Members of the GATA Family of Transcription Factors Bind to the U3 Region of Cas-Br-E and Graffi Retroviruses and Transactivate Their Expression

Abstract

Cas-Br-E and Graffi are two murine viruses that induce myeloid leukemia in mice: while Cas-Br-E induces mostly non-T, non-B leukemia composed of very immature cells, Graffi causes exclusively a granulocytic leukemia (E. Rassart, J. Houde, C. Denicourt, M. Ru, C. Barat, E. Edouard, L. Poliquin, and D. Bergeron, Curr. Top. Microbiol. Immunol. 211:201–210, 1995). In an attempt to understand the basis of the myeloid specificity of these two retroviruses, we used DNase I footprinting analysis and gel mobility shift assays to identify a number of protein binding sites within the Cas-Br-E and Graffi U3 regions. Two protected regions include potential GATA binding sites. Methylation interference analysis with different hematopoietic nuclear extracts showed the importance of the G residues in these GATA sites, and supershift assays clearly identified the binding factors as GATA-1, GATA-2, and GATA-3. Transient assays with long terminal repeat (LTR)-chloramphenicol acetyltransferase constructs showed that these three GATA family members are indeed able to transactivate Cas-Br-E and Graffi LTRs. Thus, the availability and relative abundance of the various members of the GATA family of transcription factors in a given cell type could influence the transcriptional tissue specificity of murine leukemia viruses and hence their disease specificity.

Nondefective murine leukemia viruses (MuLV) can induce a large spectrum of pathologic responses in mice, with a predominance of hematopoietic tumors. They do not carry an oncogene, and tumorigenic transformation is usually achieved by retroviral integration at the vicinity of a cellular proto-oncogene. Although MuLVs can infect many tissues and cell types, each virus will induce a specific type of tumor: T or B lymphomas, erythroleukemia, myeloid leukemia, etc. Many studies have shown that the primary determinant for this disease specificity and for tumorigenicity itself is the viral long terminal repeat (LTR) (11, 19–21, 25, 34, 35, 42, 58, 76). Moreover, a very good correlation has been demonstrated between transcriptional tissue specificity and disease specificity: viral gene expression is higher in the cell type that is the target for oncogenic transformation (6, 12, 32, 64).

The U3 region of the LTR contains the promoter and the transcriptional regulatory elements. Since retroviruses use the cellular machinery for gene expression (transcription, translation), they are also likely to use the cellular regulatory functions, including transcription factors. Indeed, dissection of retroviral promoter-enhancer regions has led to the identification of binding sites for many cellular transcription factors, some of which seem to be crucial for tissue and disease specificity and tumorigenicity (11, 32, 46, 59, 60, 75). One striking example is the core binding factor: first identified as a factor binding to the core motif present in simian virus 40, polyomavirus, and MuLV enhancers (72), this transcription factor is now involved in the regulation of a growing number of genes specifically expressed in lymphoid or myeloid lineages (28, 49, 55, 63, 71, 78, 79). The core motif was shown to play a key role in disease specificity for several MuLV (11, 59). Other cellular factors playing a role in viral gene regulation are ETS family members (62), CAAT/enhancer binding proteins, AP1, NF1, helix-loop-helix (HLH) proteins binding to E-boxes, Oct proteins, and hormone receptors (14, 38, 47). Thus, analysis of retroviral regulatory regions can lead to new insights into the control of gene expression in eucaryotes.

We have undertaken an analysis of the factors binding to the U3 region of two murine retroviruses that induce myeloid leukemia: Cas-Br-E and Graffi leukemia virus.

Cas-Br-E induces a wide variety of hematopoietic tumors in NFS/N mice; however, injection in NIH Swiss mice causes mainly a non-T, non-B leukemia composed of blasts lacking any myeloid or lymphoid markers (13, 54). Genetic analyses have shown that the determinants for leukemogenicity are dispersed in different regions of the genome, including the LTRs (36). We have shown that the virus preferentially targets two potential oncogenes, fli-1 and evi-1, in 70 and 18% of the tumors, respectively (5–7).

We have recently molecularly cloned the Graffi MuLV and shown that the molecular clones induced the same pathologic responses in BALB/c and NFS mice as the parental mixture did, i.e., a granulocytic leukemia, composed of myeloblasts and neutrophils with characteristic donut-like nuclei (54, 56). Two molecular clones, GV1.2 and GV1.4, have been characterized and shown to cause the same disease. They were highly similar, except for the presence of a perfect 60-bp duplication in the U3 region of the GV1.2 clone, which displays a shorter latency period. This correlation between the latency period and the number of enhancer repeats strongly suggest a role for the U3 region in the leukemogenic potential and disease specificity.

