T-cell Expression of the Human GATA-3 Gene Is Regulated by a Non-lineage-specific Silencer*

The GATA-3 transcription factor is required for development of the T-cell lineage and Th2 cytokine gene expression in CD4 T-cells. We have mapped the DNase-I-hypersensitive (HS) regions of the human GATA-3 gene in T-cells and non-T-cells and studied their transcriptional activities. HS I–III, located 5′ from the transcriptional initiation site, were found in hematopoietic and non-hematopoietic cells, whereas HS IV–VII, located 3′ from the transcriptional start site, were exclusively observed in T-cells. Among these hypersensitive sites, two transcriptional control elements were found, one in the first intron of the GATA-3 gene and the other between 8.3 and 5.9 kilobases 5′ from the GATA-3transcriptional initiation site. The first intron acted as a strong transcriptional activator in a position-dependent manner and with no cell-type specificity. The upstream regulatory element could confer T-cell specificity to the GATA-3 promoter activity, and analysis of this region revealed a 707-base pair silencer that drastically inhibited GATA-3 promoter activity in non-T-cells. Two CAGGTG E-boxes, located at the 5′- and 3′-ends of the silencer, were necessary for this silencer activity. The 3′-CAGGTG E-box could bind USF proteins, the ubiquitous repressor ZEB, or the basic helix-loop-helix proteins E2A and HEB, and we showed that a competition between ZEB and E2A/HEB proteins is involved in the silencer activity.

Lineage commitment and differentiation of multipotent hematopoietic stem cells occur throughout life and are mostly regulated at the transcriptional level (1). Multiple studies have now shown that lineage-restricted expression of a subset of transcription factors is essential to achieve proper development of all the hematopoietic lineages, and therefore, the knowledge of the mechanisms involved in the regulation of the expression of these lineage-restricted transcription factors is of considerable importance for further understanding of hematopoiesis.
Transcription factors of the GATA family are related by their conserved zinc-finger motif that binds to the consensus DNA sequence 5Ј-(A/T)GATA(A/G)-3Ј (2). Among the GATA factors, GATA-1, -2, and -3 are necessary for hematopoiesis, as gene disruption of any one of these factors results in major hematopoietic defects (3)(4)(5)(6). GATA-1, -2, and -3 display different lin-eage-restricted patterns of expression in hematopoietic cells. GATA-1 is abundant in the erythrocytic, mastocytic, and megakaryocytic lineages and is also present at a lower level in multipotential progenitors (7)(8)(9); GATA-2 is mostly expressed in uncommitted hematopoietic progenitors, immature erythroid cells, and proliferating mast cells (10); and GATA-3 is expressed in very immature hematopoietic progenitors and then only in the T-cell and natural killer cell lineages (11,12). GATA-3 was first shown to be abundantly expressed in Tlymphocytes, natural killer cells, and embryonic brain (13,14). More detailed studies have shown that GATA-3 gene expression occurs in numerous sites during development: placenta, kidney, and adrenal gland; the embryonic central and peripheral nervous systems; and embryonic liver and thymus (14). Contrasting with this wide expression during development, GATA-3 mRNAs are mostly detected in thymocytes and Tlymphocytes and in the central nervous system in the adult. This gene expression pattern, regulated during development and cellular differentiation, might be mediated by a complex array of cis-acting elements, as shown in the regulation of transcription factor genes in invertebrates (15). As for hematopoiesis, an extinction of GATA-3 gene expression in stable cell hybrids formed by fusion of cell lines representing the erythrocytic and the T-lymphocytic lineages indicated that GATA-3 might be repressed in hematopoietic non-T-cells (16), a kind of regulation already shown for the CD4 gene, another T-cell-specific gene (17).
