Molecular Architecture of the Basal Transcription Factor B-TFIID*

BTAF1 (formerly named TAFII170/TAF-172) is an essential, evolutionarily conserved member of the SNF2-like family of ATPase proteins and together with TATA-binding protein (TBP) forms the B-TFIID complex. BTAF1 has been proposed to play a key role in the dynamic regulation of TBP function in RNA polymerase II transcription. We have determined the structure of native B-TFIID purified from human cells by electron microscopy and by image analysis of single particles at a resolution of 28 Å. B-TFIID is 15 × 9 nm in size and is organized into a large domain of about 170 kDa, which can be subdivided into two domains. Extending from this domain is a long thumb, which in turn is divided into subdomains of about 25, 15, and 35 kDa, the largest of which is located at the end of the thumb. Immunolabeling experiments localize the extreme carboxyl terminus of BTAF1 within the 170-kDa domain, whereas the amino terminus and TBP co-localize to the end of the protruding thumb. The central portion of BTAF1 localizes to the base of the thumb. Comparison of the native B-TFIID with its recombinant form shows that both share a similar domain organization. Collectively, these data provide the first structural model of the B-TFIID complex and map its key functional domains.

The TATA-binding protein (TBP) 1 is a central player in RNA polymerase II (pol II)-dependent transcription. TBP as a component of the TFIID complex recognizes the TATA box within pol II promoters and initiates the assembly of the preinitiation complex including the recruitment of pol II (1). TBP function on promoter DNA is influenced positively and negatively by several transcriptional regulators (2,3). Among these BTAF1 and its ortholog in the yeast Saccharomyces cerevisiae, Mot1p, have emerged as essential elements in the regulation of TBP recruit-ment and activity on promoter DNA (4). BTAF1 and Mot1p belong to the SNF2-like family of ATPase proteins based on the presence of a carboxyl-terminal ATPase domain (5)(6)(7). Consistent with their role as key regulators of TBP function, a large proportion of TBP is associated with Mot1p in yeast cells (8,9) and with BTAF1 to form the B-TFIID complex in human cells (10,11). The B-TFIID complex can support pol II transcription in vitro as efficiently as TFIID (10). However, B-TFIID displays several important differences that make it distinct from TFIID. It has been shown that B-TFIIDmediated pol II transcription does not respond to a number of activators and that B-TFIID exhibits dATPase activity. Finally, B-TFIID does not form highly stable complexes with DNA in template commitment experiments (10).
The most prominent biochemical activity ascribed to BTAF1 and Mot1p is their capacity to remove TBP from TATA DNA using the energy of ATP hydrolysis. This activity explains their transcriptional inhibition properties (7,(12)(13)(14)(15). However, a large body of in vitro and in vivo data is also consistent with the idea that BTAF1 and Mot1p utilize this activity to positively regulate transcription by dynamically mobilizing TBP on promoter DNA (10, 16 -19). How BTAF1 and Mot1p use the energy of ATP hydrolysis to dissociate TBP from DNA has been extensively studied. Experiments with Mot1p suggest that disruption of TBP-DNA complexes does not involve DNA helicase activity or changes in DNA structure such as unwinding, twisting, or bending (20,21). Furthermore, an ATP-dependent DNAtracking mechanism does not appear to be involved (22). Recent models propose that Mot1p uses DNA upstream of TATA as a handle to dissociate TBP from DNA. It has been proposed that Mot1p translocates along the DNA handle as an ATPase motor and by pushing or pulling removes TBP (21). Alternatively, a domain of Mot1p could be anchored to the upstream handle, and during an ATP-driven power stroke an effector portion of Mot1p is inserted between TBP and DNA acting as a wedge to displace TBP (21). Consistent with this, distinct subdomains in the amino terminus of BTAF1 and Mot1p contact the concave surface of TBP to inhibit its binding to DNA, and preformed Mot1p-TBP complexes are unable to bind TATA DNA (21,23,24). However, very recent work has challenged aspects of these models. Fluorescence anisotropy experiments have suggested that dissociation of TBP-TATA complexes may not be reliant on Mot1p contacting upstream handle DNA but rather on Mot1pmediated changes in TBP conformation altering its contacts with TATA (15).
