infected and selected as described previously (20).
The pXM-GATA1 plasmid was a gift of J. N. Ihle (St. Jude Children's Research Hospital, Memphis, TN). The GATA1 internal deletion mutants (GATA1ΔNF and GATA1ΔCF) were generated by PCR cloning using primers upstream and downstream, respectively, of the proximal or the distal zinc finger motif. The plasmids Flag-EVI1, hemagglutinin (HA)-tagged EVI1 (HA-EVI1), HA-EVI1-Δ283, and HA-EVI1-283 (wild type or mutated) and the GATA1 deletion mutants (GATA1ΔNF and GATA1ΔCF) were all cloned into the BamHI/XhoI sites of pCMV/myc/nuc (Invitrogen/Life Technologies, Inc.). EVI1 was also cloned into the pBK-CMV vector under the regulation of the T3 promoter for in vitro translation. To generate glutathione S-transferase (GST) fusions, GATA1 wild type, GATA1ΔNF, and GATA1ΔCF were subcloned in frame to GST into the BamHI/SmaI sites of pGEX-2T (Amersham, Piscataway, NJ). The EVI1 point mutant HA-EVI1(1+6Mut) contains the H39A and C44A mutations in the first zinc finger and the C190A and C193A mutations in the sixth zinc finger. The mutations were generated by PCR. All PCRs were performed with high-fidelity Pfu DNA polymerase (Stratagene). To generate the recombinant retroviruses for BM infection, HA-EVI1 and HA-EVI1(1+6Mut) were subcloned in the EcoRI/BglII sites of the pMSCV vector (BD Biosciences-Clontech). All cloning junctions and PCR-generated fragments were verified by DNA sequencing.
The expression and purification of GST fusion proteins and their interaction with HA-EVI1 were carried out as described previously (21).
We used PCR to amplify the region from nucleotide −1050 to nucleotide +3 of the human EpoR promoter. After DNA sequencing, the fragment was cloned into the XhoI/HinDIII sites of the luciferase reporter vector pGL3-Basic (Promega). For normalization of transfection efficiency, we used the pRL-TK plasmid (Promega), which expresses Renilla luciferase. The experiments were repeated three times as we described previously (3, 21).
Nuclear extracts of COS7 cells were incubated for 30 min with a 32P-labeled double-stranded oligonucleotide probe containing either the palindromic GATA1 site of the GATA1 promoter (5′-AGTCCATCTGATAAGACTTATCTGCTGCCC-3′) or the single GATA1 site of the EpoR promoter (5′-CAGGCACTTATCTCTACCCAGGCTG-3′). The samples were loaded on a 0.5 M Tris-borate-EDTA nondenaturing 6% polyacrylamide gel, separated by electrophoresis at 120 V for 2 h, and autoradiographed overnight. For the supershift assay, 40 ng of anti-GATA1 antibody (Santa Cruz, CA) was added to the nuclear extracts before the incubation with the labeled probe.
Total cellular RNA was extracted from 32Dcl3 cells as described previously (2). The cDNA was prepared according to the First Strand cDNA synthesis kit protocol (Fermentas). Reverse transcription-PCR (RT-PCR) was performed in 150-μl reaction mixtures containing 10 μl of cDNA, 1 μl of Taq Red polymerase, and 300 nM forward and reverse primers. Fifteen microliters of the PCR products were collected after 23, 26, 29, 32, 35, 38, and 41 cycles.
Chromatin immunoprecipitation (ChIP) analysis was performed with 32Dcl3 cells, either noninfected or expressing EVI1 or EVI1(1+6Mut). For analysis of GATA1 promoter occupancy, the proteins were cross-linked to the DNA with 1% formaldehyde. An anti-GATA1 antibody (Santa Cruz, CA) was used with protein G-Sepharose to adsorb and immunoprecipitate the protein-DNA complexes. The cross-linking agent was removed as described previously (4). The DNA fragments were analyzed by PCR (30 cycles at 60°C) using two primers designed to amplify a 145-bp region of the murine EpoR promoter. The primer sequences are 5′-AACTCTGCTGTCTGCCCCAC-3′ and 5′-TGGCAGCTCCTTCCAGGGGGC-3′. To amplify a 250-bp region of the murine c-Mpl promoter, we used the primers 5′-TACCTCTGTGTCCCTGCCAA-3′ and 5′-CATGAAGAGGGCCCAAGAGG-3′. Finally, primers were designed to amplify a 150-bp region of the human EpoR promoter. Their sequences are 5′-TCTGAAGCCAGAACGGGAGC-3′ and 5′-CGAGGGGGCGGGGCCAGCA-3′.