Two of the regions found protected by DNase I footprinting analyses contain potential binding sites for GATA factors (1a). GATA factors are a family of DNA binding proteins that recognize the motif (A/T)GATA(A/G) (24, 44, 51). They possess a zinc finger of the form Cys-X₂-Cys-X₁₇-Cys-X₂-Cys (24, 65). The founding member, GATA-1, was identified as a positive regulator of globin gene transcription (23) and has since been involved in the regulation of all known erythroid specific genes (50, 74). At least six GATA family members have been identified: GATA-1, GATA-2, and GATA-3 are involved in the regulation of hematopoiesis-specific genes (reviewed in reference 50) and are required for normal hematopoietic development in mice (51, 66, 74). Given this crucial role of GATA family members in hematopoietic gene regulation, one would not be surprised if they were involved in the transcription regulation of leukemia viruses.

Here, we identify in the Cas-Br-E and Graffi U3 regions two GATA elements and show that GATA-1, GATA-2, and GATA-3 can bind to these elements. We also demonstrate that those three family members can indeed transactivate Cas-Br-E and Graffi LTR-driven expression.

MATERIALS AND METHODS

Cell lines and nuclear extracts.

Myeloid cell-derived M1, M-NFS-60, WEHI-3B, and 32Dcl3 cell lines, T-cell lines Ti-6, BW5147, EL4, and Jurkat, and the B-cell line A20 were maintained in RPMI 1640 medium supplemented with 10% fetal calf serum, and 5% WEHI-3B conditioned medium was added as a source of interleukin-3 for the M-NFS-60 and 32Dcl3 cell lines. Erythroid cell lines D1B, MEL, and CB7 were cultured in Eagle minimum essential medium supplemented with 10% fetal calf serum, and NIH 3T3, CHO, and COS-7 cells were cultured in Dulbecco modified Eagle medium plus 10% calf serum.

Nuclear extracts were prepared from 10⁹ cells as described by Wall et al. (70). The protein concentration in the extracts was between 5 and 15 μg/ml. Microextracts were prepared from 10⁶ to 10⁷ cells as described by Andrews and Faller (1). Extracts were aliquoted and stored at −80°C.

Oligonucleotides.

Oligonucleotides were synthesized on a Gene Assembler Plus apparatus (Pharmacia) by the deoxyphosphoamidite method. Their sequence is shown in Fig. 1.

FIG. 1.

Schematic diagram of a retroviral LTR showing the region analyzed in this study. Sequences from a portion of the Cas-Br-E and GV1.4 U3 regions are compared. Asterisks show differences between Cas-Br-E and Graffi sequences. The sequence of synthetic oligonucleotides C4, G4, and CG5 used in EMSA is underlined. Nucleotides are numbered from the 5′ end of U3. Conserved binding sites for other transcription factors (CBF, LVb, NF1, and E-box) are indicated.

Protein-DNA interaction analyses. (i) Electrophoretic mobility shift assays (EMSA).

A 50-ng portion of sense oligonucleotide was 5′-end labeled with T4 polynucleotide kinase and [γ-³²P]ATP and annealed with 200 ng of the complementary oligonucleotide. A 10-μg portion of nuclear extract was added to 0.8 ng of the labeled double-stranded oligonucleotide in 10 mM HEPES (pH 7.9) containing 4% glycerol, 1% Ficoll, 200 μg of poly(dI-dC) per ml, 25 mM KCl, 1 mM dithiothreitol, 0.5 mM EDTA, and 25 mM NaCl (total volume, 10 μl). After a 20-min incubation at room temperature, 1 μl of a 20% Ficoll–0.2% bromophenol blue–0.2% xylene cyanol solution was added and the samples were run on a 4% acrylamide nondenaturing gel in 0.5× Tris-borate-EDTA (TBE) at 150 V for 90 min. The dried gels were autoradiographed on Kodak X-Omat films. For competition assays, a 100-fold excess (80 ng) of cold double-stranded oligonucleotide was added before addition of the nuclear extract. For supershift assays, nuclear extracts were incubated with 1 to 2 μl of antibodies for 30 min at 0°C.

(ii) Methylation interference.

A 50-ng portion of 5′-end-labeled oligonucleotide (upper or lower strand) was partially methylated with dimethyl sulfate as described previously (57). After two ethanol precipitations, the oligonucleotide was annealed with 200 ng of the complementary oligonucleotide. About 3 × 10⁵ cpm of oligonucleotide was subjected to EMSA as described above, except that the reaction mixture was scaled up threefold. The wet gel was autoradiographed, and the free and bound oligonucleotides were eluted separately in 300 μl of elution buffer, extracted with phenol-chloroform, and ethanol precipitated. The recovered oligonucleotides were then cleaved by piperidine treatment, and analyzed on a 10% acrylamide sequencing gel.

