Transcriptional activation domain of the herpesvirus protein VP16 becomes conformationally constrained upon interaction with basal transcription factors.

The transcriptional activation domain of the herpesvirus protein VP16 resides in the carboxyl-terminal 78 amino acids (residues 413-490). Fluorescence analyses of this domain indicated that critical amino acids are solvent-exposed in highly mobile segments. To examine interactions between VP16 and components of the basal transcriptional machinery, we incorporated (at position 442 or 473 of VP16) tryptophan analogs that can be selectively excited in complexes with other Trp-containing proteins. TATA-box binding protein (TBP) (but not transcription factor B (TFIIB)) caused concentration-dependent changes in the steady-state anisotropy of VP16, from which equilibrium binding constants were calculated. Quenching of the fluorescence from either position (442 or 473) was significantly affected by TBP, whereas TFIIB affected quenching only at position 473. 7-aza-Trp residues at either position showed a emission spectral shift in the presence of TBP (but not TFIIB), indicating a change to a more hydrophobic environment. In anisotropy decay experiments, TBP reduced the segmental motion at either position; in contrast, TFIIB induced a slight change only at position 473. Our results support models of TBP as a target protein for transcriptional activators and suggest that ordered structure in the VP16 activation domain is induced upon interaction with target proteins.

The herpes simplex virus virion protein VP16 is a potent transcriptional activator that specifically activates viral immediate early gene expression (1,2). As a transcriptional regulatory protein, it contains two functional domains. The aminoterminal portion of the protein, in association with host cellular proteins, binds to specific sequences upstream of the immediate early gene core promoters (3,4). The transcriptional enhancement activity resides in the carboxyl-terminal 78 amino acids (5,6). This domain can strongly activate transcription in various systems when attached to the DNA-binding domain of a heterologous protein (7). The VP16 activation domain is rich in acidic residues and has been regarded as a prototype acidic activation domain (AAD) 1 (8). Extensive mutational studies of this domain have identified aromatic and hydrophobic amino acids critical for its activity, for example the Phe at position 442 (9,10). These studies have further suggested that the VP16 AAD contains two independent subdomains: the N-subdomain (residues 413-456) and the C-subdomain (residues 453-490) (10 -12).
The activation mechanisms of eukaryotic transcriptional activators have been the focus of many studies (13,14). In addition to alleviating chromatin-mediated inhibition (15), activators have been proposed to interact with components of the basal transcription apparatus to stimulate or stabilize the formation of the transcription initiation complex at the promoter. Biochemical approaches have identified several potential targets of activation domains, particularly for the AAD of VP16. TATA-box binding protein (TBP) was the first basal factor shown to directly bind to the VP16 AAD (16). The specificity of this interaction was demonstrated by a correlation between binding of VP16 mutants to TBP and the transcription activities of these mutants (17). Later, VP16 was shown to directly interact with another basal transcription factor, TFIIB (18), although there is some discrepancy about the specificity of this interaction (11,13,19). Recently, a specific interaction between VP16 and a subunit of transcription factor H has been reported (20), as have interactions between VP16 and putative co-activator or adaptor proteins (11,21). Direct interactions between several of these target proteins and many other activation domains have also been shown (14). Although the physical interactions have been demonstrated, their relevance and role in transcriptional activation are still largely unknown.
Despite abundant functional studies of activation domains, little is known of their structures. No activation domain structure has yet been solved by x-ray crystallographic analyses or NMR. The limited biophysical studies of several AADs suggest that isolated AADs are unstructured (22)(23)(24)(25)(26). We recently performed fluorescence analyses employing chimeric proteins comprising the GAL4 DNA-binding domain (residues 1-147) fused to the VP16 activation domain (64). Trp residues were substituted for Phe at either position 442 or 473 of VP16, thus providing unique fluorescence probes at two positions. Dynamic quenching, time-resolved fluorescence decay, and timedependent anisotropy decay studies showed that the Trp residues at either position are solvent-exposed and highly mobile. Our results, in agreement with CD and NMR studies of this domain, reveal that the isolated VP16 AAD is poorly structured. Noteworthy is that many of the biophysical studies suggest that under certain conditions (low pH, hydrophobic solvent) these AADs can acquire specific conformations such as helix and ␤-sheet. These conditions might mimic the in vivo conditions under which the AAD interacts with its target proteins. The AADs, therefore, have been hypothesized to adopt a specific conformation in the presence of their target proteins. However, no structural characterizations of these AADs have yet been carried out in the presence of their target proteins.
One major difficulty in studying protein-protein interactions by various biophysical means is that the signals from different proteins overlap and make the interpretation ambiguous. Recently, several groups reported that Trp analogs (5-hydroxytryptophan or 7-azatryptophan) can be successfully incorporated into proteins by using Trp auxotrophic Escherichia coli strains and supplementing the growth media with the relevant Trp analog (27)(28)(29). The excitation spectra of these Trp analogs are shifted to longer wavelengths compared with Trp itself. Hence, fluorescence of proteins containing these Trp analogs can be selectively excited in the presence of other proteins containing natural Trp. Here we used this strategy to study the structural features of the VP16 AAD in the presence of two basal transcription factors, TBP and TFIIB. Our results indicate that the structure of the VP16 AAD becomes considerably constrained upon its interaction with these basal factors, particularly with TBP.

