Alternative Promoters Regulate Transcription of the Mouse GATA-2 Gene*

Transcription factor GATA-2 has been shown to be a key regulator in hematopoietic progenitor cells. To elucidate how the expression of the GATA-2 gene is con-trolled, we isolated the mouse GATA-2 (mGATA-2) gene. Transcription of mGATA-2 mRNAs was found to initiate from two distinct first exons, both of which encode en-tirely untranslated regions, while the remaining five exons are shared by each of the two divergent mRNAs. Reverse transcriptase-polymerase chain reaction analysis revealed that GATA-2 mRNA initiated at the upstream first exon (IS) in Sca-1 1 /c-kit 1 hematopoietic progenitor cells, whereas mRNA that initiates at the downstream first exon (IG) is expressed in all tissues and cell lines that express GATA-2. While the structure of the IG exon/promoter shows high similarity to those of the Xenopus and human GATA-2 genes, the IS exon/ promoter has not been described previously. When we examined the regulation contributing to IS transcription using transient transfection assays, we found that sequences lying between 2 79 and 2 61 are critical for the cell type-specific activity of the IS promoter. DNase I footprinting experiments and electrophoretic mobility shift assays demonstrated the binding of transcription factors to this region. These data indicate that the proximal 80 base pair region of IS promoter Internal authenticity out DOTAP methylsulfate; Autolumat luminometer b activity co-transfected pENL plasmid SK( 20 m g plasmid di- and Kpn I and by I and Sac I antisense probe. These fragments purified and labeled with [ 32 P]dCTP with the Klenow fragment of DNA polymerase I (TaKaRa). Aliquots of the probes (2 3 10 4 cpm) were mixed with the purified nuclear extract (0.3 to 1.5 m g) in 100 m l of binding buffer containing 10 m M Tris-HCl, 75 m M KCl (pH 8.0), 2 m M dithiothreitol, 50 m g of bovine serum albumin, and 250 m g of poly(dA-dT/dA-dT) (Sigma) and incubated for 30 min at 25 °C. DNase I (0.15 unit) was then added to the protein-DNA mixture. After incubation for 2 min in a 25 °C water bath, the reaction was stopped by addition of 30 m l of stop solution (1.5 M sodium acetate, 20 m M EDTA, and 100 m g/ml tRNA). DNA fragments were analyzed on a 6% polyacrylamide, 7 M urea gel. Electrophoretic Mobility Shift Assay (EMSA)— Nuclear extracts were prepared as described in the previous section, and EMSA was carried out as described (1). Complementary oligonucleotides were annealed, and the resulting double-stranded DNA was used as probes. Probes used for DNA-binding assays correspond to 2 85 to 2 51 (5 9 -GGCAC-CCTCCTGCCCCCCTGCGGCGTTCCCTCCCC-3 9 ). Poly(dA-dT/dA-dT) was used as a nonspecific competitor, and 100-fold excesses of unlabeled double-stranded oligonucleotides were used as sequence-specific com- petitors. Sequences of the mutant oligonucleotides used as competitors were identical to those used in the LUC mutant reporter constructs. Probes were incubated for 30 min at room temperature in 20 m l of reaction mixture containing 1.5 m g of partially purified nuclear extract. The

GATA-2 is a member of the GATA transcription factor family that binds to a common consensus sequence motif (A/ T)GATA(A/G) through a highly conserved zinc finger DNA binding domain (1). The founding member of the GATA family is GATA-1 (2)(3)(4), whose expression is restricted to erythroid, megakaryocytic, mast and eosinophil hematopoietic cell lineages (4 -6), and Sertoli cells in the testis (7,8). Targeted disruption of the gene revealed that loss of GATA-1 function results in a severe deficit in both primitive and definitive erythropoiesis (9 -11).
The tissue distribution of GATA-2 is wider than that of GATA-1, i.e. GATA-2 is expressed not only in several hemato-poietic lineages that also express GATA-1 but in hematopoietic stem and progenitor cells, in the embryonic brain, and in endothelial cells, fibroblasts, kidney, liver, and cardiac muscle (1,6,(12)(13)(14). The expression profile of GATA-2 during the development of hematopoietic tissues has been studied most extensively in Xenopus and zebrafish. In Xenopus, low levels of maternal GATA-2 are present in early embryogenesis (15) while zygotic expression of GATA-2 is localized in the ventrolateral ectoderm and mesoderm, in blood cells, and in the central nervous system (16 -18). A ventralizing transforming growth factor (TGF)-␤ family member, BMP-4, induces Xenopus GATA-2 expression (19). In zebrafish, GATA-2 is initially detected in the ventral ectoderm (20). High levels of GATA-2 are expressed at the boundary of the embryo and the yolk syncytial layer, and then by the 2-5 somite stage, in the presumptive hematopoietic progenitors that border the anterior and posterior of the embryo (20). Thus, zygotic GATA-2 induction in vertebrates is coincident with the commencement of hematopoiesis, suggesting that GATA-2 is intimately involved in the development of hematopoietic cell lineages.
Analysis of chimeric mice generated from GATA-2 (Ϫ/Ϫ) embryonic stem (ES) 1 cells revealed that GATA-2 (Ϫ/Ϫ) cells do not contribute to any hematopoietic lineage (21). GATA-2-null mutant mice die of severe anemia at approximately 10 to 11 days post coitus during embryonic development (21) and manifest a broad hematopoietic deficit. These studies demonstrated that GATA-2 plays a unique and crucial role in the differentiation of hematopoietic progenitor cells.