Plasmid construction.

Standard methods (57) were used in all plasmid constructions.

LTR-chloramphenicol acetyltransferase (CAT) plasmids are derived from pSV0CAT, donated by C. Gorman (27). To construct pCasCAT, the XbaI-PstI fragment from a permuted molecular clone of Cas-Br-E (53), containing two LTRs, was inserted in the HindIII site of pSV0CAT with HindIII linkers. One LTR was then removed by KpnI digestion followed by self-ligation. The resulting plasmid contains 465 nucleotides of the envelope gene, the entire LTR, and about 100 nucleotides of the gag gene, located upstream from the bacterial CAT gene. The others constructions were derived from pCasCAT, by exchanging the HindIII-KpnI (Env-U3) fragment with a 0.6-kbp RsaI-KpnI fragment from Graffi clone GV1.2 (56) to generate pGV1.2CAT and with a 0.9-kbp RsaI-KpnI fragment from Graffi clone GV1.4 (56) to generate pGV1.4CAT.

Simian virus 40–GATA-1 was constructed by inserting a SalI-XbaI fragment containing the GATA-1 coding sequence into the pSV2 vector (27). pMT2 and pMT2–GATA-2, in which the human GATA-2 cDNA is expressed from the adenovirus major late promoter, were obtained from S. Orkin (22). Mouse GATA-3 expression vectors consisting of GATA-3 cDNA in the sense or antisense orientation under the control of Rous sarcoma virus LTR were obtained from J. Engel.

Transient-transfection assays.

COS-7 cells were transfected with DEAE-dextran, using a 10% dimethyl sulfoxide shock with 2.5 μg of LTR-CAT plasmid and 0 to 7.5 μg of GATA expression vector (made up to 7.5 μg with the empty vector when necessary). At 48 h after transfection, the cells were washed in cold phosphate-buffered saline and whole-cell extracts were prepared by three freeze-thaw cycles in 0.25 M Tris-HCl (pH 7.8). Cell extracts were clarified by a 5-min centrifugation at 10,000 × g. The protein concentration was assessed by the Bradford method (Bio-Rad), and equal amounts of total protein were used for the determination of CAT activity. CAT activity was measured by using [¹⁴C]acetyl coenzyme A and extraction with organic solvents as described previously (57). The results were expressed as the amount of radioactivity present in the organic phase, from which was subtracted the result obtained with mock-transfected cells (or cells transfected with pSV0CAT). We chose not to use a cotransfected expression vector, like Rous sarcoma virus–β-galactosidase to monitor transfection efficiency, because we have shown that it can lead to experimental bias, since the β-gal expression level can be influenced by the presence of other plasmids (8). Instead, we performed each experiment several times, at least in triplicate.

RESULTS

A distinct CG5 binding factor is present in erythroid, myeloid, and lymphoid cell extracts.

A previous analysis of the factors binding to Cas-Br-E and Graffi U3 regions by footprinting and EMSA has demonstrated the existence of an abundant, hematopoietic cell-specific factor, binding to a very well conserved, central part of U3 (200 bp from the 5′ end [Fig. 1]) (1a). We have attempted to identify this factor.

The sequence found protected in footprint analyses is present in Cas-Br-E and Graffi clone 1.4 in this region, and so one double-stranded oligonucleotide (CG5) spanning the entire protected region was used to study DNA-protein interactions. EMSA performed with nuclear extracts from several hematopoietic cell lines revealed one strong retarded band (Fig. 2). This signal was competed by a 100-fold excess of cold CG5 oligonucleotide but not by a nonspecific oligonucleotide demonstrating the specificity of the binding (Fig. 2, lanes 2 to 13). Interestingly, a slight difference in migration was observed with extracts from different hematopoietic cell types: the complex formed with all myeloid cell extracts had a slower migration than with T-lymphoid cell extracts (lanes 15 to 17), and erythroid cell extracts showed a even faster-migrating complex (lane 14). This observation was confirmed by analysis of other myeloid (32Dcl3), lymphoid (BW5147 and EL4), and erythroid (D1B and CB7) cell lines (data not shown). These results suggest that CG5 oligonucleotide can form different complexes, each specific for myeloid, erythroid, and lymphoid cell types. These complexes could be formed either by distinct proteins binding to the same sequence or by the same protein bearing different posttranslational modifications that can alter its mobility.

FIG. 2.

