Eosinophil-specific regulation of gp91(phox) gene expression by transcription factors GATA-1 and GATA-2.

The glycoprotein gp91(phox) is an essential component of the phagocyte NADPH oxidase and is expressed in eosinophils, neutrophils, monocytes, and B-lymphocytes. We previously suggested an eosinophil-specific mechanism of gp91(phox) gene expression. To elucidate the mechanism, we performed functional assays on deletion mutants of the gp91(phox) promoter in various types of gp91(phox)-expressing cells. A 10-base pair (bp) region from bp -105 to -96 of the promoter activated transcription of the gene in eosinophilic cells, but not in neutrophilic, monocytic, or B-lymphocytic cells. A 2-bp mutation introduced into the GATA site spanning bp -101 to -96 (-98GATA site) of the fragment abolished its activity. Gel shift assays using a GATA competitor and specific antibodies demonstrated that both GATA-1 and GATA-2 specifically bound to the -98GATA site with similar affinities. Individual transfection of GATA-1 and GATA-2 into Jurkat cells, which have neither endogenous GATA-1 nor GATA-2, activated the -105/+12 construct in a -98GATA site-dependent manner. Combined transfection of GATA-1 and GATA-2 activated the promoter less than transfection of GATA-1 alone. These results suggest that GATA-1 is an activator and that GATA-2 is a relative competitive inhibitor of GATA-1 in the expression of the gp91(phox) gene in human eosinophils.

The large subunit of flavocytochrome b 558 (gp91 phox ) is an essential component of the phagocyte NADPH oxidase, which produces superoxide anion to kill parasites and microorganisms. Mutations in the gp91 phox gene such as deletions (1) and substitutions (2) result in X-linked chronic granulomatous disease, which is characterized by severe recurrent infections due to the lack of superoxide generation (3). The gp91 phox gene is expressed in terminally differentiated phagocytes and B-lymphocytes (4,5), indicating the expression of the gene to be both lineage-and differentiation stage-specific.
Several transcription factors regulate gp91 phox gene expression through cis-elements in a 450-base pair (bp) 1 sequence in the 5Ј-flanking region of the gene. A CCAAT displacement protein (CDP/cut) (6,7) suppresses the expression of the gp91 phox gene by binding to four sites at bp Ϫ350, Ϫ220, Ϫ150, and Ϫ110 of the gene in immature myeloid cell lines. Overexpression of CDP inhibits the induction of the gene in myeloid differentiation (8). The down-regulation of its DNA binding activity during differentiation allows the expression of the gene (9,10). CP1, one of the CCAAT box-binding proteins, and BID proteins (binding increased during differentiation) are activators. However, these factors may work after bound CDPs are released because their binding sites overlap with sites for CDP (10,11). Interferon regulatory factor-2 may have dual functions as a transcriptional activator and repressor depending on binding sites in gp91 phox gene expression (12,13). TF1 phox , a component of BID-2, displaces CDP and interferon regulatory factor-2, resulting in increased expression of the gp91 phox gene in the myeloid cell lines (13). We have shown that PU.1 is an essential activator for the expression of the gp91 phox gene in human neutrophils, monocytes, and B-lymphocytes (14). Eklund et al. (15) have proposed a cooperative activation of the gp91 phox gene by PU.1 and interferon regulatory factor-1 in myeloid cell lines.
Eosinophils have an important role in the host defense against pathogenic parasites and microorganisms and produce superoxide anion as do other phagocytes (neutrophils, monocytes, and macrophages) and B-lymphocytes by means of NADPH oxidase activity (16). It has long been thought that the expression mechanism of gp91 phox in eosinophils is the same as that in other phagocytes. However, a restricted expression of gp91 phox in eosinophils from our X-linked chronic granulomatous disease patients (17) 2 implied a certain eosinophil-specific expression mechanism of the gene. No positive regulatory mechanisms of the gene have, however, been shown in eosinophils, although an eosinophil-specific GATA-3 suppression mechanism of the gene was recently suggested by us (18). In this report, we show the eosinophil-specific regulation of the gp91 phox gene by transcription factors GATA-1 and GATA-2 through the GATA-binding site spanning bp Ϫ101 to Ϫ96 of the gene.

