Synergistic Transcriptional Activation by hGABP and Select Members of the Activation Transcription Factor/cAMP Response Element-binding Protein Family*

The Ets-related DNA-binding protein human GA-binding protein (hGABP) α interacts with the four ankyrin-type repeats of hGABPβ to form an hGABP tetrameric complex that stimulates transcription through the adenovirus early 4 (E4) promoter. Using co-transfection assays, this study demonstrated that the hGABP complex mediated efficient activation of transcription from E4 promoter synergistically with activating transcription factor (ATF) 1 or cAMP response element-binding protein (CREB), but not ATF2/CRE-BP1. This synergy also partially occurred when hGABPα was used alone in place of the combination of hGABPα and hGABPβ. hGABP activated an artificial promoter containing only ATF/CREB-binding sites under coexistence of ATF1 or CREB. Consistent with these results, physical interactions of hGABPα with ATF1 or CREB were observed in vitro. Functional domain analyses of the physical interactions revealed that the amino-terminal region of hGABPα bound to the DNA-binding domain of ATF1, which resulted in the formation of ternary complexes composed of ATF1, hGABPα, and hGABPβ. In contrast to hGABPα, hGABPβ did not significantly interact with ATF1 and CREB. Taken together, these results indicate that hGABP functionally interacts with selective members of the ATF/CREB family, and also suggest that synergy results from multiple interactions which mediate stabilization of large complexes within the regulatory elements of the promoter region, including DNA-binding and non-DNA-binding factors.

In eukaryotes, the control of gene expression often involves regulated interactions of gene-specific transcription factors with promoters and enhancer regions. The regulatory proper-ties of DNA-binding proteins are often modulated in a combinatorial fashion by interactions among them (1). hGABP 1 has been identified as the transcription factor E4TF1 in HeLa cells because of its ability to activate transcription within the adenovirus early 4 (E4) promoter (2,3). Characterization of cDNAs of E4TF1 subunits (4) revealed that the subunits are highly homologous to the respective rat transcription factor GABP (GA-binding protein) subunits. The GABP subunits bind to the cis-regulatory DNA sequence important for immediate early gene activation of herpes simplex virus type-1 (5,6). Therefore, E4TF1 has been re-designated as hGABP (human GABP), as described in Ref. 7. The Ets-related protein hGABP␣ can bind by itself to the DNA sequence 5Ј-CGGAAGTG-3Ј in the E4 promoter, but has no effect on in vitro or in vivo transcription (3,7,8). By itself, hGABP␤, which contains four notch/ankyrin-type repeats, neither binds to a specific sequence nor stimulates transcription, but has homodimerization activity via the carboxyl-terminal leucine zipper-like domain. However, the four ankyrin repeats mediate the association of hGABP␤ with hGABP␣, leading to the formation of an ␣ 2 ␤ 2 heterotetrameric complex on the DNA, resulting in in vitro and in vivo transcriptional activation (3,(7)(8)(9). Certain transcription factors, such as NRF-2, EF-1A, XrpFI, and RBF-1, which have been independently studied as transcription factors involved in cellular or viral gene expression, have been found to be immunologically related to GABP or hGABP (10 -13). In particular, subunits of NRF-2 are identical to those of hGABP at the level of cDNA (14). The genes for the hGABP subunits, hGABP␣ and hGABP␤, have been mapped to human chromosome 21.q 21.2-q21.3 and 7.q11.21, respectively (15,16).
Members of the Ets family of DNA-binding proteins contain about an 85-amino acid region of similarity called the ETS domain, which is sufficient for direct DNA binding to a sequence containing a common 5Ј-GGA(A/T)-3Ј core motif (17,18). Recently, partnerships between the Ets-related proteins and transcription factors belonging to other structural families have been reported, and their functional protein-protein interactions were shown to be important for regulation of gene expression (19 -21). The promoter of the adenovirus E4 contains not only an hGABP-binding site, but also several binding sites for the ATF/CREB family which has been revealed to be important for its activity by deletion analysis of the promoter (22). Transcription factors belonging to the ATF/CREB family were found to be required for its efficient transcriptional activation in vitro (23). But it is unclear how hGABP and members of ATF/CREB family regulate the transcription.
In order to gain further insight into the transcriptional activator complex involved with the Ets-related protein hGABP, and to better understand the molecular basis of synergistic transcriptional regulation, we used co-transfection and biochemical assays to examine the possibility that hGABP can cooperate with some members of ATF/CREB family. Based on the studies presented here, we propose that hGABP functions as a transcriptional partner of ATF1 or CREB, leading to efficient transcription activation. Our data suggest that a functional synergy between these factors results from a multitude of DNA-protein and protein-protein interactions which stabilize the large activator complex on promoter/enhancer elements.