It should be noted that GATA-2 expression is not found in mature hematopoietic cells (22,23). In a human bone marrow cell culture system or in an in vitro ES cell differentiation, GATA-2 expression was down-regulated coordinately with the progression of erythroid and myeloid cell differentiation (12,13,24). Furthermore, forced expression of GATA-2 in an avian erythroid cell line caused the proliferation of immature cells and markedly inhibited terminal erythroid differentiation (25). These results taken together suggested that the expression of GATA-2 is related to proliferative versus differentiated capacity of hematopoietic progenitor cells and that suppression of GATA-2 activity might be essential for hematopoietic differentiation. Therefore, elucidation of the mechanisms controlling GATA-2 transcription may offer insight into the earliest aspects of hematopoietic cellular differentiation and may also give rise to a means of experimentally manipulating the proliferation versus differentiation of hematopoietic stem cells.
In this study, we have isolated and characterized the murine GATA-2 gene. We found that GATA-2 expression is regulated by two distinct promoters. One is used in general in the tissues and cell lines that express GATA-2, so we refer to this as the IG promoter. The structure of the IG promoter is homologous to that of the previously reported Xenopus and human GATA-2 gene promoters (26,27). The other promoter regulates the expression of GATA-2 specifically in hematopoietic cells, thus we refer to this as the IS promoter. Since this promoter has not been described elsewhere and displays lineage-restricted utilization, we have also characterized the regulatory domains that specify IS promoter activity.

EXPERIMENTAL PROCEDURES
Isolation and Subcloning of Genomic Clones-A C57Bl2/J mouse genomic library (CLONTECH) was screened with a random primed 0.6-kbp SacI fragment, encoding the 5Ј region of mGATA-2 cDNA, and a 0.8-kbp fragment from the HindIII site in exon VI to the poly(A) of mGATA-2 cDNA. Both fragments were derived from an mGATA-2 cDNA clone, which we previously isolated from a fetal mouse liver cDNA library (DDBJ/GenBank TM /EBI accession number AB000096). 2 Hybridization and washing were carried out as described previously (23). Positive clones were characterized by restriction enzyme mapping and Southern blot hybridization, and insert DNA in the positive phages was subcloned into pBluescript SK(ϩ) vector (Stratagene). Sequences for the exon-intron boundaries and promoter and upstream regions were determined using a Taq polymerase cycle sequencing system and an automated DNA sequencer (Perkin-Elmer Corp.).
Southern Blot Hybridization Analysis of Mouse Genomic DNA-Genomic DNA (25 g) was isolated from an adult C57Bl2/J mouse and digested with BamHI and SacI. After electrophoresis in a 0.8% agarose gel, the DNA fragments were transferred hydrodynamically to a Zeta Probe GT membrane (Bio-Rad, Hercules) using a solution containing 0.4 M NaOH and 0.6 M NaCl. The DNA fragments were cross-linked to the membrane by UV irradiation and hybridized to a 32 P-labeled mGATA-2 cDNA fragment (0.8-kbp HindIII fragment corresponding to a part of exon VI). Hybridization and washing of the membranes were performed at 65°C following the manufacturer protocol.
Cell Culture-A mouse mast cell line P815 (generously provided by Dr. A. Ichikawa) and T cell line BW5147 (from ATCC) were cultured in RPMI 1640 medium supplemented with 10% heat-inactivated fetal bovine serum (JRH Bioscience). Murine erythroleukemia (MEL) cells DS19 (from Dr. Shigeru Sassa at the Rockefeller University) were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. Mouse ProB cell line Ba/F3 (28) was cultured in RPMI 1640 medium supplemented with 10% fetal bovine serum and 10% WEHI-3B cell (29) conditioned medium.
RNA Blotting-Total RNA (20 g) was separated on a 1.0% agarose gel containing formaldehyde and transferred to a Zeta Probe GT membrane. A NotI-SacI-digested cDNA fragment (0.6 kbp, containing 118 bp of exon IS, entire exon II, and 233 bp of exon III), a PstI-SacI-digested genomic DNA fragment (0.6 kbp, containing IS), and a BamHI-SacI genomic DNA fragment (0.75 kbp, containing IG) were used as probes. After exposure, membranes were stripped and rehybridized to a probe for ribosomal RNA (from the Japanese Cancer Research Resources Bank).
RT-PCR Analysis of the Expression from IS and IG Exons-Ten g of RNA from normal mouse tissues or total RNA from 1 ϫ 10 5 cells of each FACS-sorted lineage marker-positive fraction and from 5 ϫ 10 4 cells of Sca-1 ϩ /c-kit ϩ and/or Lin Ϫ fractions were reverse transcribed by Super-script RT (Life Technologies, Inc.) with random hexamer or oligo-dT primer. Products were purified by phenol-chloroform extraction, ethanol precipitation, and one-tenth of the products was used in each PCR reaction. PCR reactions (35 cycles at 96°C for 20 s and 68°C for 60 s) were performed using primer 1 (5Ј-ACAAAAGCGGCTGTCTGCGC-GACG-3Ј), primer 2 (5Ј-CACCCCTATCCCGTGAATCCG-3Ј), and primer 3 (5Ј-AGCTGTGCTGCCTCCATGTAGTTAT-3Ј).
