Functional Interactions of Transcription Factor Human GA-binding Protein Subunits*

The transcription factor human GA-binding protein (hGABP) is composed of two subunits, the Ets-related hGABPα, which binds to a specific DNA sequence, and either one of two hGABPα-associated subunits, hGABPβ or hGABPγ. The DNA-binding protein hGABPα cannot affect transcription by itself, but can modify hGABP-dependent transcription in vitro andin vivo in the presence of its associated subunits. In this study, co-transfection assays showed that the ratio of hGABPβ to hGABPγ affected transcription from a promoter containing hGABP binding sites. Biochemical analysis showed that they bind to hGABPα competitively, indicating that the ratio of hGABPβ to hGABPγ is important for hGABP complex formation. Kinetic analysis of the protein-protein interaction using the surface plasmon resonance system showed that hGABPα binds to hGABPβ or hGABPγ with similar equilibrium constants. Kinetic analysis of the DNA-hGABP interaction showed that the binding of hGABPγ to hGABPα stabilized the interaction of hGABPα with its DNA binding site. In addition, the kinetic analysis revealed that this was due to a slower dissociation of the protein complex from the DNA. These results suggest that hGABPα-associated subunits influence the DNA binding stability of hGABPα and regulate hGABP-mediated transcription by competing with each other.

Transcription initiation by RNA polymerase II is regulated by various promoter-specific activators (1), which bind to specific DNA sequences where they interact with various components of the transcriptional machinery. To understand the molecular basis of transcription regulation, it is important to know the effects of complex formation of these factors on transcription regulation.
Human GABP (hGABP) 1 was originally identified in HeLa cell nuclear extract under the name E4TF1 and was shown to be one of the transcription factors responsible for adenovirus early region 4 gene transcription (2). Later, it was revealed to be a human homologue of murine GABP, which was identified from rat liver cells as a factor capable of binding to the herpes simplex virus immediate early gene promoter (3,4). In recent years, hGABP (or GABP) was also reported to be a transcription factor responsible for the expression of various cellular proteins such as the leukocyte-specific adhesion molecule CD18 (␤2 leukocyte integrin) (5), the tumor suppressor retinoblastoma protein (6), the cytochrome c oxidase subunit IV and Vb proteins (7,8), and male-specific steroid 16␣-hydroxylase protein (9).
Biochemical studies of hGABP revealed that hGABP was composed of a DNA-binding subunit hGABP␣ and either one of two associated subunits, hGABP␤ or hGABP␥ (10). The genes for hGABP␣ and hGABP␤ have been mapped on human chromosome 21q21.2-p21.3 and 7q11.21, respectively (11,12). hGABP␣ is an Ets-related DNA-binding protein that recognizes the specific sequence 5Ј-CGGAAGTG-3Ј, but is unable to stimulate transcription by itself (10,13,14). It can form a heterocomplex with either hGABP␤ or hGABP␥ (14). hGABP␤ and hGABP␥ share a common amino acid sequence in their amino-terminal regions including four tandem repeats homologous to the Notch/ankyrin repeat motif (10,13), which is responsible for heterodimerization. hGABP␤ contains a leucine zipper-like motif in its carboxyl terminus region which is necessary for its homodimerization and capacity to activate transcription (14,15). Via these dimerization domains, hGABP␣ and hGABP␤ can form a ␣ 2 ␤ 2 heterotetrameric complex, which is able to stimulate transcription efficiently in vitro and in vivo (14,15). On the other hand, the other associated subunit, hGABP␥, can form heterodimers, but not heterotetramers, with hGABP␣ because it lacks the leucine zipper-like motif. Also, this heterodimer cannot mediate transactivation (14,15). hGABP␤ was reported to specifically recognize only hGABP␣, but not other Ets family proteins (16). These findings suggest that hGABP␤ and hGABP␥ play important roles both in the complex formation and transcriptional regulation of hGABP␣.
In resent years, it has been revealed that some promoter specific activators with DNA binding activity need co-activators for transactivation. Therefore, it is possible that co-activators having different transactivation potentials competitively interact with a DNA-binding factor to precisely control the transcription of certain target genes. Transcriptional regulation by hGABP seems to be a good model system to study such controls mediated by competitive co-activators. In order to expand our understanding of the transcription regulation of hGABP complexes, we examined the influence of hGABP␤ and hGABP␥ on hGABP-mediated transcription regulation and determined the kinetics of the interactions between the different hGABP subunits and between hGABP and DNA.

