Self-association of the amino-terminal domain of the yeast TATA-binding protein.

The amino-terminal domain of yeast TATA-binding protein has been proposed to play a crucial role in the self-association mechanism(s) of the full-length protein. Here we tested the ability of this domain to self-associate under a variety of solution conditions. Escherichia coli two-hybrid assays, in vitro pull-down assays, and in vitro cross-linking provided qualitative evidence for a limited and specific self-association. Sedimentation equilibrium analysis using purified protein was consistent with a monomer-dimer equilibrium with an apparent dissociation constant of approximately 8.4 microM. Higher stoichiometry associations remain possible but could not be detected by any of these methods. These results demonstrate that the minimal structure necessary for amino-terminal domain self-association must be present even in the absence of carboxyl-terminal domain structures. On the basis of these results we propose that amino-terminal domain structures contribute to the oligomerization interface of the full-length yeast TATA-binding protein.

TBP functions in transcription initiation as a monomer. It binds TATA sequence DNA as a monomer (21)(22)(23)(24)(25)(26) and interacts with TAFs and general transcription factors as a monomer (13,14,(27)(28)(29). On the other hand, TBP is capable of self-association. Dimers, tetramers, octamers, and higher oligomers have been previously reported in vitro (15, 16, 30 -35), while in vivo cross-linking results are consistent with association to at least the dimer level in HeLa and yeast cells (16,36,37). On the basis of these results, it has been proposed that monomeroligomer equilibria affect the concentration of TBP monomer that is available for transport into the nucleus for transcription-regulatory functions or for degradation (15, 16, 32-34, 38, 39).
Little is known about the mechanism(s) of TBP self-association. Crystal structures of the carboxyl-terminal domain of yeast TBP reveal a saddle-shaped molecule of ϳ180 amino acid residues with a concave DNA-binding face and a convex TAFand transcription factor-binding face (9,21,23,30,31). In the absence of DNA, this crystallizes as a dimer, stabilized by extensive contacts between the concave surfaces (9,30,31). 2 Light scattering and more recent sedimentation equilibrium results indicate that dimers of the carboxyl-terminal domain are also present in solution (30). 3 On the other hand, fulllength yeast TBP has a monomer-tetramer-octamer pattern of self-association over a wide range of solution conditions (24,33,34,40). These contrasting self-association mechanisms suggest that the amino-terminal domain plays an important role in the association mechanism. Evidence of this involvement comes from significant changes in the fluorescence spectrum and anisotropy of Trp-26 (roughly the middle of the amino-terminal domain) that accompany TBP selfassociation (34,40).
Based on these observations, we have proposed that the amino-terminal domain contributes structures necessary for the monomer-tetramer-octamer association pattern (34). This might occur if some or all of an interaction surface were contained in the amino-terminal domain. If this were the case, isolated amino-terminal domains might be expected to self-associate. In the experiments described below we tested this prediction and present the first evidence, to our knowledge, of a self-association activity for the amino-terminal domain.
Strains, Plasmids, and Growth Conditions-Plasmid pKA9-TBP, encoding full-length yeast TBP was the kind gift of Dr. Michael Brenowitz, Albert Einstein University. The Escherichia coli strain used as a "re-* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
porter" for two-hybrid experiments was Bacteriomatch TM , a restrictionminus, Kan R derivative of XL1-Blue MR obtained from Stratagene. Strain XL1-Blue MR (Stratagene) was used for routine propagation of plasmids. Expression plasmids were propagated in E. coli BL21 Star TM cells (Invitrogen). Cells containing the two-hybrid bait plasmid pBT (3.2 kb, Cam R , from Stratagene) and prey plasmid pTRG (4.4 kb, Tet R , from Stratagene) were maintained on LB agar containing chloramphenicol (34 g/ml) or tetracycline (15 g/ml), respectively. Plasmids pBT-LGF2 and pTRG-Gal11 p were supplied by Stratagene.