MATERIALS AND METHODS
Cell Culture-To prepare a differentiated eosinophilic cell line (HL-60-C15E), HL-60-C15 cells (a kind gift from Dr. Y. Yamaguchi, Kumamoto University, Kumamoto, Japan), the eosinophil-committed subline of the promyelocytic cell line HL-60, were maintained for Ͼ3 months at pH 7.7 in RPMI 1640 medium supplemented with 10% fetal calf serum and 25 mM EPPS (Sigma). They were then treated with 0.5 mM n-butyrate (Sigma) for 5 days for differentiation into eosinophils (19). HL-60 (Riken Cell Bank, Tsukuba, Japan), B-lymphocytic HS-Sultan (Japanese Cancer Center Research Resources Bank, Tokyo), and T-lymphocytic Jurkat (given by Dr. K. Furukawa, Nagoya University) cells were cultured at pH 7.2 in RPMI 1640 medium with 10% fetal calf serum. HL-60 cells were differentiated into neutrophils and monocytes by treatment with 1 M all-trans-retinoic acid (Sigma) and 1 ng/ml recombinant human transforming growth factor-␤ 1 (Roche Molecular Biochemicals, Tokyo) in combination with 0.1 M vitamin D 3 (1,25dihydroxyvitamin D; a gift from Chugai Pharmaceutical Co., Ltd., Tokyo), respectively, for 3 days (20,41). COS-7 cells were maintained in Eagle's minimal essential medium supplemented with 10% fetal calf serum.
Flow Cytometry-Cells (5 ϫ 10 5 ) in 50 l of phosphate-buffered saline containing 0.5% bovine serum albumin were stained by incubation with an appropriate amount of either one of the following monoclonal antibodies or isotype-matched immunoglobulins on ice for 60 min: R-phycoerythrin-conjugated anti-interleukin-5 receptor ␣ chain (Pharmingen, San Diego, CA), fluorescein isothiocyanate-conjugated anti-CD19 (Pharmingen), and R-phycoerythrin-conjugated anti-CD14 (Dako Japan, Kyoto, Japan). These are lineage-specific to eosinophils, B-lymphocytes, and monocytes, respectively. Surface gp91 phox of the cells was stained as described above with fluorescein isothiocyanateconjugated 7D5 (22). 3 These cells were washed twice with phosphatebuffered saline containing 0.5% bovine serum albumin, and their fluorescence intensities were analyzed on a FACScan (Becton Dickinson, La Jolla, CA).
Northern Blot Analysis-Total RNA was prepared from cells with Trizol LS reagent (Life Technologies, Inc.) according to the manufacturer's protocol. The RNA (10 g/lane) was electrophoresed on formaldehyde-containing 0.9% agarose gels, transferred to Hybond TM -N ϩ nylon membranes (Amersham Pharmacia Biotech, Tokyo), and fixed by ultraviolet light. Messenger RNAs were detected by hybridization with probes labeled with [ 32 P]dCTP using a random primer labeling kit (Amersham Pharmacia Biotech). The following DNAs were used as probes for defining cell lineage: a 780-bp full-length cDNA of human major basic protein (MBP; a gift from Dr. I. Nagaoka, Juntendo University, Tokyo), a 725-bp full-length cDNA of human eosinophil cationic protein cloned by polymerase chain reaction, a 500-bp fragment of eosinophil peroxidase (kindly provided by Dr. Y. Yamaguchi), and a 285-bp cDNA for defensin (human neutrophil peptide-3; kindly supplied by Dr. I. Nagaoka). After hybridization patterns by these probes were visualized on a Molecular Imager FX (Bio-Rad), the membrane was washed to remove the first probes and hybridized with a 32 P-labeled 800-bp cDNA of rat glyceraldehyde-3-phosphate dehydrogenase as the internal control of RNA loading.