EXPERIMENTAL PROCEDURES
Chemicals and Reagents-L-tryptophan, L-5-hydroxytryptophan, and D,L-7-azatryptophan were purchased from Sigma. The E. coli tryptophan auxotrophic strain CY15077 (W3110⌬trpEA2) and the lacI q -bearing plasmid pMS421 were kindly provided by Dr. Charles Yanofsky. Plasmid pKA9 carrying the Saccharomyces cerevisiae SPT15 gene in a pET expression vector was a gift of Dr. Fred Winston. Plasmid phIIB expressing the human TFIIB in a pET vector was a gift of Dr. Danny Reinberg. Yeast nuclear extract and plasmid pCZ3GAL were gifts of Dr. Shelley Berger. HeLa cell nuclear extracts and plasmid pML were kindly provided by Chun-hsiang Chang and Dr. Zachary Burton.
Purification of 5-OH-Trp-or 7-aza-Trp-incorporated GAL4-VP16 -Expression plasmids for GAL4-VP16 fusion proteins with unique Trp codons in the VP16 activation domain have been described (64). E. coli strain CY15077 was transformed with pMS421 and with an expression plasmid for one of the various GAL4-VP16 fusion proteins. Cell growth and Trp analog incorporation procedures were followed as described (27) with some modifications. The cells were maintained under ampicillin (100 g/ml) and streptomycin (20 g/ml) selection. An overnight culture was diluted 1:100 into M9 medium supplemented with 0.1 mM CaCl 2 , 1 mM MgSO 4 , 0.5% glucose, 0.1% thiamine, 1% casamino acids, and 0.25 mM L-tryptophan. The culture was grown at 37°C to an A 550 of 0.6. The cells were then collected by centrifugation and resuspended in the original volume of M9 medium, except that 0.25 mM L-5-OH-Trp or 0.5 mM D,L-7-aza-Trp was added in place of L-Trp. After the culture was grown for an additional 20 min, expression of GAL4-VP16 fusion proteins was induced by the addition of isopropyl-1-thio-␤-D-galactopyranoside to a final concentration of 1 mM. The cells were harvested after 2 h of growth at 37°C, and the analog-labeled GAL4-VP16 proteins were purified using procedures described elsewhere (64).
Purification of Recombinant TBP-TBP was purified using a procedure from Brenowitz et al. (30) with minor modifications. E. coli BL21 (DE3) cells carrying the plasmids pLysS and pKA9 were grown at 37°C in LB medium containing 30 g/ml chloramphenicol and 25 g/ml ampicillin. TBP expression was induced by the addition of 1 mM isopropyl-1-thio-␤-D-galactopyranoside when cell density reached an A 600 of 0.4. Cells were shifted to 30°C to grow for an additional 2 h. Cells were harvested by centrifugation and resuspended in a buffer comprising 20 mM HEPES, pH 7.9, 1 mM EDTA, 1 mM EGTA, 10 mM 2-mercaptoethanol, 2 g/ml pepstatin, 2 g/ml leupeptin, 1 mM benzamidine, and 0.8 mM phenylmethylsulfonyl fluoride, in a volume of 20 ml/liter of culture. After lysis by sonication, the cell debris was removed by centrifugation. Protamine sulfate was added to the supernatant to 0.3 mg/ml. The precipitate was removed by centrifugation, and the supernatant was dialyzed against buffer D (20 mM HEPES, pH 7.9, 20% glycerol, 1 mM EDTA) plus 100 mM KCl. The crude protein fraction was loaded onto a Pharmacia Q-Sepharose Fast Flow column, from which TBP eluted mainly in the flow-through. This Q-Sepharose Fast Flow chromatography step was repeated, and the flow-through fraction was loaded onto a Pharmacia S-Sepharose Fast Flow column. The column was washed with buffer D plus 100 mM KCl and then eluted with a linear gradient from 100 mM KCl to 460 mM KCl in buffer D. Fractions containing TBP (at 95% purity or greater as analyzed by SDS-polyacrylamide gel electrophoresis) were stored at Ϫ70°C.
Purification of Recombinant TFIIB-TFIIB was purified using modified published procedures (31). E. coli BL21 cells containing phIIB were grown in LB media containing 100 g/ml ampicillin at 37°C. TFIIB expression was induced by the addition of 0.4 mM isopropyl-1thio-␤-D-galactopyranoside when an A 600 of 0.6 was reached. Cells were harvested after 2 h of additional growth. The cell pellets were resuspended (50 ml/liter of culture) in a buffer comprising 20 mM HEPES, pH 7.9, 25 mM EDTA, 10 mM 2-mercaptoethanol, 0.2 mM phenylmethylsulfonyl fluoride, 0.2 mM benzamidine, 2 g/ml pepstatin, and 100 mM KCl. The cells were broken by sonication, and the lysate was cleared by centrifugation. Polyethyleneimine (pH 7.9) was added dropwise to the supernatant to 0.1%, and the precipitate was removed by centrifugation. Ammonium sulfate was added to the supernatant to 0.258 g/ml. The precipitated proteins were resuspended in the resuspension buffer and dialyzed against buffer B (20 mM HEPES, pH 7.9, 10% glycerol, 0.2 mM EDTA, 0.2 mM EGTA, 0.2 mM phenylmethylsulfonyl fluoride, 2 mM dithiothreitol) plus 100 mM KCl. The crude protein fraction was loaded onto a Whatman P-11 column. The column was washed with buffer B plus 100 mM KCl and then washed with buffer B plus 300 mM KCl. The column was eluted with a linear gradient of 300 -800 mM KCl in buffer B. Fractions containing TFIIB of highest purity (eluting between 620 and 670 mM KCl) were pooled and dialyzed against buffer B plus 100 mM KCl and loaded onto a pre-equilibrated DE-52 column. The flowthrough contained TFIIB at greater than 95% homogeneity. The protein was stored at Ϫ70°C.
GAL4-VP16 Activity Assay-Activities of the various GAL4-VP16 fusion proteins were tested in in vitro transcription reactions using yeast nuclear extracts as described (32).