The pGEX/hGABP␣ plasmid, which encodes hGABP␣ fused with glutathione S-transferase (GST) to its amino terminus (designated GST/hGABP␣), was created by the insertion of a BamHI digested DNA fragment from hGABP␣ cDNA into pGEX/prehGABP␣ that was constructed by insertion of Klenow polymerase-treated NcoI DNA fragment from pET60 (4) into the SmaI site of the GST fusion vector pGEX-2T. The pGEX/hGABP␣ mutant plasmids, which encode various GST/hGABP␣ mutants were constructed using available restriction sites or polymerase chain reaction-mediated strategies. The pGEX/ hGABP␤ plasmids, which encode hGABP␤ fused with GST to their amino terminus (designated GST/hGABP␤), were constructed by the insertion of Klenow polymerase-treated NcoI-BamHI DNA fragments from pET53 (4) into the SmaI site of pGEX-2T. Restriction enzymedigested DNA fragments encoding the full-length of hGABP␣ and hGABP␤ were cloned into NdeI-BamHI-digested pKA, a plasmid designed for the expression of a histidine-stretch and the phosphorylation target sequence of a catalytic subunit of cAMP-dependent protein kinase-fused proteins to the amino terminus in Escherichia coli. The GST-ATF1 expression plasmid pGEX-ATF1 was described previously (24) and a series of GST-ATF1 mutant expression plasmids were constructed using available restriction sites or polymerase chain reactionmediated strategies.
Expression plasmids for full-length hGABP subunits in Drosophila melanogaster Schneider line 2 (SL2) cells (25) were constructed as described previously (7). To facilitate the construction of expression plasmids for full-length ATF1, pBS/ATF1 was constructed by insertion of a EcoRI-HindIII DNA fragment containing ATF1 cDNA isolated from pGEM-ATF1 (gift of H. C. Hurst) (26) into the sites of EcoRI and HindIII of pBlueScriptII SKϩ. The DNA fragment encoding ATF1 was isolated from pBS/ATF1 by digestion with KpnI and EcoRI, and subcloned into the sites of KpnI and EcoRI of pA5C⌬P to create an ATF1 expression plasmid, pA5C⌬PATF1. The coding region for CREB was obtained by polymerase chain reaction using pT7␤CREB (gift of H. C. Hurst) (26) as template, and the polymerase chain reaction product was then ligated into the site of SmaI of pUC119 to create pUC119/CREB. The CREB expression plasmid, pA5C⌬P/CREB was constructed by insertion of an EcoRI-BamHI DNA fragment containing CREB cDNA isolated from pUC119/CREB into the site of BamHI and EcoRI of pA5C⌬P.
GST fusion proteins were expressed in E. coli, BL21(DE3). The cells were suspended in lysis buffer (50 mM Tris-HCl pH 7.9, 1 mM EDTA, 0.5 M NaCl, 0.5% Nonidet P-40, 5% sucrose), lysed by sonication, and subsequently centrifuged at 12,000 ϫ g at 4°C for 10 min to remove cell debris. The supernatants were stored at Ϫ80°C until they were used for in vitro protein binding assays. hGABP␣ and hGABP␤ with six histidine residues and a PKA phosphorylation site fused to their amino terminus, which were used in Figs. 5 and 6, and also expressed in E. coli BL21(DE3). These fused proteins were purified from supernatants on His-Bind Resin and dialyzed against 0.05 TGKEDN. They were stocked at Ϫ80°C until they were used as 32 P-labeled proteins for GST pulldown assays. Recombinant hGABP␣ intact form used in Fig. 5B was expressed in E. coli BL21(DE3) and purified from the bacteria extract using the hGABP-binding site-immobilized latex beads as described previously (9).
Gel Shift Assay-The binding reaction was performed as described previously by Watanabe et al. (3). Electrophoresis was performed using a 1% agarose gel containing 2.5% glycerol and TGE buffer (50 mM Tris-HCl, pH 7.9, 380 mM glycine, 2 mM EDTA) at 4°C and 5 V/cm for 4 h. The DNA probes derived from adenovirus early 4 promoter was prepared by digesting pUCE4-20 (3) with EcoRI and HindIII. The fragments were isolated using 10% polyacrylamide gel and 32 P-labeled by treatment with Klenow polymerase in the presence of [␣-32 P]dATP, followed by purification using a Nick column (Amersham Pharmacia Biotech). About 2 ng of the DNA probe was used for the binding reactions.