Plasmids and Site-directed Mutagenesis-Plasmids pGVϪ986/Luc, pGVϪ301/Luc, and pGVϪ267/Luc, with promoter fragments of gp91 phox spanning from the corresponding bp positions to bp ϩ12, were prepared by the exonuclease III deletion method (21) from pGVϪ5635/Luc, which has the fragment of the gene extending from its initiation codon to the upstream bp Ϫ5635 at the NcoI site of the firefly luciferase gene in the pGVϪB2 vector (Toyoink, Tokyo). For making deletion constructs pϪ267/Luc, pϪ115/Luc, pϪ105/Luc, pϪ95/Luc, and pϪ84/Luc, we first made fragments spanning from their corresponding positions to bp ϩ12 by the polymerase chain reaction method using pGVϪ267/Luc as the template and sets of forward primers with KpnI sites and a reverse primer with a BamHI site. These polymerase chain reaction products were digested with KpnI and BamHI and subcloned into the KpnI-and BglII-digested pXP2N of firefly luciferase reporter gene vector, originating from pXP2 (a kind gift from Dr. Y. Yamaguchi). The order of the restriction sites in the multicloning site of pXP2N (5Ј-BamHI/KpnI/ SmaI/SalI/HindII/XhoI/BglII-3Ј) is different from that of pXP2 (5Ј-BamHI/HindIII/SmaI/SalI/KpnI/XhoI/BglII-3Ј). To make pϪ986/Luc and pϪ301/Luc, fragments were cut out from pGVϪ986/Luc and pGVϪ301/Luc, respectively, at the KpnI sites of their multicloning site and the bp Ϫ137 HindIII site of the gp91 phox gene. The corresponding KpnI/HindIII fragment of the pϪ267/Luc reporter construct was replaced by either one of the above fragments. For making the mutant reporter plasmid pϪ105M/Luc, a two-point mutation (GA to CT) was introduced into the bp Ϫ100 to Ϫ99 sequence of pϪ105/Luc by polymerase chain reaction using a mutated primer. Each cDNA for human GATA-1 and GATA-2 was inserted into the pEF-MCIneo vector with a forward orientation (24).
Promoter Activity Assays-For HL-60-C15E and other differentiated HL-60 cells, cells (5 ϫ 10 6 ) were incubated with 20 g of gp91 phox promoter/firefly luciferase plasmid, 10 g of carrier Bluescript II KS plasmid, and 3 g of herpes simplex virus thymidine kinase promoter/ Renilla luciferase plasmid in 0.25 ml of 20 mM HEPES/RPMI 1640 medium (pH 7.2) at room temperature for 15 min and electroporated at 200 V for 70 ms on an ElectroSquarePorator T820 (BTX, Inc., San Diego, CA). HS-Sultan cells (5 ϫ 10 6 ) were incubated with 10 g of gp91 phox promoter firefly luciferase plasmid and 50 ng of cytomegalo virus promoterRenilla luciferase plasmid as described above, but electroporated with a Gene Pulser II (Bio-Rad) at 950 microfarads and 310 V. Jurkat cells (5 ϫ 10 6 ) were electroporated as described for HS-Sultan cells, but in the presence of 5 g of firefly luciferase plasmid and 20 ng of Renilla luciferase plasmid at 950 microfarads and 280 V. After HL-60-C15E, HL-60, and HS-Sultan, Jurkat cells were incubated at 37°C under 5% CO 2 and 95% air for 6, 6, 10, and 24 h, respectively, their reporter activities were measured as described previously (14). In the case of cotransfection experiments, various amounts of human GATA expression plasmids were transfected with 5 g of reporter plasmid into 5 ϫ 10 6 Jurkat cells.
Preparation of Nuclear Extracts, Electrophoretic Mobility Shift Assays (EMSAs), and Scatchard Plots-Prior to nuclear extraction, COS-7 cells (10 7 ) were transfected with 4 g of GATA-1 or GATA-2 plasmid in 5 ml of Eagle's minimal essential medium by 20 nmol of L1-Liposome (kindly supplied by Dai-ichi Pharmaceutical Co., Ltd., Tokyo) for 8 h and cultured in 15 ml of 10% fetal calf serum/Eagle's minimal essential medium for 40 h. Preparation of nuclear protein extracts, labeling of the double-stranded oligonucleotides from bp Ϫ115 to Ϫ90 of the human gp91 phox gene for the probe, and EMSAs were performed as described previously (14). Each reaction mixture (20 l) containing 0.5-14 g of nuclear protein extracts, 10,000 cpm of each radiolabeled probe equivalent to 1.16 fmol, and 1 g of poly(dI-dC)⅐poly(dI-dC) (Amersham Pharmacia Biotech) in 20 mM HEPES (pH 7.9), 50 mM KCl, 1 mM MgCl 2 , 1 mM dithiothreitol, 0.2 mM EDTA, 0.01% Triton X-100, 5% glycerol, and 0.5 mM spermidine was incubated on ice for 15 min. In competition assays, a 300-molar excess of unlabeled competitor oligonucleotides was added prior to the addition of the probe to the mixture, which was then preincubated on ice for 15 min. For the inhibition assay with antibodies, an aliquot of nuclear extracts was incubated on ice with 2-4 g of goat IgG against human GATA-1, murine IgG against human GATA-2, control goat IgG, or control murine IgG for 1 h before the addition of the probe. Both anti-GATA antibodies were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Electrophoresis was performed on a native 6% acrylamide gel in 0.4ϫ buffer containing 36 mM Tris, 36 mM boric acid, and 8 mM EDTA (pH 8.3).
