The a and b Subunits of the GA-binding Protein Form a Stable Heterodimer in Solution ASSEMBLY*

We have studied the assembly of GA-binding protein (GABP) in solution and established the role of DNA in the assembly of the transcriptionally active GABP a 2 b 2 heterotetrameric complex. GABP binds DNA containing a single PEA3/Ets-binding site (PEA3/EBS) exclusively as the ab heterodimer complex, but readily binds as the GABP a 2 b 2 heterotetramer complex on DNA containing two PEA3/EBSs. Positioning of the PEA3/EBSs on the same face of the DNA helix stabilizes heterotetramer complex binding. These observations suggest that GABP ab heterodimers are the predominant molecular species in solution and that DNA containing two PEA3/ EBSs promotes formation of the GABP a 2 b 2 heterotet- rameric complex. We analyzed the assembly of GABP a 2 b 2 heteromeric complexes in solution by analyt- ical ultracentrifugation. GABP a exists as a monomer in solution while GABP b exists in a monomer-dimer equilibrium ( K d 5 1.8 6 0.27 m M ). In equimolar mixtures of the two subunits, GABP a and GABP b formed a stable heterodimer, with no heterotetramer complex detected. Thus, GABP exists in solution as the heterodimer previously shown to be a weak transcriptional activator. Assembly of the transcriptionally active GABP a 2 b 2 het- erotetramer complex requires the presence of specific DNA containing at least two PEA3/EBSs. The Q single-species (Equation a molecular weight equal to of 6 GABP a Q. concentration distributions GABP b fitted to a two-species associative model (Equation “Experimental with the molecular weight of the monomer equal of 6 GABP b concentration distributions GABP fitted to a two-species ideal molecular terminal leucine zipper-like domain of GABP b is obligatory for formation of GABP heterotetramer complexes.

The GABP-binding site first identified in the herpes simplex virus immediate early promoter contains two tandemly arranged PEA3/Ets-binding sites (EBS) (1,18,19). Early analysis of GABP binding to the herpes simplex virus immediate early promoter indicated that GABP forms predominantly a heterotetrameric complex composed of two ␣ and two ␤ subunits (1,2). The results of gel filtration and preparative ultracentrifugation of highly purified recombinant GABP␣ and GABP␤ proteins were further interpreted to support a model depicting GABP as a stable heterotetrameric complex in solution, suggesting that this complex binds as a single unit to DNA (1,2). The ability to form the heterotetrameric complex has been shown to be necessary for GABP-dependent transcription (20 -22). Thus, modulation of GABP tetramer formation represents a potentially important means of regulating GABP-dependent transcription.
We have previously shown that GABP-dependent transcription initiator activity required two PEA3/EBSs, and the highest activity was achieved with two sites positioned on the same face of the DNA helix (23)(24)(25). Since the ability to activate transcription was previously shown to be dependent on heterotetramer complex formation, these results favored the stable GABP heterotetramer complex model. However, electrophoretic mobility shift assays (EMSA) with DNA probes containing only a single PEA3/EBS detected a complex consistent with the mobility of the GABP␣␤ heterodimer (23,26,27). In contrast, a probe with two PEA3/EBSs formed two complexes, one co-migrating with the GABP␣␤ complex observed on a single PEA3/EBS and a slower migrating complex identified as GABP␣ 2 ␤ 2 heterotetramer. These observations are consistent with the notion that GABP binds to a single PEA3/EBS only as a heterodimer, and requires two or more PEA3/EBSs to bind DNA as a heterotetramer. Therefore, if GABP forms a stable heterotetrameric complex in solution, then dissociation of this complex would be required upon binding to a single PEA3/EBS. Alternatively, GABP may form a stable heterodimer, or may exist in heterodimer-heterotetramer equilibrium that, at concentrations typically found in EMSA assays, favors the GABP␣␤ heterodimer. Therefore, prior interaction of a GABP␣␤ heterodimer complex with DNA containing two PEA3/EBS may be required for successful assembly of a transcriptionally competent GABP␣ 2 ␤ 2 heterotetramer.