Cell Culture and Transfections-SL2 cells were maintained as described previously (7). Cells were plated onto 35-mm polystyrene dishes at a density of 1 ϫ 10 6 cells/2 ml of medium per dish 5-10 h prior to transfection. DNA/CaPO 4 precipitates were formed by the dropwise addition of 100 l of 0.25 M CaCl 2 containing the DNA to 100 l of 2 ϫ HBS (42 mM Hepes, pH 7.1, 275 mM NaCl, 1.4 mM Na 2 HPO 4 ) and added to the cells 25 min later. Transfection mixture contained 0.6 g of reporter construct, 50 ng of the ␤-galactosidase plasmid as an internal control for transfection efficiency and the indicated activators' expression plasmids. In each transfection assays, empty expression plasmid A5C⌬P was added as necessary to achieve a constant amount of transfected DNA. After addition of DNA, cells were incubated at 27°C and left undisturbed until the time of harvest 40 h later. Transfected SL2 cells were lysed and assayed for luciferase and ␤-galactosidase activity, as described previously (7).
Cell Line and Immunoprecipitations-N173 cells was established by transfection of pSV2/neo and an HA-tagged hGABP␣ expression plasmid pCHA/E4TF1-60 which was constructed to insert HA-tagged hGABP␣ cDNA into mammalian expression plasmid pCAGGS (29), and following two isolations of a colony in Dulbecco's modified Eagle's medium supplemented with 10% heat-inactivated fetal calf serum and 100 g/ml G418. The cell line was maintained in tissue culture dish containing Dulbecco's modified Eagle's medium supplemented with 10% heat-inactivated fetal calf serum and 50 g/ml G418. N173 and HeLa monolayer cells were harvested and washed twice with ice-cold phosphate-buffered saline, and cell extracts were prepared as previous described (30) with slight modification. The nuclear fraction and cytoplasm fraction were mixed and dialyzed against 0.1 TGKEDNP buffer (50 mM Tris-HCl, pH 7.9, 10% glycerol, 100 mM KCl, 1 mM EDTA, 1 mM dithiothreitol, 0.1% Nonidet P-40, 1 mM phenylmethylsulfonyl fluoride) for 6 h. Immunoprecipitation was performed by adding anti-HA antibody HA.11 (Babco) immobilized on protein A-Sepharose (Amersham Pharmacia Biotech) to 200 l of the cell extract followed by rotation at room temperature for 2 h. The immunoprecipitation pellet was washed once with 200 l of 0.1 TGKEDNP buffer, and antigens were released by five subsequent incubating for 3 min in 20 l of 100 mM glycine, pH 2.5. Samples were loaded on SDS-PAGE and examined by immunoblotting experiment using antibodies against hGABP␣, hGABP␤, CREB (Santa Cruz), and Sp1 (Santa Cruz) and the ECL Western blotting analysis system (Amersham Pharmacia Biotech).
In Vitro Binding Assay Using GST Fusion Proteins-The proteins fused with both a histidine tag and a protein kinase A site and Histagged ATF1 were phosphorylated by addition of 2 l of [␥-32 P]ATP (Ͼ7,000 Ci/mmol; Amersham Pharmacia Biotech) or 0.1 mM ATP (Amersham Pharmacia Biotech) and 5 units of PKA catalytic subunit (Promega), followed by incubation in 10 l of kinase buffer (20 mM Tris-HCl, pH 7.9, 10 mM MgCl 2 , 100 mM KCl) at 30°C for 30 min. The radiolabeled proteins were purified on His-Bind resin and subsequently loaded onto a Nick column (Amersham Pharmacia Biotech) to remove [␥-32 P]ATP and imidazole. For the in vitro binding assay, equal amounts (about 5 g) of GST fusion proteins were immobilized on 3 l of glutathione-Sepharose 4B beads (Amersham Pharmacia Biotech) using the corresponding volumes of the supernatants containing GST fusion proteins. The GST protein-immobilized beads were equilibrated in binding buffer (50 mM Tris-HCl, pH 7.9, 10% glycerol, 50 mM KCl, 1 mM EDTA, 10 mM MgCl 2 , 0.1 mM CaCl 2 , 1 mM dithiothreitol, 0.01ϳ0.05% Nonidet P-40). The beads were then incubated with about 1 ng of 32 P-labeled proteins at 4°C for 2 h and packed in a 1-ml syringe as an affinity column and then washed with 300 l of binding buffer. Bound proteins were eluted by boiling the beads in 25 l of SDS sample buffer. Released proteins were resolved by 10% SDS-polyacrylamide gel electrophoresis. After staining with Coomassie Blue, the gels were dried and subjected to autoradiography.