The human, mouse, and chicken GATA-3 genes have been cloned, and sequence analysis of their promoters revealed that they are embedded within a CpG island and share structural features often found in promoters of housekeeping genes (14,18,19). Transfection experiments of reporter genes controlled by the mouse or chicken GATA-3 promoter have failed to show any appropriate T-cell-regulated expression, which indicates that the GATA-3 regulatory elements lie 5Ј and/or 3Ј from the GATA-3 promoter (14,19). To understand the transcriptional controls that regulate GATA-3 gene expression in T-cells, we have mapped the DNase-I-hypersensitive (HS) 1 regions of the human GATA-3 gene in T-cells and in hematopoietic non-Tcells, and we have studied the role of these regions in GATA-3 gene expression in T-cells.

Nuclei Preparation and DNase-I Treatment
Approximatively 10 9 cells were washed twice in phosphate-buffered saline; resuspended in 20 ml of homogenization buffer (10 mM Tris (pH 7.4), 15 mM NaCl, 60 mM KCl, 1 mM EDTA, 0.1 mM EGTA, 0.15 mM spermine, 0.5 mM spermidine, and 5% sucrose) containing 0.05% (Jurkat), 0.2% (K562), or 0.6% (HeLa) Nonidet P-40; and broken by five strokes of a Dounce homogenizer. Nuclei were purified by centrifugation over a sucrose gradient (10% sucrose in homogenization buffer), washed twice in wash buffer ( 1. A and B, mapping of the DNase-I HS sites in the human GATA-3 locus after BglII-EcoRV or SalI digestion, respectively. The map shows the human GATA-3 gene, the BglII-ClaI and SspI-SstI probes used, and the resulting fragments observed after DNase-I digestion together with the 12-kb BglII-EcoRV or the 17-kb SalI-SalI germ line fragment. HS sites are depicted by arrows. HS V, located at the beginning of intron 3, could not be distinguished from the germ line fragment in the SalI digestion. DNA was extracted from nuclei treated with increasing amounts of DNase-I, digested with BglII and EcoRV or SalI, electrophoresed, blotted, and hybridized with the indicated probe. Jurkat cells (a T-cell line) KCl, 0.15 mM spermine, 0.5 mM spermidine, and 10% sucrose), and subjected to increasing concentrations of DNase-I (0.1-15 mg/ml; Worthington) in wash buffer plus 1 mM MgCl 2 for 10 min at 37°C. The reaction was stopped by the addition of proteinase K (0.1 mg/ml final concentration), SDS (1% final concentration), and EDTA (10 mM final concentration).

DNA Extraction and Southern Hybridization
DNA was extracted by proteinase K digestion (0.1 or 0.2 mg/ml) at 56°C overnight, followed by phenol/chloroform extraction and ethanol precipitation. 10 g of DNA were digested to completion with the indicated restriction enzymes, electrophoresed on 0.8% agarose gels, and transferred to nitrocellulose membranes (Hybond C Extra, Amersham Pharmacia Biotech) by Southern blotting. Hybridization was performed with random-primed 32 P-labeled probes at 65°C in 5ϫ SSC (0.6 M NaCl and 0.06 M sodium citrate (pH 7)), 1ϫ Denhardt's solution, 20 mM NaPO 4 (pH 6.7), and 10% dextran sulfate. Highest stringency washes consisted of 0.1ϫ SSC and 0.1% SDS at 65°C. The genomic probes used for the DNase-I studies were a 582-bp BglII-ClaI fragment and a 441-bp SspI-SstI fragment.

Construction of Plasmids
Constructs Used to Delimit the Human GATA-3 Promoter-The Ϫ96/ ϩ598 DNA fragment was cloned from a cosmid that contained the human GATA-3 gene by BamHI-BstEII digestion, followed by electrophoresis and fragment purification. The Ϫ96/ϩ44 DNA fragment was obtained by an XmnI digest of the BamHI-BstEII fragment, followed by purification of the BamHI-XmnI fragment. Mutants of the Ϫ96/ϩ598 DNA fragment were obtained by double restriction digests of the Ϫ96/ ϩ598 DNA fragment, followed by fill in with Klenow polymerase and blunt-end ligation. The constructs used for orientation and position dependence of the 3Ј-activating element were obtained by cloning a RsaI-BstEII DNA fragment containing the 3Ј-activating element 5Ј or 3Ј from the Ϫ96/Ϫ44 DNA fragment.