Upper-strand sequences for the probe and of competitors are as follows: sequence for the probe and of the wild-type competitor (bp Ϫ115 to Ϫ90 of the normal fragment of the gp91 phox gene), 5Ј-TATTAGC-CAATTTCTGATAAAAGAAA-3Ј; GATA-binding site competitor of the human T-cell receptor ␦ gene (25), 5Ј-TTATTTAAGTCTTAGTTGTAG-3Ј; GATA site mutant of the wild-type competitor (Ϫ105M), 5Ј-AATT-TCTCTTAAAAGAAAAGGAA-3Ј; and a heterologous competitor, 5Ј-CA-CAACCACATTCAACCTCTGCCACC-3Ј. The underlined sequence represents mutated bases.
Apparent dissociation constants (K d ) were determined according to the methods described by Merika and Orkin (26). The binding and running conditions were same as those described above. Each fixed 3 A. Yamauchi, unpublished data.

FIG. 1. An eosinophilic phenotype of HL-60-C15E cells. Total
RNAs for Northern blot analysis were extracted from parental HL-60 cells and from HL-60-C15E cells differentiated from HL-60-C15 cells by a combination of alkaline and butyrate treatments (see "Materials and Methods"). Probes used were cDNAs for MBP, eosinophil cationic protein (ECP), and eosinophil peroxidase (EPO). After removing the first probe from each filter, it was hybridized with the probe for glyceraldehyde-3-phosphate dehydrogenase mRNA (GPD). amount of nuclear protein extract from COS-7 cells expressing GATA-1 (0.5 g/20 l) or GATA-2 (14 g/20 l) was incubated on ice with serially diluted labeled probe and incubated for 15 min, which was confirmed to be long enough to bring the reaction to equilibrium. Quantitation of free and bound DNAs was performed with a Molecular Imager FX. Scatchard plots for GATA-1 and GATA-2 were accomplished assuming front counts and specifically retarded counts to be proportional to free and bound concentrations, respectively, and also assuming their binding to one DNA duplex to be 1.
Statistical Analyses-Each value shown as the mean Ϯ S.D. was

FIG. 2. Functional analysis of the 5-deletion of the gp91 phox promoter in eosinophilic HL-60-C15E cells.
Luciferase reporter plasmids containing the serially truncated gp91 phox promoter fragments illustrated on the left were transfected into HL-60-C15E cells. Firefly luciferase activities were normalized by the accompanied activities of the cotransfected Renilla luciferase reporter gene (see "Materials and Methods"). Each column and bar are the relative mean of three independent values and the S.D., respectively, assuming the activity of the pϪ105/Luc construct to be 100%. The difference in promoter activities between the pϪ84/Luc and pϪ95/Luc constructs and that between the pϪ95/Luc and pϪ105/Luc constructs are both statistically significant (p Ͻ 0.001).
FIG. 3. Cell lineage dependence of gp91 phox promoter activation by the region from bp Ϫ105 to Ϫ96. A, the promoter activity of pϪ105/Luc relative to that of pϪ95/Luc was measured in a variety of phagocytic cells and B-lymphocytic HS-Sultan cells (B ly. Sultan). HL-60 cells were treated with all-trans-retinoic acid (Ne. HL-60) and transforming growth factor-␤ 1 /vitamin D 3 (Mo. HL-60) to differentiate into neutrophilic and monocytic cells, respectively. HL-60-C15E is indicated by Eo. C15E. These cells were transfected individually with pϪ105/Luc and pϪ95/Luc constructs. Each column and bar indicate the mean of three or more independently assayed ratios, namely, net increases in pϪ105/Luc promoter activity divided by net increases in pϪ95/Luc promoter activity. The ratio is significantly higher than 1.0 (p Ͻ 0.001) only in HL-60-C15E cells. B, demonstration of lineage specificities of differentiated HL-60 cell lines and HS-Sultan cells. In Northern blot analyses, cDNA for neutrophil lineage-specific defensin and that for the common glyceraldehyde-3-phosphate dehydrogenase (G3PDH) were sequentially used as probes (upper panel). Flow cytometry demonstrated the surface molecules CD14, CD19, and the interleukin-5 receptor (IL5R), which are lineage-specific and exclusively expressed on monocytes, B-lymphocytes, and eosinophils, respectively (lower panel). gp91 phox is common to all these committed leukocyte cell lines.
obtained from three or more means of independent duplicates or triplicates. Values of p lower than 0.05 in mostly paired Student's t test were taken to be statistically significant.