EXPERIMENTAL PROCEDURES
Western Blotting Assay-Whole cell extracts of variable cell lines were prepared according to Manley et al.'s (17) method. Nuclear extracts were prepared from 1 ϫ 10 6 co-transfected cells as described previously (18). Whole cell extracts and equivalent amounts of nuclear extracts were loaded on an 8% SDS-polyacrylamide gel. After electrophoresis, proteins were transferred to Immobilon transfer membranes (Millipore) and hGABP proteins were detected using the ECL detection kit for rabbit antibody (Amersham Pharmacia Biotech) and polyclonal antibody against hGABP␤ and hGABP␥.
Co-transfection Assay, Luciferase Assay, and ␤-Galactosidase Assay-Co-transfection assays, luciferase assays, and ␤-galactosidase assays were carried out as described previously (6) except that the transfection scales were reduced to one third.
Preparation of hGABP Subunit Polypeptides-hGABP␣, hGABP␤, and hGABP␥ were expressed individually in Escherichia coli BL21 (DE3) as described previously (13). hGABP␣ was purified from the lysate using latex particles bound to DNA containing the hGABPbinding sequence (19 -21). hGABP␤ and hGABP␥ were prepared from the insoluble fraction of E. coli lysates. Insoluble fractions containing hGABP␤ or hGABP␥ were dissolved and denatured by the addition of 0.05TGKEDN (50 mM Tris-HCl (pH 7.9), 1 mM EDTA, 1 mM dithiothreitol, 20% glycerol, 50 mM KCl, 0.1% Nonidet P-40) with 6 M guanidine hydrochloride. The proteins were then renatured by dialysis for 6 h against 0.05TGKEDN followed by another 6 h against 0.03TKEDT (30 mM Tris-HCl (pH 7.9), 1 mM EDTA, 1 mM dithiothreitol, 30 mM KCl, 0.005% Tween 20). The concentrations of the hGABP subunits were determined by the Bradford assay (Bio-Rad). The renaturation efficiency of hGABP␤ and hGABP␥ was determined by comparing the degree of complex formation of renatured proteins with hGABP␣ to that of native proteins with hGABP␣ using the gel-shift assay. The native proteins were co-purified with hGABP␣ by affinity chromatography on latex particles containing DNA with the hGABP-binding sequence.
Gel-shift Assay and Data Evaluation-Gel-shift assays were performed as described (10,14) except that polyacrylamide gel electrophoresis (PAGE) was carried out for 30 min at 4°C. The amounts of protein-bound and free reactants separated by nondenaturing PAGE were measured using the AMBIS system (AMBIS). The equilibrium dissociation constant (K D ) of hGABP␣ for its DNA binding site was calculated by Lineweaver-Burk plot analysis. Various equilibrium states were obtained by the addition of increasing amounts of unlabeled probe DNA to the binding reactions containing either a constant amount of hGABP␣ alone or hGABP␣ and hGABP␥. The slope of the plots, 1/[DNA/hGABP␣] equilibrium against 1/[DNA] equilibrium , represents K D /[hGABP␣] input . For the analysis of the dissociation rate constant (k d ) of hGABP␣ for its binding site, the 32 P-labeled probe was incubated with hGABP␣ for 30 min at 30°C followed by the addition of a 150-fold excess of unlabeled probe DNA to the reaction. After 0 -9 min, protein-DNA complexes were separated from free DNA by PAGE and amounts of protein-DNA complex were quantified by measuring the radioactivity in each band. A single-exponential dissociation rate equation [DNA/ hGABP␣] t ϭ [DNA/hGABP␣] 0 e Ϫkdt was used to estimate the dissociation constant (k d ), which corresponded to the slope of the semilogarithmic plots of [DNA/hGABP␣] t /[DNA/hGABP␣] 0 against t, where 0 and t represent the time 0 and t after the addition of the unlabeled probe.