Constructs Used to Characterize a Human GATA-3 Gene-regulating Element-A human placental DNA library (CLONTECH) was screened with a 5Ј-probe obtained from a cosmid previously cloned, and a phage containing 12 kb of DNA 5Ј from the human GATA-3 transcriptional initiation site was isolated. A 2.4-kb BamHI-BamHI DNA fragment containing HS I and HS II (Ϫ8300/Ϫ5900 fragment) was subcloned into pBSK, and the various constructs shown in this study were obtained by subsequent digests of the BamHI-BamHI DNA fragment. All point mutations were obtained by polymerase chain reaction and subsequent cloning.
All the constructs were finally cloned 5Ј from the CAT reporter gene using the pBL-CAT-3 vector (20). All these constructs were sequenced before use. Sequence analysis was performed using a Taq DyeDeoxy Terminator Cycle Sequencing kit (Applied Biosystems, Inc., Foster City, CA) and an automatic sequencer (Applied Biosystems Model 373A).
Construct Used to Overexpress ZEB-ZEB cDNA was cloned by reverse transcription-polymerase chain reaction using oligonucleotides that encompass the ATG initiation codon and the TAA stop codon. The polymerase chain reaction product was sequenced and cloned into the pECE vector (21), where the inserted cDNA is driven by the SV40 promoter and enhancer.

Cell Culture and Transfection
Human cell lines were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum (HeLa) or in RPMI 1640 medium supplemented with 10% fetal calf serum (Jurkat and K562), L-glutamine, penicillin, and streptomycin (Life Technologies, Inc.). For transient transfection, 10 7 cells were transfected by electroporation (Bio-Rad Gene Pulser; 960 microfarads, 200 -250 V) in a volume of 175-250 l using the same molarity of reporter gene plasmids together with 2 g of pRSV-Luc plasmid. pBluescript was added to bring the total amount of DNA to 10 g.
Cells lysates were prepared 24 h after transfection by the freezethaw procedure (22). Luciferase assays were used to determine transfection efficiency, and CAT assays were performed using amounts of extract normalized for transfection efficiency (23). The percentage of acetylation for each extract was determined by quantification of the 14 C-acetylated chloramphenicol on thin-layer chromatography plates using a Molecular Dynamics PhosphorImager. Stable transfections were performed like transient transfections, except that the cells were electroporated with 20 g of linearized plasmid together with 1 g of plasmid containing the SV40 promoter-driven neomycin-phosphotransferase gene. After selection on G418, three independent pools of transfected cells were assayed for CAT activity.

Nuclear Extracts and DNA Binding Assays
Nuclear extracts were prepared from HeLa and Jurkat cell lines (24), and DNA binding assays were performed essentially as described (25). Poly(dI⅐dC) was used as nonspecific competitor, and in competition assays, 50 ng of unlabeled competitor DNA was preincubated with the nuclear extract for 5 min before the addition of the labeled probe (0.5 ng). The E5 and MEF1m oligonucleotides used have been previously described (26), and the 3Ј-CAGGTG E-box oligonucleotide is 5Ј-AGCT-TTTTACCAGGTGGTCTCTA-3Ј. Antibodies against HEB and E2A proteins were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Antibodies against human USF1 and USF2 were a kind gift from Dr. Michel Raymondjean (INSERM U.129, Paris, France), and anti-ZEB antiserum was a gift from Dr. H. Kondoh (Osaka University, Osaka, Japan).