The isolation of lineage-negative murine hematopoietic progenitors and their retroviral infection have been previously described (20). The infected hematopoietic progenitors were selected in methylcellulose cultures supplemented with granulocyte-macrophage colony-stimulating factor (GM-CSF), interleukin 3, interleukin 6, stem cell factor, and G418 (1 mg/ml). After 7 days, the cells were isolated, purified, counted, and replated with either GM-CSF or Epo. The colonies were counted 7 days later.
We reported that in vivo EVI1 represses the c-Mpl and EpoR promoters as determined by quantitative RT-PCR (2). These promoters contain DNA binding sites for and are activated by GATA1 (13, 33). Although it was reported that EVI1 binds in vitro to an artificial DNA fragment that contains 11 consecutive head-to-tail repeats of the GATA motif, such a high number of repeats is not present in the promoters of EpoR and c-Mpl (13, 33). Therefore, to determine whether the interaction between EVI1 and GATA1 could account at least in part for the transcription repression that we observed, we used co-IP assays with transiently transfected 293T cells. The results (not shown) indicated that the two proteins interact. To define the region of interaction, we constructed two EVI1 deletion mutants (Fig. (Fig.1A),1A), one of which consists essentially of the proximal zinc finger domain (HA-EVI1-283), whereas the second one encodes all the protein excluding the proximal domain (HA-EVI1-Δ283). These two deletion mutants, as well as full-length HA-EVI1, were used in co-IP assays to evaluate their capacity to bind GATA1. The results indicate that GATA1 interacts with HA-EVI1 (Fig. (Fig.1B,1B, lane 4) and HA-EVI1-283 (Fig. (Fig.1B,1B, lane 5) but not with the deletion mutant HA-EVI1-Δ283 (Fig. (Fig.1B,1B, lane 6), indicating that the major interaction surface in EVI1 is within the proximal zinc finger domain.
To identify the zinc finger motif(s) that mediate EVI1-GATA1 interaction, we disrupted the structure of the motifs in the proximal domain contained in HA-EVI1-283 (Fig. (Fig.1A)1A) by Cys-to-Ala and His-to-Ala mutation. We started by introducing mutations in the first, fifth, sixth, and seventh motifs. We found that single mutants retained GATA1 binding capacity. However, when the mutations were in the first motif or sixth motif, there was a slight decrease in the intensity of the co-IP GATA1 band (data not shown). These results suggested that perhaps EVI1 recognizes GATA1 with more than a single motif and prompted us to disrupt both the first and the sixth motifs in HA-EVI1-283. As shown in Fig. Fig.2A,2A, the results confirm that whereas wild-type HA-EVI1-283 (lane 6), the single point mutant HA-EVI1-283(6Mut) (lane 9), and the double mutant EVI1-283(1+5Mut) (lane 7) were capable of interaction with GATA1, the double mutant HA-EVI1-283(1+6Mut) (lane 8) failed to bind efficiently to GATA1 even though all the proteins were expressed at comparable levels in the cells (Fig. (Fig.2A,2A, right side).
In order to confirm that the first and sixth zinc fingers were the only domains involved in EVI1-GATA1 interaction, we introduced the mutations in full-length EVI1. Wild-type and double-mutant EVI1 were used in a co-IP assay. As shown in Fig. Fig.2B,2B, whereas wild-type HA-EVI1 and the single mutant HA-EVI1(6Mut) interacted with GATA1 (lanes 6 and 7), the double mutant HA-EVI1(1+6Mut) was not able to co-IP with GATA1 (lane 8). Note that the interaction of GATA1 with the single mutant HA-EVI1(6Mut) (lane 7) is significantly weaker than that with wild-type EVI1 (lane 6), although the proteins were expressed at similar levels in the cells (Fig. (Fig.2B,2B, right side).