Recombinant TBP and TFIIB Activity Assay-In vitro transcription assays using HeLa nuclear extracts were performed as described (33). The template plasmid pML containing the adenovirus major late promoter was linearized with SmaI. The activity of purified recombinant TBP was tested using HeLa extracts preincubated at 47°C for 15 min to inactivate endogenous TBP (34). To test the activity of recombinant TFIIB, the HeLa nuclear extract was depleted of endogenous TFIIB as follows: 0.12 ml of agarose-conjugated antibodies directed against TFIIB (Santa Cruz Biotechnology) was equilibrated with a buffer comprising 20 mM HEPES, pH 7.9, 20% glycerol, 1 mM EDTA, 0.2 mM EGTA, 0.5 mM dithiothreitol, 0.2 mM phenylmethylsulfonyl fluoride, and 500 mM KCl. 5 M NaCl and 2% Triton X-100 was added to 120 l of HeLa nuclear extract to bring the final NaCl concentration to 500 mM and the final Triton X-100 concentration to 0.02%. This extract was incubated with the equilibrated anti-hTFIIB agarose bead at room temperature for 40 min and at 4°C for an additional 1.5 h. The agarose beads were centrifuged at 2500 rpm for 5 min, and the supernatant was used as the TFIIB-depleted extract.
The steady-state fluorescence spectra were obtained with an SLM 8000 spectrofluorometer. The excitation wavelength was 309 nm. The emission spectra titration experiments of 7-aza-Trp-incorporated GAL4-VP16 were performed by recording the initial emission spectrum of the 4 M 7-aza-Trp-incorporated GAL4-VP16 and then adding small aliquots of concentrated TBP or TFIIB solution and recording emission spectra until no further change could be detected. The same amounts of TBP or TFIIB were added to the buffer control, and these blank emission spectra were also recorded. Final emission spectra were corrected for blank control and for dilution.
Steady-state fluorescence anisotropy was measured using an L-format detection configuration. The excitation bandpass was 4 nm, and the emission bandpass was 8 nm. Excitation was at 309 nm, and emission was at 360 nm. Every data point was measured at least eight times. Data were fit to the equations describing formation of the 1:1 binary complex between GAL4-VP16 and TBP (36), where r is the measured anisotropy when the fluorophore is present in both the free form (5-OH-Trp incorporated GAL4-VP16) and the bound form (complex with TBP), r B and r F are the anisotropy of the free and bound fluorophores, f B and f F refer to the fraction of the total fluorophore that is present in the bound and free forms, and I B and I F are the fluorescence intensities of the fluorophore in bound or free forms, and, . The values of fluorescence emission intensity at 338 nm (for 5-OH-Trp-incorporated proteins) or at 396 nm (for 7-aza-Trp-incorporated proteins) were corrected for dilution and for blank. Quenching data were analyzed by the classic Stern-Volmer equation for dynamic quenching, or by the single species dynamic-static quenching equation, or by the two-species quenching equation (a rearrangement of Equation 11 in Ref. 37), where F 0 and F are the fluorescence intensity in the absence and presence of quencher, respectively, [Q] is the quencher concentration, and K sv is the Stern-Volmer dynamic quenching constant. V is the static quenching constant. f a is the fractional contribution of the fluorophores that are accessible to the quencher, and K a is the Stern-Volmer constant for the accessible fraction. K sv , V, f a , and K a values were determined using least-squares regression (IGOR, Wavemetrix). The bimolecular collisional quenching constant k q is defined as follows, where ϽϾ is the mean lifetime at zero quencher. Time-resolved fluorescence was measured on a single photon count-ing fluorometer (38). Anisotropy decay curves were obtained by alternatively recording the emission oriented parallel and perpendicular to the plane of excitation. When the anisotropy decay curves for the mixture of 2 M 5-OH-Trp-incorporated GAL4-VP16 with 4 M TBP or TFIIB were recorded, those of the appropriate blank (4 M TBP or TFIIB) were also recorded at the same time. Excitation wavelength was at 309 nm, and emission wavelength was at 360 nm. Time per channel was 90 ps, and 512 channels were recorded. Data were corrected for blank control and analyzed by the global analysis (39) or the "sum and difference" method (40). Rigorous confidence limits were not measured for all parameters. Instead, several individual parameters were explored by fixing their values to bracket those recovered. Typical 1 errors are approximately 10%, and typical ␤ 2 errors are approximately 5%. As discussed in Ref. 39, 2 values that are much larger than 6 become indistinguishable without further averaging.

Incorporation of 5-OH-Trp and 7-aza-Trp into GAL4W36V-VP16
Proteins-By using a Trp auxotrophic E. coli strain, 5-OH-Trp or 7-aza-Trp were biologically incorporated into various GAL4-VP16 proteins (Fig. 1). These proteins were purified to more than 95% homogeneity and were functionally active when tested by in vitro transcription assays (data not shown). Therefore, the structural features revealed by these proteins should reflect those of the wild-type VP16 AAD. Fig. 2A shows the peak normalized absorbance spectra of GAL4-VP16 proteins with Trp or its analogs incorporated at position 442 of the VP16 AAD. The absorbance spectrum of the protein containing 5-OH-Trp demonstrated a characteristic shoulder between 290 and 320 nm, while that of the protein containing 7-aza-Trp showed extended low energy absorbance. The fluorescence excitation spectra of the same GAL4-VP16 fusion proteins ( Fig.  2B) demonstrate that the Trp analogs at position 442 can be selectively excited at 310 nm. Fig. 2C shows the normalized emission spectra of these fusion proteins. GAL4-VP16 containing 5-OH-Trp had an emission maximum at 340 nm, the same maximum observed for the free amino acid analog. GAL4-VP16 containing 7-aza-Trp showed an emission maximum centered at 396 nm, close to that of 7-aza-Trp in aqueous solution (398 nm). Absorbance spectra, excitation spectra, and emission spectra of GAL4-VP16 proteins with Trp analogs incorporated at position 473 or at position 442 in a truncated activation domain all showed similar properties, indicating that both Trp analogs were successfully incorporated into all the proteins. Moreover, the concordance of the spectra of the labeled proteins with the spectra of free amino acid analogs supports our observation that these residues of the VP16 activation domain are largely solvent-exposed (64).