RESULTS
Eosinophil-specific Transcriptional Activation of the gp91 phox Gene by the Gene Promoter Region from bp Ϫ105 to Ϫ96 -To analyze an eosinophil-specific cis-element of the gp91 phox promoter, we differentiated HL-60-C15 cells, an eosinophil-committed subline of the promyelocytic HL-60 leukemia cell line, by treatment with butyrate under alkaline pH. As shown in Fig. 1, these treated cells, but not parental HL-60 cells, expressed mRNAs for eosinophil lineage-specific MBP, eosinophil cationic protein, and eosinophil peroxidase. These results support that the cells are phenotypically eosinophilic. We named the cells HL-60-C15E and used them in further experiments as differentiated human eosinophilic cells.
Next, we transiently transfected the pϪ986/Luc construct, which contains the gp91 phox gene fragment from bp Ϫ986 to bp ϩ12 in front of a luciferase reporter gene, into HL-60-C15E cells. As shown in Fig. 2, the pϪ986/Luc construct had a significantly higher reporter activity than the promoterless pXP2N construct. This result indicates that the bp Ϫ986 fragment contains positive regulatory elements for gp91 phox gene expression in eosinophilic cells. To clarify the elements, a series of progressive deletion mutants was generated, and the promoter activity of each was examined (Fig. 2). Deletions of sequences Ϫ986/Ϫ302 (between bp Ϫ986 and Ϫ301), Ϫ301/ Ϫ116, and Ϫ115/Ϫ106 produced no significant changes in promoter activity (pϪ301/Luc, pϪ115/Luc, and pϪ105/Luc). However, a further 10-bp deletion from bp Ϫ105 to Ϫ96 significantly decreased promoter activity by 40% (p Ͻ 0.01; compare pϪ105/Luc and pϪ95/Luc). An additional 11-bp deletion resulted in a further decrease in activity (p Ͻ 0.01; compare pϪ95/Luc and pϪ84/Luc). These results suggest that positive regulatory elements exist in the regions from bp Ϫ105 to Ϫ96 and bp Ϫ95 to Ϫ85 of the gp91 phox promoter.
To determine the lineage specificity of the first region from bp Ϫ105 to Ϫ96, we examined the relative promoter activity of pϪ105/Luc versus pϪ95/Luc in non-eosinophilic gp91 phox -ex- FIG. 4. Functional analysis of the Ϫ98GATA site of the gp91 phox promoter in HL-60-C15E cells. Either the pϪ105/Luc construct, its mutant construct (pϪ105M/Luc) with the mutation GA to CT at the Ϫ98GATA-binding site, or the pϪ95/Luc construct was transfected into HL-60-C15E cells as described under "Materials and Methods." The relative mean luciferase activity of the wildtype construct was arbitrarily set to 100% in three independent experiments. Each column and bar are the mean and S.D., respectively. The mutation abrogates the promoter activity dependent on the bp Ϫ105 to Ϫ95 fragment. pressing cells as well as eosinophilic HL-60-C15E cells (Fig.  3A). The ratio of the promoter activity of pϪ105/Luc to that of pϪ95/Luc was significantly higher than 1.0 in HL-60-C15E cells (1.7 Ϯ 0.19; p Ͻ 0.001), but not in neutrophilic HL-60, monocytic HL-60, or B-lymphocytic HS-Sultan cells. Therefore, at least one eosinophil-specific positive cis-element should exist in the region from bp Ϫ105 to Ϫ96 of the gp91 phox promoter. On the other hand, the Ϫ95/Ϫ84 region had no specificity for eosinophilic cells (data not shown). The lineage specificities of these gp91 phox -expressing cells were confirmed by particular expressions of CD14, CD19, and the interleukin-5 receptor on surfaces of transforming growth factor-␤ 1 /vitamin D 3 -treated HL-60, HS-Sultan, and HL-60-C15E cells, respectively, and the only significant expression of defensin mRNA in all-trans-retinoic acid-treated HL-60 cells (Fig. 3B).