These results were confirmed by using two hematopoietic cell lines, 32Dcl3 and AML14.3D10, that express endogenous GATA1. By retroviral infection and selection, we generated 32Dcl3 and AML14.3D10 cell lines that express the murine stem cell virus retrovirus or a recombinant retrovirus in which either EVI1 or EVI1(1+6Mut) was cloned and used these modified cell lines to confirm the co-IP results obtained with 293T cells. We found that in both 32Dcl3 and AML14.3D10, endogenous GATA1 was able to co-IP with wild-type HA-EVI1 (Fig. (Fig.2C,2C, lane 2, and D, lane 2) but not the double mutant HA-EVI1(1+6Mut) (Fig. (Fig.2C,2C, lane 3, and D, lane 3).
To map the GATA1 region necessary for the interaction with EVI1, we utilized internal deletions of GATA1, removing either the NF (GATA1ΔNF) or the CF (GATA1ΔCF). The CF of GATA1 is essential for DNA binding, whereas the NF strengthens DNA binding and is required for interaction with FOG-1 (5, 12, 29). We tested the two GATA1 internal deletion mutants (HA-GATA1ΔNF and HA-GATA1ΔCF) and wild-type GATA1 in co-IP assays to evaluate their capacity to interact with Flag-EVI1. The results indicate that EVI1 interacts with the deletion mutant GATA1ΔNF (Fig. (Fig.3A,3A, lane 7) and the GATA1 wild type (Fig. (Fig.3A,3A, lane 8) but not with the deletion mutant GATA1ΔCF (Fig. (Fig.3A,3A, lane 6), suggesting that the C-terminus zinc finger of GATA1 is the major region required for EVI1 binding. The right panel of Fig. Fig.3A3A shows that all the proteins were expressed at comparable levels.
A method that is often used to confirm the interaction between two proteins is to pull down one of them with the second protein expressed in Escherichia coli as a fusion with GST, which can be immobilized on a solid support and purified. We used this technique to verify the EVI1-GATA1 interaction and to confirm that the C-terminus zinc finger is necessary for EVI1 interaction. The recombinant GST fusion proteins GST-GATA1, GST-GATA1ΔNF, and GST-GATA1ΔCF were expressed in E. coli cells, purified using glutathione-Sepharose beads as described previously (21), and incubated with nuclear extracts expressing HA-EVI1. After extensive washes, the proteins in the GST fusion complexes were separated and identified by Western blot analysis. The results, shown in Fig. Fig.3B,3B, indicate that GATA1 and GATA1ΔNF (lanes 4 and 8) interact with EVI1, whereas GATA1ΔCF (lane 6) does not. Proteins of untransfected cells did not bind the GST-GATA1 fusion proteins (lanes 3, 5, and 7), confirming the specificity of the assay. After stripping, the same membrane was hybridized to anti-GST antibody. The results confirm the expression of GST-GATA1, GST-GATA1ΔNF, and GST-GATA1ΔCF (Fig. (Fig.3B,3B, bottom, lanes 3 to 8) and GST (Fig. (Fig.3B,3B, bottom, lanes 1 and 2). The right panel of Fig. Fig.3B3B shows that EVI1 was expressed at comparable levels in all extracts. The GST fusion pull-down assay was also used to determine whether the EVI1-GATA1 interaction was direct. Purified GST and GST-GATA1 fusion protein expressed in E. coli were incubated with in vitro-translated EVI1 labeled with [35S]methionine. After extensive washing, the Sepharose beads were subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis, and the separated proteins were analyzed by autoradiography. The results show evidence of direct association of EVI1 with GST-GATA1 (Fig. (Fig.3C,3C, lane 5) but not with GST (Fig. (Fig.3C,3C, lane 3). The right panel of Fig. Fig.3C3C shows the Coomassie blue staining of the total cell lysates and the expression of GST and GST-GATA1. Taken together, these results confirm that GATA1-EVI1 interaction occurs through the C-terminal zinc finger region of GATA1 and suggest that the interaction is probably direct.