Mapping of the DNase-I-hypersensitive
Sites of the Human GATA-3 Locus-Hypersensitivity of chromatin to DNase-I digestion has been used to identify regulatory elements of numerous genes (27,28). To determine which regions of the human GATA-3 locus might be implicated in the regulation of its expression in T-cells, we mapped the DNase-I HS sites in 35 kb of human genomic DNA encompassing the human GATA-3 transcription unit in a T-cell line (Jurkat) that expresses human GATA-3 and in a non-hematopoietic cell line (HeLa) and an erythrocytic cell line (K562) that do not express human GATA-3. Fig. 1A shows the mapping of DNase-I HS sites after EcoRV-BglII digestion and hybridization with a 5Ј-BglII-ClaI probe. In addition to the 12-kb germ line fragment, two fragments of 7 and 2.5 kb were observed exclusively in the Jurkat cell line, and a fragment of 1.1 kb was detected in the three cell lines studied. The 2.5-kb fragment results from DNase digestion around the transcriptional initiation site of the human GATA-3 gene; the 7-kb fragment located an HS site in intron 3; and the 1.1-kb fragment identified a region located 1.5 kb 5Ј from the transcriptional initiation site of the human GATA-3 gene. Using SalI digestion and a 3Ј-SspI-SstI probe, we then mapped two HS sites, located between exons 5 and 6 and found only in Jurkat T-cells (Fig. 1B). Using other digests, we finally mapped the seven HS sites that are shown in Fig. 1C. Interestingly, the 5Ј-HS sites were found in all the cell lines tested, whereas the 3Ј-HS sites were found only in the Jurkat cell line.
Human GATA-3 Contains a Minimal Promoter That Extends in the First Intron-To look for any regulatory function of the HS sites previously mapped, we first delimited the human GATA-3 minimal promoter. A segment of the human GATA-3 gene encompassing the presumptive minimal promoter, from nucleotides Ϫ96 to ϩ44, was first studied ( Fig. 2A). After transient transfection into a T-cell line (Jurkat) and into two cell lines that do not express human GATA-3 (HeLa and K562), we detected a weak transcriptional activity that did not display any cell-type specificity ( Fig. 2A). By primer extension, we showed that the transfected constructs were correctly initiated in the three cell lines (data not shown). The addition of 5Ј- , an activating cis-acting sequence is located within the first intron of the human GATA-3 gene. Sequential 3Ј-deletions of the Ϫ96/ϩ598 human GATA-3 DNA fragment were cloned 5Ј from the CAT reporter gene as indicated under "Materials and Methods" and transfected in the three cell lines previously described. C, the sequence located at the 3Ј-end of the first human GATA-3 intron acts in a position-dependent manner. The ϩ208/ϩ598 3Ј-activating sequence was cloned in both orientations upstream or downstream from the Ϫ96/ϩ44 DNA fragment, and the resulting constructs were transfected in the three cell lines studied. The indicated values represent the average of three independent transfections. sequence (up to Ϫ2500) did not change the transcriptional activity of the Ϫ96/ϩ44 DNA fragment or bring any T-cell specificity to the constructs used (data not shown). However, the addition of 554 nucleotides located 3Ј from the Ϫ94/ϩ44 DNA fragment resulted in a 6 -10-fold enhancement of transcriptional activity in the three cell lines ( Fig. 2A). This 554nucleotide fragment contained both the first exon and most of the first intron of the human GATA-3 gene and, together with the Ϫ96/ϩ44 DNA fragment, defined the Ϫ96/ϩ598 human GATA-3 minimal promoter. To delimit the regions involved in the activity of this promoter, we subjected this fragment to 3Ј-deletion analysis. As shown in Fig. 2B, this analysis defined a 123-bp DNA fragment, located at the end of the first intron, as necessary for efficient activity of the human GATA-3 promoter. We then looked for any position dependence of this 3Ј-element. A 390-bp DNA fragment containing the human GATA-3 promoter 3Ј-element was cloned in both orientations 5Ј or 3Ј from the Ϫ96/ϩ44 DNA fragment (Fig. 2C). Only the constructs that contained the 390-bp DNA fragment 3Ј from the Ϫ96/ϩ44 DNA fragment were transcriptionally active in the three cell lines tested, indicating that the 3Ј-element acted as a strong transcriptional activator in a position-dependent manner and with no cell-type specificity.