The expression vectors used for protein purification and pull-down assays were pPROTet.E (2.2 kb, Cam R , from Clontech) encoding an amino-terminal (His-Asn) 6 affinity tag for in-frame fusion with target genes and pFLAG.mac (5 kb, Amp R from Sigma) expressing an aminoterminal FLAG epitope tag (Asp-Tyr-Lys-Asp-Asp-Asp-Asp-Lys) for inframe fusion with target genes. These were maintained on LB agar containing 34 g/ml chloramphenicol and 50 g/ml ampicillin, respectively.
The LB selection agar for two-hybrid experiments contained 30 g/ml carbenicillin, 15 g/ml tetracycline, 34 g/ml chloramphenicol, and 50 g/ml kanamycin (CTCK), simultaneously selecting for protein-protein interaction and the presence of pBT and pTRG plasmids. In some cases these selection conditions were modified by the inclusion of 10 M isopropyl-1-thio-␤-D-galactopyranoside (IPTG, conditions designated CTCK ϩ IPTG).
Cloning and Constructs-The DNA sequence encoding the aminoterminal domain (amino acid residues 1-62) of TBP was PCR-amplified from plasmid pKA9-TBP using primers P1 and P3 (Table I). Reaction products were cut with endonucleases BamHI and XhoI and ligated into the prey plasmid pTRG (linearized with the same enzymes). Independently constructed isolates are designated pTRGN1, pTRGN2, and pTRGN5 (Table II). Amplification of pKA9-TBP with primers P2 and P3 and cleavage with EcoRI and XhoI gave products for ligation with the bait plasmid pBT (previously linearized with the same enzymes) designated pBTN1 and pBTN2. Primer P2 also encodes a flexible (Gly 4 Ser) 3 linker connecting the plasmid-encoded repressor protein to the TBP amino-terminal (bait) domain.
Constructs for protein purification and pull-down assays were prepared as follows. Primers P6 and P7 (Table I) were used for PCR amplification of TBP amino-terminal sequences from pTRGN1. After cleavage with SalI and NotI, reaction products were ligated into pPRO-Tet.E (linearized with the same enzymes). The product was designated pPROTet.E.N1. TBP amino-terminal domain sequences amplified from pTRGN1 with primers P8 and P9 were digested with XhoI and BglII and ligated into pFLAG.mac (linearized with the same enzymes). The product was designated pFLAG.mac.N1. E. coli cells were transformed by the CaCl 2 /heat shock method (41), and clones were selected on LB agar containing the appropriate antibiotics. All plasmid constructs were verified by PCR analysis and DNA sequencing. Primers P3 and P4 amplified pBT inserts, and primers P3 and P5 amplified pTRG inserts. Primers P7 and P10 amplified pPROTet.E inserts, and primers P9 and P11 amplified pFLAG.mac inserts.
Expression of TBP amino-terminal domain sequences in candidate cells was analyzed by SDS-PAGE with immunoblot detection; anti-TBP polyclonal antibody (kindly supplied by Frank Pugh, Penn State University, University Park) was used to detect expression of TBP sequences from pBTN1, pBTN2, pTRGN1, pTRGN2, and pPROTet.E133.N1 vectors. Anti-FLAG M2 monoclonal antibody (Sigma) was used to detect FLAGtagged proteins expressed from pFLAG.mac.N1.
E. coli Two-hybrid Assays-The method used was based on the work of Dove et al. (42,43) and is shown schematically in Fig. 1. Aliquots of challenge bait plasmids (pBTN1 and pBTN2), challenge prey plasmids (pTRGN1 and pTRGN5), and control vectors pBT (with no insert), pTRG (with no insert), pBT-LGF2, and pTRG-Gal11p were co-transformed as pairs (Table II) into freshly prepared competent cells. Transformations were carried out in triplicate by the CaCl 2 /heat shock method (41). Following heat shock, samples were incubated on ice for 10 min and diluted with NZYM medium (41), and aliquots of each transformation mixture were plated in duplicate on CTCK and CTCK ϩ IPTG agar media. Colonies were counted after growth at 30°C for 48 h.
