Activation of the interleukin-5 promoter by cAMP in murine EL-4 cells requires the GATA-3 and CLE0 elements.

Interleukin-5 (IL-5) plays a central role in the growth and differentiation of eosinophils and contributes to several disease states including asthma. Accumulating evidence suggests a role for cAMP as an immunomodulator; agents that increase intracellular cAMP levels have been shown to inhibit production of cytokines predominantly produced by T helper (Th) 1 cells such as IL-2 and interferon (IFN-). In contrast, the production of IL-5, predominantly produced by Th2 cells, is actually enhanced by these agents. In this report, we have performed transient transfection experiments with IL-5 promoter-reporter gene constructs, DNase I footprinting assays, and electrophoretic mobility shift assays to investigate the key regulatory regions necessary for activation of the IL-5 promoter by dibutyryl cAMP and phorbol esters in the mouse thymoma line EL-4. Taken together, our data demonstrate the critical importance of two sequences within the IL-5 5′-flanking region for activation by these agents in EL-4 cells: one, a highly conserved 15-base pair element present in genes expressed by Th2 cells, called the conserved lymphokine element 0 (CLE0; located between −53 and −39 in the IL-5 promoter), and the other, two overlapping binding sites for the transcription factor GATA-3 (but not GATA-4) between −70 and −59. Taken together, our data suggest that activation via the unique sequence combination GATA/CLE0 results in selective expression of the IL-5 gene in response to elevated levels of intracellular cAMP.

Interleukin-5 (IL-5) plays a central role in the growth and differentiation of eosinophils and contributes to several disease states including asthma. Accumulating evidence suggests a role for cAMP as an immunomodulator; agents that increase intracellular cAMP levels have been shown to inhibit production of cytokines predominantly produced by T helper (Th) 1 cells such as IL-2 and interferon ␥ (IFN-␥). In contrast, the production of IL-5, predominantly produced by Th2 cells, is actually enhanced by these agents. In this report, we have performed transient transfection experiments with IL-5 promoter-reporter gene constructs, DNase I footprinting assays, and electrophoretic mobility shift assays to investigate the key regulatory regions necessary for activation of the IL-5 promoter by dibutyryl cAMP and phorbol esters in the mouse thymoma line EL-4. Taken together, our data demonstrate the critical importance of two sequences within the IL-5 5-flanking region for activation by these agents in EL-4 cells: one, a highly conserved 15-base pair element present in genes expressed by Th2 cells, called the conserved lymphokine element 0 (CLE0; located between ؊53 and ؊39 in the IL-5 promoter), and the other, two overlapping binding sites for the transcription factor GATA-3 (but not GATA-4) between ؊70 and ؊59. Taken together, our data suggest that activation via the unique sequence combination GATA/CLE0 results in selective expression of the IL-5 gene in response to elevated levels of intracellular cAMP.
Interleukin-5 (IL-5) 1 is the key cytokine that regulates the biological functions of eosinophils and contributes to several human disease states, including asthma (1,2). In both atopic and non-atopic asthma, elevated IL-5 has been detected in peripheral blood and the airways (3)(4)(5). There is considerable interest in the identification of the transcriptional mechanisms controlling the synthesis of this cytokine. IL-5 is predominantly produced by activated T helper 2 (Th2) lymphocytes, although mast cells and eosinophils have been also shown to produce this cytokine (2, 6 -8). Murine T helper clones are classified into two distinct subsets (Th1 and Th2) on the basis of their patterns of lymphokine secretion. Whereas IL-5 gene expression is restricted to the Th2 subset of CD4 ϩ cells, which also express IL-4 and IL-10 (but not IL-2 or interferon-␥, which are produced by Th1 cells), GM-CSF is produced by both Th1 and Th2 cells (9,10).
In contrast to cytokines such as IL-2, IL-5 is strongly induced by factors that raise intracellular cAMP levels, such as IL-1␣, prostaglandin E 2 , and forskolin (11)(12)(13)(14). In murine Schistosomiasis mansoni infection, vasoactive intestinal peptide released from eosinophils induces adenylate cyclase in T cells via vasoactive intestinal peptide receptors, resulting in IL-5 production (15). In transient transfection assays, dibutyryl cAMP (Bt 2 cAMP)-induced activation of the IL-5 promoter was mimicked by transfection of an expression plasmid encoding the catalytic subunit of protein kinase A, suggesting that cAMP stimulates IL-5 transcription via the protein kinase A signaling pathway (16). The observation that cAMP increases the expression of cytokines such as IL-5, while suppressing that of other cytokines such as IL-2 and interferon-␥ (IFN-␥), suggests a possible regulatory role for this second messenger (9,10,13,17,18).