RNase Protection Assay-Total RNA was extracted from cultured cell lines and normal mouse tissues using RNAzol B (TEL-TEST) or ISO-GEN (Wako Pure Chemicals, Osaka, Japan). Two mGATA-2 genomic fragments were subcloned into pBluescript KS(ϩ); one was a 610-bp PstI-SacI fragment containing the 5Ј flanking sequence, IS exon, and the sequence 3Ј to the IS exon, and the other was a 760-bp BamHI-SacI fragment containing the 5Ј flanking region, IG exon, and the sequences 3Ј to the exon IG. The constructs were linearized by SpeI and BamHI, respectively, and a labeled antisense RNA was synthesized from the T7 promoter of the plasmid using T7 RNA polymerase (Promega, Madison). Total RNA samples (100 g) from P815 and Ba/F3 cells were hybridized to the labeled transcript for 12 h at 45°C. The RNA samples were digested with an RNase mixture (Ambion), and the sizes of the RNA hybrids were determined by denaturing polyacrylamide gel electrophoresis. Yeast tRNA (Sigma) and total RNA from DS19 cells were used as negative controls to detect nonspecific bands. Radiolabeled RNA size markers were synthesized using an RNA marker template set (Wako Pure Chemicals).
Primer Extension Assay-An oligonucleotide, 5Ј-CCGGCTCCGGC-CCCTCTGCATCCTC-3Ј, was labeled with [ 32 P]ATP using T4 polynucleotide kinase (Toyobo, Osaka). The labeled primer (1 pmol) was annealed to 50 g of total RNA from P815 cells or tRNA for 90 min at 65°C and then slowly cooled to room temperature in 150 mM KCl, 10 mM Tris-Cl (pH 8.3), and 1 mM EDTA. Reverse transcription was initiated using 10 units of avian myeloblastosis virus reverse transcriptase (Promega) in a solution containing 50 mM KCl, 23 mM Tris-Cl (pH 8.3), 20 mM MgCl 2 , 5 mM dithiothreitol, 0.15 mM dNTPs and 150 g of actinomycin-D/ml and incubated for 60 min at 42°C. The template was then digested with RNase A (2 g) for 15 min at 37°C followed by phenolchloroform extraction, ethanol precipitation, and size fractionation on a denaturing polyacrylamide gel. The TaqTrack sequencing system (Promega) was used to generate a sequence ladder of the IG exon with the same primer.
Rapid Amplification of cDNA Ends Assay-Poly(A) ϩ RNA was prepared from P815 cells with Oligotex-dT30 (TaKaRa). 5Ј-Rapid amplification of cDNA ends (5Ј-RACE) assay was performed utilizing the AmpliFINDER 5Ј-RACE system (CLONTECH). Two g of poly (A) ϩ RNA was reverse transcribed, and mGATA-2 cDNA with the transcription initiation site was amplified using oligonucleotide antisense primer mG23 (5Ј-AGCGGCGTGGCTGGTCGGCC-3Ј, corresponding to nucleotides 211 to 192 of the IS exon) and primers for the tailed linker sequence. Another antisense primer used for PCR amplification was mG26 (5Ј-CGCAGGCAGCCGCTTTTGTC-3Ј, corresponding to nucleotides 124 to 105 of the IS exon). PCR products were subcloned into a pGEMT vector (Promega), and their sequences were determined.
A DNA fragment spanning from 79 bp upstream of the transcription start site of IS exon to the NcoI site at the translation start site in the second exon was ligated to KpnI-XhoI double digested pGL2-Basic, resulting in pGL-79ISIG-II. The NcoI site was blunted by mung bean nuclease (TaKaRa) so that the construct does not contain the mGATA-2 translation start site. Similarly, DNA fragments from the NotI site in the IS exon, from the KpnI site at Ϫ2.0 kbp upstream from the IG exon, and from the BamHI site at Ϫ482 bp 5Ј to the IG exon to the NcoI site were ligated to the KpnI-XhoI double digested pGL2-Basic (pGLNot-II, pGL-2.0IG-II, and pGL-482IG-II). pGL-79IS⌬IG-II was generated from pGL-79ISIG-II by deleting SacI fragment. pGL-128IS, pGL-115IS, pGL-79IS, and pGL-61IS, containing the sequences from Ϫ128, Ϫ115, Ϫ79, or Ϫ61 to ϩ177 of IS exon, respectively, and pGL-125IG, pGL-2 N. Suwabe, N. Minegishi, and M. Yamamoto, unpublished data.
103IG, pGL-31IG, and pGL-19IG, containing the sequences from Ϫ125, Ϫ103, Ϫ31, or Ϫ19 to ϩ37 of IG exon, respectively, were prepared by PCR using sequence-specific oligonucleotides. Internal promoter mutants were constructed by substituting the sequence between Ϫ80 and Ϫ35 of pGL-145IS with a series of overlapping oligonucleotides containing mutations. The authenticity of these mutations were confirmed by DNA sequencing.
P815 cells were transfected with plasmid DNAs by the DEAE-dextran method as described (30). Transfection of plasmid DNAs into Ba/F3 and BW5147 cell lines was carried out by electroporation following methods previously described (31) with slight modifications. DS19 cells were transfected using DOTAP (N-[1-(2,3-dioleoyloxy)propyl]-N,N,N-trimethylammonium methylsulfate; Boehringer-Mannheim). LUC activity in the total cellular extract was determined using an Autolumat luminometer (Bethorude), and the activity was normalized to the ␤-galactosidase activity of a co-transfected pENL control plasmid (32). pGL2-Control (Promega) was used for a positive control. The mean and standard deviation of at least three independent experiments in duplicate were determined.