To address the mechanism of GABP assembly and the role of DNA binding in this process, we have utilized EMSAs and analytical ultracentrifugation. In this report we show that GABP exists exclusively as a stable heterodimeric complex (GABP␣␤). Even at exceedingly high concentrations little or no significant quantities of GABP heterotetramer complex is observed. These observations support the hypothesis that GABP exists as a stable heterodimer in solution which assembles into the heterotetrameric complex upon binding to target DNA containing two PEA3/EBSs. We further demonstrate that heterotetramer DNA binding is stabilized by positioning the two PEA3/EBSs on the same face of the DNA helix, and that the two motifs can be separated by up to three helical turns. These results are consistent with our earlier results demonstrating that GABP-dependent initiators were most efficient when two PEA3/EBSs were positioned on the same face of the DNA helix, and confirms the importance of the GABP heterotetramer complex in transcription initiator and activator activities of this important regulatory factor (23 Cloning and Expression of Recombinant Proteins in Escherichia coli-DNAs encoding GABP␣ and GABP␤ proteins were amplified by polymerase chain reaction from cDNAs kindly provided by C. C. Thompson and cloned into pET15b (Novagene) as described previously (26). The individual recombinant (rGABP) His 6 -tagged proteins (Fig. 1A) were expressed in E. coli BL21 strain and recovered from cell extracts by nickel chelating or cobalt immobilized affinity chromatography under denaturing conditions in the presence of 6 M urea. rGABP␣ and rGABP␤ proteins were precipitated with ammonium sulfate at 20 and 14% saturation, respectively. Precipitated material was resuspended in 50 mM potassium phosphate buffer, pH 7.4, containing 100 mM potassium chloride, 6 M urea, and 5 mM dithiothreitol and subjected to several steps of dialysis in the same buffer with progressively lower concentrations of urea (5 M, 3 M, etc.). The extinction coefficients of rGABP␣ (⑀ 280 ϭ 62,796 M Ϫ1 cm Ϫ1 ), rGABP␤ (⑀ 280 ϭ 18,262 M Ϫ1 cm Ϫ1 ), rGABP␣ c Q (⑀ 280 ϭ 26,008 M Ϫ1 cm Ϫ1 ), and rGABP␤ 334 (⑀ 280 ϭ 15,422 M Ϫ1 cm Ϫ1 ) proteins were calculated from their amino acid compositions. The rGABP␣ c Q protein containing four Cys-Ser substitutions, described previously (27), was expressed at high levels in E. coli and was less prone to aggregation than the wild-type GABP␣ c protein.
Analytical Ultracentrifugation Analysis-Prior to sedimentation analysis, the samples were clarified by centrifugation at 45,000 rpm for 30 min in a TL-100 centrifuge (Beckman Instruments, Palo Alto, CA). Equilibrium sedimentation experiments with His 6 -tagged rGABP␣ and rGABP␤ were performed in a Beckman Optima XL-I ultracentrifuge equipped with AN50ti rotor (Beckman Instruments, Palo Alto, CA). Six-channel cells with 12-mm optical path length were used. Data were collected at 20°C at rotor speeds between 6,000 and 28,000 rpm. Sample loading concentrations ranged between 2 and 13 M in 50 mM phosphate buffer, pH 7.4, containing 100 mM KCl, 10% glycerol, and 1 mM dithiothreitol. The distribution of solutes in the cell was monitored by absorbance at 280 nm. The resulting distributions at equilibrium were subjected to global least-square analysis employing several models: a single exponential (monomeric ideal, Equation 1), two exponential (dimeric ideal, Equation 2), or two exponential (associative, Equation 3) (28) using the program Origin 5.0 (MicroCal Software, Inc. MA), Where C 10 , C 20 , and C m are the absorbance of the first, second, and monomeric species, respectively, at a reference point. M 1 , M 2 , and M m are the molecular weights of the first, second, and monomeric species, respectively. The angular velocity, , was calculated from the rotor speed. The value for partial volume, v, at 20°C was calculated from amino acid composition using SEDNTERP software. R is the universal gas constant. T is the absolute temperature equal to 293.15 K. The parameter C x is the baseline that was either determined from the absorbance near the meniscus after sedimentation of the samples at 20,000 rpm for 12 h or from global least-square analysis.