Characterization of a DNA Fragment That Confers T-cell Specificity to the Human GATA-3 Minimal Promoter-All the DNA fragments that encompassed the different HS sites mapped in the human GATA-3 locus were cloned 5Ј from the Ϫ96/ϩ598 human GATA-3 promoter, and the resulting constructs were stably transfected into Jurkat, K562, and HeLa cell lines. None of the 3Ј-T-cell-specific HS sites provided any specificity to the Ϫ96/ϩ598 human GATA-3 promoter (data not shown), and thus, we focused our study on the 5Ј-region of the human GATA-3 gene. Starting from a construct that contained all the 5Ј-HS sites, we performed deletion analysis to study, by stable transfections, the three 5Ј-HS sites (Fig. 3). The 8.3-kb DNA fragment that contained all the 5Ј-HS sites displayed a T-cell specificity. The 6.5-kb DNA fragment containing only HS II and HS III showed a higher transcriptional activity in Jurkat cells than in HeLa or K562 cells, but the difference between these three cell lines was smaller than that observed with the 8.3-kb DNA fragment, as the transcriptional activity of this 6.5-kb DNA fragment was present in HeLa and K562 cells (Fig.  3). The 3-kb DNA fragment that contained only HS III displayed no cell-type specificity and was as active as the human GATA-3 minimal promoter in all three cell lines (Fig. 3). To demonstrate that HS I and HS II were sufficient for T-cell specificity, a 2.1-kb DNA fragment that contained these two HS sites was cloned 5Ј from the human GATA-3 minimal promoter, and this construct was stably transfected into Jurkat, K562, and HeLa cells. As shown in Fig. 3, this 2.1-kb DNA fragment conferred T-cell specificity to the human GATA-3 minimal promoter, indicating that HS I and HS II were necessary and sufficient for T-cell specificity of the human GATA-3 minimal promoter in the assay we used.
HS I Acts as a Strong Silencer in Non-T-cells-To characterize the function of HS I and HS II, a 966-bp DNA fragment (Ϫ8025 to Ϫ7059) containing only HS I was cloned 5Ј from the human GATA-3 promoter, and the resulting construct was stably transfected into HeLa, K562, and Jurkat cells. This fragment repressed human GATA-3 promoter activity in the HeLa and K562 cell lines, but not in the Jurkat T-cell line. The repression level was identical to the one obtained with the HS I-HS II DNA fragment (Fig. 4), and interestingly, this silencer did not reduce the transcriptional activity of the human GATA-3 minimal promoter in T-cells (data not shown). The orientation dependence of this negative regulatory element was tested by inserting the 966-bp DNA fragment in reverse orientation 5Ј from the human GATA-3 promoter, and indeed, the transcriptional inhibition obtained with this construct was the same, indicating that this silencer was orientation-independent (Fig. 4). We then cloned a 1.1-kb DNA fragment (Ϫ7059 to Ϫ5900) containing only HS II 5Ј from the human GATA-3 minimal promoter and stably transfected the resulting construct into HeLa, K562, and Jurkat cells. We obtained a 5-6fold decrease in the transcriptional repression in HeLa and K562 cells, and we got a 2-3-fold increase in the transcriptional activity of the human GATA-3 promoter in Jurkat cells (Fig. 4). These results indicate that HS I contains a major regulatory element that confers T-cell specificity to the human GATA-3 minimal promoter. We thus performed a detailed analysis of this silencer.
A 5Ј-and a 3Ј-Element Are Necessary for HS I Silencing-5Јand 3Ј-deletion analysis of this silencer revealed two cis-acting elements that were necessary for efficient silencing of the human GATA-3 promoter activity. The 5Ј-element was located between Ϫ7828 and Ϫ7746, and the 3Ј-element was located between Ϫ7197 and Ϫ7121 (Fig. 5). This analysis was done by stable transfection, and we next examined whether these elements were sufficient for efficient repression. We cloned the various deleted regions 5Ј from the human GATA-3 minimal promoter and transfected these constructs into HeLa or Jurkat cells, but never obtained any efficient repression of the human GATA-3 promoter (data not shown), suggesting that the silencer identified requires multiple elements to be functional. The sequence of the 707-bp DNA fragment (Ϫ7828 to Ϫ7121) containing HS I is shown in Fig. 6. The 3Ј-DNA fragment necessary for efficient repression in non-T-cells contained a YY1-binding site adjacent to a CAGGTG E-box and a TC-CTCCT motif already shown to be required for neuronal expression of the zebrafish gata-2 gene (29), and the 5Ј-DNA fragment characterized also contained a CAGGTG E-box. The presence of this same motif in the 5Ј-and 3Ј-regions of the silencer prompted us to analyze its function.