In these studies, we sought to define precisely the cis-activating elements that regulate inducible murine IL-5 transcription in EL-4 cells in response to Bt 2 cAMP and PMA, both of which are required for optimal stimulation of the IL-5 promoter in these cells (16,26). Our data suggest that activation of the IL-5 promoter by Bt 2 cAMP and PMA in EL-4 cells requires sequences within the CLE0 element and also a region located between Ϫ70 and Ϫ59 that binds the transcription factor GATA-3. We speculate that activation via this unique sequence combination confers the specificity needed for selective expression of the IL-5 gene in response to elevated levels of intracellular cAMP.

MATERIALS AND METHODS
Cell Culture-EL-4 cells were obtained from American Type Culture Collection (ATCC) and maintained in suspension culture in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) supplemented with 10% horse serum at 37°C with 5% CO 2 .
Plasmid Preparation (Deletion Constructs)-A murine IL-5 promoter plasmid, p4k-pUC18, was a kind gift from Dr. Honjo. The luciferase * This work was supported by a National Institutes of Health Grant AI31137 and a special fellowship from the Leukemia Society of America (to A. R.), National Institutes of Health Grant HL52014 (to P. R.), and a research grant from the American Lung Association (to M. S.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. This paper is dedicated to the memory of Professor Igor Tamm, a mentor, adviser and friend.
reporter plasmid pXP1 was obtained from ATCC. Deletion constructs were prepared by polymerase chain reaction, using the following primers for the upstream sequence: p545IL5luc (ATGGATCCTGTACCTC-CCACATCTGCTG), p118IL5luc (ATGGATCCGGGCACTGGAAACC), p91IL5luc (ATGGATCCCTCGCCTTTATTAGG), p76IL5luc (ATG-GATCCTGTCCTCTATCTGATTGTTAGC), p66IL5luc (ATGGATCCCT-GATTGTTAGCAA), and the primer (AAAAAGCTTCTGGCCTTCAG-CAAAG) for the downstream sequence. These primers incorporated a restriction site for BamHI and HindIII in the upstream and downstream sequences, respectively. These polymerase chain reaction products were cloned into pXP1. Another deletion construct p168IL5luc was prepared by digesting p545IL5luc with HpaI and HindIII and ligating back into pXP1.
Site-directed mutants were generated following the method of Kunkel using a kit from Bio-Rad (27). All constructs were confirmed by DNA sequencing.
Transfection-EL-4 cells were washed once in serum-free Dulbecco's modified Eagles' medium and resuspended in the same medium. ϳ1 ϫ 10 7 cells in 0.8 ml were combined with approximately 40 g of DNA (5 g of reporter plasmid, 1-2 g of cytomegalovirus-␤-galactosidase plasmid, and approximately 33 g of the carrier plasmid pGEM7Z), and electroporation was carried out with the Gene Pulser (Bio-Rad) at 0.25 kV, 960 microfarads. The cells were then immediately returned to 4.8 ml of the same growth medium and incubated at 37°C, 5% CO 2 . 16 h after transfection, the cells were split into two aliquots. The first aliquot received no stimulation, while the second aliquot was stimulated with 25 ng/ml PMA and 1 mM Bt 2 cAMP for 8 h.
Luciferase and ␤-Galactosidase Assays-The luciferase assay was performed essentially according to manufacturer's instructions as described previously (28). Briefly, cells were harvested 8 h after stimulation and washed twice in phosphate-buffered saline. They were then lysed with 100 l of lysis buffer for 10 -15 min, and debris was removed by centrifugation. 20 l of supernatant was combined with 100 l of luciferase assay reagent and immediately analyzed in a Berthold luminometer. In order to adjust luciferase values for relative transfection efficiency, ␤-galactosidase assays were performed as described previously (29). Since the basal adjusted relative light units (ϳ250 -300) were slightly higher than the background readings of reagent blanks (ϳ150 -200), the background readings were not subtracted from each experimental value to calculate the -fold induction.
DNase I Footprinting Assays-Nuclear extracts were prepared from ϳ1-1.5 ϫ 10 7 EL-4 cells. Cells were washed in ice-cold phosphatebuffered saline and incubated for 10 min on ice in 200 l of buffer A containing 10 mM Hepes, pH 7.9, 1.5 mM MgCl 2 , 10 mM KCl, 0.5 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride. At the end of the incubation, a 0.1 volume of 1% Nonidet P-40 was added and the lysates were immediately centrifuged at low speed (ϳ350 ϫ g) at 4°C for 2 min. The supernatants were discarded, and the pelleted nuclei were resuspended in 50 l of a lysis buffer containing 20 mM Hepes, pH 7.9, 420 mM NaCl, 1.5 mM MgCl 2 , 0.2 mM EDTA, 25% glycerol, 0.5 mM phenylmethylsulfonyl fluoride, and 0.5 mM dithiothreitol. After incubation for 15 min at 4°C with vigorous shaking, the lysed nuclei were centrifuged at 16,000 ϫ g for 10 min to clear the debris. The supernatants were aliquoted, snap-frozen, and stored at Ϫ70°C until further use.