Protein Pull-down Interaction Assay-E. coli BL21 Star cells were transformed with the plasmid pPROTet.E (empty) or with pPRO-Tet.E.N1 (encoding the TBP amino-terminal domain) and grown with agitation at 30°C. Protein expression was induced with anhydrotetracycline (final concentration, 0.4 g/ml) when cultures reached A 600 ϭ 0.3. After incubation for 8 h at 30°C, cells were collected by centrifugation and lysed at 4°C with lysozyme (0.75 mg/ml) followed by sonication (3 ϫ 30 s with a pause for 30 s on ice between each burst). The (His-Asn) 6 -tagged TBP amino-terminal domain was purified by absorption on cobalt affinity resin (Talon, Clontech) followed by elution with 50 mM sodium acetate, 300 mM NaCl, pH 5.0 as specified by the manufacturer. Samples were concentrated using Centricon YM3000 filter devices (Millipore), and pH was adjusted to 7.0 with sodium phosphate buffer.
E. coli BL21 cultures transformed with pFLAG.mac (empty) or with pFLAG.mac.N1 (encoding the TBP amino-terminal domain) were grown at 30°C to A 600 ϭ 0.5, and protein expression was induced with 1 mM IPTG. Growth was continued at 30°C for an additional 8 h. Cells were harvested by centrifugation and lysed by incubation with lysozyme followed by sonication as described above.
Samples were prepared by mixing 20 g of purified (His-Asn) 6 tagged amino-terminal protein with crude cell lysate proteins (1 mg) containing FLAG-tagged amino-terminal domain in a reaction volume of 800 l. The incubation buffer was 94 mM sodium phosphate (pH 7.0 at 20°C), 6.6 mM sodium acetate, and 222 mM NaCl. Samples were incubated at 4°C for 3 h. Cobalt affinity resin was equilibrated with 50 mM sodium phosphate (pH 7.0 at 20°C), 300 mM NaCl. Samples were incubated (batchwise for 1 h at 4°C) with cobalt affinity resin (1-ml packed volume). The resin was washed with 5 volumes of 50 mM sodium phosphate (pH 7.0 at 20°C), 300 mM NaCl and then eluted with 5 volumes of the same buffer adjusted to pH 5.0. Eluates were concentrated using Centricon YM3000 concentrators and analyzed by SDS-PAGE with immunoblot detection using anti-FLAG and anti-TBP antibodies.
Protein Cross-linking-Extracts were prepared from E. coli BL21 Star BEM cells expressing (His-Asn) 6 -tagged TBP amino-terminal domain pROTet.E.N1 or from cells containing the empty control vector  pPROTet.E by the lysozyme-sonication method. (His-Asn) 6 -tagged proteins were partially purified by cobalt affinity chromatography as described above. The cross-linkers used were DFDNB (spacer length, ϳ3 Å) and ANB-NOS (spacer length, ϳ7.7 Å). Cross-linking reactions (30 l) contained 0.3 mg/ml protein and 5 mM cross-linker in buffer consisting of 110 mM sodium phosphate, 10 mM sodium acetate, and 115 mM NaCl, pH 7.0 at 20°C. Samples were incubated at room temperature in the dark for 30 min. Samples containing ANB-NOS were then exposed to light from a short wavelength UV light box (Fotodyne) for 1 min. Proteins were resolved by SDS-PAGE, and TBP amino-terminal domain peptides were detected by immunoblotting using anti-TBP polyclonal antibody. Sedimentation Equilibrium Analysis-The (His-Asn) 6 -tagged aminoterminal domain was purified to homogeneity (by SDS-PAGE criteria) by immobilized metal and ion-exchange chromatography. Aliquots were transferred into 10 mM Tris (pH 7.0), 100 mM KCl buffer using a centrifugal concentrator (Centricon YM3000). Samples were brought to sedimentation equilibrium at 4°C in a Beckman XL-A analytical ultracentrifuge equipped with an AN-60 rotor. Absorbance values were measured at 280 nm as functions of radial position. Five scans were averaged for each sample at each rotor speed. The approach to equilibrium was considered to be complete when replicate scans separated by Ն6 h were indistinguishable. Solvent densities were calculated using the formula and tables published by McRorie and Voelker (44).