Binding Detected by the SPR-based Biosensor System-The SPRbased biosensor system, BIAcore2000 system (BIACORE), was used to detect in real-time the association and dissociation reactions between hGABP and DNA or between hGABP subunits. A continuous flow of running buffer 1 (30 mM Tris-HCl (pH 7.9), 30 mM KCl, 3.4 mM EDTA, 0.005% Tween 20) was maintained at 15 l/min for all of the experiments. hGABP subunits as analytes were diluted to various concentrations with running buffer 1 and were then injected so that they passed over the sensor chip surface containing the immobilized ligands. Running buffer 1 was automatically replaced with the analyte solution with no intermediate delay. The sensor chip surfaces were regenerated at the end of each detection by injecting a pulse of 2.0 M KCl for immobilized oligonucleotides or 6.0 M guanidine HCl for immobilized hGABP subunits. The data was described in the form of sensorgrams (plots of resonance units (RU) versus time), and could be used to estimate association and dissociation kinetics.
Preparation of the Sensor Surface-A Sensor Chip SA5 (BIACORE) containing streptavidin covalently preimmobilized to dextran was used for the immobilization of the biotinylated oligonucleotides. The oligonucleotides containing the core DNA sequence of the hGABP binding site, 5Ј-CGGAAGTG-3Ј (DNA1), or mutated DNA sequences (DNA2) as shown in Table I were synthesized by an Oligo 1000 M DNA synthesizer (Beckman). The complementary oligonucleotides were annealed prior to biotinylatiton with Biotin-21-UTP (CLONTECH) and Klenow fragment. Nonincorporated Biotin-21-UTP was removed by passing through a Nick Column (Amersham Pharmacia Biotech). They were diluted to 0.02 mg/ml in 20 mM Tris-HCl (pH 7.9), 0.3 M NaCl, 0.5 mM EDTA and applied to the sensor surface at a constant flow rate of 10 l/min so that the sensor surface could capture the canonical or mutant oligonucleotides via the biotin-streptavidin interaction. For the immobilization of hGABP subunits to the sensor surface of a Sensor Chip CM5 (BIACORE), the amine-coupling kit (BIACORE) was used according to manufacturer's directions. Briefly, carboxymethyl dextran on the sensor surface was activated by injecting 30 l of a mixture of 0.05 M EDC and 0.2 M N-hydroxysuccinimide, and then hGABP subunits, which were diluted to 20 g/ml with 10 mM sodium citrate (pH 3.6), were applied to the sensor surface to create covalent links with carboxymethyl dextran. The remaining active esters were blocked and deactivated by the additional injection of 35 l of 1.0 M ethanolamine.
Data Evaluation of the SPR System-The rate constants of DNAprotein and protein-protein interactions were calculated by a nonlinear analysis of the association and dissociation curves using the SPR kinetic evaluation software BIAevaluation 2.1. (BIACORE). The kinetic data were interpreted in the context of a first order kinetic model: A ϩ B ϭ AB (22,23). From the analysis of the sensorgrams of the dissociation phase, dissociation rate constants (k d ) were calculated: ln(R t /R 0 ) ϭ Ϫk (t Ϫ t 0 ), where R t and R 0 represent the SPR signal expressed in RU at time t and at the starting time of dissociation, t 0 , respectively. The slope value in a plot of ln(dR/dt) against t is expressed as Ϫ k s . Based on this value, the association rate constant (k a ) was calculated: k s ϭ Ck a ϩ k d , where C denotes the concentration of analyte. The value of the equilibrium dissociation constant was calculated from the value of the association rate constant and a single averaged dissociation rate constant was derived according to the thermodynamic relationship: The self-consistency of the data obtained from the SPR system was confirmed when these K D values were found to be approximately equal to those of the equilibrium constants, K Deq , evaluated by fitting data of the steady state response level, R sat , to the equation: 1/R sat ϭ (K Deq ϩ C)/(R max C), where R max denotes the SPR signal corresponding to complete saturation of immobilized acceptor with analyte.

hGABP␤ and hGABP␥ Are Co-expressed in Various Cell
Lines at Different Ratios-We previously reported the cloning of two types of cDNAs for hGABP␣-associated factors, hGABP␤ and hGABP␥, from a HeLa cDNA library. To examine whether hGABP␤ and hGABP␥ are co-expressed in various cell lines, Western blotting analysis was carried out using polyclonal antibody directed against each factor. As shown in Fig. 1, all of the extracts from HeLa, Jurkat, and 293 cells contained both hGABP␤ and hGABP␥. However, the ratio of hGABP␤ to hGABP␥ varied depending on the cell type. Duplex bands of hGABP␤ and hGABP␥ correspond to variants of these proteins containing a 12-amino acid insert at a position after 195 amino acids (24). These results suggest that the ratio of hGABP␥ to hGABP␤ may influence the transcription regulation mediated by hGABP␣.