A fragment containing sequences between Ϫ168 and ϩ24 of the IL-5 promoter was used for footprinting assays using techniques previously described (30). Labeled fragment (ϳ30 fmol) was incubated at room temperature for 15 min in 100 l of reaction buffer containing 10 mM Tris-HCl, pH 7.9, 0.5 mM EDTA, 5% glycerol, 0.05% Nonidet P-40, 1 mM MgCl 2 , 0.2% polyvinyl alcohol, 10 g/ml poly(dI)-poly(dC), in the presence or absence of nuclear extract prepared from EL-4 cells. At the end of the incubation, 100 l of a salt mixture (10 mM MgCl 2 , 5 mM CaCl 2 ) was added to each reaction. DNase I (Worthington) was added to a final concentration of 2.5 or 0.025 unit/ml to tubes with or without nuclear extracts, and digestion was carried out for 1 min at room temperature. The reactions were stopped with 200 l of stop buffer containing 0.2 M NaCl, 0.04 M EDTA, 1% SDS, 125 g/ml tRNA, and 100 g/ml proteinase K. The samples were extracted with a mixture of phenol:chloroform, and the DNA was precipitated with ethanol and electrophoresed on 8% polyacrylamide, 8.3 M urea gels. For accurate reading of the footprints, a TϩC sequencing reaction of the labeled strand was electrophoresed in parallel.
Electrophoretic Mobility Shift Assays-The probes in the EMSAs were three 20-bp double-stranded oligonucleotides containing sequences between (i) Ϫ57 and Ϫ34 (containing the CLE0 element), (ii) Ϫ93 and Ϫ54, (iii) Ϫ73 and Ϫ54 in the IL-5 gene, and (iv) a fourth 27-bp oligonucleotide containing two consensus GATA elements (Santa Cruz Biotechnology). For the CLE0 element containing probe, the binding reaction contained 10 mM Hepes, pH 7.9, 50 mM NaCl, 1 mM dithiothreitol, 1 mM EDTA, 0.5 mM MgCl 2 , 10% glycerol, 1 mM phenylmethylsulfonyl fluoride, 50 g/ml poly(dI)-poly(dC), and 200 g/ml bovine serum albumin. The buffer composition was the same for the other two probes, except that the MgCl 2 was added at 1 mM for the CLE0 probe. The competitor oligonucleotides were added at a 100-or 200-fold molar excess. Antibodies to the Jun, NF-B, and GATA-3 proteins were purchased from Santa Cruz Biotechnology, while the anti-Fos antibody was obtained from Oncogene Science. The anti-GATA-3 antibody was raised against a glutathione S-transferase (GST)-human GATA-3 (amino acids 1-264) fusion protein (31). Anti-GATA-4 antibody, kindly provided by Dr. David Wilson, was rabbit anti-mouse GATA-4 antiserum that does not cross-react with GATA-1, -2, or -3 (32). Fig. 1A depicts the IL-5 promoter and potential transcription factor binding sites based on sequence homology. Considerable information exists regarding the role played by NF-AT in the transcriptional regulation of IL-2 (33). CLE1 represents a "conserved lymphokine element" found in the 5Ј upstream sequence of a variety of cytokine promoters, including GM-CSF, IL-3, and IL-5 (21,22). The role it plays in the transcriptional regulation of these genes is essentially unknown. The CLE0 element appears to represent a composite element consisting of a consensus sequence for binding of Ets family proteins, more specifically Elf-1, and an AP-1-like binding site, which in the GM-CSF gene appears to bind both JunB and c-Fos (25). Both the Elf-1-and the AP-1binding sites are crucial for transcriptional activation of the GM-CSF promoter (24,25). The IL-5 CLE0 element differs from the GM-CSF CLE0 element by only 1 base pair within the AP-1-like region (Fig. 1C).

Activation of the IL-5 Promoter Requires the AP-1-like Site and Additional Upstream Sequences-
In order to determine precisely the region in the IL-5 promoter responsible for transcriptional regulation, EL-4 cells were transiently transfected with luciferase reporter plasmids incorporating IL-5 gene sequences ranging from 545 to 66 bp upstream from the transcription start site (Fig. 1B). An 8-h stimulation with Bt 2 cAMP and PMA caused an 8 -10-fold increase in transcriptional activity in the 545-bp construct. While the 168-, 118-, and 91-bp constructs also gave a similar response, ϳ40 -50% activity was induced in the construct containing 76 bp of sequence upstream of the transcription start site. However, the 5Ј-flanking sequence deleted to Ϫ66 was unresponsive to the stimuli (Fig. 1C).