At sedimentation equilibrium, the absorbance at a specified wavelength and position in the solution column is given by Equation 1 (44,45).
Here A(r) is the absorbance at radial position r; the summation is over all species, n; ␣ n,0 is the absorbance of the nth species at the reference position r 0 ; n ϭ M n (1 Ϫ v n ) 2 /2RT where M n is the molecular weight of the nth species, v n is the partial specific volume (0.699 for TBP amino-terminal domain), is the solution density, is the rotor angular velocity, R is the gas constant, and T is the absolute temperature. The base-line offset term compensates for slight position-independent differences in the optical properties of different cell assemblies. Global analysis of data obtained at different rotor speeds was performed with the program NONLIN. 4

RESULTS
Yeast TBP Amino-terminal Domains Interact in Living E. coli Cells-The output signal of the classical yeast two-hybrid assay (cell survival and growth) results from enhanced transcription of one or more indicator genes (49). Interactions between yeast transcription factors may be difficult to study by this assay because two-hybrid constructs containing these proteins may be capable of activating yeast transcription in the absence of the protein-protein interaction of interest (false positive results). In addition, other yeast transcription factors may compete with or inhibit the interaction of interest, giving rise to false negative results. To minimize the potential for these complications we have used an E. coli two-hybrid assay (Refs. 42 and 43, summarized schematically in Fig. 1). While this strategy does not prevent all false positive and false negative results it has the advantage that the TBP amino-terminal domain is an exogenous protein so any interactions that it may have with the E. coli transcription machinery are likely to be adventitious.
In the two-hybrid method that we have used, the proteins of interest are expressed as carboxyl-terminal fusions with the cI protein (bait constructs in plasmid pBT) or with the ␣ subunit of RNA polymerase (prey constructs in plasmid pTRG). Thus, appropriate controls for the absence of interaction are plasmid pairs in which bait and prey contain only cI protein and ␣ subunit of RNA polymerase, respectively (since no additional sequences are added, we refer to these as "empty" constructs). The two-hybrid data shown in Fig. 2 N5). Triplicate transformations were grown on CTCK and CTCK ϩ IPTG media (columns labeled "ϩI"). Negative control pairs, pBTN1 with empty pTRG (N1:E) and empty pBT with pTRGN1 (E:N1), represent background levels of growth with unpaired expression of aminoterminal domain bait (pBTN1) and prey (pTRGN1) constructs. All values are the average number of colonies on three test plates normalized to the average obtained with control cells carrying two empty vectors (pBT:pTRG). The error bars represent S.D. in colony numbers propagated through this normalization. nated N1:E). A similar control for the interaction of the TBP amino-terminal domain with cI protein alone is the pairing of an empty bait construct with the TBP amino-terminal domain RNA polymerase ␣ subunit construct (designated E:N1). Both of these controls gave levels of carbenicillin resistance that were within error the same as that found with the negative control in which both vectors are empty. This indicates that the expression of either amino-terminal bait or amino-terminal prey protein alone is not sufficient to give enhanced carbenicillin resistance. It also indicates the absence of endogenous E. coli proteins that combine strongly with bait or prey proteins to activate transcription from the test promoter.
Greater numbers of carbenicillin-resistant colonies were obtained when both bait and prey plasmids encoded TBP aminoterminal domain protein. Although there was some variance, growth of cells containing amino-terminal domain bait and amino-terminal domain prey constructs in the absence of IPTG gave an average 240% increase in the number of surviving colonies, while growth in the presence of IPTG increased the average number of survivors by 310%. This IPTG effect supports the interpretation that the enhanced carbenicillin resistance of challenge cells reflects enhanced expression of bait and prey gene products.
As a positive control, we examined the carbenicillin resistance of cells containing the plasmids pBT-LGF2 and pTRG-Gal11 p .