To account for nonspecific aggregation at high protein concentrations or for deviations from stoichiometry in multi-component mixtures, a third term was introduced into equation III, producing Equation 4.

GABP Requires Two PEA3/EBSs for Heterotetramer Complex Formation on DNA-
The GABP-binding site originally identified in the HSV IE promoter contained two tandemly arranged PEA3/EBSs (1,2,18). When two PEA3/EBSs (dPEA3-0) are present on the target DNA, approximately equal amounts of both heterodimeric and heterotetrameric complexes are readily observed (Fig. 1B). However, on a probe containing only a single PEA3/EBS (PEA324), GABP forms exclusively the heterodimeric complex (Fig. 1B). No heterotetrameric complex was observed on the PEA324 probe containing a single PEA3/ EBS even at 10-fold higher GABP protein concentrations. 2 These results indicate that at the concentration used in these EMSA experiments (ϳ10 nM), GABP requires two PEA3/EBSs to efficiently assemble into a heterotetrameric complex on DNA.
Deletion of the Amino Terminus of GABP␣ Enhances Heterotetramer Complex Formation on DNA-Previous studies indicated that an NH 2 -terminal truncation mutant of the GABP␣ protein (GABP␣ c Q) exhibited enhanced heterotetramer complex DNA in the presence of GABP␤ protein (26,27). GABP␣ c Q protein (Fig. 1A) readily forms both heterodimeric and heterotetrameric complexes with GABP␤ protein on DNA containing a single PEA3/EBS (Fig. 1B). When two PEA3/EBSs are present on the target DNA, only the heterotetrameric complex is observed. Thus, enhanced GABP heterotetramer complex DNA binding is achieved by deletion of the NH 2 -terminal two thirds of the GABP␣ protein. Enhanced DNA binding of the heterotetramer complex requires the COOH-terminal leucine zipper-like domain in GABP␤. Deletion of this region in the GABP␤ 334 mutant abolishes the formation of detectable heterotetrameric complex with either full-length GABP␣ or GABP␣ c Q proteins on probes containing either one or two PEA3/EBSs. GABP Heterotetramer Complex DNA Binding Is More Stable When Two PEA3/EBSs Are Positioned on the Same Face of the DNA Helix-We have previously shown that two PEA3/EBSs function as a GABP-dependent initiator element, and that maximal initiator activity occurs on templates containing two PEA3/EBSs on the same face of the DNA helix (23). In addition, GABP activation has been shown previously to depend on the ability of GABP to form the heterotetrameric complex (20 -22). Therefore, we sought to determine whether the preference of GABP-dependent initiator elements for templates containing two PEA3/EBS on the same face of the DNA helix correlates with the ability of GABP heterotetrameric complex to form on these initiator elements. We performed EMSA analysis of GABP binding to DNA probes containing two PEA3/EBSs sep-arated by a 0, 10-, 16-, 22-, or 26-base pair linker DNA, corresponding to 0.5, 1.5, 2.0, 2.5, and 3.0 helical turns between PEA3/EBS. Approximately equivalent amounts of heterotetramer complex formed on each of the DNAs analyzed ( Fig. 2A), suggesting that the ability to assemble into a heterotetrameric complex is not affected by the helical spacing between two PEA3/EBS elements.