DNase I footprinting was performed as described previously (35). The IS promoter and upstream promoter regions (Ϫ145 to Ϫ11) were first subcloned into pBluescript SK(ϩ), and 20 g of the plasmid was digested by EagI and KpnI for sense probe and by XhoI and SacI for antisense probe. These fragments were purified and labeled with [ 32 P]dCTP with the Klenow fragment of DNA polymerase I (TaKaRa). Aliquots of the probes (2 ϫ 10 4 cpm) were mixed with the purified nuclear extract (0.3 to 1.5 g) in 100 l of binding buffer containing 10 mM Tris-HCl, 75 mM KCl (pH 8.0), 2 mM dithiothreitol, 50 g of bovine serum albumin, and 250 g of poly(dA-dT/dA-dT) (Sigma) and incubated for 30 min at 25°C. DNase I (0.15 unit) was then added to the protein-DNA mixture. After incubation for 2 min in a 25°C water bath, the reaction was stopped by addition of 30 l of stop solution (1.5 M sodium acetate, 20 mM EDTA, and 100 g/ml tRNA). DNA fragments were analyzed on a 6% polyacrylamide, 7 M urea gel.
Electrophoretic Mobility Shift Assay (EMSA)-Nuclear extracts were prepared as described in the previous section, and EMSA was carried out as described (1). Complementary oligonucleotides were annealed, and the resulting double-stranded DNA was used as probes. Probes used for DNA-binding assays correspond to Ϫ85 to Ϫ51 (5Ј-GGCAC-CCTCCTGCCCCCCTGCGGCGTTCCCTCCCC-3Ј). Poly(dA-dT/dA-dT) was used as a nonspecific competitor, and 100-fold excesses of unlabeled double-stranded oligonucleotides were used as sequence-specific competitors. Sequences of the mutant oligonucleotides used as competitors were identical to those used in the LUC mutant reporter constructs.
Probes were incubated for 30 min at room temperature in 20 l of reaction mixture containing 1.5 g of partially purified nuclear extract. The mixture was then electrophoresed in 4% polyacrylamide gels. For supershift and knockout analysis, 2 l of anti-Sp1 or anti-AP-2 antibody (both from Santa Cruz Biotechnology) was added to the mixture 1 h before the start of incubation.

RESULTS
Structure and Organization of mGATA-2 Gene-To begin to analyze the regulatory mechanisms controlling transcription in the mGATA-2 gene, genomic DNA fragments containing the gene were cloned and characterized. Using the 5Ј end region and the 3Ј untranslated region (UTR) of an mGATA-2 cDNA clone, which was previously isolated from a fetal liver cDNA library, a mouse genomic library was screened, and the recovered recombinants were plaque-purified. Twelve independent mGATA-2 bacteriophage genomic clones were characterized by restriction enzyme site mapping and Southern blot hybridization. The genomic phage clones were found to be largely overlapping and to cover the entire coding region of the mGATA-2 locus of approximately 14 kbp (Fig. 1A).
We had previously isolated human GATA-2 cDNA and genomic clones and also determined the gene structure including a first exon (23). 3 We found that the 5Ј end sequence of the human GATA-2 cDNA, corresponding to the first exon, did not show high similarity with that of mouse GATA-2 cDNA, whereas other regions show remarkable sequence similarity. Indeed, the structures of the Xenopus (26) and human GATA-2 genes (27) were recently reported, and these were found to be highly homologous to each other. However, we did not find strong similarity in either of those structures to the mouse first exon sequence. For this reason, we speculated that there might be an alternative first exon for the mGATA-2 gene in addition to the one we had identified in both genomic and cDNA clones.
To test this hypothesis, we first carried out Southern blot analysis of the genomic phage clones using the 5Ј end of the human cDNA as a probe. This test was successful and identified a separate first exon to the one already identified. The two alternative first exons and their associated promoters are apparently both functional since mRNAs containing one or the other first exon were detected by RNA blot analyses (see below). We also searched for the human IS promoter/exon by Southern blotting of the hGATA-2 genomic phage clones, using the mouse IS promoter/exon fragment as a probe. However, we could not find the IS exon in our hGATA-2 genomic phage clones (data not shown), probably because those genomic clones contained only 4-kbp sequences upstream of the IG exon and, therefore, insufficient 5Ј sequences to include the IS promoter and exon. C, genomic Southern blot analysis of mGATA-2 gene. High molecular weight DNA was extracted from the liver of adult C57Bl/2J mouse and digested with BamHI and SacI. A HindIII fragment (0.8 kbp) containing exon VI was used as a probe. (Fig. 1A). The two distinct first exons are named IS and IG for the upstream and downstream exons, respectively, and both encode only 5Ј UTR sequence. The common second exon contains the initiation codon for translation of the mGATA-2 protein. The amino and carboxyl zinc fingers are encoded in exons IV and V, respectively. The basic amino acid tail region (36), 3Ј UTR, and the two polyadenylation signals (37) are located within exon VI. The exon-intron boundaries conform to the general splicing consensus (Fig. 1B). Southern blot analysis of genomic DNA confirmed the presence of bands of the predicted size from the restriction map of the phage clones (Fig. 1C).
Tissue-specific Utilization of the IS and IG Exons/Promoters-To elucidate how the promoters preceding the two individual first exons might be differentially utilized, we performed a series of expression analyses. First, RNA blot analysis showed that, of the cell lines examined, GATA-2 mRNA was expressed abundantly in P815 mastocytoma (Fig. 2, lane 1) and Ba/F3 proB lymphocyte (lane 2). In contrast, GATA-2 mRNA levels were undetectable in the DS19 murine erythroleukemia (lane 3) and BW5147 T lymphocyte (lane 4) cell lines. When we carried out additional analyses using IS or IG exon-specific probes, the experiment showed that IS mRNA was transcribed abundantly in P815 cells but only poorly in Ba/F3 cells, whereas IG mRNA is expressed in both P815 and Ba/F3 cells (Fig. 2). RNA blot analysis using these three probes all showed the presence of two mRNAs for mouse GATA-2. In this regard, we previously found the presence of two GATA-2 mRNAs in human and presented evidence indicating that the two bands reflect alternative usage of two polyadenylation consensus signals (23). Since the two polyadenylation consensus signals are also conserved in the mouse GATA-2 gene (see above), we assume that the usage of multiple polyadenylation signals is also true for the mouse GATA-2 gene.