Analysis of the CG5 binding factors by EMSA. End-labeled oligonucleotide CG5 was incubated with nuclear extracts (N.E.) from the following cell lines: MEL (lanes 2 to 4 and 14), Ti-6 (lanes 5 to 7 and 14), M-NFS-60 (lanes 8 to 10 and 16), and M1 (lanes 11 to 13 and 17). Competition (Comp.) was performed with a 100-fold excess of cold CG5 (lanes 3, 6, 9, and 12) or nonspecific (NS) oligonucleotide (lanes 4, 7, 10, 13). Lane 1 contained no nuclear extract.

Methylation interference analysis indicates a contact with two putative GATA binding sites.

To gain more information about the precise sequence recognized by the CG5 binding factor(s), methylation interference analyses were performed with myeloid (M-NFS), erythroid (D1B), and lymphoid (Ti-6) cell extracts. The results (Fig. 3) show similar interference patterns with the three nuclear extracts and indicate that one or more factors make a strong contact with two G residues (positions 10 and 21 from the 5′ end) on the upper strand and a weaker contact with G24 on the lower strand. These nucleotides are part of two potential binding sites for members of the GATA family. Thus, the factors present in myeloid, lymphoid, and erythroid cells recognize exactly the same sequence, consisting of two imperfect GATA sites.

FIG. 3.

Methylation interference analysis of the CG5 binding factors. The analysis was performed on the CG5 probe labeled on either the upper (coding) or lower (noncoding) strand, with M-NFS-60, Ti-6, or D1B nuclear extract. C, control (oligonucleotide not incubated with nuclear extract); F, free probe; B, probe recovered from the DNA-complex band. The sequence of the double-stranded oligonucleotide is shown below. Residues whose methylation interferes partially or totally with the binding are indicated by open and solid circles, respectively. Potential GATA sites are underlined.

CG5 binding factors are members of the GATA family.

Although the identified GATA sites do not match perfectly with the consensus sequence WGATAR (44), we hypothesized that the CG5 binding factors could be different members of the GATA family, which would be consistent with the variation observed in their mobility. GATA-1 was likely to be found in erythroid cells, whereas myeloid and T-lymphoid cell types could contain GATA-2 and GATA-3, respectively.

Indeed, supershift assays with anti-GATA-1 antibodies showed a supershift of the erythroid cell-specific complex only (Fig. 4A, lanes 2 and 4). This identification of the erythroid cell-specific CG5 binding factor as GATA-1 was confirmed by transfecting a GATA-1 expression vector in CHO cells: EMSA performed with extracts from GATA-1-transfected but not mock-transfected CHO cells showed a complex similar to that obtained with D1B nuclear extracts (Fig. 4A, lanes 5 and 6). Similarly, the signal observed with myeloid cell extracts could be supershifted by an anti-GATA-2 antiserum but not by a preimmune serum (Fig. 4B, lanes 1 to 7); furthermore, nuclear extracts of COS-7 cells transfected with a GATA-2-expressing plasmid elicited a very similar signal, which could also be supershifted by anti-GATA-2 antibodies (Fig. 4B, lanes 8 to 11). Lastly, the complex elicited by T-cell extracts contains GATA-3, since it could be supershifted only by anti-GATA-3 antibodies (Fig. 4C). No cross-reaction was observed between the anti-GATA-1, anti-GATA-2, and anti-GATA-3 antisera, confirming our identification of three different GATA family members binding to the same sequence (data not shown).

FIG. 4.

Identification of the CG5 binding factor. (A) GATA-1. EMSA was performed with labeled CG5 and nuclear extracts (N.E.) from CB7 (lanes 1 and 2), D1B (lanes 3 and 4), or CHO transfected with vector pSV2 (lane 5) or with pSV2 GATA-1 (lane 6). Supershift assays were performed with an anti GATA-1 antiserum (lanes 2 and 4). The supershift is indicated by an arrow. (B) GATA-2. CG5 was incubated with the following cell extracts: M-NFS-60 (lanes 2 to 4), M1 (lanes 5 to 7), or COS-7 transfected with vector pMT2 (lane 8) or with pMT2-GATA-2 (lanes 9 to 11). Supershift assays were performed with a preimmune serum (p) (lanes 3, 6, and 10) or an anti-GATA-2 antiserum (G2) (lanes 4, 7, and 11). (C) GATA-3. CG5 was incubated with Ti-6 nuclear extracts and a preimmune serum (p) (lane 2) or an anti-GATA-3 antiserum (G3) (lane 3). (D) GATA factors bind preferentially to the more distal GATA motif. Oligonucleotides bearing mutations destroying the proximal (Mut G1) or distal (Mut G2) GATA sites were used in EMSA with D1B nuclear extracts. Supershift assays were performed with an anti-GATA-1 antiserum. The position of the GATA-1-containing complex is shown, and supershift is indicated by an arrow. The sequence of mutated oligonucleotides is shown below.