In Vitro Binding Assay by SPR-Protein interactions detected by surface plasmon resonance were performed using a BIACORE instrument BIACORE2000 (Biacore AB). Sensor Chip CM5 and GST Kit for fusion capture (Biacore AB) were used. For capture of GST fusion proteins, polyclonal goat anti-GST antibody was covalently coupled on the sensor chip surface using standard amine-coupling conditions. The supernatant of E. coli lysates containing GST fusion proteins were used for their immobilization on the sensor chip via anti-GST antibody. Sensor chips containing about 2,500 resonance units of GST fusion protein and running buffer 3 (50 mM Hepes, pH 7.5, 1 mM EDTA, 120 mM KCl, 10 mM MgCl 2 , 0.01% Tween 20) were used for experiments. Different concentrations of analyte proteins were injected over the surface in a total volume of 250 l of running buffer 3 at 30 l/min continuous flow. After each protein injection, the sensor chip surfaces were washed with an injection of 2 M KCl at 30 l/min continuous flow for 1 min to dissociate proteins interacting with the surface-bound GST fusion protein. All experiments were performed at 24°C. Kinetics constants of the interactions were calculated by the analysis software BIAevalution, adapted to the modified sensorgrams.

Proteins of the hGABP and ATF/CREB Family Bind
Simultaneously to Promoter DNA-The E4 promoter has ATF/CREBbinding sites and an hGABP-binding site. To investigate whether hGABP and ATF/CREB family proteins exist on the promoter simultaneously to regulate its activity, gel shift assays were performed with probes derived from the E4 promoter (Fig. 1). This probe DNA contains an hGABP-binding site and a major and a minor recognition site for ATF/CREB family. As shown in Fig. 1, purified hGABP, containing both ␣ and ␤ subunits, formed ␣ 2 ␤ 2 heterotetramers with the probe as described previously (lanes 1 and 8). Members of the ATF/CREB family were prepared by purification from HeLa nuclear extract using DNA affinity beads (28), and ATF/CREB family proteins formed DNA-protein complexes with higher mobility (lane 2). When hGABP␣/hGABP␤ heterotetramers were incubated with purified ATF/CREB family proteins, a new complex with the least mobility appeared (lane 3). Three retarded bands appeared when HeLa nuclear extract was used (lane 7). The mobility of the three bands were similar to those of the three bands obtained by combination use of the purified proteins of hGABP and ATF/CREB family. In either case when the purified proteins or HeLa nuclear extract were used, the DNAprotein complex with the least mobility was competed out with both cold DNA fragments containing hGABP and ATF/CREB family recognition sequences, while the two complex with faster mobility disappeared by adding either competitor containing hGABP or ATF/CREB-binding sites (lanes 4, 5, 8, and 9). This result indicates that hGABP and the ATF/CREB family can bind to the E4 promoter simultaneously with sequencespecific interaction. To further examine whether the slowest migrated DNA-protein complex by HeLa nuclear extract contains hGABP and the ATF/CREB family, antibodies were used in gel shift assay. The most slowly migrated band was obviously reacted against both anti-hGABP␤ and anti-CREB antibodies (lanes 13 and 15). These results show that hGABP and some members of the ATF/CREB family simultaneously bind to the adenovirus E4 promoter in a sequence-specific manner in vitro.
Synergistic Transactivation by hGABP and Select Members of ATF/CREB Family-To study which members of the ATF/ CREB family functionally interact with hGABP, we performed co-transfection assays using D. melanogaster Schneider (SL2) cells with pE4-luci, a luciferase reporter gene under control of the adenovirus E4 promoter (7). In this study, activating transcription factor 1 (ATF1) (31), cAMP response element-binding protein (CREB) (32), and cAMP response element-binding protein 1 (CRE-BP1 (33) which is identical to ATF2 except for two amino acids) were examined as representative members of the ATF/CREB family because these three factors belonging to ATF/CREB family were reported to be involved in the regulation of adenovirus E4 gene (26,33). Transfected ATF1 increased luciferase activity only slightly above background levels, even when large quantities were transfected ( Fig. 2A, solid   FIG. 1. hGABP  columns). In contrast, co-transfections of increasing amounts of transfected ATF1 with constant amounts of hGABP␣ and hGABP␤ led to a marked synergistic transactivation in a dose-dependent manner ( Fig. 2A, white columns). Qualitatively similar results were obtained with the use of transfected CREB as substitute for transfected ATF1 (Fig. 2B). But ATF2/CRE-BP1 transfected together with hGABP did not lead to a synergistic activation of transcription (Fig. 2C). These results show that hGABP functionally interacts with ATF1 and CREB, but not ATF2/CRE-BP1, resulting in stimulation of synergistic transcription.