We next examined the expression of IS and IG mRNAs by RT-PCR. We used three primers for this purpose; primers 1 and 2 were specific for the sense strand sequence of the IS and IG exon, respectively, and primer 3 for the antisense strand sequence of the common exon II (Fig. 3A). RNA from P815 cells was used as a positive control for the expression of both types of mGATA-2 mRNAs (Fig. 2B). This analysis revealed that the IG mRNA is expressed widely in tissues that express GATA-2, whereas IS mRNA could not be detected in these same tissues (Fig. 3B). Since we isolated cDNA clones containing the IS exon from a fetal liver cDNA library, we tested the possibility that the IS exon might be used more often in differentiating hematopoietic cells.
To examine in further detail the possibility that IS and IG mRNAs might be differentially expressed during hematopoietic differentiation, we examined FACS-sorted hematopoietic cell fractions using these same RT-PCR methods. We used antibodies against several lineage markers (Lin) and Sca-1 and c-kit for this purpose (29). Both the IS and IG mRNAs were found to be expressed in the immature hematopoietic Lin Ϫ cell population as well as in the even more immature c-kit ϩ /Sca-1 ϩ /Lin Ϫ cells (Fig. 3C). In addition, IS mRNA was weakly expressed in NK cells. These results show that the GATA-2 IS promoter is active in immature (Lin Ϫ ) hematopoietic cell lineages.
Transcription Initiation Sites for the Alternate mGATA-2 First Exons-To identify the transcription initiation sites of the IS and IG exons, RNase protection assays were performed using RNA samples recovered from P815, Ba/F3, and DS19 cells. We found that both the IS and IG probes were protected in P815 cells (Fig. 4A) and in Ba/F3 cells (upon longer exposure; data not shown). The sizes of the major protected fragments of the IS and IG probes were estimated to be 280 bp (Fig. 4A, lane  4) and 216 bp (lane 11), respectively. There were several extra bands protected by both probes, suggesting the existence of minor transcription start sites; in particular, the IG protection probe gave rise to bands around 270 bp (lane 11).
The assignment of the start site of the IG exon was confirmed by primer extension analysis (Fig. 4B). The result coincided well with the results from the RNase protection assay. Our assignment of the mouse IG start site is close to the reported transcription initiation sites of Xenopus and human GATA-2 mRNAs (26,27). The transcription initiation site of the IS exon was confirmed using 5Ј RACE, employing poly (A) ϩ RNA recovered from P815 cells. Six independent transformants were selected from the RACE cloning products, and their sequences were determined (data not shown). All 5Ј ends of the selected clones were consistent with the major transcription initiation site determined from the RNase protection assay (6/6). The IS and IG exon transcription initiation sites, superimposed on the exons (boxed areas) and variable amounts of promoter sequence, are shown in Fig. 5 (see below).
Structure of the Promoter and Upstream Regions of the IS and IG Exons-We next determined the sequences of the promoter and upstream regions of both the IS and IG exons (Fig.  5). We analyzed these sequences by TFSEARCH 4 which searches within sequenced fragments versus the TFMATRIX transcription factor binding site profile data base by E. Wingender, R. Knueppel, P. Dietze, and H. Karas (GBF-Braunschweig). We found a number of possible transcription factors binding sites, including some expressed uniquely in hematopoietic cells, in this search and in other literature searches (described under "Discussion"). These positions of these transcription factor binding sites are indicated in Fig. 5.
In the IS promoter, a TATA box-like AAATAAAAA element was found between position Ϫ33 to Ϫ25 relative to the major IS start site. Four consensus binding sequences for GATA factors can be found in this region, with three of them clustered at a position around Ϫ1,050. Of the GATA motifs, two are in a palindrome position separated by only 5 bp relative to one another, whereas two other GATA sites (at Ϫ1,095 and Ϫ645) are palindromic with incomplete GATA motifs: AGATAC and GGATAA, respectively (see underlined sequences in Fig. 5, left). The AGATAC motif, in particular, has been shown to be a preferential binding site for GATA-2 and GATA-3 over a GATA-1 preference (38).
Since it had not been previously reported, we also determined the mouse IG exon/promoter sequence. The sequence is highly conserved with that of human GATA-2 (Fig. 5, right) (27). In particular, the double CCAAT boxes that were shown to be important for the expression of Xenopus GATA-2 during early embryonic development are also found in the mouse IG promoter region (26).
Trans-activation Activity of the IS Promoter in P815 Cells-We next attempted to localize cis-acting elements involved in the determination of the hematopoietic cell-specific expression of the IS promoter. We first examined six constructs bearing rather large deletions of 5Ј sequence information containing the IS promoter. The largest construct in this analysis is pGL-6.0IS, containing 6.0 kbp upstream from the transcription initiation site and 177 bp of exon IS (to a NotI site in the IS exon). The shortest construct we examined was pGL-50IS, in which the LUC reporter gene was driven by a proximal promoter fragment extending only to Ϫ50 relative to the beginning of exon IS. These reporters were transfected into P815 cells. LUC activity was determined using total cellular extracts and normalized for transfection efficiency based on a cotransfected pENL ␤-galactosidase control plasmid. Fig. 6A shows that all the constructs, except pGL-50IS, yielded high and reproducible LUC activity in cells which normally express GATA-2. Deletion of the upstream sequence from Ϫ6.0 kbp to Ϫ145 bp resulted in no significant decrease in promoter activity. In contrast, deletion of an additional 95 bp from the pGL-145IS reporter gene resulted in more than a 5-fold reduction of promoter activity. These results seem to exclude our anticipated possibility that the three (complete and incomplete) palindromic GATA binding motifs in the upstream region might significantly influence (either positively or negatively) GATA-2 transcriptional activity.