Despite these studies the molecular mechanism of Mot1p and BTAF1 action remains poorly understood. In particular, the molecular details of BTAF1 and Mot1p action have not been defined to an extent that the various models can be integrated , and the Association pour la Recherche sur le Cancer. 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. or distinguished to explain how BTAF1 and Mot1p cause TBP-TATA dissociation. Key to defining a mechanism of action for BTAF1 and Mot1p will be a description of their structure in complex with TBP. In this study we have employed electron microscopy (EM) coupled with image processing to examine the  molecular organization of the native and recombinant human  B-TFIID complexes. B-TFIID resembles a hand-shaped molecule containing a large globular domain of about 170 kDa from which extends a long, curved thumb-like structure, which can be subdivided into three domains. The relative positions of TBP and the carboxyl-terminal, the amino-terminal, and the central portions of BTAF1 were localized within the structure of the native B-TFIID by immunolabeling experiments. These observations provide the first structural data of the B-TFIID complex.

EXPERIMENTAL PROCEDURES
Purification of B-TFIID Complexes-Phosphocellulose B fractions containing native B-TFIID were obtained by chromatography of a HeLa whole-cell extract on a phosphocellulose column (P-11, Whatman) according to Samuels et al. (25). The B fraction was purified by Q-Sepharose and MonoS chromatography as described previously (10). Subsequently, B-TFIID-containing fractions were purified by immunoaffinity chromatography as follows. 2C1 anti-TBP (kind gift of L. Tora, Institut de Génétique et de Biologie Moléculaire et Cellulaire) was coupled to 0.5 ml of protein G-Sepharose, and the slurry was incubated with the MonoS B-TFIID fractions for 12-14 h at 4°C. The resin was washed with IP500 buffer (25 mM Tris-HCl, pH 7.9, 5 mM MgCl 2 , 10% glycerol, 0.1% Nonidet P-40, 0.5 M KCl, 0.5 mM phenylmethylsulfonyl fluoride, 1 g/ml leupeptin, 1 g/ml aprotinin, 1 g/ml pepstatin) and IP100 (as above but containing 100 mM KCl). Elution was performed by adding 1 volume of IP100 containing 6 mg/ml PB242 peptide (SPQGAMTPGIPIFSPMMPYGTC) and 0.5 mg/ml insulin for 4 h at 4°C, and the elute was collected by a brief spin at 1200 rpm. A second elution was performed for 2 h at 4°C. Peak B-TFIID eluates were applied to a MonoS HR5/5 column (Amersham Biosciences), and bound B-TFIID was resolved with a linear gradient in A buffer (20 mM HEPES-KOH, pH 7.9, 20% glycerol, 0.5 mM EDTA, 1 mM dithiothreitol, and protease inhibitors as above) to 1 M KCl. Peak B-TFIID fractions were pooled, concentrated on a YM-30 centricon filter column (Millipore), and stored at Ϫ80°C.
For the production of recombinant B-TFIID, a SalI-SpeI fragment from pTRE-BTAF1 was cloned into the SmaI and SpeI sites of pTM3 (26) (kind gift of B. Moss, National Institutes of Health) to create pTM3BTAF1i. Subsequently, a SpeI-NotI fragment from pTREBTAF1 was cloned into the SpeI and NotI sites of pTM3BTAF1i to create pTM3BTAF1. pTM3hisTBP was created by cloning an NcoI-DraI fragment from pEThisTFIID (27) into the NcoI and StuI sites of pTM3. The pTM3BTAF1 and pTM3hisTBP plasmids were used to construct recombinant vaccinia viruses vv-TBP and vv-BTAF1 as described previously (28). Recombinant viruses were selected, amplified, and titered according to Walhout et al. (28). For purification of recombinant B-TFIID, 4 liters of HeLaS3 cells were infected as described previously (28) with vv-TBP, vv-BTAF1, and vvTF7-3 expressing T7 RNA polymerase (kind gift of R. Padmanabhan, University of Kansas) at a multiplicity of infection of 2 plaque-forming units/cell. After 20 h at 37°C infected cells were harvested, and extracts were prepared as above and were applied to a P-11 column. The phosphocellulose B fraction was adjusted to buffer T (20 mM Tris-HCl, pH 7.9, 20% glycerol, 0.5 mM EDTA, 1 mM dithiothreitol, and protease inhibitors as above) containing 50 mM KCl and purified over a MonoQ HR10/10 column (Amersham Biosciences), and bound B-TFIID was resolved with a linear gradient in T buffer to 300 mM KCl. Peak B-TFIID fractions were applied to a 2.5-ml nickelnitrilotriacetic acid-agarose column (Qiagen) equilibrated in T buffer containing 50 mM KCl, 0.1 mM EDTA, and 1 mM imidazole. Bound B-TFIID was eluted with buffer T containing 50 mM KCl, 0.1 mM EDTA, and 0.25 M imidazole. Peak B-TFIID fractions were chromatographed on a Sephacryl S300 26/60 column (Amersham Biosciences) equilibrated in A buffer with 0.01% Triton X-100 and 300 mM KCl. B-TFIIDcontaining fractions were adjusted to A buffer containing 50 mM KCl and loaded on a MonoS HR5/5 column (Amersham Biosciences), and bound B-TFIID was resolved by a linear gradient with A buffer to 300 mM KCl. Peak fractions were pooled and stored at Ϫ80°C. The composition and purity of B-TFIID preparations were assessed by SDS-PAGE followed by silver staining and immunoblot analysis with C-BTAF1 blanc (␣-BTAF1) and 1F8 (␣-TBP). B-TFIID preparations were quantitated by their TBP content as determined by comparison against known amounts of recombinant hisTBP (29) on immunoblots.