To further define the key cis-regulatory elements in the IL-5 promoter, transient transfection experiments using site-directed mutants were performed (Fig. 1C). Mutations in NF-AT and CLE1 decreased transcriptional activity modestly (Fig.  1B). It has been well established that activation of NF-AT occurs via a Ca 2ϩ -dependent mechanism (33). Previous work indicated that elevations in [Ca 2ϩ ] i play a minimum role in modulating the transcriptional regulation of the IL-5 gene, suggesting that NF-AT probably does not play a major role in the transcriptional activity of IL-5 (34 -36). It is important to note, however, that activation of the NF-AT element in EL-4 cells by PMA alone, in the absence of a Ca 2ϩ -activated pathway, is probably due to a constitutive increase in [Ca 2ϩ ] i in EL-4 cells (37). In these studies, the preservation of near full activity in the deletion construct p91IL5luc, compared to the 545-bp construct, further supports the absence of a critical contribution by NF-AT-activated pathways in the activation of IL-5 gene transcription (34 -36). Fig. 1B shows that a mutation in the AP-1 sequence within the CLE0 element eliminated activity of the IL-5 promoter. This observation indicated that transcription factors binding to the AP-1 or overlapping sequences within the CLE0 element are critical (but not sufficient) for transcriptional activity of IL-5 in response to Bt 2 cAMP and PMA.
Stimulation of EL-4 Cells with Bt 2 cAMP and PMA Protects the IL-5 Promoter from Digestion with DNase I-In order to identify potential sites for transcription factor binding in response to stimulation by Bt 2 cAMP and PMA, footprinting experiments utilizing DNase I digestion were performed (Fig. 2). Footprints that spanned almost the entire sequence from Ϫ88 to Ϫ41 were detected (Fig. 2). They included one clearly distinguishable footprint spanning nucleotides Ϫ68 to Ϫ41 and a second less clear footprint across the region Ϫ88 to Ϫ77 (Fig. 2). While the footprint between Ϫ53 and Ϫ39 was due to proteins binding to the CLE0 element, the region between Ϫ88 and Ϫ54 appeared to be binding sequences for as yet uncharacterized proteins. Thus, regions of the murine IL-5 promoter that showed functional activity in the transient transfection experiments also demonstrated DNase I protection, presumably reflecting the binding of transcription factors important for the transcriptional activation of the IL-5 promoter in response to Bt 2 cAMP and PMA. Although the footprints were more marked with nuclear extracts prepared from Bt 2 cAMP ϩ PMA-activated cells, nuclear proteins isolated from unstimulated cells also showed some binding activity corresponding to these regions (Fig. 2). A third footprint between Ϫ30 and Ϫ16 included the TATA box and adjacent sequences. No footprints were identified in the region corresponding to the NF-AT and CLE1 elements.
Bt 2 cAMP and PMA Induce Specific Transcription Factor Binding to the Murine IL-5 Promoter CLE0 Element-In order to identify specific transcription factor binding to the murine IL-5 promoter, EMSAs were performed. Using an oligonucleotide containing the IL-5 CLE0 element as probe, at least three complexes were readily detected with nuclear extracts from stimulated EL-4 cells (Fig. 3A, lane 2). Although the ratio of protein:DNA used in these experiments was less than that used in the DNase I assays, nuclear extracts prepared from unstimulated cells also displayed some DNA binding activity (Fig. 3A, lane 1). An identical unlabeled oligonucleotide, used in 100-fold molar excess competed for binding (lane 3); in contrast, an oligonucleotide containing the consensus sequence for AP-1 competed partially only for formation of complexes I and II but not at all for complex III (lane 4). A mutation within the IL-5-AP-1-like sequence totally abolished the ability of an oligonucleotide to compete for formation of any complex (lane 6). Taken together, the results depicted in lanes 4 and 6 suggest that the AP-1 sequence within the IL-5 CLE0 site is critical for binding but proteins that bind to this element probably cannot be fully accounted for by factors which can bind a consensus AP-1 sequence. An oligonucleotide incorporating a mutation within the Elf-1-like site within the IL-5 CLE0 element competed completely for binding, suggesting that the Elf-1-binding sequence is not involved in binding of transcription factors to the IL-5 CLE0 element in these cells (lane 5).