LGF2 is the dimerization domain of the yeast Gal4 transcription activator, while Gal11 p is a domain of the yeast Gal11 protein carrying a point mutation that enables it to interact with Gal4 protein. The resulting complex is quite stable (K d Յ 3 ϫ 10 Ϫ7 M) under representative in vitro conditions (50). As summarized in Fig. 2, the expression of these constructs in our test E. coli gave an average 390% increase in the number of surviving colonies, while growth in the presence of IPTG increased the average number of survivors to 470%. These values are greater than those found with challenge pairs expressing the TBP amino-terminal domain, suggesting that the amino-terminal domain interaction may be weaker in the context of the E. coli cytoplasm than that of LGF2 with Gal11 p .
Following the two-hybrid test, bait and prey plasmids were reisolated from colonies growing on test plates. DNA sequencing confirmed the presence of unmodified sequences encoding TBP amino-terminal bait and prey proteins (result not shown). Together these results are consistent with an interaction between yeast TBP amino-terminal domain bait and prey proteins in the environment of the E. coli cytoplasm. (46) were carried out to further test the self-association of TBP aminoterminal domains. FLAG-and (His-Asn) 6 -tagged amino-terminal proteins were prepared from E. coli transformed with pFLAG.mac.N1 and pPROTet.E133.N1. (His-Asn) 6 -tagged TBP amino-terminal domain (partially purified by cobalt affinity chromatography) was incubated with E. coli cell extract containing FLAG-tagged TBP amino-terminal protein at 4°C as described under "Materials and Methods." The sample was divided, and one part was subjected to cobalt affinity chromatography, then to PAGE with immunoblot detection using polyclonal anti-TBP antiserum. As shown in Fig. 3, lane a, this mixture contained comparable amounts of (His-Asn) 6 and FLAG-tagged amino-terminal domains. The electrophoretic mobilities of these species give apparent molecular masses of 8.9 and 8.2 kDa, respectively, in good agreement with values predicted from protein sequence.

Association of TBP Amino-terminal Domains Detected by Pull-down Assay-Pull-down interaction assays
The second aliquot was applied to a short (1-ml bed volume) column containing cobalt affinity resin, and the eluate was subjected to PAGE with immunoblot detection using anti-FLAG monoclonal antibody (Fig. 3, lane b). A band with apparent molecular weight equal to that of FLAG-tagged aminoterminal domain is evidence that this protein is retained by the cobalt affinity resin in the presence of (His-Asn) 6 -tagged TBP amino-terminal domain. FLAG-tagged amino-terminal domain was not retained by the resin when applied alone (Fig. 3, lane  c). As a second control, an extract from cells containing pPRO-Tet.E (empty vector) was combined with one from cells containing pFLAG.mac.N1 (expressing FLAG-tagged amino-terminal domain). No FLAG-tagged amino-terminal domain was detectable in the eluate of the cobalt affinity column to which this mixture was applied (result not shown). Together these results are consistent with the interpretation that FLAGtagged TBP amino-terminal domain forms stable complexes with (His-Asn) 6 -tagged TBP amino-terminal domain in the presence of components of E. coli cell extract under our buffer conditions. TBP Amino-terminal Domains Can Be Cross-linked in Vitro-Chemical cross-linking was performed to confirm the in vitro interaction of amino-terminal domains and to obtain an estimate of the degree of oligomerization that these molecules experience. Two chemically distinct cross-linkers (DFDNB and ANB-NOS) were tested. Samples of cell extract from E. coli BL21 Star cells expressing (His-Asn) 6 -tagged TBP amino-terminal domain (0.3 mg/ml total protein) were incubated with cross-linkers as described under "Materials and Methods" and resolved by SDS-PAGE (Fig. 4). Reaction with cross-linkers resulted in a shift of the amino-terminal domain to a gel mobility corresponding to an M r ϳ 17,000, slightly less than twice the molecular weight expected for dimeric amino-terminal domain (M r ϭ 17,800). Higher molecular weight species were undetectable despite nearly quantitative conversion of the monomer into the cross-linked form. Thus, if higher oligomers are formed, they must be present at low concentration relative to the dimer, or they must be inefficiently cross-linked by the reagents that we tested. The TBP Amino-terminal Domain Sediments as a Mixture of Monomers and Dimers-Samples of the (His-Asn) 6 -tagged amino-terminal domain were brought to sedimentation equilibrium at 20,000, 25,000, 30,000, and 40,000 rpm (Fig. 5). The solid curve through the data is the global least squares fit of the expression for a monomer-dimer mixture (Equation 1 with species of monomer and dimer molecular weights) to four data sets (one concentration, four rotor speeds) obtained using the program NONLIN. The uniform distribution of residuals around zero indicates the compatibility of this model with the data. Several other models were tested including monomer only, monomer-trimer, monomer-tetramer, monomer-n-mer (with stoichiometry a parameter of the fit), and monomerdimer-trimer; all gave larger residuals than the monomerdimer model, and the residuals had systematic deviation with radius (results not shown). This outcome indicates that if the amino-terminal domain can self-associate beyond the dimer, these species are not present in significant concentration in our samples. Fits using the monomer-dimer model returned a monomer molecular weight of 9,650 Ϯ 1,050 in good agreement with the value predicted from the sequence.