Effect of the Quantitative Ratio of hGABP␤ and hGABP␥ on hGABP-mediated Transcription-Our previous study using in vitro and in vivo assays showed that the hGABP␣⅐hGABP␤ heterotetramer was able to stimulate transcription, whereas the hGABP␣⅐hGABP␥ heterodimer was not. To examine the possibility that the ratio of hGABP␤ to hGABP␥ plays a role in the control of transcription from promoters with hGABP binding sites, co-transfection assays were performed using Drosophila melanogaster Schneider line 2 (SL2) cells ( Fig. 2A). The plasmid pE4-luciferase has the luciferase gene under the control of the adenovirus early 4 promoter which contains a hGABP binding site. Various amounts of hGABP␥ expression plasmid were transfected into the cells along with constant amounts of pE4-luciferase and expression plasmids for hGABP␣ and hGABP␤. Luciferase activities were observed to decrease as a function of the increase in the amount of transfected hGABP␥ expression plasmid. When equal amounts of hGABP␤ and hGABP␥ expression plasmids were transfected, the luciferase activity was reduced by half, compared with when no hGABP␥ expression plasmid was employed (lane 3 and 6). Conversely, when various amounts of the hGABP␤ expression plasmid were transfected with constant amounts of hGABP␣ and hGABP␥ expression plasmids, transactivation activities were found to increase as a function of the augmentation in the amount of transfected hGABP␤ expression plas-mid. To check the expression level of hGABP␤ and hGABP␥, we employed a Western blotting assay using nuclear extracts prepared from the corresponding transfected cells and polyclonal antibody against hGABP␤ and hGABP␥. The results revealed that the amounts of hGABP␤ and hGABP␥ in the nuclei of the transfected cells were correlated to amounts of the corresponding expression plasmids (Fig. 2B). To examine hGABP complex formation on the adenovirus early 4 promoter in the presence of hGABP␣, hGABP␤, and hGABP␥, gel-shift assays were carried out using the recombinant proteins. As shown in Fig. 3, the amounts of the hGABP␣⅐hGABP␤ heterotetramer and hGABP␣⅐hGABP␥ heterodimer roughly correlated with the amounts of hGABP␤ and hGABP␥, respectively. This result showed that hGABP␤ and hGABP␥ bind to hGABP␣ competitively, and that hGABP␥ can drive all of the hGABP␣⅐hGABP␤ heterotetramer complex into the hGABP␣⅐hGABP␥ heterodimer complex, and vice versa. Furthermore, luciferase activities were correlated with the amount of the hGABP␣⅐hGABP␤ heterotetramer. Taken together, these results suggested that hGABP-mediated transcription activation was controlled by the quantitative ratio of hGABP␤ to hGABP␥ in the cells.
Measurement of the Kinetic Parameters of the Interactions between hGABP Subunits-Previous studies of the interactions between hGABP␣ and its two associated subunits hGABP␤ and hGABP␥ were performed using gel-shift assays (10). However, no kinetic estimates of the affinities involved in these interactions were obtained using this technique. To examine for an eventual difference in the affinity between hGABP␤ and hGABP␥ for hGABP␣, we employed the SPR system. First, using hGABP␥ immobilized on the sensor chip surface, the binding reaction between hGABP␣ and hGABP␥ was studied by injecting hGABP␣ at various concentrations (Fig. 4A). The values of the k d , k a , and K D were estimated from the experimental curves as being 3.6 Ϯ 0.86 ϫ 10 Ϫ4 s Ϫ1 , 5.7 Ϯ 0.18 ϫ 10 5 M Ϫ1 s Ϫ1 , and 6.4 Ϯ 1.7 ϫ 10 Ϫ10 M, respectively (Fig. 4B). The data used to construct the curves was corrected by subtracting the responses of an empty sensor chip surface from the responses of a hGABP␥-immobilized sensor chip surface. Kinetic values were also calculated using a sensor surface with immobilized hGABP␣. Analysis of the experimental curves yielded kinetic constants of 6.5 Ϯ 0.97 ϫ 10 Ϫ4 s Ϫ1 , 4.0 Ϯ 0.60 ϫ 10 5 M Ϫ1 s Ϫ1 , and 17 Ϯ 5.0 ϫ 10 Ϫ10 M for the k d , k a , and K D , respectively  (Table II). Thus, there exist discrepancies in these values, depending on which protein is bound to the sensor chip. This may reflect the negative charges of dextran on the sensor chip surface or the heterogeneity in the immobilizing sites introduced by the immobilization procedure.