EMSA performed with 32Dcl3 cell extracts showed the presence of two specific retarded bands with mobilities similar to those of GATA-1 and GATA-2. The 32Dcl3 cell line is a very immature myeloid cell line, which can be induced to differentiate along the erythroid, monocytic, or granulocytic lineage by using various differentiation-inducing agents (68). These cells were shown to express both GATA-1 and GATA-2 in their undifferentiated stage, as a lot of immature progenitors do (18, 45). Supershift assays confirmed the presence of both GATA-1- and GATA-2-containing complexes (data not shown). The GATA-1 signal was always stronger, suggesting either the presence of more GATA-1 in these cells or a better affinity of this family member for the GATA elements in oligonucleotide CG5.

The presence of only one retarded band in EMSA suggests that GATA factors bind only one of the two GATA elements present in oligonucleotide CG5 at a time, even if the two sites were found occupied in methylation interference analysis. This could be due to steric hindrance or to a much higher affinity of GATA factors for one of the two sites. To determine which of the two GATA motifs was preferentially used, mutations destroying either the proximal (mutG1) or distal (mutG2) GATA element were introduced into oligonucleotide CG5. EMSA and supershift assays performed with these mutated oligonucleotides and D1B nuclear extracts show that destroying the proximal element does not impair the binding of GATA-1, but it is abolished if the distal element is destroyed (Fig. 4D). However, the proximal element must participate in the observed complex, since oligonucleotide mutG1 does not compete as well as the wild-type CG5 for GATA-1 binding (data not shown). Essentially the same results were obtained with extracts from myeloid and T-lymphoid cell lines (data not shown), suggesting that GATA-2 and GATA-3 also have a better affinity for the second GATA element.

GATA-1 binds at least another site in the Cas-Br-E and Graffi U3 regions.

The finding that GATA factors can bind to at least two sites in Cas-Br-E and Graffi U3 regions prompted us to look for other GATA elements, especially in regions found protected by DNase I footprinting (1a). Indeed we identified two other putative GATA sites: one at the 5′ end of the LTR (CCATC, nucleotides 14 to 18), present only in Cas-Br-E LTR, and one next to the core, shared by the two retroviruses (GGATAT [Fig. 1]). Supershift assays failed to identify any GATA-related factor binding to the 5′-end site (data not shown). On the other hand, oligonucleotides C4 and G4, representing nucleotides 135 to 160 in Cas-Br-E and 132 to 157 in GV1.4, respectively (Fig. 1), were shown upon EMSA analysis to bind not only the core binding factor (CBFα/β), present in myeloid and lymphoid cell extracts, but also an erythroid cell-specific factor. This erythroid cell-specific complex was not dependent on an intact core element, suggesting the presence of another binding site in that region (1a). Antibodies against GATA-1 were able to supershift the complex formed with oligonucleotide C4 or G4, although not to the same extent as with oligonucleotide CG5 (Fig. 5, lane 3). Furthermore, when extracts from CHO cells transfected with a GATA-1 expression construct were used in EMSA, a complex similar to the erythroid cell-specific complex was observed (Fig. 5, lanes 4 and 5). Thus, GATA-1 is able to bind the GATA element located near the core but is probably not the only factor responsible for the complex observed with erythroid cell extracts. Antibodies against GATA-2 or GATA-3 failed to show any supershift with erythroid, myeloid, or lymphoid cell extracts.

FIG. 5.

GATA-1 also binds to region IV. EMSA was performed with oligonucleotide C4 and extracts from the following cell lines: D1B (lanes 2 and 3), CHO (lane 4), and CHO transfected with the pSV2 GATA-1 (lane 5). Lane 1, no nuclear extract. Lane 3 shows the supershift assay with an anti-GATA-1 antiserum. The supershift is indicated by an arrow.

GATA factors are able to transactivate expression from Cas-Br-E and Graffi LTRs.

As in any retroviruses, cis-acting elements important for Cas-Br-E and Graffi transcription regulation are likely to be located in the U3 region of the LTR. Since GATA factors are able to bind to more than one site in this region, we were interested in whether they could influence transcription driven from Cas-Br-E and Graffi promoter-enhancer regions. To this end, the entire Cas-Br-E LTR was cloned upstream from the bacterial CAT gene. From this LTR-CAT construct, the Cas-Br-E U3 region was removed and replaced by the same fragment from Graffi LV clone GV1.2 or GV1.4. Each of those three reporter plasmids was cotransfected in COS-7 cells, with expression vectors carrying GATA-1, GATA-2, or GATA-3 coding sequence, and CAT activity was measured 48 h posttransfection. These transient-transfection assays were repeated several times in triplicate, and representative results are shown in Fig. 6.

FIG. 6.