The Two CAGGTG E-boxes Are Necessary for HS I Silencing Activity-As the Ϫ7828/Ϫ7121 DNA fragment has the same transcriptional activity as the initial Ϫ8025/Ϫ7059 repressor, the role of the two CAGGTG E-boxes in HS I silencing activity was tested in this DNA fragment beginning with the 3Ј-region. Mutations that deleted or disrupted the 3Ј-CAGGTG E-box were shown by gel-shift analysis to prevent binding of any protein (data not shown), and stable transfections of the resulting constructs showed that they were completely unable to silence the human GATA-3 promoter activity in non-T-cells and did not modify this promoter activity in T-cells (Fig. 7A). A similar mutation was performed on the 5Ј-CAGGTG E-box, and stable transfections of the mutated silencer showed that it could not repress the human GATA-3 promoter activity in non-T-cells (Fig. 7B). These results indicate that the HS I silencing activity needs these two CAGGTG E-boxes.
The 3Ј-CAGGTG E-box Could Bind ZEB, USF, and E2A Proteins-To characterize the proteins that can bind the 3Ј-CAGGTG E-box, we first performed gel-shift analysis using Jurkat or HeLa nuclear extract. As shown in Fig. 8A, two complexes (C1 and C2) were obtained with Jurkat or HeLa nuclear extract, whereas a third complex (C3) was obtained only with Jurkat nuclear extract. To define the proteins present in these three complexes, we first used competition with oligonucleotides known to bind E2A or related proteins (E5 oligonucleotide) or the ubiquitous ZEB repressor (MEF1m oligonucleotide) (26). These experiments showed that the C1 complex had the same migration as a ZEB complex (Fig. 8B, lanes  4 and 7) and that the 3Ј-CAGGTG E-box oligonucleotide efficiently competed the ZEB binding on the MEF1m oligonucleotide (lanes 7 and 8). The C3 complex migrated like the E2A complex bound on the E5 oligonucleotide (Fig. 8B, lanes 3 and  4), and the 3Ј-CAGGTG E-box and the E5 oligonucleotides  2 and 3 and lanes 4 and  6, respectively).
To characterize the proteins present in the C1, C2, and C3 complexes, we used antibodies against ZEB, E2A, HEB, USF1, and USF2 for supershift assays. As shown in Fig. 8C, the addition of anti-ZEB antibodies supershifted the C1 complex, whereas the addition of control serum had no effect on this complex (compare lanes 1 and 2). Increasing amounts of anti-ZEB antibodies completely abolished the formation of the C1 complex, but also abolished the supershift shown in Fig. 8C  (lane 2) (data not shown). To identify the proteins present in this C3 complex, antibodies against E2A or HEB protein were used for supershift assays, and both antibodies completely supershifted the C3 complex, indicating that this complex contains a heterodimer of HEB and E2A proteins (Fig. 8D, lanes 2  and 3). Finally, antibodies against human USF1 or USF2 basic helix-loop-helix protein showed that the C2 complex contained homo-and heterodimers consisting of these two proteins (Fig.  8E, lanes 2-4). In conclusion, the 3Ј-CAGGTG E-box oligonucleotide could bind ZEB and USF proteins in HeLa cells, whereas it bound ZEB, USF, and an E2A/HEB heterodimer in Jurkat T-cells.