Analysis of heterotetramer complex DNA binding revealed that GABP heterotetramer complexes bound to DNA containing two PEA3/EBSs on the same face of the DNA helix were significantly more stable than complexes bound to DNA containing two PEA3/EBSs on the opposite sides of the DNA helix The GABP␣ c Q protein contains four Cys to Ser substitutions that allowed enhanced expression of this truncated protein in E. coli. This protein was previously shown to have DNA binding characteristics similar to that of the wild-type protein (26,27). Amino acid residues 1-156 of GABP␤ and GABP␤ 334 include the four ankyrin repeats involved in association with GABP␣, and also found in several transmembrane proteins including the product of the Notch gene of Drosophila melanogaster (1)(2)(3). Residues 334 to 382 of GABP␤ include the leucine zipper-like homodimerization domain, which has been deleted in the GABP␤ 334 protein. B, recombinant GABP␣, GABP␤, or mutant proteins in various combinations were incubated with 32 P-labeled probes containing either one (PEA324) or two (dPEA3-0) PEA3/EBSs and analyzed by EMSA as described under "Experimental Procedures." In all cases, the GABP␣ or GABP␣ c Q component was present in excess to allow visualization of the monomeric GABP␣ and GABP␣ c Q protein-DNA complexes. The positions of GABP wild-type and truncated protein-DNA complexes are indicated. To conserve space, complexes containing the GABP␣ c Q protein are labeled "␣ c " in the figure.
FIG. 2. The role of spacing between PEA3/EBSs in heterotetramer complex DNA stability. A, rGABP␣ and rGABP␤ proteins were mixed in EMSA buffer and incubated with 32 P-labeled probes containing either one (PEA324) or two PEA3/EBSs separated by various distances, indicated as multiples of helical turns, and the resulting complexes resolved EMSA as described under "Experimental Procedures." GABP␣␤ complexes, monomeric GABP␣, and unbound probe are indicated on the left. B, rGABP␣ and GABP␤ subunits were mixed with 32 P-labeled probes as in panel A, incubated for 15 min, and the resulting DNA-protein complexes challenged with 500-fold molar excess of unlabeled double-stranded oligonucleotide containing a single PEA3-binding site (PEA324). Aliquots of the binding reactions were taken at various times and applied to a running polyacrylamide gel. Following fractionation, gels were autoradiographed and quantified by PhosphorImager analysis using ImageQuant (Molecular Dynamics). The data are expressed as the Log 10 % Bound, representing the fraction of heterotetramer complex that bound to each probe at the zero time point, which remained bound at each subsequent time point. Data for the PEA324 is derived from analysis of the stability of the heterodimer complex formed with this probe. (Fig. 2B). GABP heterotetramer complexes bound to radiolabeled probes were challenged with an excess of unlabeled competitor DNA and the rate of heterotetramer complex decay measured over time. Heterotetramer complexes bound to probes containing two PEA3/EBSs separated by 2.0 and 3.0 helical turns were substantially more resistant to challenge by the unlabeled competitor DNA than were complexes bound to probes containing two PEA3/EBSs separated by 0.5, 1.5, and 2.5 helical turns. Heterotetramer complexes bound to all probes were substantially more stable than the heterodimer complex bound to DNA containing a single PEA3/EBS. Thus, the ability to assemble into a heterotetrameric complex during DNA binding is independent of helical spacing or distance (up to 4 helical turns), 2 but complex stability is enhanced when two PEA3/EBSs are positioned on the same face of the DNA helix. These observations explain, in part, our previous results cor- FIG. 3. Sedimentation equilibrium analysis of individual GABP␣, GABP␣ c Q, GABP␤, and GABP␤ 334 proteins. Analysis was performed at 20°C in a Beckman Model XL-I ultracentrifuge at speeds of 6,000, 9,000, and 12,000 rpm (GABP␣), 12,000, 16,000, and 20,000 rpm (GABP␣ c Q), or 9,000, 12,000, and 15,000 rpm (GABP␤, and GABP␤ 334 ), and at initial concentrations ranging between 