The GATA-binding Site Is Essential for Activation of the gp91 phox Promoter by the Ϫ105/Ϫ96 Region in HL-60-C15E Cells-To determine cis-elements in the Ϫ105/Ϫ96 region of the gp91 phox promoter, we searched transcription factor-binding sites in the region. A sequence between bp Ϫ101 and Ϫ96 (5Ј-TGATAA-3Ј) perfectly matched with the GATA consensus sequence ((A/T)GATA(A/G)) (27). To determine whether this putative GATA-binding site (Ϫ98GATA site) actually contributed to activation of the gp91 phox promoter by the Ϫ105/Ϫ96 region, we examined the effect of a GATA site mutation (GA to CT) that abolishes the ability to bind GATA proteins from a GATA site (26) on activation in HL-60-C15E cells. As shown in Fig. 4, the mutation completely abolished all the activity dependent on the region, clearly demonstrating that the Ϫ98GATA site is the essential cis-element for the eosinophilspecific activation of the gp91 phox promoter by the Ϫ105/Ϫ96 region.
GATA-1 and GATA-2 Bind to the Ϫ98GATA Site in the gp91 phox Promoter-To identify nuclear protein complexes that specifically bind to the Ϫ98GATA site in HL-60-C15E cells, we performed EMSAs with the probe from bp Ϫ115 to Ϫ90 encompassing the Ϫ98GATA site (Fig. 5). Two nuclear protein complexes (C1 and C2) specifically bound to the probe because binding was abolished by a 300-fold molar excess of wild-type competitor with sequence identical to that of the probe, but not by an excess of the heterologous sequence oligonucleotide (lanes 1-3). Binding was abolished to a similar extent even by a 120-fold molar excess of wild-type competitor (data not shown). An excess of GATA site-mutated oligonucleotide spanning bp Ϫ105 to Ϫ85 of the promoter (Mt), which has the same mutation as pϪ105M/Luc in Fig. 4, failed to abolish the binding of the C1 and C2 complexes (lane 4), indicating that these protein complexes bind to the Ϫ98GATA site. An excess of a GATA-binding sequence derived from the T-cell receptor ␦ gene completely abolished the binding of these complexes to the probe (lane 5), as did the wild-type competitor, suggesting that the C1 and C2 complexes include members of the GATA family. To further characterize these complexes, we performed gel shift immunoassays with anti-GATA-1 and anti-GATA-2 antibodies because mRNAs of GATA-1 and GATA-2 are abundantly expressed in HL-60-C15E cells as well as in human peripheral eosinophils (28) To characterize the individual binding of GATA-1 and GATA-2 to the Ϫ98GATA site in the gp91 phox promoter, EMSA was performed with nuclear extracts from COS-7 cells that transiently expressed either human GATA-1 or GATA-2 cDNA and the probe spanning bp Ϫ115 to Ϫ90 of the gp91 phox promoter (Fig. 6). The binding of GATA-1 (G1) and GATA-2 (G2) was abolished by the unlabeled wild-type competitor (Wt; lanes 3 and 10) and the GATA-binding site of the T-cell receptor ␦ gene (GA; lanes 5 and 12), but not by a heterologous sequence (He; lanes 4 and 11) or the GATA site-mutated competitor (Mt; lanes 6 and 13). Furthermore, the GATA-1 complexes were supershifted by anti-GATA-1 antibody (SS; lane 8), and the GATA-2 complexes were abolished by anti-GATA-2 antibody (lane 15), contrasting with no effects by corresponding control antibodies (lanes 7 and 14). Nuclear extracts from untransfected COS-7 cells made no such complexes (lane 1). These results confirm that GATA-1 and GATA-2 specifically bind to the Ϫ98GATA site and support the use of these nuclear extracts in the kinetic analysis of the interaction between the site and these GATA proteins.
To determine the relative binding affinities of GATA-1 and GATA-2 for the Ϫ98GATA site, we determined the apparent K d values from Scatchard plots obtained from EMSA data (Fig. 7). The amount of bound probe increased with increases in the concentration of the probe (panels A1 and B1). Scatchard plots for GATA-1 (panel A2) and GATA-2 (panel B2) revealed their FIG. 7. Binding affinities of GATA-1 and GATA-2 for the Ϫ98GATA site of the gp91 phox promoter. Various concentrations of the labeled probe (bp Ϫ115 to Ϫ90 fragment of the gp91 phox promoter) were incubated with a fixed concentration of nuclear extract of COS-7 cells that had been transiently transfected with either GATA-1 (panel A1) or GATA-2 (panel B1) as described under "Materials and Methods." G1, G2, and Free indicate the positions of GATA-1, GATA-2, and free probes, respectively. Scatchard plots of GATA-1 (panel A2) and GATA-2 (panel B2) were obtained from the data in panels A1 and B1, respectively, as described under "Materials and Methods." B/F, bound/free. apparent K d values to be 0.13 and 0.10 nM, respectively, suggesting that both GATA proteins have similar affinities for the Ϫ98GATA site. Accordingly, GATA-1 and GATA-2 may efficiently compete with each other for binding to the site.