FIG. 4. Transcription initiation site of the mGATA-2 IS and IG exons. A,
RNase protection mapping of the IS and IG exons. RNA samples used were from DS19, Ba/F3, and P815 cells. Sizes of the protected bands were estimated by the DNA sequence ladder (lanes 7-10) and by the RNA marker (lanes 5 and 6). Lane 6 was loaded with one-tenth of the amount of the marker that was loaded on lane 5. tRNA was used as a negative control (lanes 1 and 14). Major protected bands in P815 cells are estimated to be 280 bp (IS, lane 4) and 216 bp (IG, lane 11), respectively. Several extra bands suggest the presence of infrequent start sites. Although an RNA from Ba/F3 cells (lanes 3  and 12) show no obvious bands, longer exposure of the gel revealed that the RNA sample does protect the IS and IG probes. The sizes of the protected bands are identical to those in lanes 4 and 11 (data not shown). RNA from DS19 cells that do not express a significant amount of mGATA-2 did not protect both of the probes (lanes 2  and 13). B, primer extension analysis of the IS and IG exons. Lanes 1-4 represent the sequence ladders of IG exon that were generated with the primer used in the primer extension experiment (see "Results"). Use of P815 RNA template gave a single band (lane 5), which was not observed with tRNA template (lane 6).
Since the IS promoter appeared to be significantly influenced by local promoter sequences, we next examined finer mutations in the LUC reporter plasmid pGL-145IS using a PCR method and again transiently transfected these constructs into the same cells. Among these constructs, pGL-79IS showed the strongest transcriptional activity of all the plasmids, but additional deletion of only 18 bp from the pGL-79IS resulted in a consistent 4-fold decrease in LUC reporter activity (Fig. 6B), and similar results were obtained from experiments using Ba/F3 cells (data not shown). These data strongly suggest that a positive regulatory element for IS-directed transcription resides in that 18-bp region.
To refine the position of the important cis-acting positive regulatory elements within the IS promoter region, we prepared a set of substitution mutants and analyzed their effect on expression using P815 as the substrate recipient cells (Fig. 6C). Replacement of the sequences Ϫ36 to Ϫ39, Ϫ42 to Ϫ48, and Ϫ51 to Ϫ54 reduced the transcriptional activity to 52, 29, and 40% of the wild-type pGL-145IS, respectively, indicating that important cis-acting elements exist in this region (proximal to Ϫ54). Substitution of the sequence of Ϫ55 to Ϫ59, Ϫ62 to Ϫ68, and Ϫ74 to Ϫ78 did not affect transcriptional activity. Replacement of the sequence between Ϫ70 and Ϫ73 again affected IS promoter activity, resulting in a reduction to 40% in comparison with the wild-type plasmid. Conversion of the Ϫ55 to Ϫ78 region of the pGL-145IS to GGTACC (shown as del55-78 in Fig. 6C) caused a reduction to 32% of wild type. Therefore, there is a possibility that the Ϫ55 to Ϫ78 deletion might reflect a function that is elicited through that specific cis-acting element (the Ϫ70 to Ϫ73 mutation).
IS and IG Promoter Activities in P815 Cells-We performed a second set of transient transfection assays using constructs containing sequences lying between the IS promoter and the common second exon transfected into P815 cells (Fig. 7A). Transfection of reporter plasmid pGL-79IS⌬IG-II, containing the IS promoter (79 bp), IS exon, and a part of the second exon, resulted in almost identical LUC activity to that generated by pGL-482IG-II, indicating that genuine IS promoter activity is comparable with the IG promoter activity in P815 cells. pGL-79ISIG-II typically generated lower LUC activity than did pGL-482IG-II. Thus the region upstream from the Ϫ482 to IS exon has no stimulating activity in P815 cells (Fig. 7A).
We also prepared a series of deletion mutants of IG promoter and transfected the constructs into P815 cells. This analysis revealed that the region spanning Ϫ103 to Ϫ31 is important for IG promoter activity. The 72-bp region contains double CCAAT boxes that are conserved between the human (27) and Xenopus (26) GATA-2 gene promoters, both of which appear to correspond to the mouse IG promoter. Therefore, we introduced a mutation into the proximal CCAAT box of mGATA-2 IG promoter and found that the mutation resulted in approximately 70% decrease of the promoter activity (data not shown). This result is consistent with that of Xenopus GATA-2 promoter analysis and indicates that the CCAAT box is essential for full IG promoter activity.
Cell-type Specificity of Minimal IS Promoter Activity-To test for cell-type specificity of IS promoter activity, the reporter gene activities of pGL-145IS and pGL-50IS were determined in several culture cell lines. LUC activity generated by pGL-145IS in DS19 and BW5147 cells were 27 and 11% of that generated in P815 cells, respectively, whereas LUC activity in Ba/F3 cells was 99% of that in P815 cells (Fig. 8). The former two cell lines do not express GATA-2 mRNA to a detectable level (see Fig. 2). These differences were not observed when we transfected pGL-50IS construct into the same cell lines. These results indicate that the IS promoter indeed contributes primarily, if not exclusively, to the determination of cell-type specificity of mGATA-2 expression from the IS exon.