Similarly, the binding of hGABP␣ to immobilized hGABP␤ was also analyzed using the SPR system. The kinetic constants were determined to be 4.3 Ϯ 0.14 ϫ 10 Ϫ4 s Ϫ1 , 5.5 Ϯ 0.26 ϫ 10 5 M Ϫ1 s Ϫ1 , and 7.8 Ϯ 0.63 ϫ 10 Ϫ10 M for the k d , k a , and K D , respectively (Table II). These results suggest that hGABP␤ and hGABP␥ bind to hGABP␣ with similar affinities.
Analysis of the Role of hGABP␥ in hGABP␣⅐hGABP␥ Complex Binding to DNA-To examine whether the binding of hGABP␤ or hGABP␥ to hGABP␣ affects the DNA binding activity of hGABP␣, gel-shift assays were performed. When increasing amounts of hGABP␤ or hGABP␥ were added to the binding reactions containing constant amounts of 32 P-labeled DNA probe and hGABP␣ protein, the level of free DNA was lowered compared with when hGABP␣ associated factors were not added (Fig. 5). This suggests that the binding of hGABP␣associated factors enhances the stability of the interaction between hGABP␣ and DNA. To further analyze the phenomenon kinetically, the dissociation rates of both hGABP␣⅐hGABP␤ and hGABP␣⅐hGABP␥ complexes from the probe were measured and compared with that of hGABP␣ by the gel-shift assay.
In both cases, about a 10-fold stabilization was observed in the dissociation rate constant compared with the hGABP␣-DNA interaction whose equilibrium dissociation constant (K D ) and dissociation rate constant (k d ) were found to be 2310 Ϫ10 M and 1.910 Ϫ3 s Ϫ1 (Fig. 6), respectively.
To further confirm that hGABP␣-associated factors enhance the affinity of hGABP␣ for DNA, the SPR system was also employed to measure the dissociation rate constant, as rebinding reactions between hGABP complex and DNA and between hGABP subunits can be avoided using this system. The effect of hGABP␥ on hGABP␣-DNA binding was taken to be representative of those of hGABP␤ and hGABP␥, as they share the same binding mechanism and affinity for hGABP␣. Oligonucleotides with the hGABP-binding sequence (DNA1; Table I) or mutated sequence (DNA2; Table I) were immobilized on the surface of a sensor chip SA5 via biotin-avidin interactions. First, repetitive injections of various concentrations of hGABP␣ were carried out for precise estimations of the kinetic parameters (Fig. 7A). The response was only detected when the sensor surface contained DNA1, while there was no response using sensor surfaces lacking oligonucleotides or containing DNA2. The analyses of the experimental curves gave a dissociation rate constant (k d ), an association rate constant (k a ), and an equilibrium dissociation constant (K D ) between hGABP␣ and its DNA binding site of 4.4 Ϯ 0.64 ϫ 10 Ϫ3 s Ϫ1 , 3.0 Ϯ 0.090 ϫ 10 6 M Ϫ1 s Ϫ1 , and 15 Ϯ 2.6 ϫ 10 Ϫ10 M, respectively. These values were similar with those evaluated by the gel-shift assay. Next, the analyte mixtures containing a constant concentration of hGABP␣ and increasing concentrations of hGABP␥ were passed over the DNA1 immobilized on the sensor surface (Fig. 7C). In the dissociation phase of each of the sensorgrams, the response level was normalized at the point where the dissociation reaction began (Fig. 7D). The results showed that the dissociation rate of hGABP␣ from the immobilized DNA became gradually lower as the concentration of hGABP␥ was increased in the analyte mixture. This result is consistent with the results obtained by the gel-shift assay. Using these results, we estimated the dissociation rate constant of the interaction between the hGABP binding site and the hGABP␣⅐hGABP␥ complex. We used the experimental curve of dissociation obtained by injecting a mixture of 20 nM hGABP␣ and 80 nM hGABP␥ because factors binding to DNA were thought to contain little hGABP␣ alone at the beginning of dissociation. We analyzed this curve as the sum of two dissociation modes. One dissociation pathway was that of the hGABP␣⅐hGABP␥ complex from the DNA. The other one was the dissociation pathway of hGABP␥ from hGABP␣ and the following dissociation of hGABP␣ from the DNA. The dissociation rate constant of hGABP␣⅐hGABP␥ complex from the DNA was calculated to be 1.6 Ϯ 0.022 ϫ 10 Ϫ3 s Ϫ1 , which is about 2.8-fold less than the k d value of hGABP␣ alone. We confirmed that the experimental dissociation curves almost coincided with ideal interaction curves which were simulated with kinetic constants. The result suggests that the binding of hGABP␥ to hGABP␣ stabilizes the DNA-hGABP␣ interaction by imparting a lower dissociation rate constant.