LTR transactivation by GATA factors. LTR-CAT constructs containing Cas-Br-E, GV1.2, or GV1.4 U3 regions were cotransfected with GATA-expressing vectors in COS-7 cells as described in Materials and Methods. CAT activity was measured and compared with the activity observed with an empty expression vector, which was given the arbitrary value of 1. The mean values and standard deviations for a typical triplicate experiment are shown. (A) GATA-1. The LTR-CAT plasmid (2.5 μg) was cotransfected with 0, 2.5, 5, or 7.5 μg of pSV2 GATA-1. The total amount of expression vector was completed to 7.5 μg with the empty vector pSV2. (B) GATA-2. The LTR-CAT plasmid (2.5 μg) was cotransfected with 5 μg of the empty vector pMT2 or 5 μg of pMT2 GATA-2. (C) GATA-3. The LTR-CAT plasmid (2.5 μg) was cotransfected with 5 μg of pRSV GATA-3 in the sense or antisense orientation.

These results suggest that GATA-1 is able to transactivate Cas-Br-E, GV1.2, and GV1.4 LTRs in a dose-dependent fashion with a maximum at 5 μg of expression vector for 2.5 μg of reporter plasmid. At this ratio, Cas-Br-E, GV1.4, and GV1.2 promoter were transactivated 2-, 2.5-, and 2.8-fold (Fig. 6A). Transactivation was also observed with GATA-2, varying from 3.6-fold for GV1.4 to 5.11-fold for GV1.2 (Fig. 6B). Lastly, cotransfection of a GATA-3 expression vector in the sense orientation led to a 1.5- to 2.5-fold transactivation of the viral LTRs compared to the antisense orientation (Fig. 6C). The transactivation rate varied from one experiment to another but was always significant. Thus, these results suggest that three members of the GATA family are able to transactivate Cas-Br-E and Graffi expression. The Graffi clone GV1.2 LTR, which has the strongest activity in COS-7 cells, also showed the greatest transactivation by the three GATA factors. Clone GV1.4 and Cas-Br-E displayed roughly the same transactivation pattern, ranging from 1.5- to 4-fold. The transactivations observed with GATA-1, GATA-2, and GATA-3 cannot be compared since their expression was driven by different promoters.

DISCUSSION

The goal of this work was to identify some of the factors binding to the promoter-enhancer region of two MuLV, Cas-Br-E and Graffi MuLV. The results presented above indicate that at least three members of the GATA family of transcription factors, GATA-1, GATA-2, and GATA-3, are able to bind to three GATA elements and to transactivate LTR-driven expression.

The three GATA motifs identified here do not perfectly match the consensus WGATAR described for GATA binding factors (23, 44). However, analyses of the GATA protein binding specificity by PCR-mediated random-site selection have shown binding to sequences quite different from this consensus: in particular, CGATAT was shown to bind GATA-1 and GATA-2 and TGATGG could weakly bind GATA-2 and GATA-3 (44). Our results indicate that the motifs AGATGG and/or AGAGAT strongly bind GATA-1, GATA-2, and GATA-3 and that GGATAT binds at least GATA-1.

Two putative GATA elements are present in oligonucleotide CG5 (Fig. 1). Methylation of the G residue in the two elements interferes with the binding of the factors present in erythroid, myeloid, and T-lymphoid cell extracts (Fig. 3), which were identified as GATA-1, GATA-2, and GATA-3, respectively (Fig. 4). Hence, the two GATA elements seem capable of binding each GATA protein; however, only one complex, corresponding to one bound factor, was observed in EMSA, even when large amounts of nuclear extracts were used (data not shown). Moreover, mutations destroying the second site clearly abolish the complex formation while mutations destroying the first site only slightly diminish the affinity of the CG5 oligonucleotide for GATA factors, since the mutated oligonucleotide does not compete as well as the wild type (Fig. 4D and data not shown). Thus, the more distal motif AGATAT must be responsible for the complex observed with oligonucleotide CG5. This result is surprising since the same motif is present in oligonucleotide C4, which binds GATA factors only poorly. This difference is probably due to the presence of other sites in oligonucleotide C4, which may interfere with the binding of GATA factors. Also, the motif AGATGG is a better match to the consensus sequence for GATA-1 and GATA-2, determined by PCR-mediated random-site selection (44). However, this site is overlapping with an E-box motif (CAGATG) named E_GRE, well conserved among MuLV (Fig. 1). A factor binding to this motif has been identified and named ALF1. This class A basic HLH (b-HLH) transcription factor has been shown to transactivate the LTRs of several MLVs including Akv, SL3-3, Moloney MuLV, and spleen focus-forming virus (46). Also, this site is included in a perfect consensus site (10-of 10-nucleotide match) for the TAL1/SCL transcription factor (33), a class B b-HLH protein (16, 17). The tal-1 gene was found rearranged in 25% of cases of human acute T-cell lymphoblastic leukemia (4, 16), and knockout studies have identified it as a key factor in the regulation of hematopoiesis (10). Although we have not been able to demonstrate the binding of a TAL-related factor by supershift assays, it has been shown that TAL1 indeed can form heterodimers with ALF1 that bind specifically to this site in the Akv or Moloney MuLV U3 region and can modulate the transcriptional activation of MuLV by ALF-1 (48). Thus, at least three transcription factors can bind to the same region: GATA factors, ALF-1 homodimers, and TAL-1 in heterodimers with class A b-HLH proteins. Interestingly, TAL1 was shown to participate in a large oligomeric complex with E47, GATA-1, Lmo2, and Ldb/NL1, which bind a bipartite DNA motif comprising an E-box followed 9 bp downstream by a GATA site (69). This bipartite motif is reminiscent of the E-box/GATA motif present in oligonucleotide CG5, although the spacing between the two motifs is different. Thus, tal-1 could interact with GATA-1 and other, unknown factors to form a large transactivating complex.