The presence of 5-OH-Trp or 7-aza-Trp in the GAL4-VP16 proteins enables the fluorescence of the fusion proteins to be selectively excited at 310 nm in the presence of other Trpcontaining proteins. Recombinant basal transcription factors TBP and TFIIB were purified from E. coli, and their transcriptional activities were confirmed using specifically depleted nuclear extracts (data not shown). As expected, these proteins were not efficiently excited using 310-nm light; the fluorescence observed for a 2-fold molar excess of TBP or TFIIB when excited at 310 nm amounted to less than 10% of the signal observed for GAL4-VP16 proteins bearing Trp analogs in the presence of a 2-fold molar excess of TBP or TFIIB.
Interaction between TBP and VP16 AAD Changes the Polarity of the Environments Surrounding 7AW-442 and 7AW-473-The emission spectrum of 7-aza-Trp is very sensitive to the polarity of the environment (29). In aqueous solution, the emission maximum is near 400 nm, but in hydrophobic environments a maximum at 370 nm is observed. To test whether TBP or TFIIB could change the polarity around residues 442 and 473 in the VP16 AAD, increasing amounts of TBP or TFIIB were added to GV-7AW442 or GV-7AW473, and emission spec-tra were recorded. In the absence of either basal transcription factor, both GV-7AW442 and GV-7AW473 showed the characteristic 396-nm emission maximum of exposed 7-aza-Trp. With the addition of increasing amounts of TBP to either labeled protein, the relative intensity around the 370-nm region increased gradually and eventually reached saturation (Fig. 3, A  and B). To quantitate the spectra shift, the ratio of the emission intensities at the two wavelengths (F 376 /F 396 ) was calculated at each concentration of TBP or TFIIB (Fig. 3, C and D). These ratios increased from 0.81 to 0.96 with the addition of TBP. Thus, residues at both positions are found in more hydrophobic environments in the presence of TBP. In contrast, the addition of TFIIB did not increase the relative intensity around the 370-nm region of these proteins. The F 376 /F 396 ratio was unchanged by the addition of increasing amounts of TFIIB (Fig. 3, C and D), suggesting that even if TFIIB interacts with the activator, the polarity of the environments surrounding both 442 and 473 remains the same.
It should also be noted that the quantum efficiency of GV-7AW442 and GV-7AW473 increased modestly in the presence of TBP. The relative intensity around 370 nm region increased; however, the emission maximum is still at 396 nm. These fluorescence properties most closely match those of the model compound 7-azaindole in alcohol rather than those of 7-azaindole in aprotic solvents such as acetonitrile (41). Hydroxyl groups in alcohols induce tautomerization of 7-azaindole, resulting in two populations of fluorescing molecules (29,41). The results of this study imply that the surroundings of both 7AW-442 and 7AW-473 become more hydrophobic; however, either the solvent is not totally excluded from these residues or there are nearby polar residues that hydrogen bond to the 7AW to Interaction between Basal Factors and VP16 AAD Alters the Solvent Accessibility of Residues at Amino Acid Positions 442 and 473-Acrylamide quenching experiments were performed to test whether the presence of TBP or TFIIB affected the solvent accessibility of the fluorophores at residues 442 and 473. Activator proteins labeled with 5-OH-Trp were mixed with saturating amounts of TBP or an equivalent amount of TFIIB in the presence of increasing concentrations of acrylamide. The Stern-Volmer plots of these quenching experiments are shown in Fig. 4, and the best fit parameters are summarized in Table I. In the absence of basal transcription factors, the Stern-Volmer plots of both GV-5HW442 (Fig. 4A) and GV-5HW473 (Fig. 4B) showed upward curvature. These data were best fit to a model invoking both dynamic and static quenching, yielding Stern-Volmer constants (K sv ) of 6.  (64), these results suggest that residues 442 and 473 are solvent-exposed.
In the presence of saturating amounts of TBP, the response of both GV-5HW442 and GV-5HW473 to acrylamide changed significantly. The Stern-Volmer plots of both proteins were linear, and the data were best fit to the purely dynamic quenching model, with a K sv of 8.2 M Ϫ1 and 6.3 M Ϫ1 , respectively. In this case, the dynamic quenching rate constant was 2.6 M Ϫ1 ns Ϫ1 for GV-5HW442 and 2.2 M Ϫ1 ns Ϫ1 for GV-5HW473. In the presence of a similar amount of TFIIB, the Stern-Volmer plots of both proteins showed upward curvature, as observed for those of the labeled fusion proteins alone. The analysis gave a K sv of 5.7 M Ϫ1 and 3.9 M Ϫ1 and a static quenching constant V of 1.0 M Ϫ1 and 1.6 M Ϫ1 , respectively. The dynamic quenching rate constant was 2.2 M Ϫ1 ns Ϫ1 for GV-5HW442 and 1.5 M Ϫ1 ns Ϫ1 for GV-5HW473. Although both TBP and TFIIB altered the quenching effect of acrylamide for both probes, the nature of the effect is very different in the two cases. The presence of TBP eliminated the static quenching process, whereas it did not change the dynamic quenching process significantly. In contrast, in the presence of TFIIB both static and dynamic quenching remain, albeit somewhat altered. Moreover, the addition of TBP to the truncated activator N-5HW442 altered the quenching rate by acrylamide as it did to the full-length AAD, whereas the addition of TFIIB had no effect on accessibility (data not shown). Thus, if there is any interaction between TFIIB and the activators, the effect of that interaction on the structure of the VP16 AAD is apparently different from that seen with TBP.