DISCUSSION
In this report, we studied the role of hGABP␣-associated proteins hGABP␤ and hGABP␥ in hGABP-mediated transcriptional activation and their effect on the DNA binding activity of hGABP␣. We also sought to determine the interaction affinity of hGABP␣ for hGABP␤, hGABP␥, and DNA.
Many Ets family proteins have been shown to possess associated factors that result in the formation of large complexes with various levels of transcription activity (25,26). We previously showed that hGABP␤ and hGABP␥ are functional partners of Ets-related hGABP␣. Here, we show that they have similar affinities for hGABP␣ and that they compete with one another for binding to hGABP␣ to regulate transcription activity. These findings are supported by functional domain analysis (14), which has revealed that hGABP␤ and hGABP␥ have identical domains required for binding to hGABP␣. The similar K D values of hGABP␤ and hGABP␥ for hGABP␣ would also appear to reflect the ratio of the hGABP␣⅐hGABP␤ complex to the hGABP␣⅐hGABP␥ complex in cells which is dependent on the respective intracellular concentrations of hGABP␤ and hGABP␥. The transcription activation of hGABP is dependent on the ratio of hGABP␤ to hGABP␥. These hGABP complexes have been shown to have different transcription activities in vivo and in vitro, depending on whether they are made up of hGABP␣⅐hGABP␤ complex or the hGABP␣⅐hGABP␥ complex (14,15). Our results by Northern blot assay show that the expression levels of hGABP␥ mRNA were varied in human tissues examined, compared with that of hGABP␤ (data not shown). Therefor, the ratio of the hGABP␥ mRNA to hGABP␤ mRNA was different in various human tissues. For example, the ratios in skeletal muscle, liver and testis were 0.26, 0.43, and 1.2, respectively. This result is reminiscent of a similar observation previously reported by M. Marchioni et al. (32) who found that the pattern of immunologically related hGABP␤ subunit expression changes during Xenopus embryonic development while the pattern of hGABP␣-like protein expression remains almost the same. These suggest that the ratio of hGABP␤ to hGABP␥ may contribute to the characters and functions of the tissues. Taken together, the results of the co-transfection assays and the kinetic studies suggest that the quantitative ratio of hGABP␤ to hGABP␥ directly influences the relative amounts of hGABP␣⅐hGABP␤ complex and FIG. 6. Analysis of the interaction between hGABP␣ and DNA using the gel-shift assay. A, the gel-shift assay was performed as described under "Experimental Procedures." The indicated amounts of nonradiolabeled DNA were used with 0.5 ng of 32 P-labeled DNA in each reaction containing 2 ng (lanes 1-5) or 5 ng (lanes 6 -10) of hGABP (except for lane 11). B, Lineweaver-Bark plot of the gel-shift assay shown in panel A. The radioactivity in each band corresponding to both free DNA and DNA⅐hGABP␣ complex was measured by the AMBIS system (AMBIS) and plotted. 2 ng (open circles) or 5 ng (closed diamonds) of hGABP␣ were used in the assay. C, gel-shift assays were performed as described under "Experimental Procedures." 5 ng of hGABP␣ and 0.2 ng of 32 P-labeled DNA probe were mixed. After the incubation, 30 ng of nonradiolabeled DNA probe was added to the reaction (lanes 2-6). After the indicated time, the reaction mixtures were applied to PAGE. D, semi-log plots of the relative amounts of the DNA⅐hGABP␣ complex. After the gel-shift assay shown in panel C, the radioactivity of each band corresponding to the DNA⅐hGABP␣ complex was measured by the AMBIS system (AMBIS) and plotted. The slope represents Ϫ k d . 2 ng (open squares) or 5 ng (closed diamonds) of hGABP␣ were used in the gel-shift assay.