Another putative GATA site is present in oligonucleotide C4 (Cas-Br-E) or G4 (Graffi). A GATA-related complex was observed only with erythroid cell nuclear extracts. This complex was slightly supershifted by anti-GATA-1 antibodies, and extracts from GATA-1-expressing CHO cells gave rise to a complex with identical mobility (Fig. 5). These observations indicate that GATA-1 is able to bind to this site, perhaps with a weak affinity. The anti-GATA-1 antiserum was not able to supershift the entire band, suggesting that this retarded band is composed of more than one complex with similar mobilities, one of which contains GATA-1. This GATA element is located next to the core element and overlaps with an Ets binding site (EBS), which is well conserved among MuLV (26). Ets family members can bind to the EBS motif (also called LVt) in the Moloney MuLV enhancer (30, 43) and transactivate LTR-driven expression (62). We have already shown that when myeloid or lymphoid cell extracts are used the complex observed is formed by the heterodimer CBFα/β (1a). Although the EMSA studies were done with an excess of probe, binding of GATA family members seems to be observed only with extracts containing small amounts of CBF, like erythroid nuclear extracts, and we were unable to show any binding of Ets family members. Experimental conditions could be responsible for these observations. Indeed, the type of gel used in EMSA has been shown to greatly influence the complex observed: complexes involving either CBF or LVt are observed with the Mo-MuLV enhancer, depending on the buffer used in the gel (43). Possibly, GATA-2- or GATA-3-containing complexes would be observed with myeloid and lymphoid cell extracts under different EMSA conditions. Nevertheless, at least three different transcription factors are able to bind to a region as small as 15 bp. The exact combination of bound factors is probably determined by the availability and relative abundance of each factor in the infected cell, their respective affinity, as well as interactions with adjacent complexes.

Our results demonstrate that the binding of GATA factors may have a functional significance, since they are indeed able to activate the transcription from the viral promoter (Fig. 6). This transactivation, although not very strong (1.6- to 5-fold), was reproducibly observed with GATA-1, GATA-2, and GATA-3 on the three viral promoters studied: Cas-Br-E, GV1.2, and GV1.4. A similar activating potential has been reported for GATA factors: a threefold activation was demonstrated for GATA-1 on the platelet glycoprotein IIb (GpIIb) (41) and the SCL (39) gene promoters and for GATA-2 on the endothelin-1 promoter (37, 40). This modest effect is easily explained by the fact that a multitude of transcription factors cooperate to activate LTR-driven transcription, so the effect of one factor is assessed by comparison with an already high transcription level due to the presence of other factors. Our data do not allow the evaluation of the effect of each GATA element on the transactivation. However, the observed transactivation is always stronger for the GV1.2 clone of Graffi virus than for the GV1.4 clone. The GV1.2 U3 region possess two copies of a 50-bp direct repeat and hence two copies of the more proximal GATA element; thus, the stronger activation observed indicates that this proximal GATA site might be involved in the activation. Mutagenesis studies are required to precisely define which GATA elements are used in vivo.