0.1 and 0.7 A 280 nm . Concentration distributions were subjected to global least square analysis using Origin 5.0 software (Microcal). The quality of fit is illustrated by the random distribution of residuals, i.e. the differences between calculated and observed values. A, the concentration distributions for the GABP␣ protein were treated using an ideal single-species model (Equation 1, "Experimental Procedures"), with the molecular weight equal to that of the His 6 -tagged GABP␣ protein. B, the concentration distributions for the GABP␣ c Q protein were treated using an ideal single-species model (Equation 1, "Experimental Procedures") with a molecular weight equal to that of His 6 -tagged GABP␣ c Q. C, the concentration distributions for GABP␤ were fitted to a two-species associative model (Equation 3, "Experimental Procedures") with the molecular weight of the monomer equal to that of His 6 -tagged GABP␤. D, the concentration distributions for GABP␤ 334 were fitted to a two-species ideal model (Equation 2, "Experimental Procedures") with the molecular weight of the first species equal to that of His 6 -tagged GABP␤ 334 . The molecular mass of the second species was 1,417.503 Ϯ 69.435 kDa. relating helical spacing between PEA3/EBSs with GABP initiator activity.

GABP␣ and GABP␣ c Q Proteins Are Monomeric in Solution-
The association states of GABP␣, GABP␤, and the various mutant proteins (Fig. 1) were analyzed by sedimentation equilibrium. The distribution of GABP␣ (Fig. 3A) is satisfactorily described by a single species model ("Experimental Procedures," Equation 1), in which the molecular weight is set to the calculated value of 53,500 for the His 6 -tagged GABP␣ protein.
We attribute the discrepancies between the calculated and experimental values observed at the highest protein concentration and rotor speeds, to nonspecific aggregation of GABP␣. These data suggest that GABP␣ exists as a monomeric species under our experimental conditions. GABP␣ c Q protein behaved similarly (Fig. 4B), although no aggregation was observed. The radial distribution for GABP␣ c Q (Fig. 4B) is accommodated by an ideal single species model in which the molecular mass is fixed at 18,551 Da, the calculated molecular mass of His 6 -tagged GABP␣ c Q. In contrast to the full-length GABP␣ protein, we observe no indication of higher molecular weight aggregates.
GABP␤ Weakly Self-associates in Solution Through the COOH-terminal Leucine Zipper-like Domain-In contrast to GABP␣, we see evidence for dimerization, albeit weak, of GABP␤. The distribution of GABP␤ in sedimentation equilibrium experiments is consistent with a monomer-dimer equilibrium associative model (Equation 3) (Fig. 3C). The mass of the monomeric species was fixed at 43,422 Da, equal to that of His 6 -tagged GABP␤, and the dissociation constant was treated as a variable parameter. The optimal value for the homodimer dissociation constant, K d␤ , was determined to be 1.8 Ϯ 0.27 M, indicating that GABP␤ associates only weakly under these conditions. At physiologically relevant concentrations (0.1-10 nM), GABP␤ would be almost exclusively monomeric.
Deletion of the COOH-terminal portion of GABP␤ (GABP␤ 334 ), which includes the leucine zipper-like domain, eliminates the tendency to self-associate. The radial distribution of GABP␤ 334 is consistent with an ideal two-species model (Equation 2) with the molecular mass of the first species equal to that of His 6 -tagged GABP␤ 334 protein (37,691 Da), and the molecular mass of the second species equal to 1,417,503 Ϯ 69,435 Da (Fig. 3D). GABP␤ 334 protein tends to precipitate at high protein concentrations suggesting that the second high molecular weight species may be attributed to nonspecific aggregation of the GABP␤ 334 protein at high concentrations. Consistent with this assumption, the molecular mass of the second species varied from preparation to preparation but was consistently higher than 600 kDa. These results are consistent with the previously proposed role of the COOH-terminal leucine zipper-like domain of GABP␤ in self-association (1-3, 20, 29).