Target-induced Structure in VP16 Activation Domain
The activator proteins containing 7-aza-Trp were also tested in acrylamide quenching assays (Fig. 5, A and B). Stern-Volmer plots of acrylamide quenching of GV-7AW442 and GV-7AW473 were linear, yielding a K sv of 2.3 M Ϫ1 and 3.3 M Ϫ1 , respectively (Table II). Acrylamide was a less efficient quencher for 7-aza-Trp than for Trp or 5-OH-Trp. The presence of TBP altered the solvent accessibility of both residue 442 and 473. The downward curves of the Stern-Volmer plots were best fit to a twospecies model, with approximately 40% of the probe molecules being inaccessible to acrylamide (assumed K sv of 0) and an accessible fraction of approximately 60% having a K a of 4.0 M Ϫ1 or 7.0 M Ϫ1 for GV-7AW442 or GV-7AW473, respectively. In contrast, the presence of TFIIB with GV-7AW442 did not change the quenching mechanism, nor did it change significantly K sv . However, TFIIB did alter the solvent accessibility of GV-7AW473. Its Stern-Volmer plot was downward curved; in a two-species model, approximately 30% of the probe was inaccessible, and the accessible fraction had a K a of 4.9 M Ϫ1 . In this case, TFIIB caused a change similar to that seen with TBP.
Steady-state Anisotropy Analysis and Dissociation Constants for Binding of TBP and VP16 AAD-The steady-state fluorescence anisotropy of the 5-OH-Trp-labeled GAL4-VP16 proteins was measured at 360 nm in the presence of TBP or TFIIB (Fig. 6). The addition of TBP to GV-5HW442 resulted in a large saturable increment in its anisotropy (Fig. 6A), indicating a large decrease in the mobility of the 5-OH-Trp fluorophore. Assuming a 1:1 stoichiometry for the GAL4-VP16⅐TBP complex, the dissociation constant for the interaction was calculated to be 3.3 (Ϯ 1.7) ϫ 10 Ϫ7 M. The magnitude of the anisotropy change cannot be attributed solely to the increase in mass of the complex. Therefore the large (70%) increase in anisotropy in the presence of TBP suggests that the local motion of the fluorophore is significantly reduced upon interaction with TBP.
In contrast, the addition of TFIIB to GV-5HW442 had little or no effect on anisotropy (Fig. 6A). We infer either that GV-5HW442 did not form a complex with TFIIB, or that even if a complex formed, TFIIB did not change the segmental motion of the fluorophore in the GV-5HW442. Similar effects of TBP and TFIIB were observed for the truncated fusion protein bearing 5-OH-Trp at position 442 of N-5HW442 (Fig. 6B). The calculated dissociation constant for the interaction between TBP and N-5HW442 is 3.8 (Ϯ 1.4) ϫ 10 Ϫ7 M. When the analog was incorporated at position 473 of the full-length activation domain (GV-5HW473), the anisotropy also increased rapidly as TBP was added to the system, and the anisotropy reached a limiting value (Fig. 6C). The calculated dissociation constant for this interaction is 2.6 (Ϯ 0.6) ϫ 10 Ϫ8 M. Again, the addition of TFIIB to the same protein caused no significant anisotropy change.
Interaction between TBP and VP16 Restricts the Segmental Motion in the AAD-The findings from steady-state anisotropy analysis were further confirmed by time-resolved anisotropy decay measurements. The anisotropy decay curves are shown in Fig. 7, and the fitted parameters are summarized in Table III. The anisotropy decays of the activator proteins (tested in the absence of target proteins) were best fit to two components: a subnanosecond fast decay component representing segmental motion around the 5-OH-Trp fluorophore and the slower decay component in the range of 2-6 ns. In all of these proteins, the segmental motion contributed at least 60% of the anisotropy decay. The extent of these segmental motions were comparable with those of the known most flexible proteins (42). These results are consistent with our previous results using GAL4-VP16 bearing natural Trp and taken together suggest that both VP16 subdomains are very mobile (64). The rotational correlation times ( 2 ) for the slower decay components reflect a molecular size smaller than that predicted for a globular GAL4-VP16 protein and presumably represent the size of the activation domain itself tethered to the GAL4 DNA binding

Target-induced Structure in VP16 Activation Domain
domain by a highly flexible linker. The anisotropy decay of these proteins was then measured in the presence of a 2-fold molar excess of TBP or TFIIB. Under these conditions, binding between GAL4-VP16 and TBP reached saturation as indicated by steady state anisotropy titration experiments. In the presence of TBP, the anisotropy decay of GV-5HW442 is greatly slowed (Fig. 7A). As for the activator fusion protein alone, data were also best fitted to two decay components. However, the contribution of the segmental motion (␤ 1 ) was dramatically reduced from 75 to 40% of the total, and the slow component dominated this decay process (Table III). If one assumes segmental motion can be reconciled with the "wobble in cone" model (i.e. the localized motion of Trp is a wobbling of its transition moment within a cone), the extent of this motion can be described by the cone semiangle magnitude (43,44). The cone semiangle for the GV-5HW442 alone (52°) is larger than or comparable with those of many known flexible polypeptides such as apocytochrome C with a reported semiangle of 47° (42). However, the presence of TBP reduced the calculated cone semiangle to 35°. This result indicates that the segmental motion around residue 442 in the VP16 AAD is restricted in the presence of TBP. Moreover, the rotational correlation time of the slow component ( 2 ) increased, reflecting the change in molecular mass from the activation domain alone to a complex containing GAL4-VP16 and TBP together. In contrast, TFIIB did not cause any change in the anisotropy decay of GV-5HW442 (Fig. 7A). The lack of any effect on 2 suggests that TFIIB did not interact with the VP16 AAD or that if any interaction does exist the mode of association has no effect on the overall rotation of the AAD.