Functional Interaction of hGABP with ATF1 and CREB-To explore whether the synergy between hGABP and ATF1 or CREB entirely depends on the interactions to their binding sites on DNA, we used four luciferase reporter gene constructs. The p4(hGABP-CRE)luc reporter plasmid was constructed to be under control of an artificial promoter including four tandem repeats of an E4 promoter-derived DNA sequence containing both an hGABP-binding site and a CRE site (ATF/CREB family binding site). It is noted that the CRE sequence 5Ј-TGACGAAA-3Ј is not a very good CRE site with some deference compared with the consensus one 5Ј-TGACGTCA-3Ј. The p4(hGABPmt-CRE)luc, p4(hGABP-CREmt)luc, and p4(hGABPmt-CREmt)luc reporter plasmids were also constructed to have mutations in their hGABP-binding sites, CREs, or both sites, respectively, resulting in the inability of each factor to bind to their reporter genes. Lack of the hGABP-binding motif resulted in no activation by transfected hGABP␣ and hGABP␤, even when large amounts of the expression plasmids were used (Fig. 3A, compare solid columns 2-5 with white columns [2][3][4][5]. But p4(hGABPmt-CRE) luc did not abrogate the ability of hGABP to synergistically activate transcription in the presence of co-transfected ATF1 or CREB (Fig. 3A, solid columns 6 -15). Importantly, the magnitude of the synergistic effect was reduced when compared with the reporter constructs p4(hGABP-CRE)luc (Fig. 3A, compare solid columns 6 -15 with white columns 6 -15). In the case of the reporter construct p4(hGABP-CREmt)luc, lack of a CRE motif did not abolish the ability of ATF1 or CREB to synergistically transactivate in the presence of co-transfected hGABP␣ and hGABP␤ (Fig. 3B, lanes 1-7). However, the synergism was not observed when p4(hGABPmt-CREmt)luc was used as a re-porter gene (Fig. 3B, lanes 8 -12). These results suggest that the synergism requires the binding of proteins to their corresponding binding sites on the promoter DNA.
To further investigate which subunit of hGABP effects synergistic transactivation, either hGABP␣ or hGABP␤ were cotransfected together with ATF1 (Fig. 4A). Although synergistic transactivation was not detected when hGABP␤ was co-transfected with ATF1 (solid columns), transfected hGABP␣ synergistically activated the transcription from the E4 promoter in the presence of ATF1 (white columns). When CREB was substituted for ATF1 as shown in Fig. 4B, transfected hGABP␣ similarly showed synergistic transactivation with CREB. A weak synergistic transactivation was detected when a large amount of hGABP␤ was co-transfected together with CREB. The magnitude of the synergy of hGABP␣ with ATF1 or CREB decreased compared with that of the synergy of when hGABP␣ and hGABP␤ were used in combination (compare Figs. 4, A and B, with 2, A and B).
Physical Interaction between hGABP␣ and ATF1-The results of our co-transfection assay could be explained if hGABP somehow interacted with ATF1 and CREB on the promoter. To study potential physical interaction between hGABP and ATF1 or CREB, we performed co-immunoprecipitation experiments using the whole cell extract prepared from a HeLa cell transformant N173 cell line which constitutive expressed HA-tagged hGABP␣. The immunoprecipitate with anti-HA monoclonal antibody was loaded on 10% SDS-PAGE and examined by immunoblotting assay with antibodies against hGABP␣, hGABP␤, and CREB. Immunoreacted bands corresponding to hGABP␤ and CREB were present in the immunoprecipitate as well as that to HA-tagged hGABP␣, whereas they were not detected in an immunoprecipitate from HeLa cell extract using anti-HA antibody (Fig. 5A). We also did not detect DNA-binding transcription factor Sp1 in the immunoprecipitate by HA-antibody (data not shown).
Synergism and in vitro association of hGABP␣ with CREB indicated that hGABP␣ may interact directly with ATF1 and CREB. We examined this possibility by in vitro binding assays using GST fused proteins. When 32 P-labeled recombinant ATF1 and CREB were mixed and captured with GST-hGABP subunits bound to glutathione-coated Sepharose beads, about 8% of input ATF1 and CREB bound to GST-hGABP␣ beads (Fig. 5B, lanes 5 and 6), and they were not retained to a significant degree when GST-hGABP␤ immobilized beads were used (lanes 8 and 9). GST-hGABP␤ used in the assay appeared to retain their native structure as judged by their ability to bind to hGABP␣ (lane 7). To test if the observed interaction of hGABP␣ with ATF1 and CREB depends on contaminated DNA, the in vitro interaction assays were performed in the presence of ethidium bromide (34). When 32 P-labeled recombinant hGABP␣ was mixed and captured with GST-ATF1 bound to glutathione-coated Sepharose beads, about 15% input of hGABP␣ was retained on GST-ATF1 beads, either in the presence or absence of 100 g/ml ethidium bromide (Fig. 5A, lanes 10  and 11). On the other hand, hGABP␣ was not bound to a significant degree onto GST immobilized beads (lanes 12 and 13).