GABP Exists Predominantly as a Heterodimer in Solution in the Absence of DNA-The virtual absence of heterotetrameric GABP bound to a single PEA3/EBS in EMSAs suggests that any heterotetrameric complex formed in solution must have a low affinity for a single PEA3/EBS. Alternatively, GABP heterodimers may have very little tendency to assemble into the heterotetrameric complex on DNA containing a single PEA3/ EBS. To distinguish between these possibilities, we performed analytical ultracentrifugation to determine the relative amounts of GABP heterodimer and heterotetramer complexes in solution in the absence of DNA binding.
When stoichiometric amounts of GABP␣ and GABP␤ are combined and centrifuged to equilibrium, the resulting radial distribution is consistent with exclusive heterodimer formation. Optimal agreement between the calculated and observed values was obtained with an ideal two species model (Equation 2), in which the molecular mass for the major species was fixed at 96,949 Da (the sum of molecular masses of GABP␣ and GABP␤). The mass for the second species was allowed to vary and converged to 619,573 Ϯ 13,597 Da (Fig. 4A). The presence of this material, which accounted for less than 5% of the total protein, is attributed to nonspecific aggregation. We were unable to fit the concentration distributions to a hetero-associative model, suggesting a very strong GABP␣-GABP␤ interaction. Based on the instrument detection limit, the K d for the GABP ␣␤ heterodimer was estimated to be below 10 Ϫ8 M. This estimated K d for the GABP ␣␤ heterodimer is consistent with the reported K d (7.8 Ϯ 0.63 ϫ 10 Ϫ10 M) as measured by surface plasmon resonance (30). These results demonstrate that GABP under the experimental conditions and concentration range used in these experiments, exists exclusively as a stable heterodimer. Attempts to fit the data to a "stable heterotetramer" model ( Fig. 4B, left panel) or to a two-species associative model (Fig. 4B, right panel) were unsuccessful.
Deletion of the NH 2 -terminal Portion of GABP␣ Promotes Heterotetramer Formation in Solution-Although our data indicate that GABP␤ weakly dimerizes in solution, the GABP␣␤ heterodimer exhibits no tendency to associate, even at concentrations above 2.5 M. To resolve this apparent contradiction, we studied the association of GABP␣ c Q and GABP␤ in solution.
In EMSA experiments (Fig. 1), GABP␣ c Q and GABP␤ predominantly form heterotetrameric complex when incubated with probe containing two PEA3/EBSs and displays a limited tendency to bind as a heterotetramer to DNA containing a single PEA3/EBS. To determine the associative state of GABP␣ c Q and GABP␤ when free in solution, a slight excess of GABP␣ c Q protein was combined with GABP␤ protein and centrifuged to equilibrium. The solute distribution was satisfactorily described by a heterodimer-heterotetramer equilibrium plus a third nonassociating species (Fig. 5, Equation 4). The molecular mass of the heterodimer was fixed at 61,525 Da, equal to the molecular mass of the His 6 -tagged GABP␣ c Q⅐GABP␤ complex.
The molecular mass of the third species was fixed at 18,551 Da, equal to that of the His 6 -tagged GABP␣ c Q monomer, which was present in excess. Least-squares minimization yielded a dissociation constant for GABP␣ c Q-GABP␤ dimer-tetramer equilibrium equal to 0.17 Ϯ 0.045 M, significantly lower than that obtained for GABP␤ alone. These results suggest: 1) the COOH terminus of GABP␣ promotes GABP␤ dimerization and 2) that the NH 2 terminus of GABP␣ may interfere with heterotetramer complex assembly in solution. The latter conclusion explains, in part, the failure of the full-length GABP␣ protein to form heterotetrameric complex with GABP␤ protein in solution or bound to DNA containing a single PEA3/EBS.