GATA-1 and GATA-2 Individually Transactivate the gp91 phox Promoter through the Ϫ98GATA Site-The above results indicate that both GATA-1 and GATA-2 similarly bind to the Ϫ98GATA site of the gp91 phox promoter. To determine whether either one or both of the GATA proteins transactivate the gp91 phox promoter, we individually cotransfected various amounts of their expression vectors with wild-type pϪ105/Luc into Jurkat cells, which express neither of the endogenous GATA proteins (Fig. 8). Up to 4 g cotransfected GATA-1 and GATA-2 cDNAs dose-dependently and significantly (p Ͻ 0.01 and 0.05, respectively) increased gp91 phox promoter activity. At doses higher than 6 g, the activity decreased similarly in both plasmids (panel A). The GA-to-CT mutation at the Ϫ98GATA site of pϪ105/Luc significantly decreased the GATA-1-and GATA-2-mediated promoter activities, to 30.5 Ϯ 13.6 and 16.4 Ϯ 16.2%, respectively (panels B and C). These results indicate that individually expressed GATA-1 and GATA-2 transactivate the gp91 phox promoter mostly through the Ϫ98GATA site.
GATA-2 Is a Relative Competitive Inhibitor of GATA-1 for Transactivation of the gp91 phox Promoter-To clarify each role of GATA-1 and GATA-2 in cells expressing both GATA proteins as eosinophils, GATA-1 and GATA-2 plasmids were cotransfected individually or in combination with the wild-type pϪ105/ Luc reporter construct into Jurkat cells (Fig. 9). The amount of GATA-1 plasmid and that of GATA-2 were fixed to 2 g, which gave 85 and 80%, respectively, of their maximal activities (Fig.  8A). Transiently expressed GATA-1 (Fig. 9A, second bar) produced a (15.5 Ϯ 1.5)-fold increase above vehicle activity (p Ͻ 0.001 versus the first bar), and transiently expressed GATA-2 (third bar) produced a (1.0 Ϯ 0.2)-fold increase (p Ͻ 0.001 versus the first bar). The combined expression of the two GATA proteins (fourth bar) exhibited an (8.3 Ϯ 0.6)-fold increase, which is significantly lower than the increase obtained by GATA-1 alone (p Ͻ 0.005), suggesting that GATA-2 inhibits the transactivation of the gp91 phox promoter by GATA-1. This was confirmed in Fig. 9B. The amount of GATA-1 plasmid was also fixed here to 2 g, but the amount of GATA-2 plasmid was increased from 0 to 4 g to reach its maximal activity (Fig. 8A). Total amounts of GATA plasmids were 6 g or less for avoiding their inhibitory effects. The combined expression of increased amounts of GATA-2 along with a fixed amount of GATA-1 impaired the ability of GATA-1 to transactivate the gp91 phox promoter in a dose-dependent fashion. Therefore, GATA-1 is the principal activator for the gp91 phox promoter, and GATA-2 is a relative competitive inhibitor in the presence of both GATA proteins, as in eosinophils. DISCUSSION In a previous report, we suggested the existence of an eosinophil-specific mechanism of gp91 phox gene expression (17). In this study, we have shown that the Ϫ98GATA site works as a positive cis-element for the gp91 phox promoter in an eosinophil lineage, but not in other lineages expressing gp91 phox . EMSAs have shown that GATA-1 and GATA-2 are the only specific DNA-binding proteins for the GATA site in eosinophilic cells. These results suggest that expression of the gp91 phox gene is regulated by the combination of GATA-1 and GATA-2 through the Ϫ98GATA site, particularly in eosinophils. We also found that GATA-1 and GATA-2 individually bind to the Ϫ98GATA site with similar apparent dissociation constants and can individually work as positive regulators for the gp91 phox gene. However, GATA-2 inhibits the transactivation of the gene by GATA-1 in coexisting conditions, as in eosinophils. Taken together, our findings suggest that GATA-1 works as a principal activator and that GATA-2 works as a competitive inhibitor of GATA-1 for the expression of the gp91 phox gene in eosinophils. The negative regulation of the gene by GATA-2 can be done simply by competitive binding to the Ϫ98GATA site with GATA-1 with an assumption that GATA-2 is one order less effective than GATA-1 for the transactivation of the gene. However, inhibition of the interaction between GATA-1 and some protein factors by GATA-2 may also contribute to decreasing the GATA-1-dependent promoter activity of the gene.