The three members studied here are essential to the normal development of the hematopoietic system and subsequent hematopoiesis (9, 52, 66, 73). GATA-1 is expressed in all erythroid cell lineages, megakaryocytes, mast cells, and mutipotent stem cells and has been involved in the regulation of most erythroid genes (45; reference 50 and references therein). GATA-3 is expressed at all stages of T-cell development (80). GATA-2 has a much broader expression pattern including all myeloid and erythroid cell lineages and endothelial cells (50), and is essential for multilineage hematopoiesis (67). It is likely to act as a general transcription factor (40). It is not surprising that those three key factors in hematopoiesis should be involved in the transcriptional regulation of retroviruses that target primarily hematopoietic cells and cause leukemia. Cas-Br-E and Graffi MuLV both induce myeloid leukemia; therefore, the target cell is most probably a myeloid progenitor which expresses GATA-2 and possibly GATA-1. The GATA elements described here are well conserved in other murine retroviruses that induce erythroleukemia (Friend MuLV) or T-cell lymphoma (Moloney MuLV, SL3-3, radiation leukemia virus, etc.) (Fig. 7). In these cases, the target cell for transformation is likely to express GATA-1 or GATA-3, respectively; therefore, transcription of these viruses may also be regulated by GATA factors. The different members of the GATA family seem to be highly interchangeable (51); for example, GATA-3 and GATA-4 can compensate for a GATA-1 defect (9). Indeed, our results show that the three members GATA-1, GATA-2, and GATA-3 are able to activate transcription from retroviral promoters. Hence, GATA factors alone are probably not responsible for the tissue and disease specificity of Cas-Br-E, Graffi, or other MuLV. On the contrary, they could ensure a strong expression in a very broad spectrum of lineages: instead of one ubiquitous factor, a set of different but very similar factors could be used in different cell types. One could hypothesize that GATA elements could participate in tissue specificity by cooperating with more specific elements. Cooperation of GATA factors with other transcription factors has been described in several cases: GATA-1 cooperates with Ets to activate the platelet glycoprotein Ib and IIb (GpIb and GpIIb) promoters (31, 41) and also with SP1 and EKLF (29); GATA-2 cooperates with the AP-1 complex to activate the endothelin 1 promoter (37); cooperation with CREB and with SP1 has also been described (39, 61). In some cases, a direct protein-protein contact has been demonstrated (29). Such a cooperation could be possible between GATA factors and CBF, since the more proximal GATA site is located next to the core motif. CBF is known to act almost always in cooperation with other transcription factors. A cooperation between CBF and Ets family members to activate the Moloney MuLV enhancer has been demonstrated, with a simultaneous binding of CBF to the core and an Ets protein to the EBS, which overlaps the GATA motif (62). Also, CBF can cooperate with Myb to transactivate the SL3-3 enhancer (77). It is tempting to hypothesize that CBF can cooperate with either Ets factors or GATA factors, depending on the relative amounts of these factors in the infected cell. However, we did not observe any cooperation between GATA and CBFα in cotransfection experiments. Moreover, there was no additive effect of GATA over CBF transactivation and vice versa (data not shown). This suggests that either CBF or GATA factors can regulate MuLV transcriptional activity, depending on their relative abundance in a given cell type.

FIG. 7.

Conservation of GATA binding sites among various MuLV. The sequence of the GATA element in oligonucleotide C4 and the proximal (p) and distal (d) oligonucleotide CG5 are shown. Data from references 26, 54, and 56.

A transactivation of other MuLV by GATA-1 has been reported: in transient transfection in CV-1 cells, GATA-1 was shown to transactivate Friend MuLV and Moloney MuLV LTRs 15- and 4-fold, respectively (15). In COS-7 cells, we observed a more modest fourfold induction of the Friend MuLV LTR by GATA-1 (data not shown), which is comparable to the induction of 2- and 2.8-fold obtained with Cas-Br-E and Graffi, respectively. However, a strong activation of the Friend MuLV LTR by GATA-1 would not be surprising since it possesses additional perfect GATA sites. Since Friend MuLV is a truly erythroid cell-specific virus, it would be interesting to assess its transcriptional activation by other, nonerythroid members of the GATA family like GATA-3, to determine if the specificity of this virus for erythroid cell types is related to a specific transcriptional regulation by GATA-1.

Interestingly, we found that in 70% of tumors induced by Cas-Br-E, the fli-1 gene is activated by a promoter insertion mechanism (5). All the proviruses were found integrated in the same transcriptional orientation as fli-1, within 30 bp at the end of exon 1 (2, 54). We showed recently the presence of a EBS-GATA dual binding site in this exon 1 which binds GATA-1 and the Ets family member Spi1/PU-1, and we hypothesized that Spi1 could negatively regulate fli-1 (3). The integration of Cas-Br-E provirus downstream of this site bypasses this regulation; however, fli-1 may still be regulated by GATA and Ets factors, this time through the LTR, leading to overexpression.

The implication of GATA factors in the transcription regulation of MuLV adds even more complexity to the already complex model array of cis-acting elements in the regulatory region and stresses the importance of interactions between several regulatory elements to achieve a fine regulation of transcription.