The physical interaction of hGABP␣ with ATF1 or CREB was hereafter studied using the binding activity of ATF1 as a representative of the ATF/CREB family because the primary structures of these proteins are very similar to each other. To study the kinetics of the interaction between hGABP␣ and ATF1, an in vitro binding assay was performed with the BIA-CORE system (Biacore AB), using surface plasmon resonance (SPR), a technique able to direct protein-protein interactions in real-time manner. A GST-ATF1 fusion protein that had been immobilized to the sensor surface, via anti-GST antibody coupled to the surface via their amino group, was exposed to purified recombinant hGABP␣, and the association and dissociation of hGABP␣ were recorded in real time. As shown in Fig.  5C, the interaction of hGABP␣ with GST-ATF1 was detected in a dose-dependent manner when increasing concentrations of hGABP␣ were used. As a result of analyzing the sensorgrams, the dissociation and association rate constants of the interaction were measured as 2.4 ϫ 10 Ϫ3 s Ϫ1 and 1.2 ϫ 10 5 M Ϫ1 s Ϫ1 , respectively, which indicates the dissociation constant as 2.0 ϫ 10 Ϫ8 M. These results show that ATF1 and CREB interact physically with hGABP␣ in cell extract and in vitro, but not with hGABP␤ or hGABP␥.
Complex of hGABP and ATF1-We examined the possibility that hGABP␤ could influence the ATF1-hGABP␣ interaction, for hGABP␤ was originally copurified with hGABP␣ from HeLa cell nuclear extract (3,4). hGABP␣ was mixed alone, or in the presence of hGABP␤ with GST-ATF1 fusion protein immobilized on beads, and the captured proteins were detected by autoradiography (Fig. 5D). Approximately 8% of the input hGABP␣ was bound to the beads in the absence of hGABP␤, the same result as in Fig. 5 (lane 3). When increasing amounts of hGABP␤ were mixed together with constant amounts of hGABP␣, the amounts of the captured hGABP␣ progressively increased (lanes 4 -6). hGABP␤ was also progressively captured, together with hGABP␣, via an hGABP␣-hGABP␤ interaction (lanes 4 -6), while hGABP␤ were not captured to a significant degree by the beads in the absence of hGABP␣, as a negative control (lanes 7-9). These results show that the ATF1-hGABP␣ and hGABP␣-hGABP␤ interactions were mutually permissible, and that the association of hGABP␤ with hGABP␣ led to a more stable interaction of hGABP␣ with ATF1.
hGABP␣ Interacts with the DNA-binding Domain of FIG. 3. hGABP and either ATF1 or CREB affect transcription from artificial promoters containing four tandem repeats of both the hGABP-binding site and CRE or CRE alone (A), the hGABP-binding site alone (B), but not the binding site. A, SL2 cells were transfected with 0.6 g of either p4(hGABP-CRE)luc or p4(hGABPmt-CRE)luc, each 0.03, 0.1, 0.3, and 1.0 g (columns 2-5, 7-10, and 12-15, respectively), of hGABP␣ and hGABP␤ expression plasmids, along with 0.3 g of ATF1 (columns 6 -10), or CREB (columns 11-15) expression plasmid. B, SL2 cells were transfected with 0.6 g of p4(hGABP-CREmt)luc, 0.03, 0.1, and 0.3 g of ATF1 (columns 2-4, respectively) or CREB (columns 5-7, respectively) expression plasmid, along with/without each 0.3 g of hGABP␣ and hGABP␤ expression plasmids (lanes 1-7). SL2 cells were transfected with 0.6 g of p4(hGABPmt-CREmt)luc, 0.3 and 1.0 g of ATF1 (columns 8 and 9, respectively) or CREB (columns 11 and 12, respectively) expression plasmid, along with/without each 0.3 g of hGABP␣ and hGABP␤ expression plasmids (lanes 8 -12). Co-transfection assays and their analyses were carried out as described in the legend Fig. 2. All results shown represent mean Ϯ S.E. of three separate experiments. ATF1-To define the domains responsible for complex formation between hGABP␣ and ATF1, a series of ATF1 deletion mutants, with GST fused to their amino termini, were used for the in vitro binding assay (Fig. 6A). Roughly equal amounts of GST-ATF1 deletion mutants bound to glutathione-coated beads were incubated with 32 P-labeled full-length hGABP␣. Fig. 6B shows the result of the deletion analysis performed to identify the hGABP␣ interaction region of ATF1. Incubation with GST alone served as a negative control (Fig. 6B, lane 7). The results show that the amino-terminal deletion mutants tested were sufficient to interact with hGABP␣ ( lanes 5 and 6). Removal of the basic region-leucine zipper (bZip) domain, known as a DNA-binding motif, led to a complete loss of the interaction activity with hGABP␣ (lanes 2-4). These results demonstrate that hGABP␣ interacts with the DNA-binding bZip domain of ATF1.