GABP Heterotetramer Complex Formation Is Dependent on the COOH-terminal Leucine Zipper-like Domain in GABP␤-
Since the COOH-terminal leucine zipper-like domain has been shown previously to be essential for GABP␤ homodimerization and for heterotetramer complex DNA binding (Fig. 1) (2,3,18,29), we studied the association of GABP␤ 334 , lacking the COOH-terminal homodimerization domain, with full-length GABP␣ and GABP␣ c Q in solution. The solute distribution in a GABP␣ c Q/GABP␤ 334 mixture could be accommodated by an ideal single-species model (Equation 1) with a molecular mass fixed at 56,241 Da, which equals the sum of the His 6 -tagged GABP␣ c Q and GABP␤ 334 molecular masses (Fig. 6A). Similar results were obtained when the concentration distribution of a mixture of full-length GABP␣ and GABP␤ 334 proteins was analyzed (Fig. 6B). These data are consistent with an ideal two-species model (Equation 2) with a molecular mass of one species fixed at 91,218 Da (His 6 -tagged GABP␣ ϩ His 6 -tagged GABP␤ 334 ) and the other fixed at 37,691 Da (uncomplexed His 6 -tagged GABP␤ 334 ). The apparent absence of tetrameric species in these two experiments suggests that the COOH- terminal leucine zipper-like domain of GABP␤ is obligatory for formation of GABP heterotetramer complexes. DISCUSSION Since the discovery of GABP (1,2,18), it has been widely accepted that GABP exists in solution predominantly as a heterotetramer (␣ 2 ␤ 2 ), due to stable ␤-␤ interactions (2,3). This model was supported by subsequent observations that heterotetramer formation was required for full transcriptional transactivation by GABP (20,21). Although attractive, this model is not consistent with the absence of heterotetramer complex in EMSA experiments performed with a DNA probe containing a single PEA3/EBS (Fig. 1). DNA containing two PEA3/EBSs is required to observe significant amounts of heterotetramer complex in EMSA experiments, although equivalent amounts of heterodimer complex is observed in these experiments. These results are inconsistent with a stable GABP heterotetramer complex binding as a single unit to DNA containing one or two PEA3/EBSs. Cellular proteins such as bcl3 can promote GABP heterotetramer formation in EMSA experiments (31), suggesting that the absence of the heterotetrameric complex does not merely reflect its instability under EMSA conditions, but that its formation may be a regulated process. Our observations suggest that the GABP␣␤ heterodimer is the major species found in solution and that efficient heterotetramer formation requires specific DNA for assembly.
To directly address the associative behavior of GABP in solution, we have performed sedimentation equilibrium analysis of purified GABP␣, GABP␤, and several deletion mutants in various combinations (27,32,33). Our observation that GABP␤ exists in a "monomer to dimer" equilibrium (K d ϭ 1.8 Ϯ 0.4 M), is incompatible with the "stable heterotetramer" model previously suggested to be required for heterotetramerization (1)(2)(3). Concentration distributions of an equimolar mixture of GABP␣ and GABP␤, however, best fit to a two-species model consisting of the GABP␣␤ heterodimer and a 619,573 Da species, which we attribute to the aforementioned nonspecific aggregation of GABP␣. We have been unable to satisfactorily model the GABP␣␤ concentration distribution with an ␣ ϩ ␤ 7 ␣␤ 7 (␣␤) 2 equilibrium model. These results suggest (i) that GABP ␣ and ␤ subunits form a stable heterodimer in solution with an estimated K d less than 10 Ϫ8 M, and (ii) that, under our experimental conditions, no GABP␣ 2 ␤ 2 heterotetramer is observed in solution. Our estimation for a GABP␣␤ dissociation constant is consistent with the apparent K d determined by Suzuki et al. (30) using surface plasmon resonance (K d ϭ 7.8 Ϯ 0.63 ϫ 10 Ϫ10 M). Our results are also consistent with EMSA experiments suggesting that efficient heterotetramer formation in the absence of any exogenous regulatory influences occurs only on DNA with two or more PEA3/EBSs (Fig. 1).