The anisotropy decay curves of the chimeric protein containing the VP16 AAD N-subdomain in the presence of TBP and TFIIB are shown in Fig. 7B. TBP formed a complex with N-5HW442, evidenced by the increase in the rotational correlation time ( 2 ) of the slow decay component (Table III). TBP also reduced the amplitude (␤ 1 ) of the segmental motion around residue 442, although the magnitude of the effect is less than that seen for the full-length protein (GV-5HW442). In contrast, TFIIB had much less of an effect on the anisotropy decay of N-5HW442, consistent with the results seen for TFIIB with the full-length protein labeled at position 442.
The anisotropy decay curves of GAL4-VP16 labeled at position 473 of the full-length activation domain (GV-5HW473) in the presence of TBP or TFIIB are shown in Fig. 7C. As observed with the probe at amino acid 442, TBP greatly restricted the segmental motion around residue 473 (␤ 1 , Table III). In addition, the apparent size of the segment associated with this fluorophore at position 473 became larger, as shown by the increment of the rotational correlation time of the fast decay component ( 1 ). The rotational correlation time of the slow component ( 2 ) also increased, albeit not to the extent seen with the probe at position 442. In this case, the slow component may represent the "freezing" of a subdomain surrounding position 473 rather than the size of the entire GAL4-VP16⅐TBP complex. This result suggests that binding of TBP may have different effects on the flexibility of the two subdomains of the VP16 AAD.
In this experiment, the presence of TFIIB also affected the anisotropy decay, reflected in an increase in the rotational correlation time of the slow component ( 2 ). The magnitude of this parameter was smaller than that expected for the TFIIB⅐GV-5HW473 complex, and thus probably represents only a subdomain of that complex. TFIIB also moderately restricted the motion surrounding residue 473 (␤ 1 ). The extent of restriction was much smaller than that caused by TBP; TBP reduced the cone semiangle of local motion from 47°to 31°, whereas TFIIB only reduced it to 43°. DISCUSSION Previous structural characterization of the AADs of VP16, GAL4, GCN4, NF-B p65, and glucocorticoid receptor by CD and NMR studies revealed that these domains were unstructured in aqueous solution under neutral pH (22)(23)(24)(25)(26). However, the AADs of VP16, NF-B p65, and glucocorticoid receptor all form an ␣-helix conformation in less polar solvent (22,23,25,26). The AADs of GAL4 and GCN4 form ␤-sheet in lower pH solution or in a hydrophobic solvent (24). Authors of these reports all speculate that in the process of transcriptional activation, the AADs adopt higher-order structure upon contacting their target molecules by an "induced fit" mechanism. The present report provides biophysical evidence to support that speculation, in that the local structure surrounding key residues of the VP16 AAD was significantly constrained upon interaction with TBP, and to a less extent, with TFIIB. The induced conformations in transcription factors have been previously shown only in the DNA binding basic region of leucine zipper proteins and the arginine-rich RNA binding domain of human immunodeficiency virus Tat protein (45)(46)(47). The finding of the induced ordered structure in the VP16 activation domain will likely lead to a more refined analysis of the specific secondary and tertiary structures induced by its target proteins.
In this study, Trp analogs with unique fluorescent properties were incorporated at key positions of the VP16 transcriptional activation domain. These spectrally enhanced proteins were used to study the interactions between this activation domain and the basal transcription factors TBP and TFIIB. In the absence of these factors, studies of the VP16 AAD containing 5-OH-Trp or 7-aza-Trp at positions 442 or 473 showed that both residues are solvent-exposed and are associated with highly mobile protein segments, consistent with our previous fluorescence analyses of the VP16 AAD containing natural Trp at these positions. The presence of TBP induced a significant change in the VP16 AAD, with a more ordered or constrained structure becoming apparent using fluorescent probes at either position. In contrast, effects of TFIIB interaction were observed only for probes at position 473 of the VP16 AAD, and those effects were weaker than those induced by TBP. Probes placed at positions 442 and 473 showed similar changes in the presence of basal transcription factor TBP. Probes at position 442 either in the full-length AAD or in the truncated subdomain also showed similar changes upon interaction with TBP.