We earlier showed that EVI1 represses the activation of EpoR in vivo and proposed that this repression strongly contributes to the severe anemia of EVI1-positive mice (2). It is well known that GATA1 regulates its own expression, and it is therefore possible that inhibition of GATA1 expression by interaction with EVI1 could be the major reason for or a contributing factor to the decrease of EpoR that we observed in vivo. Therefore, to exclude this possibility, we first used reporter gene assays with NIH 3T3 cells, which do not express endogenous GATA1. To determine whether the interaction between EVI1 and GATA1 is the major factor in repression of GATA1-dependent genes, we transiently transfected GATA1, EVI1, or EVI1(1+6Mut) with EpoR-luciferase, a plasmid we made that contains the fragment −1050 to +3 of the EpoR promoter linked to the luciferase gene. At nucleotide −180, this promoter has a single GATA-binding site (Fig. (Fig.4A).4A). The results show that wild-type EVI1 and the double mutant EVI1(1+6Mut) had very little effect on the basal activity of the promoter by themselves (Fig. (Fig.4B,4B, bars 3 to 8). As expected, significant promoter activation was noted in the presence of GATA1 (Fig. (Fig.4B,4B, bar 2). This activation significantly decreased in the presence of increasing amounts of EVI1 (Fig. (Fig.4B,4B, bars 9 to 11). However, when the double mutant EVI1(1+6Mut) was used, there was minimal repression of GATA1 activation (Fig. (Fig.4B,4B, bars 12 to 14). Because in the transfected plasmid the expression of GATA1 is driven by the cytomegalovirus promoter, the results suggest that the interaction between EVI1 and GATA1 (rather than GATA1 downregulation) is the major factor of gene repression. The expression of GATA1 (1 μg of transfected plasmid), EVI1 (6 μg of transfected plasmid), and EVI1(1+6Mut; 6 μg of transfected plasmid) was evaluated by Western blotting and is reported in Fig. Fig.4C.4C. To confirm that the EpoR promoter also is repressed by EVI1 in a hematopoietic cell line, we repeated the reporter gene assays with the GATA1-expressing cell lines AML14.3D10 and 32Dcl3 with similar results (Fig. 4D and E). The proteins expression in each cell line are shown in the bottom panels of Fig. Fig.4D4D and Fig. Fig.4E4E.
To determine whether the DNA binding of EVI1 to GATA-DNA sites contributes to the repression of activation by GATA1, we used EMSA with a probe containing a single GATA1 consensus corresponding to that of the EpoR promoter and a probe with two consensus sites extensively used by others to evaluate DNA binding by GATA1 (26, 28). Nuclear extracts of COS7 cells transfected with GATA1, EVI1, EVI1(1+6Mut), or a combination of GATA1-EVI1 and GATA1-EVI1(1+6Mut) were tested separately with the two probes. As shown in Fig. Fig.5A,5A, GATA1 efficiently binds to both DNA probes (lanes 4 and 16). In contrast, neither EVI1 nor EVI1(1+6Mut) significantly interacts with these probes (Fig. (Fig.5A,5A, lanes 2 and 3 and lanes 14 and 15). We also found that the coexpression of GATA1 and EVI1 significantly reduced the amount of GATA1 bound to the probes (Fig. (Fig.5A,5A, lanes 5 and 17). This reduction was not observed with EVI1(1+6Mut) (Fig. (Fig.5A,5A, lanes 6 and 18). The specificity of the GATA1-DNA complex was confirmed by adding an excess of unlabeled DNA (Fig. (Fig.5A,5A, lanes 7 to 9 and 19 to 21). To further confirm the EMSA results, we used anti-GATA1 antibody to visualize the supershift of the protein-DNA complex. The results (Fig. (Fig.5A,5A, lanes 10 to 12 and 22 to 24) clearly show supershifted bands for GATA1 alone or with EVI1(1+6Mut) only. No band is observed when EVI1 is present. The level of protein expression in each cell extract was comparable, as assessed by Western blotting (Fig. (Fig.5B).5B). These data suggest that the EVI1-GATA1 association interferes with binding of GATA1 to the DNA, leading to the displacement of GATA1 from its consensus sequence.