ZEB and E2A/HEB Regulate the Silencing Activity of HS I-To demonstrate that ZEB was involved in human GATA-3 gene repression, we cotransfected a ZEB expression vector together with the human GATA-3 promoter linked to the Ϫ7828/Ϫ7121 repressor or to the mutated Ϫ7828/Ϫ7121 fragment that could not bind ZEB (mutant CAGGTG 3 CAGGCC described in Fig. 7A). Overexpression of ZEB resulted in significantly reduced transcriptional activity of the Ϫ7828/Ϫ7121 construct in Jurkat T-cells (65% repression), whereas the mutated version of this DNA fragment was not sensible to this ZEB overexpression (Fig. 9A). These results indicate that ZEB is involved in the repression obtained with the Ϫ7828/Ϫ7121 DNA fragment through the 3Ј-CAGGTG E-box.
We then performed several point mutations of the 3Ј-CAG-GTG E-box, which can bind ZEB, E2A/HEB, and USF proteins. We transformed this 3Ј-E-box to a MEF1m sequence, i.e. a sequence that can weakly bind ZEB and neither E2A/HEB nor USF (Fig. 8B, lane 7) (26); to the sequence AGTTCAGGTGT-GTT located at Ϫ361 in the human ␣ 4 -integrin promoter and shown to bind only ZEB and USF proteins (30,31); or to a E2 sequence known to bind only ZEB and E2A/HEB proteins (32).

Stable transfections of the resulting constructs in HeLa and
Jurkat T-cells showed that the mutated E2 sequence had the same effect as the wild type 3Ј-E-box, whereas the MEF1m or Ϫ361 sequence induced a repression of the human GATA-3 promoter in Jurkat T-cells (Fig. 9B). These results indicate that the E2A/HEB heterodimer, but not the USF proteins, can relieve the repression mediated by ZEB on the 3Ј-E-box. DISCUSSION Among the GATA transcription factors, GATA-3 displays a peculiar expression, as it is expressed in many different tissues (i.e. central and peripheral nervous systems and embryonic liver, kidney, and adrenal medulla) during development and only in the placenta, the central nervous system, very immature hematopoietic cells, and T-cells in the adult. To define regulatory sequences that control GATA-3 gene expression in T-cells, we first investigated the DNase-I hypersensitivity of the human GATA-3 gene as DNase-I hypersensitivity has been associated with a wide range of cis-regulatory sequences and is usually indicative of protein-DNA interactions. Although no T-cell-specific HS site was found 5Ј from the human GATA-3 gene initiation site, three T-cell-specific HS sites were discovered in the 3Ј-direction. We assayed these different T-cellspecific HS sites, linked to the endogenous or to a heterologous promoter, in both transient and stable transfections, but we were unable to show any function of these HS sites in transcriptional regulation. These results indicate a possible requirement for other regulatory elements or that the assays we used could not detect the function of these DNase-I HS sites.
We first determined the cis-acting sequences, located near the transcriptional initiation site and involved in the human GATA-3 gene activation. We did not find any sequence involved in the T-cell-specific activity of the human GATA-3 gene, but mapped a sequence, located in the first intron between ϩ475 and ϩ598, required for high transcriptional activity. This sequence did not display classical enhancer activity, as it functioned, in both sense and antisense orientations, only when located 3Ј from the GATA-3 minimal promoter. Similar results have been obtained using the mouse GATA-3 promoter (14), and comparison of the mouse and human sequences located at the end of the first intron showed little homology except at their 3Ј-ends, where the sequence 5Ј-CAGGTCTC(C/T)-3Ј lies (in the human and murine introns) one base 5Ј from the 3Ј-splicing sequence.