GABP␣␤ heterodimers assemble into the heterotetramer complex on DNA containing two PEA3/EBSs separated by as many as 4.0 helical turns (Fig. 2). 2 The stability of the heterotetramer complex is greatly affected by the spacing between PEA3/EBSs such that, stability is enhanced on DNAs containing two PEA3/EBSs positioned on the same face of the DNA helix. These results are consistent with our previous data showing that GABP-dependent initiator activity is enhanced when two PEA3/EBSs are positioned on the same face of the DNA helix (23). Similarly, GABP␤ isoforms or mutants lacking the COOH-terminal homodimerization domain are reportedly unable to promote heterotetramer complex formation and fail to activate transcription of linked promoters (20 -22). The demonstration that heterotetramer complex stability correlates with GABP-dependent initiator activity further advances the notion that the heterotetramer complex is the functionally active form of GABP.
The NH 2 -terminal deletion mutant of GABP␣ (GABP␣ c Q) forms significant amounts of heterotetramer complex even on DNA with a single PEA3/EBS. This finding suggests that the NH 2 terminus of GABP␣ may interfere with heterotetramer complex formation. In agreement with this observation, the concentration distribution of a GABP␣ c Q/GABP␤ equimolar mixture best fits a two-species associative model (K d ϭ 0.17 Ϯ 0.045 M) with the molecular weight of the monomeric species equal to that of the His 6 -tagged GABP␣ c Q/GABP␤ dimer. Therefore, we have demonstrated that GABP exists as a stable heterodimer, and that multiple PEA3/EBSs are required to promote formation of the heterotetramer complex. Our observation that the NH 2 -terminal portion of GABP␣ subunit exerts an inhibitory effect on GABP␣ 2 ␤ 2 heterotetramer formation, and the observations of Shiijo et al. (31) demonstrating the bcl3 enhances GABP␣ 2 ␤ 2 heterotetramer DNA binding, suggest that heterotetramer assembly is likely a regulated process that depends on specific promoter organization and/or specific regulatory proteins. , "Experimental Procedures") with a molecular weight equal to the sum of the molecular weights of the His 6 -tagged GABP␣ c Q and GABP␤ 334 proteins. B, concentration distributions of a mixture of the GABP␣ and GABP␤ 334 proteins containing slight excess of the GABP␤ 334 protein, were fitted to an ideal two-species model (Equation 2, "Experimental Procedures"), with the molecular weight of one species equal to the sum of the molecular weights of the His 6 -tagged GABP␣ and GABP␤ 334 proteins, and the molecular weight of the other species equal to that of the His 6 -tagged GABP␤ 334 protein.
Homo-and heterodimerization is often used by DNA-binding proteins in order to increase specificity and affinity for a cognate binding site. Although preassembled multiprotein complexes may possess higher overall affinity, they also will have higher nonspecific affinity for DNA, and at limiting concentrations, a protein complex could become kinetically trapped at random positions in the genome (34,35). Therefore, a high "off-rate" of a monomeric protein, compared with the fully assembled complex, will ensure, that only high affinity sites will be occupied for a sufficient amount of time to enucleate the assembly of a higher order protein-DNA complex. This mechanism, designated as the "monomeric pathway," has been proposed for the basic helix-loop-helix protein max, the bZIP transcription factor ATF-2, and the Arc repressor (34,36). Although the GABP␣ subunit is capable of DNA binding, in the presence of sufficient GABP␤, a highly stable heterodimer complex is formed. Thus, the GABP␣␤ heterodimer serves as a functional "monomeric" DNA binding entity. We have demonstrated that the presence of two PEA3/EBSs promotes further high affinity multimerization, suggesting that GABP assembly may proceed along the monomeric pathway.