Whereas positive EMSA results are a good indication of a protein's ability to recognize and bind to a DNA consensus in vitro, a better assay to evaluate DNA occupancy in vivo is ChIP analysis. For these assays, we selected the hematopoietic 32Dcl3 and AML14.3D10 cell lines we generated that express either EVI1 or EVI1(1+6Mut) (Fig. 2D and C), and we tested the enrichment of the EpoR promoter with murine and human cell lines and the enrichment of the c-Mpl promoter with the murine cell line. The preparation of the chromatin and the IP with anti-GATA antibody are described in Materials and Methods. The ChIP results (Fig. (Fig.6A)6A) show that whereas anti-GATA1 antibody easily precipitates endogenous GATA1-chromatin complex in the absence of EVI1 (Fig. (Fig.6A,6A, lane 6), such a complex is not recognized by the antibody in the presence of EVI1 (Fig. (Fig.6A,6A, lane 7). However, when EVI1(1+6Mut) is expressed, there is no significant decrease in GATA1 binding to the endogenous EpoR or c-Mpl promoters (Fig. (Fig.6A,6A, lane 8). To confirm that the chromatin-immunoprecipitated DNA amounts were comparable for all lanes, equal amounts of nonimmunoprecipitated DNA were used as a control (Fig. (Fig.6A,6A, lanes 1 to 3). Immunoglobulin G was used as control antibody. Altogether, these results clearly show that EVI1 affects the occupancy of GATA1 on the EpoR and c-Mpl promoters in vivo.
We used semiquantitative RT-PCR analysis to evaluate the expression of EpoR, c-Mpl, and c-Myb, which are all regulated by GATA1, in the presence of EVI1 or EVI1(1+6Mut). In addition, because GATA1 regulates its own expression, we also evaluated the level of GATA1 in presence of EVI1. The results (Fig. (Fig.6B)6B) indicate that the transcription of the four genes is not significantly different in 32Dcl3 vector cells and in cells that express EVI1(1+6Mut), suggesting that this double mutant has a limited effect on GATA1-dependent activation in this cell line. However, EVI1 is able to downregulate these genes, although to a different extent. Those that are most repressed by EVI1 are EpoR and c-Mpl (Fig. (Fig.6B,6B, panels 1 and 2). In contrast, c-Myb is moderately reduced (Fig. (Fig.6B,6B, panel 3). Even though we did not detect a significant reduction at the protein level (Fig. 2C and D), GATA1 transcription appears to be moderately reduced in EVI1-positive 32Dcl3 cells as well (Fig. (Fig.6B,6B, panel 4). The results indicate that while EVI1 definitively affects the expression of specific genes regulated by GATA1, the effect is less significant for others, suggesting that cooperation/interaction of GATA1 with other cofactors could considerably protect it from the ability of EVI1 to dislodge the protein from promoters.