The AP-1-like Site in the IL-5 CLE0 Element Binds Transcription Factors Consistent with JunB and JunD-In order to characterize the proteins that bound to the IL-5 CLE0 element, antisera against the AP-1 family and NF-B proteins were used in the same EMSA (Fig. 3A, lanes 7-12). An antiserum that recognizes all members of the Jun family significantly affected formation of complexes I and II and only reduced complex III to the level seen in uninduced cells (lane 7). When antisera against the individual Jun proteins were used, antic-Jun inhibited the formation of either complex I or II minimally (lane 8). Antisera against Jun B and JunD appreciably inhibited the binding activity in complexes I and II (lanes 9 and  10). An antiserum that recognizes all members of the Fos FIG. 1. Transcriptional activation of the murine IL-5 promoter constructs. A, identification and localization of candidate cis-activating elements in the murine IL-5 promoter, located in the region immediately 5Ј to the transcription start site. Potential binding sites for NF-AT, CLE1, and CLE0 (which is a composite of AP-1-and Elf-1-binding sites) sites are indicated. B, a series of luciferase reporter plasmids containing deletion constructs and site-directed mutations of the IL-5 promoter are shown, with boxes representing the location of the potential cis-element identified in A. Deletions include from Ϫ545 to Ϫ66 bp of the murine IL-5 promoter. Single mutations include the potential NF-AT, CLE1, and AP-1 sites and are described in C. The average -fold luciferase induction for each of the reporter plasmids is shown in the bar graph, with results representing the average of multiple experiments and normalized for ␤-galactosidase activity. The absolute basal luciferase activity was ϳ250 -300 relative light units in these experiments and was similar for all of the promoter constructs. The deviations were no more than 10% between experiments. C, the locations of the two overlapping GATA sites (the site that fits the consensus sequence is boxed) and the three site-directed mutations in the IL-5 promoter are identified. Base pair changes are identified with lowercase letters. A comparison between the human GM-CSF and the murine and human CLE0 sequences is presented showing a single base pair difference between the two promoters in the area underlined.
family abolished most of the binding activity in complex II, but only weakly affected formation of complex I. The effect on complex III was similar to the effect of the anti-Jun antiserum (lane 11). An antibody against an unrelated transcription factor (p50 subunit of NF-B) had no effect on any of the complexes (lane 12). Taken together, these results suggest that the major inducible CLE0 binding activity in complex II contains JunB, JunD, and a Fos family member. Complex I appears to be largely composed of JunB and JunD. However, since there was a small amount of residual DNA binding activity in both of these complexes that was not abolished by any of these antisera, we cannot rule out the possibility that there are less abundant complexes composed of related proteins. The identity of complex III (inducible or constitutive) is not known. Therefore, Fos and Jun proteins partially account for protein binding to the IL-5 CLE0 element in EL-4 cells stimulated with Bt 2 cAMP and PMA.
Bt 2 cAMP and PMA Induce Transcription Factor Binding to the Region Spanning Nucleotides Ϫ93 to Ϫ54-The region of the IL-5 promoter upstream of the CLE0 element contained critical functional activity and corresponded to a footprint within the region from Ϫ41 to Ϫ68 and a second weaker footprint between Ϫ88 and Ϫ77 (the former includes the AP-1-like sequence; Figs. 1C and 2). We used a radiolabeled 40-bp oligonucleotide containing the entire sequence between Ϫ93 and Ϫ54 (but excluding the CLE0 element) in the IL-5 promoter to investigate specific binding of proteins to this region in EMSAs (Fig. 3B). Two DNA-protein complexes were readily detected using this particular probe. The faster migrating complex, which could be detected using extracts from both unstimulated and stimulated cells, was clearly nonspecific in nature (Fig.  3B). However, the complex displaying slower electrophoretic mobility was less readily detectable when nuclear extracts were prepared from unstimulated cells, but its formation was significantly induced following stimulation of the cells for 8 h with Bt 2 cAMP and PMA (Fig. 3B, lane 1). A third complex (see Figs. 4 and 5), which was also largely inducible, was detected with excess nuclear extract. The complex shown in Fig. 3B represented a sequence-specific DNA binding activity because it was efficiently competed for by excess of the same unlabeled competitor oligonucleotide (Fig. 3B, lane 3) but not by oligonucleotides corresponding to the IL-5 CLE0 or the IL-5 NF-AT/ CLE1 sites (lanes 4 and 7, respectively) or by the nonspecific competitors containing consensus AP-1 and NF-B sites (lanes 5 and 6, respectively). Antibodies against the Jun and Fos family members also did not interfere with binding, further demonstrating that the transcription factor(s) binding to this region in the IL-5 gene was not a member of the AP-1 family.