The Functional Cis-elements Are Protected in DNase I Footprinting Analysis-We next tried to identify transcription factors that might bind to the IS promoter. To this end, we first carried out DNase I footprinting analysis of these domains using nuclear extracts prepared from P815 (GATA-2 high) or DS19 (GATA-2 low) cells. As shown in Fig. 9A, two regions (downstream from Ϫ54 and from Ϫ61 to Ϫ79) were protected using a sense probe in combination with P815 nuclear extracts (lanes 4 -6). Several hypersensitive sites, another hallmark of specific protein-DNA interactions, were also observed at Ϫ51, Ϫ68, Ϫ75, and Ϫ80 in this experiment. In contrast, the DS19 nuclear extract protected only the region downstream from Ϫ54 and did not produce strong hypersensitive sites (lanes 9 -11). When we carried out the footprinting experiment with the antisense probe, the P815 nuclear extract protected the region proximal of Ϫ54 and Ϫ65 to Ϫ75 from DNase I digestion, whereas the DS19 nuclear extract did not protect these regions effectively (Fig. 9B). Since the DS19 nuclear extract was purified by the same procedure as the P815 nuclear extract and since we used identical amounts of nuclear protein in comparable lanes in this experiment, these results suggest that a transcription regulatory complex is formed more efficiently with the P815 nuclear extract than that with DS19 nuclear extract. The two regions protected by the P815 nuclear extract FIG. 6. Transcriptional activation by the IS promoter and upstream regions. A and B, various GATA-2 IS promoter-LUC reporter constructs were prepared and transfected in P815 cells. The promoter and upstream regions integrated in each reporter are indicated as a number of nucleotides from the transcription start site of IS exon. C, mutational analysis of the GATA-2 IS promoter region. LUC activities are normalized with the co-transfected pENL ␤-galactosidase activity (32) and are presented relative to that of wild-type pGL-145IS. Averages of three independent experiments in duplicate are presented with standard deviation.

FIG. 7. IS and IG promoter activity in transient transfection assay.
A and B, transient transfection analysis into P815 cells was performed with various LUC reporter constructs containing IS and/or IG promoter sequences and intron and intergenic sequences. B, the regulatory region integrated in each reporter is indicated as a number of nucleotides from the transcription start site of IG exon. LUC activity of pGL2-Control is set as 100, and relative LUC activity of each construct is shown. Note that LUC activity generated by pGL-79IS⌬IG-II, which represents genuine IS promoter activity, is comparable with that of pGL-482IG-II containing 482 bp of IG promoter and upstream regions.
(i.e. proximal to Ϫ54 and Ϫ61 to Ϫ79) show very good coincidence with the regions that are important in the transient transfection assays (see above).
Competitive Transcription Factor Binding to GC-rich Sequences in the IS Promoter-To further investigate transcription factor binding to the IS promoter region, we next performed EMSA using the P815 and DS19 nuclear extracts. A probe spanning Ϫ85 to Ϫ51 (Fig. 10A) formed several major bands (bands a to d, see lane 1 of Fig. 10B and lane 8 of Fig.  10C). These bands could be competed with a 100-fold excess of unlabeled competitor (lane 2 of Fig. 10B), indicating that the binding of transcription factors to this probe is specific.
Competition experiments with a series of mutant oligonucleotides suggested that several independent transcription factors bound to this region since no single mutant oligonucleotide competed for all the bands (Fig. 10B, lanes 3 to 6). The mutations at Ϫ70 to Ϫ73 were previously shown to be important for IG promoter activity in transient transfection assays (see above). Competition with the oligonucleotide containing this mutation left bands b and d. EMSA with DS19 cell nuclear extracts exhibited a different pattern of retarded bands (Fig.  10B, lane 7), without obvious bands corresponding to b and d. Addition of an anti-Sp1 antibody to the reaction resulted in supershift of a component of band a (lane 9), whereas addition of an anti-Ap2 antibody did not influence the banding pattern (lane 10). The intensity and mobility of bands b and d were not influenced by the addition of these antibodies, indicating that these bands did not contain Sp1, AP-2, or other transcription factors cross-reactive to these antibodies. These results indicate that transcription factors in bands b and/or d may be directing the IS promoter function in hematopoietic cells. The identity of these two factors remains to be clarified. DISCUSSION Although transcription factor GATA-2 is essential for hematopoiesis (21), its expression is not restricted to hematopoietic lineages. In hematopoietic cells, the expression level of GATA-2 changes dramatically during progressive cell differentiation (12,39). This suggested that mechanisms could exist for regulating the expression of GATA-2 specifically in the hematopoietic lineage. Consistent with this expectation, we report here the presence and activity of an independent hematopoietic promoter in the mGATA-2 gene. The IS promoter is used specifically in hematopoietic progenitors, whereas the IG promoter seems to be responsible for the expression of GATA-2 in various tissues but also including hematopoietic cells. While we found the structure of the IG promoter to be quite homologous to the previously reported human and Xenopus GATA-2 promoters (26,27), the IS promoter was identified for the first time here.
Determination of the complete structure of the mGATA-2 gene revealed an organization resembling that of the other GATA-1/2/3 factors so far defined (23,26,27,40,41) (Fig. 11). The restriction enzyme sites within the mGATA-2 gene determined in this study coincide completely with the map previously reported by Tsai and colleagues (21). Genes for vertebrate GATA-1/2/3 factors are generally composed of six exons. In addition, both GATA-1 (7) and GATA-2 (this study) genes contain two alternative first exons (Fig. 11). The chicken GATA-5 gene, which belongs to another subfamily of GATA factors  6) and DS19 cells (lanes 9 to 11) were incubated with the sense probe (A) or the antisense (B) probe. In both A and B, lanes 4 and 9 contain 1.5 g, lanes 5 and 10 contain 0.75 g, and lanes 6 and 11 contain 0.35 g of the nuclear extracts. L shows DNA footprint reaction without nuclear extract (lanes 2,  3, 7, 8, and 12). G shows a sequence ladder of G reaction (lanes 1 and 13). Hypersensitive sites (closed triangles) and protected areas (open boxes) are schematically shown.