In Vitro Transcriptions-Transcription reactions were performed as described previously (24). Reactions contained 200 ng of supercoiled template pAdML(C 2 AT)⌬400; the amounts of TBP or B-TFIID indicated in Fig. 2; and 60 ng of TFIIB, 50 ng of TFIIE, 40 ng of TFIIF, and 0.5 l of affinity-purified calf thymus pol II. Template commitment assays using 200 ng of supercoiled template pAdML(C 2 AT)⌬400 and 200 ng of pAdML(C 2 AT)⌬200 were performed as described previously (10). Transcription reaction products were quantitated using a Storm 820 Phos-phorImager (Amersham Biosciences) and Image QuaNT version 4.2a software.
Electron Microscopy, Immunoelectron Microscopy, and Image Processing-The purified B-TFIID fractions were diluted to a concentration of 20 g/ml in EM buffer (20 mM Tris-HCl, pH 7.4, 150 mM NaCl, and 20% glycerol). Ten l of this preparation was placed on a carbon film treated by a glow discharge in air, adsorbed for 2 min, and negatively stained with a 2% (w/v) uranyl acetate solution. The images were formed on a Philips CM120 transmission electron microscope operating at 100 kV with a LaB6 filament. Areas covered with individual molecules were recorded under low dose condition (less than 20 electrons/Å 2 ) at a nominal magnification of ϫ45.000 on a Pelletier cooled slow scan CCD camera (Model 794, Gatan, Pleasanton, CA). The image processing was performed using the IMAGIC software package (31) (Image Science Software, Berlin, Germany) as described previously (32). For immunoelectron microscopy a 3-to 5-fold molar excess of antibodies was incubated 30 min at 20°C with purified B-TFIID at a final protein concentration of 20 g/ml.

Purification and Functional Characterization of B-TFIID
Complexes-To study the molecular organization of B-TFIID, the native and recombinant complexes were purified from HeLa cells using a combination of ion exchange and immunoaffinity chromatography and were functionally characterized. BTAF1 and TBP displayed strict co-elution after the P-11 step (data not shown). SDS-PAGE and silver-staining analysis of peak fractions from the final column demonstrated the presence of both proteins in the highly purified B-TFIID preparations (Fig. 1B). Due to the presence of a hexahistidine tag, TBP in the recombinant preparation showed slower migration compared with its native counterpart on SDS-PAGE gels (Fig. 1B). Identity of BTAF1 and TBP in the purified B-TFIID preparations was confirmed by immunoblotting (Fig. 1C).
These protein preparations were analyzed in various biochemical assays for B-TFIID activity. First, we determined whether the purified B-TFIID possessed ATP-dependent inhibitory activity to dissociate TBP-TATA complexes by analyzing the ability of B-TFIID to interact with a radiolabeled DNA probe encompassing the AdMLP TATA in the presence of ATP or its non-hydrolyzable analog AMP-PNP by electrophoretic mobility shift assay. TBP alone formed a complex with DNA in the absence of ATP, and the presence of ATP or AMP-PNP had no detectable effect on this interaction ( Fig. 2A, lanes 2 and  5-8). Native and recombinant B-TFIID formed a slower migrating complex in the absence of ATP ( Fig. 2A, lanes 3 and 4). Identity of this species as a BTAF1-TBP-DNA complex was confirmed by supershifting with BTAF1 antibodies (data not shown). In the presence of ATP, formation of the B-TFIID-DNA complex was severely reduced ( Fig. 2A, lanes 9, 10, 13, and 14). However, in the presence of AMP-PNP, no significant disruption of the complex was observed ( Fig. 2A, lanes 11, 12, 15, and  16). Thus, the native and recombinant B-TFIID preparations possess ATP-dependent inhibitory activity.