Interaction of the VP16 AAD with TBP-In the presence of basal transcription factor TBP, 7-aza-Trp residues at either position 442 or 473 of the VP16 AAD showed an increased fluorescence emission at 376 nm, indicating that these residues were present in a more hydrophobic environment than in unbound VP16. Acrylamide quenching results for VP16 AADs labeled with either 5-OH-Trp or 7-aza-Trp are consistent with the spectral shift results. In the presence of TBP, the mechanism of the quenching process seems to be qualitatively different, suggesting that the microenvironments of residues at positions 442 and 473 are distinct from those in the absence of TBP. For example, with 5-OH-Trp at either position in the VP16 AAD, the presence of TBP apparently eliminated the static quenching, presumably by blocking formation of the ground state complex between the quencher and the fluorophore. Moreover, qualitative changes in quenching mechanism were also seen using AADs labeled with 7-aza-Trp. In this case, the presence of TBP resulted in quenching curves that best fit a two-species model, with a sizable fraction of the probe being inaccessible to the quenching agent. Together, these results suggest that when the VP16 AAD interacts with TBP, both residues 442 and 473 become more shielded from solvent. Whether these residues became buried within a folded VP16 domain or became embedded as part of the binding interface with TBP is not revealed by these experiments. However, TBP does not completely block access to either residue 442 or 473. Further, the spectra of AADs containing 7-aza-Trp were not completely blue-shifted to the 370-nm region, nor was the quenching rate by acrylamide changed to that typical of com- factors TBP and TFIIB Anisotropy decay curves were recorded for GAL4-VP16 proteins containing 5HW either alone or in the presence of a 2-fold molar excess of TBP and TFIIB. Anisotropy decay data were analyzed by the "sum and difference" method (40). The rotational correlation times ( i ) and the corresponding preexponentials (␤ i ) were derived from the best fit. The apparent limiting time zero anisotropy (r 0 (app)) is defined as ⌺␤ j . This r 0 does not include ␤ j for motions faster than 300 ps. r 0 (app) was used to calculate the cone semiangles (⌰) (43,44). Target-induced Structure in VP16 Activation Domain pletely buried residues. The steady-state anisotropy of GV-5HW442, N-5HW442, and GV-5HW473 increased substantially in the presence of TBP. Dissociation constants were calculated from these analyses. Dissociation constants between TBP and GV-5HW442 or N-5HW442 were both in the range of 3 ϫ 10 Ϫ7 M, while that between TBP and GV-5HW473 was in the range of 3 ϫ 10 Ϫ8 M. The differences in these dissociation constants may correspond to differences in transcriptional activities as a result of the Phe 3 Trp mutations at positions 442 and 473. The substitution mutant F442W retains 70% activity as a full-length AAD, whereas the F473W mutation has a negligible effect on activity (10). An affinity capture method had previously yielded an apparent dissociation constant of 2 ϫ 10 Ϫ7 M between the VP16 AAD and 35 S-labeled yeast TBP (17). The 10-fold difference in the results may be due to inherent differences between spectroscopic and capture-type assays, or to differences in the fusion protein constructs used in these experiments.
Time-resolved anisotropy decay measurements demonstrate that the mobility of protein segments surrounding positions 442 and 473 is markedly reduced in the presence of TBP ( Fig.  7 and Table III). When the VP16 AAD was labeled with 5-OH-Trp at either position, the fraction of the anisotropy associated with fast decay (␤ 1 ) was reduced by roughly 50% by binding to TBP, while the fraction associated with slow decay (␤ 2 ) was increased. Assuming that segmental motion can be correlated with the fluorophore wobbling within a cone (43,44), the calculated cone semiangle (Q) is reduced from approximately 50°t o approximately 30°, representing a considerable constraint on the segmental motion. Moreover, the increase in the rotational correlation time for the slow decay component ( 2 ) in the presence of TBP indicates that this component is moving with a much greater mass. For the probe at position 442, this mass may approach that of the GAL4-VP16⅐TBP complex altogether, whereas for the probe at position 473 the increase is less dramatic and likely represents a somewhat smaller subdomain of the complex. The rotational correlation time for the fast decay component ( 1 ) also increased for the probe at position 473 (but not for the probe at position 442), which may indicate that the fast decay component results from a larger peptide segment surrounding 473 being induced by the binding of TBP. Curiously, a subtle difference can be observed when the probe at position 442 is examined in the full-length and truncated versions of the AAD. TBP apparently caused a greater restriction of the segmental motion in the full-length AAD than in the N-subdomain (compare calculated cone semiangles () for the two AADs in the absence and presence of TBP). Nonetheless, the rotational correlation times for the slow decay component ( 2 ) of the truncated AAD increased, suggesting binding between VP16 N-subdomain and TBP. Together, these results suggest that the N-subdomain (surrounding Phe 442 ) is the major targeting site of TBP, but the C-subdomain still has some impact on this TBP-activator interaction, either by providing a second, weaker binding site or by modulating the TBP/N-subdomain interaction.
Interaction of the VP16 AAD with TFIIB-The effects of a second basal transcription factor, TFIIB, on the fluorescence of the VP16 AAD were both qualitatively and quantitatively different from those induced by TBP. Most notably, the presence of TFIIB had little or no effect on the fluorescence of Trp analogs incorporated at position 442 in the N-subdomain. No change was seen in the emission spectrum of GV-7AW442 (Fig.  3), nor in the type of acrylamide quenching observed for GV-7AW442. TFIIB did not change the steady-state anisotropy of GV-5HW442 and N-5HW442 (Fig. 6), nor did it change any aspect of time-resolved anisotropy decay of GV-5HW442 and N-5HW442 (Fig. 7). Altogether, there is no evidence of any structural change in the N-subdomain caused by TFIIB.
In contrast to the lack of effect on the N-subdomain, TFIIB did induce some changes in the fluorescence of the VP16 AAD with probes in the C-subdomain (position 473). TFIIB altered the quenching of GV-7AW473 by acrylamide (Fig. 5B), such that the quenching curves are best fit to a two-species model similar to that proposed for the effect of TBP. However, no shift in the emission spectrum of GV-7AW473 was observed in the presence of TFIIB (Fig. 3). TFIIB also partially protected both GV-5HW442 and GV-5HW473 from acrylamide quenching (Fig. 4), although the quenching mechanisms apparently retain both static and dynamic components, in contrast to the effect of TBP. This protection was C-subdomain-dependent, as it disappeared for N-5HW442 (data not shown). The differences in acrylamide quenching results for AAD proteins bearing 5-OH-Trp and 7-aza-Trp may be due to intrinsic differences in the quenching characteristics of these analogs. In sum, these results suggest that TFIIB may only sterically alter the accessibility of the quenching reagent without net changes in the polarity of the environment around residue 473, and thus no change in the emission spectrum of GV-7AW473 is induced.