The ability of EVI1 to affect the growth of Epo-responsive cells was previously reported by us and others (2, 8). To determine whether the mutant EVI1(1+6Mut) retains the ability to block erythroid differentiation in response to Epo, we stably expressed EVI1 and EVI1(1+6Mut) in murine BM cells by recombinant retrovirus insertion. Lineage-negative murine BM cells were infected in vitro with EVI1, the mutant EVI1(1+6Mut), or the empty retrovirus murine stem cell virus. After G418 selection, an equal number of cells for each BM sample was plated in the presence of Epo or GM-CSF to determine the progenitors' potential to form colonies in response to these cytokines. The results are summarized in Fig. Fig.7A,7A, in which the number of vector colonies is arbitrarily taken as 100. As we previously reported (2), EVI1 confers a significant increase in cell clonogenicity in response to GM-CSF. In agreement with these previous findings, we observed a much larger number of colonies for EVI1-positive cells than for control cells (Fig. (Fig.7A,7A, compare bars 1 and 3). However, when Epo instead of GM-CSF was added to the medium, there was a complete inhibition of colony formation in EVI1-positive cells (Fig. (Fig.7A,7A, bars 2 and 4). The results with EVI1(1+6Mut) were quite different. We found that the disruption of the two motifs restores the response to Epo to a high degree, and the number of colonies formed was about 65% of that for the normal control (Fig. (Fig.7A,7A, bars 2 and 6). The expression of EVI1 and EVI1(1+6Mut) in the BM cells evaluated by Western blotting is reported in Fig. Fig.7B.7B. The appearance of the BM colonies after Epo stimulation is shown in Fig. Fig.7C.7C. The EVI1-positive colonies were rare, weak, and had fewer cells. In contrast, EVI1(1+6Mut) colonies looked similar to those obtained with the empty vector, although somewhat smaller. The cells were recovered from the colonies, and their morphology was analyzed after Wright-Giemsa staining. Because of the in vitro culture conditions, it was not possible to detect the late stages of erythroid maturation (polychromatophilic erythroblasts or reticulocytes), and all the control cells had the appearance of normal erythroblasts without significant atypical aspects (Fig. (Fig.7D,7D, panels A and B). In contrast, virtually all the EVI1 cells examined showed a massive impairment of differentiation, with all of the cells arrested at the early stage of basophilic erythroblast. In addition, about half of the entire population showed severe dysplastic aspects, including nuclear-cytoplasmic maturative asynchronization, an irregular ratio between the nucleus and the cytoplasm that often was completely inverted, showing cells with a block of cytoplasmic development at a very early proerythroblast stage, or cells with normal cytoplasmic development and a nucleus with very immature chromatin. About 30% of the cells arrested in the early stage of basophilic erythroblast had abnormalities in the number of nuclei (bi- or tetranucleated cells) (Fig. (Fig.7D,7D, panels C and D) and in chromatin maturation (chromatin bridges, budding nuclei; Fig. Fig.7D,7D, panels E and F). These dysplastic features, common in MDS, were less prominent in EVI1(1+6Mut) cells (Fig. (Fig.7D,7D, panel G) and were observed in only a minority of cells rather than the entire cell population, as for EVI1 cells. EVI1(1+6Mut) cells were faintly delayed in maturation, as indicated by a prevalence of basophilic and orthochromatic erythroblasts (Fig. (Fig.7D,7D, panel H). In contrast to EVI1 cells, the majority (85 to 90%) of EVI1(1+6 Mut) cells did not have dysplastic aspects, and only 10 to 15% of EVI1(1+6 Mut) cells showed double nuclei and sometimes light chromatin irregularities (Fig. (Fig.7D,7D, panel I).