To test whether the B-TFIID preparations support basal transcription we compared their activity with recombinant TBP in basal transcription assays reconstituted with highly purified components (Fig. 2B). Comparable inputs of TBP were present in the transcription reactions as judged by immunoblotting (Fig. 2B, bottom, lanes 1-12). Basal transcription with TBP was ϳ2to 3-fold higher than transcription with B-TFIID (Fig. 2B, top, lanes 1-4). Previous analyses reported that B-TFIID supports basal transcription with equal efficiency as TBP (10). This may reflect differences in the activity of the recombinant TBP preparations used for this study. Neverthe-less, native and recombinant B-TFIID displayed comparable transcription efficiencies (Fig. 2B, top, lanes 5-12).
A specific feature of B-TFIID is that partially purified preparations in reconstituted transcription systems do not form highly stable complexes with template DNA (10). To determine whether this characteristic can be assigned solely to B-TFIID, we compared our highly purified B-TFIID preparations with recombinant TBP in template commitment experiments using our in vitro transcription system reconstituted with highly purified components (Fig. 2C). In the case of TBP a typical commitment pattern was observed with preference for transcription from the preincubated template (Fig. 2C, lanes 1-4).   9 -12) were preincubated with 200 ng of pAdML(C 2 AT)⌬400 or 200 ng of pAdML(C 2 AT)⌬200 as indicated above the lanes. The missing templates and remaining reaction components were then added. The arrows indicate the correctly initiated RNA products. The ratio of RNA products produced from the two DNA templates is indicated below the panel.
In contrast, in transcription reactions with native or recombinant B-TFIID both templates were transcribed at a comparable level whether they were preincubated or not (Fig. 2C, lanes [5][6][7][8][9][10][11][12]. Collectively, these data indicate that the native and recombinant B-TFIID complexes are identical and functionally active. Three-dimensional Model of B-TFIID-To obtain insight into the structure of B-TFIID, we analyzed the purified B-TFIID preparations by negative staining and electron microscopy. When processed for EM, B-TFIID appeared as a homogeneous dispersion of slightly elongated particles about 15 ϫ 9 nm in size (Fig. 3A). A total of 1981 molecular images (Fig. 3B) were recorded and numerically analyzed to obtain characteristic noise-free views of the particle (Fig. 3C). When viewed through a particular orientation, the particle was composed of a globular domain about 10 ϫ 6 nm in size from which an elongated thumb was extending at the end of which a smaller domain, about 7 ϫ 5 nm in size, could be observed. A total of 96 different views of the B-TFIID molecule were used to reconstruct a three-dimensional model of the particle. The orientation plot shows that the viewing directions were equally distributed, and the resolution tests gave values of 28 or 26 Å for the 0.5 Fourier shell correlation and 3 criteria, respectively (data not shown). The comparison between the input molecular views (Fig. 3C) and the projections of the three-dimensional model along the same directions (Fig. 3D) shows that the two data sets are consistent.
The three-dimensional structure of the B-TFIID complex is composed of a large domain 10 ϫ 6 ϫ 4 nm in size from which projects an extended thumb 12 nm long and 3-4 nm wide (Fig.  3E). The density threshold was set to delimit a volume of 248 kDa (assuming a protein density of 1.4 g/cm 3 ), which corresponds to the molecular masses of BTAF1 (210 kDa) and TBP (38 kDa). The large domain represents about 68% of the volume of the particle, which would correspond to a mass of about 170 kDa. The thumb is composed of three subdomains of about 25, 15, and 35 kDa, respectively, the largest subdomain being placed at the end of the thumb. The thumb is slightly curved and folds back over the large domain forming a 2.5-nm groove. The elongated shape of the B-TFIID complex is consistent with the hydrodynamic properties of the B-TFIID and Mot1p-TBP complexes as measured by gel filtration and rate zonal sedimentation analyses (8,10,20). Independently, we determined the structural organization of the recombinant B-TFIID complex. The size and overall shape of the recombinant particle, as determined by direct EM observation, were similar to that of the endogenous complex and in particular did not show any aggregation (data not shown). A total of 3323 molecular images were recorded and numerically analyzed to obtain 117 characteristic noise-free views of the recombinant B-TFIID molecule (data not shown) and to reconstruct a three-dimensional model (Fig. 3F). As expected the three-dimensional structure of the recombinant B-TFIID complex shares a strikingly similar domain organization with the native complex (Fig. 3, E and F). In the recombinant B-TFIID particles the large globular domain 10 ϫ 6 ϫ 4 nm in size and the 12-nm-long and 3-4-nm-wide protein thumb are clearly distinguished. Three peaks of density can be discerned within the extended thumb corresponding to the previously identified three subdomains within the native B-TFIID thumb (Fig. 3E).