TFIIB caused a modest change in the anisotropy decay of GV-5HW473 ( Fig. 7C and Table III), although the effects were less striking than those seen for TBP, and no noticeable effect was observed on the anisotropy decay of GV-5HW442. In particular, the cone semiangle () reduction caused by TFIIB is much smaller than that caused by TBP. The magnitude of the effect on the rotational correlation times for both the fast and slow decay components was approximately half that observed with TBP, implying that the sizes of the domains responsible for these components were not dramatically altered. The lack of any significant change in steady-state anisotropy in the presence of TFIIB (Fig. 6C) might further suggest that the VP16-TFIIB interaction is weak.
Taken together, these results indicate that the interaction of the VP16 AAD with TBP is very different from its interaction with TFIIB. TBP altered the fluorescence of probes at both 442 and 473, whereas TFIIB affected only probes at 473. Moreover, the magnitude of the effects induced by TBP was also consistently greater than those induced by TFIIB. While these results do not rule out the ability of TFIIB to interact with the VP16 AAD entirely, it is striking that few if any effects are observed on the properties of amino acids at or near positions critical to the transcriptional function of the VP16 AAD.
Comparisons with Other Model Systems-The disordered structure of acidic activation domains of transcriptional activators and their structural transitions in the presence of target binding proteins have precedents in other biological systems. For example, the carboxyl-terminal regions of several isoforms of tubulin are very acidic, and most experimental results suggest that these regions are extended and unstructured. However, ␣-helical structure was observed in the presence of hydrophobic solvents or low pH, consistent with secondary structural predictions (48). No evidence yet indicates whether such a secondary structure is induced when these regions bind to other proteins.
A second analogy is with the binding of the basic pancreatic trypsin inhibitor to trypsin and to trypsinogen, with dissociation constants of 10 Ϫ13 and 10 Ϫ5 M, respectively (49). X-ray crystallographic analysis shows that trypsinogen in complex with basic pancreatic trypsin inhibitor complex acquires a trypsin-like conformation (i.e. with a rigidly structured binding domain). The reduced affinity of trypsinogen for basic pancreatic trypsin inhibitor is a consequence of the energy required to order the binding domain. Thermodynamic studies and struc-tural comparisons have demonstrated a large negative heat capacity change associated with local or more extensive folding when a protein binds its ligand (or another protein) (50). In such systems, the binding energy from the interaction creates part or all of the binding sites or even drives folding beyond the interface.
Binding of a target protein to a flexible segment such as the VP16 AAD requires the reduction of its conformational entropy at the expense of association energy. Therefore, this kind of interaction, in which the flexible segment must be stabilized before it can provide optimal noncovalent interaction, is weaker than interaction with a rigid, stereochemically complementary surface. Nonetheless, an unstructured charged domain may have many advantages over a specific structured domain (51). At neutral pH in aqueous solution, charge repulsion between the many ionized residues in these domains may inhibit formation of specific structure. These domains are therefore flexible and extend away from the proteins. The flexible and extended nature of these domains may increase the possibility of encountering the target proteins, and the charged amino acid side chains may provide a suitable force for promoting macromolecule association. Target proteins such as TBP and TFIIB present many surface-exposed basic amino acids that might serve to neutralize the acidic residues of the VP16 AAD and thereby help induce a specific conformation.
In addition to the charged or strongly polar amino acids commonly found in transcriptional activation domains, hydrophobic (and particularly aromatic) residues are often critical for the function of transcriptional activators (14). Aromatic residues have been shown in several cases to provide the binding docking force for protein-ligand interactions (52)(53)(54) or to be directly involved in protein-protein interactions (55)(56)(57). We speculate that the critical aromatic residues in the VP16 AAD participate directly in the binding of target proteins, providing some degree of binding stability and specificity.
The unusual potency of VP16 as a transcriptional activator has been attributed to its ability to bind to a number of target proteins in the transcriptional apparatus (58), which may allow VP16 to act during multiple steps of preinitiation complex assembly (59). The results of this report do not contradict this multiple-target model. Although our results demonstrate most clearly a specific interaction between TBP and VP16 AAD, a weaker and more limited interaction with TFIIB was also observed. Interestingly, a TBP mutant deficient in interacting with TFIIB was shown to be deficient in GAL4-VP16 activated transcription (60). This result suggests that in addition to interacting with TBP directly, the AAD interacts with the TBP⅐TFIIB⅐promoter complex (61). Thus, the weak intrinsic interaction between VP16 AAD and TFIIB may be strengthened in the presence of TBP.
Transcriptional activation is likely not to result from simple static interactions of activators with basal transcription factors but rather may involve the dynamic exchange of interactions among activation domains, basal factors, and coactivators. Recent studies show that distinct regions of the large subunit of RNA polymerase II share features in common with either acidic or proline-rich activators (62,63). The activation domains and the RNA polymerase II domains may compete for interaction with the same basal transcription factors or coactivators. If these interactions were to occur between rigid, stereochemically complementary protein surfaces, the binding might be so strong that exchange of such tight interactions would be difficult. In contrast, interaction of a target protein with flexible segments is weaker since association energy must be spent to compensate for the reduction of the conformational entropy. Thus, the transitions between ordered and disordered structures in activation domains (and their cognates in RNA polymerase II) may be a means to facilitate the dynamic interaction exchanges and hence to regulate the activation process.