The Region between Ϫ70 and Ϫ59 Is the Major DNA-binding Sequence in the 40-bp Oligonucleotide-To further define the sequences that bound protein(s) within the 40-bp oligonucleotide, competitor oligonucleotides were derived by subdividing the 40-bp region into shorter overlapping sequences using the DNase I footprints as guides. One of these oligonucleotides contained sequences between Ϫ73 and Ϫ54, whereas the other included sequences between Ϫ91 and Ϫ74. Once again, in EMSAs with the 40-bp oligonucleotide as the probe, the upper complex shown in Fig. 3B (complex I in Fig. 4) was detected. A second specific complex (II), less abundant but largely inducible, was also detected (Fig. 4). The formation of both complexes was competed for by 100-fold excess of the 40-bp unlabeled oligonucleotide (lane 3) and the 20-bp oligonucleotide containing the sequence between Ϫ73 and Ϫ54 (lane 5). Neither of the oligonucleotides containing sequences between Ϫ91 and Ϫ74 or between Ϫ82 and Ϫ67 displayed any competition (lanes 4 and 9, respectively). This competition profile was completely compatible with the DNase I footprint shown in Fig. 2. The nature of proteins that generate the weaker footprint across Ϫ88 to Ϫ77 is unclear.
A close inspection of the sequence between Ϫ73 and Ϫ54 revealed the sequence AGATAA, between nucleotides Ϫ70 and Ϫ65, which fit the consensus sequence WGATAR (W ϭ A/T and R ϭ A/G) for the GATA family of transcription factors. We noticed an overlapping TGATTG sequence on the complementary strand, which also was in close agreement with the GATA consensus sequence except for T at the ϩ1 position. To investigate whether this region contributed to the formation of the DNA-protein complex, mutations were introduced in three different regions (m1, m2, and m3; see Fig. 4). m1 contained mutations in the distal GATA sequence, m2 in the proximal sequence, while m3 was mutated in both the sequences. Used at a 100-fold molar excess, m2 appeared to compete better than the other two. When we repeated the experiment with 100-and 200-fold excess of the same oligonucleotides, the rank order of the ability to compete was m2 Ͼ m1 Ͼ m3 (Fig. 5). We also used consensus binding sequences for three additional transcription factor family proteins, C/EBP, Oct, and CREB/ATF, none of which, even at 200-fold excess, competed for binding to the probe (Fig. 5).
We next used oligonucleotides containing wild-type or mutant GATA sequence to investigate whether complexes I or II contained GATA proteins. As illustrated in Fig. 6, while an oligonucleotide containing wild-type GATA sequence displayed impressive competition, mutations within the GATA sequence abolished this competition. Since, among the GATA factors, GATA-3 has been reported to be predominantly expressed by T lymphocytes (with some expression also in the developing central nervous system), we used a GATA-3-specific antibody (which does not cross-react with either GATA-1 or 2), to further characterize the protein that bound to the AGATAA sequence in the IL-5 promoter. The anti-GATA-3 antibody but not the control NF-B-specific antibody supershifted both the complexes (Fig. 6, lane 6). We also performed the reciprocal experiment using the oligonucleotide containing the wild-type GATA sequence as the probe. As is evident in lane 12, the IL-5 sequences between Ϫ73 and Ϫ54 efficiently competed for binding of GATA-3 to the consensus GATA sequence. The specific localization of the GATA-binding sequence to nucleotides between Ϫ70 and Ϫ59 in the IL-5 promoter also explained the complete loss of inducibility of the promoter when deleted to Ϫ66 (Fig. 1B).
While these studies were in progress, Yamagata et al. (43) reported the involvement of GATA-4 in the activation of the human IL-5 gene in the ATL-16T cell line. We, therefore, performed EMSA using a GATA-4-specific antibody that does not cross-react with any of the presently characterized GATA proteins (32). As illustrated in Fig. 7, while the anti-GATA-3 antibody again supershifted the specific DNA-protein com-plexes (lane 3), the anti-GATA-4 antiserum did not supershift or inhibit formation of either complex (lane 5). In the reciprocal situation as well, although anti-GATA 4 antiserum specifically inhibited complex formation by affinity-purified GST-GATA-4 (lane 8), anti-GATA-3 antibody had no effect on complex formation (lane 9). Taken together, these data suggest that GATA-3 (but not GATA-4) and Fos/Jun (JunB and JunD but not c-Jun) proteins, that bind to the GATA and CLE0 elements, respectively, play crucial roles in the induction of the IL-5 promoter in response to Bt 2 cAMP and PMA in EL-4 cells. DISCUSSION The results presented in this study demonstrate that sequences within the CLE0 element are critical but not sufficient for activation of the IL-5 promoter in response to cAMP agonists and phorbol esters. We have identified a second region between Ϫ70 and Ϫ59, which contains two overlapping GATA sites, disruption of which (as in the Ϫ66 deletion construct) abrogates activation of the promoter.