Immunolocalization of TBP and BTAF1-To reveal details of the internal organization of native B-TFIID, a series of immunolabeling experiments was conducted using subunit-specific antibodies (Fig. 1A). We first employed the monoclonal antibody 1F8 directed against amino-terminal residues 56 -97 of TBP (Fig. 1A). Antibody-labeled B-TFIID molecules were identified by EM, and images of 560 immune complexes (ICs) were recorded and analyzed. The IC images were aligned against projections of the three-dimensional model of B-TFIID to determine the viewing direction of each molecule. The aligned data set was then partitioned to calculate noise-free views and to identify the site labeled by the antibody. The TBP-specific antibody bound to the end of the thumb as evidenced from two independent views of B-TFIID (Fig. 4A). TBP was shown previously to interact with the amino terminus of BTAF1 (24). To map this part of the molecule within the three-dimensional model of B-TFIID, polyclonal antibodies were raised against residues 230 -249 of BTAF1 (Fig. 1A) and used to map this epitope within the B-TFIID model. The analysis of 614 ICs shows, in two different views, that this antibody also binds to the end of the thumb close to the site labeled with TBP-specific antibodies (Fig. 4B). The resolution of the mapping is, however, not sufficient to distinguish the TBP from the BTAF1 sites. To further localize the position of distinct domains of BTAF1, polyclonal antibodies were raised against residues 631-650 within the central domain of BTAF1 linking the amino-and carboxyl-terminal regions (Fig. 1A). Upon analysis of 630 ICs, the binding of this antibody was detected in the large domain, at the base of the thumb (Fig. 4C). Finally, a polyclonal antibody was raised against the carboxyl terminus of BTAF1 (amino acids 1833-1849) (Fig. 1A), and the analysis of 651 ICs showed that the large domain is labeled on the side facing the end of the thumb (Fig. 4D). Altogether these results show that TBP is located at the end of the thumb close to the amino terminus of BTAF1 and that BTAF1 is linearly organized from the end of the thumb (amino acids 230 -249) toward the base of the thumb (residues 631-650) to the end of the large domain (carboxyl terminus). DISCUSSION BTAF1 is an essential, evolutionarily conserved protein and together with TBP forms the complex B-TFIID (6,10,14). BTAF1 and its yeast ortholog Mot1p have been proposed to play a central role in the dynamic regulation of TBP function in pol II transcription by using the energy of ATP hydrolysis to dissociate TBP from DNA (4,7,13,14). However, the molecular details of the mechanism of BTAF1-mediated disruption of TBP-DNA complexes are unknown. To understand the mechanism of action of BTAF1 we have examined the molecular architecture of the functionally active native and recombinant B-TFIID complex by EM-based structural analysis.
The three-dimensional model of B-TFIID was determined at a resolution of 28 Å from negatively stained, isolated particles. B-TFIID is a slightly elongated, hand-shaped molecule, about 15 by 9 nm in size, and it contains a large globular domain of about 10 ϫ 6 ϫ 4 nm and 170 kDa, which appears split into two subdomains. A remarkable feature of the three-dimensional model is the presence of a long, curved protein thumb protruding from this globular domain. This protein thumb measures 12 ϫ 3-4 nm, can be subdivided into three domains of 25, 15, and 35 kDa, and folds back over the large domain forming a 2.5-nm groove.
From immunolabeling experiments we are able to locate within the three-dimensional model of the B-TFIID complex the TBP and BTAF1 subunits and specific domains of BTAF1.