Since its first identification as a murine gene that induces myeloid tumors, there have been many reports linking the inappropriate expression of EVI1 to human leukemia. An attempt to generate a murine germ line deletion of EVI1 by homologous recombination resulted in an allele that was truncated after the first zinc finger domain rather than in the complete deletion of the gene. This work was still very informative, because the gene alteration resulted in embryonic lethality due to widespread defects in many organs, suggesting that during embryonic development EVI1 could have an essential, pleiotropic effect rather than being required for a specific lineage commitment (7, 16). While the normal function of EVI1 remains unclear, this gene is not normally expressed in hematopoietic cells, and only recently there have been reports on the potential effect of EVI1 in the regulation of cellular genes (30, 31). The report by Yuasa et al. (31) is of particular interest because it points to the effect of EVI1 on the early hematopoietic regulator GATA2. During the last 20 years, classic cytogenetic and fluorescent in situ hybridization analyses have shown that chromosomal translocations at 3q26 leading to very high EVI1 expression are a dominant factor in the pathogenesis of a very specific group of fatal hematopoietic disorders collectively known as 3q26 syndrome (15, 25). This disease is characterized by severe anemia and megakaryocyte defects, suggesting that the inappropriate expression of EVI1 in hematopoietic cells preferentially disrupts these two lineages. In vitro studies (8, 22) and in vivo murine models (2, 11) confirm that EVI1 alters erythro- and megakaryopoiesis, two lineages strictly regulated by the transcription factor GATA1. More precisely, these studies show that EVI1 blocks erythroid differentiation in vitro by impairing cell responsiveness to erythropoietin (8) and represses the expression of EpoR and c-Mpl in vivo (2). Therefore, there is a general consensus that EVI1 exerts an antagonistic effect on GATA1 regulation, and Kreider et al. proposed that the mechanism by which EVI1 antagonizes GATA1 regulation involves competition for DNA-binding sites. Here we have looked again at the basis of EVI1-GATA1 inhibition and found that rather than binding to the DNA, EVI1 displaces GATA1 from its DNA consensus by direct protein-protein association, leading to the drastic downregulation of some genes required for normal hematopoiesis, such as those encoding EpoR and c-Mpl, and to a more moderate repression of other GATA1-regulated genes, such as those encoding c-Myb and GATA1 itself, which is autoregulated (26, 28). An important question is whether, independently of GATA1 functional repression of other genes, the downregulation of GATA1 itself contributes further to the repression of the erythroid program by EVI1. The reporter gene studies with NIH 3T3 cells in which GATA1 is transfected and expressed from the cytomegalovirus promoter suggest that there is a dose response, and therefore, it is likely that any downregulation of GATA1 itself could amplify the loss of activation of its targets. It is interesting that not all GATA1-regulated genes appear to be repressed to the same extent. At this time, we propose that the variability of the repression could depend on the cofactors and transcription factors with which GATA1 interacts and forms DNA-bound complexes at promoter sites. It is not unlikely that the stability of such complexes or the ability of cofactors to protect the CF of GATA1 could contribute to the smaller reduction of repression. The regulation of GATA1 activity has been extensively studied, and different mechanisms have been hypothesized for its inhibition by other factors, among them PU.1 (19, 24). This factor, involved in myeloid- and B-cell differentiation, inhibits the function and controls the activity of GATA1 by binding to it, and some investigators (32) propose that this interaction leads to disruption of the GATA1-DNA complex and to a decrease of GATA1-dependent erythroid differentiation, as we propose for EVI1. It is of interest that the interaction between either PU.1 or EVI1 with GATA1 is mediated by the C-terminus zinc finger of GATA1, which is essential for DNA binding, and therefore, it is not unreasonable to propose that EVI1 acts by sequestering GATA1 from its target genes and impairing in this way a normal erythroid program.
Whereas several investigators have proposed that EVI1 itself interacts with DNA through the proximal zinc finger domain, at this time only a few physiological targets have been identified, most notably GATA2 (30, 31). The proximal zinc finger domain appears to perform several other functions, including binding to histone deacetylases (3), mediating the ability to abrogate the growth-inhibitory effect of transforming growth factor beta (9, 23), and physically interacting with Jun N-terminal protein kinase, leading to inhibition of Jun N-terminal protein kinase (10). It is conceivable that an EVI1-induced block of erythroid differentiation and increased cell proliferation might be the result of multiple mechanisms, including inhibition of GATA1 and GATA1-activated genes and direct activation of genes essential for hematopoietic progenitor proliferation, such as that encoding GATA2, which is required to maintain the pool of early hematopoietic cells but is not essential for the terminal differentiation of erythroid cells and macrophages (27). In our view, however, the ability of the proximal zinc finger domain of EVI1 to inhibit GATA1 activity remains a highly relevant defect, and our finding that the disruption of well-studied structures, such as zinc fingers, could in large part repair this defect suggests that the interaction surface between EVI1 and GATA1 might be an attractive target for the development of competing small molecules to be used as a treatment in EVI1-associated leukemia.
This work was supported by NIH R01 grants HL72691, HL82935, HL79580, and CA96448 (G.N.).
We thank S. Buonamici for her help with cloning.
Published ahead of print on 5 September 2006.