In agreement with a previous report by Lee et al., Bt 2 cAMP and PMA synergistically activated the IL-5 promoter (16). In a recent report, Lee et al. have implicated the NF-AT element in the IL-5 5Ј-flanking region in induction of the gene by Bt 2 cAMP and PMA. In their studies, mutations in the NF-AT site led to an ϳ80% reduction in activation of the IL-5 promoter by these stimuli. Activation of NF-AT is coupled to increased levels of [Ca 2ϩ ] i (33,38,39). This requirement, however, can be bypassed in EL-4 cells due to constitutive increases in [Ca 2ϩ ] i (37). Our studies indicate that, although mutations in the NF-AT site in the context of the 545-bp promoter lead to ϳ50% reduction in activity of the promoter (which is approximately equivalent to an 80% reduction when compared to the 1.2kilobase pair construct), the deletion construct with the end point just 3Ј to this site (at Ϫ91) retains full transcriptional activity compared to the 545-bp construct, leading to a conclusion that the NF-AT site is not critical for induction of the promoter (Fig. 1B). In support of this, our DNase I footprinting FIG. 3. Binding of inducible EL-4 cell nuclear proteins to potential cisactivating elements of the murine IL-5 promoter. A, a radiolabeled oligonucleotide containing the IL-5 CLE0 sequence (see Fig. 1C) was used in EMSAs with nuclear extracts prepared from EL-4 cells left unstimulated or stimulated for 8 h with Bt 2 cAMP (1 mM) and PMA (25 ng/ml). Bands representing CLE0 binding activity are identified with arrows. Competitive binding assays were performed with unlabeled CLE0, classical AP-1, mutant CLE0 (mElf-1), and mutant CLE0 (mAP-1) probes. The antisera that were used were against the Jun family, cJun, JunB, JunD, the Fos family, and the p50 subunit of NF-B. B, a radiolabeled oligonucleotide incorporating the region from Ϫ93 to Ϫ54 bp of the IL-5 promoter, exclusive of the potential NF-AT, CLE1, and CLE0 binding sites was used in EMSAs with nuclear extracts as described in A. An arrow identifies a band representing specific binding activity within this sequence. Competition with unlabeled "self," IL-5 CLE0, classical AP-1, NF-B, and muIL-5 NF-AT/CLE1 oligonucleotides is shown. Antisera against the Jun and Fos families were utilized. studies did not reveal any binding of proteins across the NF-AT site in the IL-5 promoter, although distinct footprints were generated on the AP-1 and the GATA sites (Fig. 2). CsA blocks the nuclear translocation of NF-AT and has been shown to inhibit the expression of genes such as IL-2 and IL-4, which require NF-AT for induction of gene expression (33,38,39). Although Lee et al. (40) used CsA to demonstrate inhibition of binding of proteins to the NF-AT site in the IL-5 promoter, they did not investigate whether CsA indeed inhibits IL-5 gene expression in functional (transfection) experiments. Recently, Lacour et al. (41) reported inhibition of NF-AT induction by cAMP. Data previously reported by the Arai group also demonstrated that the NF-AT site in the IL-2 promoter is a target for inhibition of IL-2 promoter activation by cAMP, presumably via inhibition of calcineurin activity (42). However, this is only possible if distinct NF-AT species simultaneously control inhibition of the IL-2 gene and activation of the IL-5 gene by cAMP in the same cells (EL-4), as the authors also suggest (40).
Our studies establish a critical role for the GATA sequence in  Fig. 3B. A 100-fold molar excess of the following unlabeled oligonucleotides were used as competitors: self (lane 3), two shorter oligonucleotides derived from this 40-bp oligonucleotide containing the region between Ϫ91 and Ϫ74 and between Ϫ73 and Ϫ54 (lanes 4 and 5, respectively), an overlapping oligonucleotide containing the sequence between Ϫ82 and Ϫ67 (lane 9), and three mutant oligonucleotides m1, m2, and m3 (lanes 6 -8; the specific mutations are indicated by lowercase letters). All other conditions were as described in the legend to Fig. 3. GATA-3 and GATA-4 are members of the GATA family of transcription factors that bind to the consensus sequence Ϫ4 WGATAR ϩ2 via a highly conserved C 4 zinc finger domain (44 -46). Four GATA members, 1 through 4, have been described in vertebrates. The transcription factor GATA-3 is expressed most abundantly in T-lymphocytes and the developing central nervous system (46 -50). In contrast, GATA-4 is predominantly expressed in the heart, gut epithelium, and reproductive organs (51)(52)(53)(54). GATA-3 has been shown to play a crucial role in the transcriptional regulation of T cell receptorrelated genes (46 -49). DNA-binding studies with bacterially expressed GATA proteins and oligonucleotides containing randomized GATA sequences indicate that GATA sequences containing G at the ϩ2 position, such as the one found in the IL-5 promoter between Ϫ70 and Ϫ65, may have a relatively lower binding affinity for the GATA-binding protein than sequences containing A at the ϩ2 position (45). However, the data of Ko and Engel (45) indicate that the lower binding affinity of a site can be compensated for if this site overlaps another GATA site. Indeed, as shown in Fig. 1, the IL-5 gene does have overlapping GATA sites, one between Ϫ70 and Ϫ65 (which fits the consensus sequence) and another located between Ϫ65 and Ϫ59. Although the latter site has an intact GAT core but has a T instead of A in the ϩ1 position, it appears that this substitution can still be selected by GATA-3, particularly within overlapping GATA sites (45). Overlapping GATA sites have been also previously identified in many erythroid expressed genes such as the chicken ␣-globin promoter (55). Overlapping and or multiple GATA sites appear to confer increased GATA binding activity and may play a key role in promoting full transcriptional activity (55)(56)(57).