At least six members have been identified in the GATA transcription factor family, which recognizes the consensus motif ((A/T)GATA(A/G)) through a highly conserved zinc finger were too small to be illustrated. Total amounts of plasmids were adjusted to 13 g by pEF-MCI in all plots. B and C, five g of pϪ105/Luc construct or its mutant (pϪ105M/Luc) was cotransfected with 2 g of GATA-1 or GATA-2 expression plasmid, respectively, into Jurkat cells. In three sets of independent paired experiments, the net increase in the wild-type promoter activity by GATA was assumed to be 100% (B and C).
DNA-binding domain (27). In hematopoietic tissues, GATA-1, GATA-2, and GATA-3 are expressed in distinct but overlapping patterns (31): GATA-1 in erythrocytes, megakaryocytes, mast cells, basophils, and eosinophils; GATA-2 in all these cells and neutrophils; and GATA-3 in T-lymphocytes, mast cells, and eosinophils. Only eosinophils among gp91 phox -expressing lineages express GATA-1. We can therefore conclude that the positive regulation of the gp91 phox gene by GATA-1 is an eosinophil-specific mechanism. We can also conclude that the negative regulation by GATA-2 is unique in eosinophils, which are the only cells that express both GATA-1 and GATA-2 in gp91 phox -expressing cells. Recently, mechanisms for positive regulation by GATA-1 have been detected in two eosinophilspecific genes, the chicken EOS47 gene (32) and the human MBP gene (30). The MBP gene is, in particular, regulated positively by GATA-1 and negatively by GATA-2, as is the mechanism presented here for the gp91 phox gene (30), although GATA-2 by itself cannot activate the MBP gene. Different from the EOS47 and MBP genes, the gp91 phox gene is not specifically expressed in eosinophil lineages, but its expression is definitely supported by GATA-1, which is particular to eosinophils in gp91 phox -expressing cells. The gp91 phox gene is highly expressed in terminally differentiated peripheral eosinophils, 2 which still express GATA-1 at a high level (28), but the MBP gene at a residual level (33,34). Some critical factor(s) for the expression of the MBP gene, but not that of the gp91 phox gene, might have disappeared in the late stage of terminal differentiation to peripheral eosinophils. The expression of the gp91 phox gene is regulated by various endogenous and exogenous mediators, including interferon-␥, tumor necrosis factor-␣, and lipopolysaccharide (35). Some factors related to allergy and parasitic infections may specifically regulate the gp91 phox gene in eosinophils through the GATA-1 system.
Expression of myeloid-specific genes is regulated by a combination of transcription factors generally expressed in hematopoietic cells (c-Myb, AML-1/CBF␤, and Ets-1) and lineagespecific or -restricted factors (GATA-1, PU.1, and C/EBP) (23). Recently, Yamaguchi et al. (30) demonstrated the transactivation of the MBP P2 promoter by C/EBPs, especially by C/EBP␤. C/EBP␤ may also play a role in the activation of the gp91 phox promoter in eosinophils because possible binding sites of C/EBP are found in the promoter. Previously, we reported the deficient expression of gp91 phox in the neutrophils, monocytes, and B-lymphocytes, but not in the eosinophils, of an X-linked chronic granulomatous disease patient (17). Further analyses suggested that the deficient binding of PU.1 to the gp91 phox promoter due to the point mutation ( Ϫ53 C to T) at its PU.1 motif results in the deficient expression of the gp91 phox gene in the patient (14). These findings suggest that eosinophils have their own gp91 phox gene expression mechanism independent of PU.1. The GATA-mediated mechanism presented here is a strong candidate for it because the transactivation of the gp91 phox gene by a combination of GATA-1 and GATA-2 has occurred in Jurkat cells, which have no PU.1. FIG. 9. Competitive inhibition of GATA-1dependent transactivation of the gp91 phox promoter by GATA-2 through the Ϫ98GATA site in Jurkat cells. GATA-1 and GATA-2 were transfected individually or in combination with the wild-type pϪ105/Luc reporter construct into Jurkat cells. A, promoter activities of pϪ105/Luc cotransfected with and without 2 g each of GATA-1 and GATA-2 plasmids; B, effect of the increase in the dose of cotransfected GATA-2 plasmid on the transactivation of GATA-1. Individual doses and total doses of GATA proteins were set to give less activity than each maximal one and to give submaximal to maximal activities, respectively (see Fig.  8A). Total amounts of plasmids were adjusted to 9 and 11 g in A and B, respectively, by the pEF-MCI vector plasmid.