ATF1 Interacts with the Amino-terminal Region of hGABP␣-To map the ATF1-binding region of hGABP␣, GST-hGABP␣ mutants as shown in Fig. 6C were applied to the in vitro binding analysis using GST fused proteins. As shown in Fig. 6D, deletion of the amino-terminal portion (amino acids 1-399) of hGABP␣ had a severe impact on its ability to bind to ATF1 (lane 5). The amino-terminal region (amino acids 1-319) and the Ets DNA-binding domain (amino acids 294 -406) of hGABP␣ were sufficient to form a complex with ATF1 (lanes 3  and 4). Similar results were obtained by analyses using the BIACORE system and indicated that the GST fused N-region (1-319) and Ets domain (294 -406) bound to ATF1 with low efficiency compared with GST-hGABP␣, while the GST fused C-region detected the same change of SPR as when the GST protein was used alone (results not shown). These results appear that ATF1 associates with the amino-terminal region (amino acids 1-399) of hGABP␣, which is not always necessary for the interaction with hGABP␤ (8). These results are consistent with the previous results that ATF1-hGABP␣ and hGABP␣-hGABP␤ interactions are not exclusive to each other (Fig. 5D).

DISCUSSION
In this report, we have demonstrated that hGABP interacts functionally with selective members of the ATF/CREB family, ATF1 and CREB, which results from formation of a large transcriptional activator complex on the E4 promoter.
In this study, the adenovirus E4 promoter was used as model promoters to demonstrate the synergy of hGABP with ATF1 and CREB. We do not have any direct evidence that the syn- FIG. 5. hGABP␣ physically binds to ATF1 and CREB. A, co-immunoprecipitation of CREB with HA-tagged hGABP␣ from whole extracts of N173 cells. Immunoprecipitations using anti-HA antibody were performed as described under "Experimental Procedures." Whole cell extracts of N173 (lanes 1-4) and HeLa cells (lanes 5-8) were used. 10 and 50 l of the immunoprecipitates (lanes 4 and 8) were separated by 10% SDS-PAGE and examined by immunoblotting assay for detecting hGABP subunits and CREB, respectively. 10 l of input extracts (lanes 1 and  5), flow-though fractions (lanes 2 and 6), and wash fractions (lanes 3 and 7) were also examined. Note that there are two forms of hGABP␤ in HeLa cells (7,14). B, in vitro binding assay using GST-fused proteins. GST-fused proteins used are indicated above. Arrowheads indicate the binding of ATF1, CREB, and hGABP␣ onto GST-fused protein resin. GST pulldown assays were performed as described under "Experimental Procedures." The bound fractions to each GST-fused protein resin were analyzed by 10% SDS-PAGE and stained for autoradiography. C, the in vitro binding assay using SPR. Sensorgrams indicate the real-time interactions between hGABP␣ and GST-ATF1 immobilized on sensor chip surface via anti-GST antibody. The sensorgrams measured by BIACORE were modified by subtraction of the background SPR trace using GST proteins as the negative control. D, hGABP␣-hGABP␤ and hGABP␣-ATF1 interactions were not exclusive each other. In vitro binding assay using GST-ATF1 were performed as described under "Experimental Procedures." GST-ATF1 immobilized onto glutathione Sepharose was mixed first with (lanes 3-6) or without (lanes 7-9) 5 ng of 32 Plabeled hGABP␣, followed by the addition of 32 P-labeled hGABP␤ (lanes 4 -6 and 7-9; 5, 15, and 50 ng, respectively) to the binding reaction. Lanes 1 and 2 show 0.75 ng of 32 P-labeled hGABP␣ and 7.5 ng of 32 P-labeled hGABP␤, respectively. ergy functions on the promoters in vivo, and need additional experiments to study which gene expression the synergy functions for in cells. Our transient transfection assays also demonstrated the synergistic transcription activation on minimal promoter of retinoblastoma gene (data not shown). The likelihood of synergy contributing to adenovirus E4 gene and retinoblastoma gene expression appears high, based on our results presented here (Figs. 1 and 2) and our previous report (2,7,12,35,36).