To estimate the association constant for the amino-terminal domain dimerization, we used a variant of Equation 1 in which the reference absorbance of the dimer is expressed in terms of an equilibrium constant and the reference monomer absorbance (␣ dimer,0 ϭ KЈ(␣ monomer,0 ) 2 ).
This yields an apparent association constant KЈ scaled in absorbance units. To convert this to the familiar molar scale we used the equation, in which ⑀ is the molar extinction coefficient and l is the optical path length of the centrifuge cell. The molar extinction coefficient was estimated to be 5,500 on the basis of amino acid content (48). This analysis returned a value of the apparent equilibrium constant K molar ϭ 1.20 Ϯ 0.23 ϫ 10 5 M Ϫ1 (equivalently K d ϭ 8.36 Ϯ 1.99 ϫ 10 Ϫ6 M). This binding affinity is sufficient to ensure at least partial self-association of aminoterminal domains if the bulk concentration of TBP in yeast nuclei is ϳ6 M as estimated previously (33,34). In addition, it places the midpoint of the assembly reaction at ϳ10 Ϫ5 M aminoterminal domain; previous analytical ultracentrifugation studies have shown that full-length TBP self-associates to form tetramers and octamers over a similar range of [protein] under comparable solution conditions (33,34). 5 However these comparisons come with important caveats, which are discussed below.

DISCUSSION
Here we have tested the ability of the amino-terminal domain of yeast TATA-binding protein to self-associate. Selfassociation might be expected if this domain contributed surfaces to the oligomerization interface(s) of the full-length TATA-binding protein. The four assays that we used test for interaction under different but representative sets of conditions. The two-hybrid assay seeks interactions in the ionic conditions and highly crowded environment of the E. coli cytoplasm. The pull-down assay tests for interaction under conditions of lower macromolecular concentrations but allows the ionic and small molecule compositions of the reaction mixture to be controlled by the experimenter (50 mM sodium phosphate (pH 7.0 at 20°C), 300 mM NaCl in our experiments). The cross-linking assay provides independent evidence of interactions that might be detected in the pull-down assay. In addition, it has the potential to detect interactions that are too labile to be observed by the pull-down assay or that involve partners that would not be detected by immunoblot with anti-FLAG antiserum. Finally sedimentation equilibrium analysis (carried out in 10 mM Tris (pH 7.0), 100 mM KCl) shows that the minimum model for this self-association is monomer-dimer. While the interaction conditions that these assays test are unlikely to be identical to those found in yeast nuclei, together they span a wide range in total macromolecule concentration 5 The solution conditions used in previous studies were performed at 4°C in 20 mM HEPES/KOH (pH 7.9), 1 mM EDTA, 1 mM dithiothreitol, 120 mM KCl; the present studies were carried out at 4°C in 20 mM Tris (pH 7.0), 100 mM NaCl.

FIG. 5. Purified amino-terminal domain sediments as a mixture of monomers and dimers.