TABLE I Oligonucleotides and sequences
The hGABP-binding site and mutation are designated by boldface letters and small letters, respectively.
hGABP␣⅐hGABP␥ complex in the cell, resulting in fine regulation of promoters having binding sites for the hGABP complex. However, certain promoters may have a mechanism to select for the binding of either the hGABP␣⅐hGABP␤ complex or the hGABP␣⅐hGABP␥ complex, where other DNA-binding factors on other regions of the promoter may play an important role in recruiting either hGABP complex on the promoter as a transcriptional regulation partner. Such a mechanism of transcription regulation is analogous to that of the positive factor TFIIA and the repressors, Dr1 and Dr2. These factors bind to the basic region of the TATA box-binding protein competitively, resulting in transcription regulation (27,28). As in the case of hGABP␥ and the hGABP complex, some gene-specific transcription factors with DNA binding activity, including Ets family proteins, may have partners which antagonize their interaction with associated factors necessary for their transactivation activity.
The K D value of hGABP␤ and hGABP␥ for hGABP␣ leads to the interpretation that most of the hGABP␣ protein present in the cell forms a complex with hGABP␤ or hGABP␥, if their intracellular concentrations are taken into account. This notion is supported by the observation that these three factors can be co-purified from HeLa nuclear extracts using Sepharose beads bound to DNA containing the hGABP-binding sequence (10), and that hGABP␤ and hGABP␥ are responsible for efficient hGABP␣ migration into the nucleus as observed by an immunofluorescence assay of the co-transfected cells (15).
The DNA-protein interactions of hGABP were also characterized kinetically using the SPR system and the gel-shift assay as summarized in Table II. The binding affinity of hGABP␣ for its target sequence (K D of 15 Ϯ 2.6 ϫ 10 Ϫ10 M) is equivalent to that of p42/Ets-1 (K D of 36 ϫ 10 Ϫ10 M) reported previously (29). This may reflect the conservation of the highly homologous DNA binding domain, termed the Ets domain, in both proteins. The DNA binding affinity of gene-specific transcription factors generally ranges from 10 Ϫ8 M to 10 Ϫ11 M (equilibrium dissociation constant). The affinity of their Ets domains for target sequences is lower than that of the D. melanogaster transcription factor, fushi-tarazu, which has a homeodomain for DNA binding, and whose K D was reported to be approximately 2.5 ϫ 10 Ϫ11 M by the gel-shift assay (30). We have also shown that the binding of hGABP␥ to hGABP␣ stabilizes the DNA binding affinity of hGABP␣ because it imparts a slower dissociation rate. The interaction may lead to conformational change(s) in hGABP␣. We observed that the binding did not result in faster association of hGABP␣ with its target  7. Analysis of the interaction between hGABP and DNA using the SPR system. A, representation of sensorgrams using sensor chip surfaces with and without 300 RU of DNA. 120 l of hGABP␣ solutions of various concentrations were injected at a flow rate of 15 l/min. The concentrations of hGABP␣ are indicated above each line. B, plots of k s versus the concentration of hGABP␣ for the interaction between hGABP␣ and DNA. The slope is equivalent to the association rate constant. C, the experimental curves of the interaction between the hGABP␣⅐hGABP␥ complex and DNA. These curves were obtained by subtracting background responses from the raw interaction curves. 210 l of mixtures of 20 nM hGABP␣ and the indicated concentrations of hGABP␥ were injected onto the same sensor chip surface as described in Fig. 2 at a flow rate of 15 l/min. The mixture was incubated at 25°C for 30 min before injection. D, the arranged curves of the dissociation phase by normalizing to the initial point of the dissociation phases.
sequence. Recently, Batchelor et al. (31) reported structural analysis of the GABP␣ and GABP␤ complex and showed that Lys 69 of GABP␤ makes an indirect contact with the DNA sugar-phosphate backbone via Glu 321 of GABP␣. This may explain the slower dissociation rate. As the result of such stability the hGABP complex may reside for longer periods at promoters containing its recognition site than hGABP␣ alone. These findings suggest the importance of the stability of DNA binding activity in addition to interactions between the transactivation domain and the RNA polymerase machinery for transcription activation.