It remains undetermined from our data if both or one of the two potential GATA-3 sites are critical for protein binding. The finding that the m3 mutation, which disrupted both GATA sites, showed the least competition in our gel shifts suggests but does not prove that both sites may be critical for optimal binding. Both the IL-5 GATA sequence-containing probe and the oligonucleotide containing the GATA consensus sequence formed two complexes I and II with EL-4-nuclear extracts. It is possible that complexes I and II represent monomeric and dimeric forms (the latter induced by stimulation of cells) of the same protein. The requirement of the CLE0 element for activation of the promoter suggests that the GATA region alone is not sufficient for IL-5 promoter activation. In a similar fashion, in the T cell receptor-␤ gene, GATA-3 needs to interact with proteins bound to nearby sites for enhancer function (58).
Among the cytokine genes, potential GATA binding sequences are present in the 5Ј-flanking regions of the IL-3, IL-4, GM-CSF, and IFN-␥ genes (but not in the IL-2 gene). There is no previous evidence for functional importance of this site in any of the other genes. In the case of the IFN-␥ gene, deletion into the GATA sequence had little effect on activity of the promoter (59). To the best of our knowledge, this is the first report of a functional role for GATA-3 in the transcriptional regulation of a T cell cytokine gene. It will be interesting to determine in future studies with T cell clones and primary T lymphocytes whether human T cells utilize GATA-3 for inducible expression of the IL-5 promoter.
Naora et al. (60) have suggested that the TCATTT element, which overlaps with the AP-1-like element within the CLE0 element in the IL-5 promoter, is important for induction of IL-5 gene expression in response to mitogens and PMA. The TCA sequence in the TCATTT element overlaps with the AP-1 site, while the TTT sequence overlaps with the Elf-1-binding site within the CLE0 element. In EMSAs reported by Naora et al. (60), mutation of the TCATTT element to cgAaTT (which also mutated the AP-1 element) abolished protein binding. However, the authors did not use antibodies against the AP-1 proteins to determine whether this complex contained AP-1 family proteins (60). It is possible that the TTT sequence, immediately adjacent to the AP-1 sequence, has a permissive role only (61,62). Wang et al. (25) observed that the inducible transcription factors that bound to the GM-CSF CLE0 element contain JunB and c-Fos. Our gel shift assays show that specific antisera to the AP-1 family members JunB and JunD and an antiserum that recognizes members of the Fos family significantly inhibit formation of complexes I and II with the IL-5 CLE0 element. However, since there was a small amount of residual DNA binding activity in both of these complexes that was not abolished by any of these antisera, we cannot rule out the possibility that there are less abundant related proteins present in both the complexes. Complex III may comprise proteins similar to the constitutive proteins that Wang et al. had detected with the GM-CSF CLE0 probe (25).
Although the 5Ј-flanking region of the IL-5 gene bears some homology to the promoter of other cytokines such as GM-CSF and IL-4, the divergent regulation of these cytokines suggests that different mechanisms control their synthesis, particularly in response to elevations in intracellular cAMP levels (12,13,18,63). Identification of a distinct combination of regulatory elements, i.e. GATA/CLE0 for cAMP activation of the IL-5 promoter, is therefore consistent with the differential effect of FIG. 7. GATA-3 and not GATA-4 is the IL-5 DNA-binding protein in EL-4 nuclear extracts. 1 l each of preimmune or immune serum was used in lanes 4 or 7 and lanes 5 or 8, respectively, while 0.2 g of affinity-purified GST-GATA-4 protein was used in lanes 6 -9. All other conditions were as described in the legend to Fig. 3. cAMP on expression of the IL-5 gene. Ultimately, these findings may begin to explain how IL-5 gene expression is uniquely regulated by cAMP in T lymphocytes and how it becomes dysregulated in several common disease states.