Purified (His-Asn) 6 -tagged aminoterminal domain protein was brought to sedimentation equilibrium at 40,000 rpm and 4°C as described under "Materials and Methods." Absorbance measurements were made at 280 nm. The smooth curve represents the global fit of the monomer-dimer version of Equation 1 to data sets obtained at 20,000, 25,000, 30,000, and 40,000 rpm. This analysis returned a monomer molecular weight of 9,650 Ϯ 1,050 in good agreement with the value predicted from sequence. The symmetric residuals demonstrate the compatibility of this model with the data. and modest but useful ranges in the concentrations of buffer components. The fact that the self-association of the TBP amino-terminal domain can be detected by these disparate methods indicates that the interaction is quite robust and that it is not an artifact of preparative biochemistry or choice of reaction conditions.
The ability of amino-terminal domains to interact in the two-hybrid assay despite the availability of a high concentration of cytoplasmic protein competitors suggests that the interaction of amino-terminal domains is quite specific. This notion is supported by the pull-down assay results that indicate that any competing interactions are not of sufficient strength to displace the FLAG-tagged amino-terminal domain from its (His-Asn) 6 -tagged partner. Fluorescence and protein footprinting data obtained with full-length yeast TBP are consistent with models in which the amino-terminal domain adopts specific secondary structures and participates in specific tertiary and/or quaternary interactions (24,34,40,47). Our results support this view. The specific self-association of amino-terminal domains in the presence of high concentrations of competing proteins (as seen in the two-hybrid and pull-down assays) is evidence of a functional complementarity that is not likely to be present in an unstructured domain. That these interactions can take place between isolated amino-terminal domains indicates that tertiary contact with the carboxyl-terminal domain is not required for maintenance of conformation(s) that allows this self-association.
The sedimentation equilibrium results are consistent with a monomer-dimer pattern of self-association over the protein concentration range 0.5 M Յ [amino-terminal domain] Յ 18 M. However, limitations in the amount of amino-terminal domain protein currently available have prevented us from reaching greater protein concentrations where higher order associations might be detectable. In the absence of a more stringent test, we regard the monomer-dimer pattern as a minimal model of the self-association mechanism. The crosslinking results, performed at total protein concentrations similar to those of the sedimentation experiment, are consistent with this view. The detection of cross-linked dimers, but not higher oligomers, indicates that if higher order structures are formed they are either present at concentrations too low for detection by our Western blot procedure or that they do not offer the juxtaposition of reactive groups necessary for cross-linking.
The dimer dissociation constant estimated from sedimentation equilibrium data (K d ϭ 8.36 Ϯ 1.99 ϫ 10 Ϫ6 M) indicates that amino-terminal domains have the potential to be at least partially associated in vivo if the bulk TBP concentration within yeast nuclei is ϳ6 M as predicted (33,34). This conclusion comes with a number of significant caveats. It does not take into account the currently unknown effects of the TBP carboxyl-terminal domain on association of the amino-terminal domain. It does not take into account the possible effects of competing interactions with other proteins. Finally it does not take into account the crowding effect exerted by the high concentrations of macromolecules present in the nucleus. Each of these effects has the potential to bias the dissociation constant by orders of magnitude. Thus, while the numerical value of K d indicates that some degree of self-association is possible, it does not provide enough information to allow prediction of the association state of the amino-terminal domains of TBP in vivo.
At present we do not know whether the interactions of isolated amino-terminal domains resemble those that take place in full-length TBP. The recent protein footprinting results of Brenowitz and colleagues (47) suggest that the amino-terminal domain makes extensive contact with the convex face of the carboxyl-terminal domain. These interactions have the potential to compete with the self-association of amino-terminal domains. On the other hand, several TAFs and transcription factors interact with the convex face of the carboxyl-terminal domain near surfaces that may be occupied by the aminoterminal domain. These interactions may be modulated by the interaction between amino-and carboxyl-terminal domains. In this context, the self-association of amino-terminal domains may compete with amino-terminal-to-carboxyl-terminal domain interaction and allow transcription factors and TAFs access to the surface of the carboxyl-terminal domain. A large body of circumstantial evidence (outlined in the Introduction) implicates the amino-terminal domain in the oligomerization mechanism of the full-length TATA-binding protein. The results described here raise the possibility that the amino-terminal domain may form part of the oligomerization interface itself.