We then studied the effects of the 5Ј-HS sites on the transcriptional activity of the human GATA-3 promoter and found that a DNA fragment, located between Ϫ8025 and Ϫ5900, could confer T-cell specificity to the human GATA-3 promoter. This T-cell specificity is mainly mediated by a strong silencer active in non-T-cells and located between Ϫ7828 and Ϫ7121. Previous studies have shown an extinction of the GATA-3 gene in both murine and human erythroid ϫ T-cell hybrids (16) and during the commitment of Th0-like cells to Th1 effector cells (33). These extinctions could be direct silencing or loss of positive regulators, and our data strongly support the first hypothesis. Transcriptional regulation by negative regulatory elements is now well documented, and studies using transgenic mice or transfection assays have identified silencers of transcription in numerous tissues and developmentally specific genes (34 -37). The position of silencers, like the position of enhancers, varies depending on the gene studied and can be located adjacent to or far away from the regulated promoter (38,39). The human GATA-3 silencer was found 7 kb 5Ј from the promoter and thus belongs to the second type of silencer. The molecular mechanisms that mediate transcriptional silencing are not completely understood, but several DNA-binding proteins, like COUP-TF or YY1 (40,41), seem clearly implicated in silencing. Deletion and point mutation analyses of the human GATA-3 silencer revealed two CAGGTG E-boxes that are necessary but not sufficient for efficient silencing. Point mutation that destroyed the 5Ј-CAGGTG E-box resulted in an absence of silencing, indicating the importance of this DNA-binding motif in human GATA-3 gene regulation. Electrophoretic mobility shift assays done with this E-box as a probe revealed two specific complexes present in Jurkat, HeLa, and K562 cells (data not shown). These complexes did not contain ZEB, USF, or E2A protein, as no supershift could be obtained with antisera against all members of the USF, E2A, and ZEB proteins (data not shown). Further characterizations of these complexes are now needed to understand this 5Ј-CAGGTG E-box function. Electrophoretic mobility shift assays done with the 3Ј-CAGGTG E-box revealed three specific complexes, two present in T-and non-T-cells (C1 and C2) and one present only in T-cells (C3). These three complexes contained the ubiquitous repressor ZEB (C1), the USF basic helix-loophelix proteins (C2), and an E2A/HEB heterodimer (C3). Overexpression of ZEB in T-cells showed that ZEB acted as a repressor, and several point mutations indicated that E2A and HEB, but not USF proteins, were able to relieve ZEB repression in T-cells.
ZEB and E2A/HEB have been shown to regulate genes during hematopoiesis (26,42), but no competition between these proteins for DNA binding has yet been reported. ZEB is an active transcriptional repressor that has been shown to regulate both muscle differentiation and ␣ 4 -integrin gene expression during hematopoiesis (30,31). Interestingly, ZEB seems to compete with muscle-specific basic helix-loop-helix proteins for the regulation of muscle differentiation (31), whereas it impairs the myb-ets synergy during hematopoiesis (30). As for human GATA-3 gene regulation, our results suggest that ZEB represses the human GATA-3 promoter activity and that the E2A/HEB heterodimer can displace ZEB in T-cells. This E2A/ HEB heterodimer also positively regulates CD4 gene expression by regulating the activity of its enhancer (43), and thus, E2A/HEB heterodimers seem to regulate early and late genes activated during T-cell differentiation. Interestingly, HEB or E2A mutant homozygous mice seem to have a developmental delay in the transition from the double-negative stage to the double-positive stage (42), and studies on chimeras indicated that the differentiation of GATA-3 Ϫ/Ϫ T-cells is also blocked at the double-negative stage of development (6). These results, together with the data presented in this paper, strengthen the link between HEB and E2A proteins and GATA-3 gene regulation. Furthermore, as E2A, HEB, and ZEB proteins are not T-cell-specific, their relative amounts might switch on or switch off the human GATA-3 gene, and thus, these non-lineage-specific proteins might be involved in T-lymphoid determination.
We have previously shown that a cosmid that contains the human GATA-3 transcription unit extended by 4 kb of 3Јflanking sequence and 3 kb of 5Ј-flanking sequence is sufficient for directing T-cell expression (18). This cosmid does not contain the silencer described in this paper, and thus, the human GATA-3 gene seems to be under the control of discrete regulatory elements that can enhance and/or silence the transcriptional activity of a minimal promoter. Furthermore, a recent publication has characterized an enhancer sequence that regulates mouse GATA-3 gene expression in the brachial arch (44). This enhancer, located between nucleotides Ϫ2832 and Ϫ2642, does not confer any T-cell specificity, and thus, the GATA-3 gene seems to be regulated by discrete regulatory elements required for its complex expression pattern. Such a modular cis-regulatory organization has been described for many genes encoding transcription factors in Drosophila (15) and shows the actual complexity of cis-regulatory elements that regulate these genes.