Models for the Mechanism of the Synergistic Transactivation-An important question in understanding the role of the promoter/enhancer region is how gene-specific transcription factors can regulate transcription efficiency. Our results indicate an architecture of gene-specific activators in the context of the adenovirus E4 promoter, and we suggest a feasible model for synergistic transcription activation. The theory is that the synergistic effect is a result of a large number of individual contacts between, at least, promoter DNA, hGABP subunits, and ATF1 (or CREB). Direct interaction between hGABP␣ and ATF1 (or CREB) may contribute to the stability of their binding to the promoters. Furthermore, this possible effect could increase since hGABP␤ binding to hGABP␣ enhances the affinity of the interaction between hGABP␣ and ATF1 as shown in Fig.  5D. However, the direct interaction between hGABP␣ and ATF1 was not so stable that the transcription factors would not form a complex without tethering them on the promoter DNA in vivo. Studies of the mechanism by which ATF1 and CREB activate transcription reveals that the CREB-binding protein (CBP) act as their coactivator and their interactions are in a phosphorylation dependent manner (37,38). Recently, Bannert (39) reported that hGABP control the interleukin 16 promoter activity in concert with CBP which interacted with hGABP␣. In the case of the simultaneous and stable existence of ATF1 and hGABP on the promoter, the complex can have multiple CBPdocking sites, which contribute to a great advantage to competition for limiting the level of CBP, the subsequent recruitment onto the promoter, and the maintenance of CBP on the promoter. This would allow the formation of a large activator complex that would be important for the stimulation of transcription. For synergistic transcription regulation, it is understandable that CBP have multiple interfaces to be occupied directly and simultaneously by different DNA-binding activators in some unique fashions (40). Another possible mechanism that explains the synergistic transactivation, although neither negated by the model described above nor substantiated by our results, is that the interaction of hGABP␣ with ATF1 and CREB may facilitate phosphorylation of them. It was written previously that ATF1 and CREB must be phosphorylated to interact with CBP (37,38). This possibility was supported by our preliminary findings that hGABP␣ enhanced the phosphorylation of ATF1 in vitro using HeLa nuclear extract as a source of kinases and that hGABP␣ preferred to interact with unphosphorylated ATF1 to the phosphorylated form by protein kinase A in vitro (results not shown). This model is different from the previously reported model, which suggests that multiple direct interactions of activators with specific TAFs, components of the RNA polymerase II complex, may account for synergism and proper gene expression (41). In any case, multiple binding of these activators appears to lead to multiple contact with cofactors or basal transcription factors. This enables coordination and efficient recruitment of the complex containing RNA polymerase II to form a stable and active initiation complex. Although we did not show here any evidence that CBP plays a functional role in the synergism, this proposal should stimulate further experimental tests to study the possibility and which step makes the most important contribution to the synergistic transactivation in vivo.
The Transcription Factor Network between Ets and bZip Family Transcription Factors-As reviewed in the Introduction, some of Ets family members are often found as subunits of larger transcription complexes and are involved in the regulation of viral and cellular function via promoter/enhancer sequences. In particular, bZip family transcription factors which include the ATF/CREB subfamily and the Fos/Jun subfamily, were previously reported to act as the interaction partners of Ets proteins. Our results show a large transcription enhancer complex including Ets transcription factor hGABP and strengthen the notion that the Ets family proteins functionally and physically interact with bZip family proteins to regulate gene expression. Transcriptional activity within the Ets/bZip transcription factor network represents both repression, in the case of MafB and Ets-1 for erythroid differentiation (20), and stimulation, in the case of the Ets protein Pointed and Jun in R7 Drosophila eye cells (19). Also, Ets-1 and ATF2/CRE-BP1 promotes activity through the TCR␣ enhancer (21). Synergistic activity of hGABP with ATF1 and CREB within the Ets/bZip network is an example of the latter case. It is possible that a selective partnership generating synergistic transcriptional activation does not depend only on the degree of the interaction affinity among Ets proteins and bZip proteins, cooperative DNA binding activity, a common coactivator and recruitment of a kinase (discussed above), but on unknown important regula- FIG. 6. Mapping of the protein interaction domains in vitro. A, schematic structures of GST-fused ATF1 variants. B, in vitro binding assay using GST-ATF1 variants. An arrowhead indicates the bound hGABP␣ to GST-fused ATF1 variant immobilized on glutathione-Sepharose. Binding assays and the analysis were carried out as described in the legend to Fig. 5. Coomassie Blue staining and autoradiography of the same gel are shown. C, schematic structures of GSTfused hGABP␣ variants. D, identification of hGABP␣ domain that interacts with ATF1 in vitro. Binding assays using GST-hGABP␣ variants were performed as in panel B. An arrowhead indicates the bound 32 P-labeled ATF1 to GST-hGABP␣ variants immobilized on glutathione-Sepharose. Twenty percent of the total tions at several points. It is likely that elucidation of such signal pathways which regulate the formation and activity of the large transcription complex will establish a general idea for gene expression control.