(GATA-4/5/6), also contains two alternative untranslated first exons (42). The use of multiple first exons/promoters has been found in a number of transcription factor genes that are essential for hematopoietic cell development. For instance, the SCL/ TAL-1 gene contains Ia and Ib first exons/promoters, and the Ia promoter has erythroid-specific activity (31,43). The p45 NF-E2 gene also has two promoters, i.e. fetal (f) and adult (a), although use of these promoters is not exclusively restricted (44). We found that GATA-1 expression is regulated by both an erythroid-specific (IE) and Sertoli-specific (IT) first exons/promoters (7,32,45). These observations show that, in many cases, animals have repeatedly adopted the same simple strategy to achieve specific expression of lineage-restricted transcription factors: the use of alternative promoters.
GATA-2 cDNAs were originally cloned from a chicken erythroid cell cDNA library (1), from a human endothelial cell cDNA library (36,46), and from a stage 17 Xenopus embryonic cDNA library (16). In contrast, we have cloned the IS promoter-derived cDNA from a hematopoietic fetal liver cDNA library. We would speculate, based on the expression studies presented here, that the IS-type cDNA clone may not exist in the former libraries. Our preliminary mRNA in situ analysis demonstrates that GATA-2-positive cells increase temporally in the fetal liver and neonatal spleen, both of which are hematopoietic organs. Thereafter, however, GATA-2-positive cells are only rarely detected in various mouse tissues, suggesting that the level of GATA-2 mRNA is high when hematopoietic progenitors are rapidly expanding. 5 An assessment of the trans-activation activity in a transgenic mouse system, using constructs containing both the mGATA-2 promoter driving the expression of the Escherichia coli ␤-galactosidase gene, is now ongoing in this laboratory. Our preliminary data show that the reporter gene expression in the hematopoietic stem cell fraction of bone marrow cells and in aorta/gonad/mesonephros region of mouse embryo is mainly driven by IS promoter. 5 The basic transcriptional activity of the IS promoter was found to be localized within the proximal 80 bp. Specific transcription factor interactions to the Ϫ85 to Ϫ50 region of the IS promoter were identified through DNase I footprinting analysis and EMSA using nuclear extracts from P815 cells. Whereas EMSA competition experiments with mutated oligonucleotides showed that transcription factors binding to the Ϫ73 to Ϫ70 sequence are responsible to the IS promoter activity, no consensus DNA-binding sequence of known transcription factors has been identified for this sequence.
In this regard, we noticed that transcription factors Sp1, MZF-1 (47), CTCF (48), or MAZ/ZF87 (49) has the possibility to bind to the Ϫ58 to Ϫ48 sequence. Mutational analysis of the IS promoter revealed that this region is also responsible for mGATA-2 IS promoter activity. It is interesting to note that MZF-1 is supposed to regulate the expression of c-myb and CD34, both of which are expressed in undifferentiated hematopoietic cells (47,50), and overexpression of MZF-1 inhibits hematopoietic cell development from ES cells (50). These properties are similar to those of GATA-2. We also have noticed the presence of three CCCTC sequences in the IS promoter (Ϫ81 to Ϫ87, Ϫ58 to Ϫ54, and Ϫ50 to Ϫ45). These motifs resemble the cis-elements in the c-myc promoter, to which several transcription factors, such as Sp1, ZF78, MAZ, hnRNPK, and CNBP, can bind (48,51,52). As c-myc controls differentiation, proliferation, and apoptosis of hematopoietic cells (53), factors binding to the CCCTC sequence may also be related to mGATA-2 gene expression in hematopoietic progenitors.
A 31-bp pyrimidine-rich sequence (5Ј-TCTGCGCCGCTTT-CTGCCCCCTCCTGCCCTC-3Ј, Ϫ3831 to Ϫ3800 relative to translation start site) in the zebrafish GATA-2 gene was recently reported to be responsible for neuronal expression of the gene (54) FIG. 11. Organization of GATA factor genes. Organization of the mGATA-2 gene (this study) is highly conserved with human (27) and Xenopus GATA-2 (26) genes. The gene structure is also conserved with that of the mGATA-1 gene (40) and mGATA-3 gene (41). All these genes consist of 5 translated exons and first exons encoding 5Ј UTR. Chicken GATA-5 gene consists of six translated exons and two alternative first exons (42).
transcription initiation sites has not been firmly identified in the zebrafish GATA-2 gene. Interestingly, when the plasmid containing the 7.3-kbp region upstream from the translation start site is linked to a modified green fluorescent protein (GFP), the plasmid expresses GFP in the intermediate cell mass, the early hematopoietic tissue of the zebrafish. Deletion of 2.5 kbp from the 5Ј end of the 7.3-kbp plasmid nearly abolishes the GFP expression specifically in the posterior end of the intermediate cell mass. Since the IS promoter is active in early hematopoietic progenitors in mouse bone marrow, we speculate that the upstream 2.5-kbp region may contain a first exon homologous to the mouse IS exon.
We demonstrated here the presence of two mGATA-2 gene promoters. Of the two promoters, the IS promoter is expressed specifically in hematopoietic lineage cells. Detailed analysis of the cis-acting elements in transfection assays and protein-DNA interaction assays indicated that transcription factors binding to the IS promoter play an important role in determining its specificity. Since GATA-2 is essential for the development of all hematopoietic lineage cells, detailing the mechanisms that regulate GATA-2 gene expression may allow new insights and perspectives into hematopoietic stem cell biology.