The immunolabeling experiments show that the most carboxylterminal 17 residues of BTAF1 are located within the second, smaller subdomain at the extreme end of the 170-kDa globular domain on the side facing the end of the thumb. Given that the conserved BTAF1 ATPase motifs reside within carboxyl-terminal residues 1285-1774 (54 kDa) (6) our immunolabeling data strongly suggest that the ATPase domain of BTAF1 resides within the distal half of the large globular domain. In addition, we can infer that the central BTAF1 residues (amino acids 460 -1285; 91 kDa) linking the ATPase motif and the aminoterminal TBP-binding region would correspond to the larger subdomain of the globular domain. Consistent with this, our immunolabeling data place amino acids 631-650 in the large domain close to the base of the long thumb. BTAF1 aminoterminal residues 230 -249 reside within the 35-kDa domain at the end of the protruding protein thumb. The distribution of TBP antibody overlaps with the labeling pattern of the BTAF1 amino terminus indicating close proximity of TBP to the aminoterminal part of BTAF1. This structural organization is consistent with previous biochemical analysis, which mapped the TBP-binding region of BTAF1 and Mot1p to the first 460 residues (23,24). The quaternary organization of B-TFIID domains derived from our immunolabeling experiments is in agreement with the linear arrangement of domains described previously for BTAF1 and Mot1p (6,7,23,24).
Our immunolabeling data of Fig. 4 mapped TBP in the outermost subdomain of the thumb, and we tried to fit the atomic structure of human core TBP into this region (33). Core TBP was found to be too large to be located in the first subdomain solely, but the shape of TBP fits perfectly between the first and second subdomains. However, human TBP has a long aminoterminal extension, and the 1F8 antibody used to map TBP recognizes an epitope (residues 56 -96) within its flexible extension. Therefore, the exact position of TBP is quite hypothetical since the core TBP may be placed closer to the large BTAF1 domain, whereas the amino-terminal extension is placed at the tip of the thumb. The extension also allows localization of core TBP between the second and third subdomains. These two ways of docking the core TBP structure into the envelope of B-TFIID are shown in the Supplementary Figure.  Several structural characteristics can be suggested in light of our domain mapping within the B-TFIID model. Previous in silico analyses suggest that the amino termini of BTAF1 and Mot1p contain clusters of tandemly arrayed HEAT/ARM repeats (23,24,34,35). It has been proposed that these repeats form an extended conformation that provides a large surface for the interaction of BTAF1 and Mot1p with TBP and components of the transcriptional apparatus (4,23,24). Given the assignment of the amino terminus of BTAF1 to the long thumb of B-TFIID, the extended nature of this region may reflect the presence of clustered HEAT/ARM repeats. Strikingly, the curved, elongated structure of the thumb is consistent with the crescent-shaped, elongated HEAT repeat structures of the PR65/A subunit of protein phosphatase 2A and eIF4 (36,37). A second prominent feature of B-TFIID is the presence of a 2.5-nm solvent-accessible groove formed by the bending of the thumb over the large globular domain. Although the function of the groove is unknown, its size could accommodate the binding of a DNA molecule. Our TBP docking results would suggest orientation of TBP toward the groove, but it does not exclude other DNA conformations. Alternatively, DNA may be wrapped along the external part of the thumb in close proximity to TBP and the BTAF1 amino terminus. How DNA is recognized by B-TFIID will be crucial to our molecular understanding of BTAF1 action and awaits further EM structural analysis using promoter DNA-bound B-TFIID.
The morphology of B-TFIID determined in this report bears some similarity with the three-lobed structure of yeast and human TFIID. These lobes represent large globular domains and are connected by narrow regions around a solvent-accessible groove that is likely to present a DNA-binding surface (32,38,39). The position of TBP in the B-TFIID and TFIID models is similar in that TBP faces the solvent-exposed groove (32,38,39). Nevertheless, comparison of the three-dimensional models of B-TFIID and TFIID indicates distinct structural organizations, which reflect the differences in their subunit composition and biochemical function (4,10,40).
A distinct characteristic of BTAF1 and Mot1p is their capacity to dissociate TBP from TATA box DNA using the energy of ATP hydrolysis. However, exactly how BTAF1 and Mot1p use ATP hydrolysis to elicit this function remains an open question. Several well characterized ATPases translate nucleotide binding and/or hydrolysis into major conformational changes (41,42). It is likely that ATP binding or hydrolysis also drives a conformational change in BTAF1 that has important implications for its molecular action (20,24,43). The elongated architecture of the thumb suggests molecular flexibility. One possibility is that ATP binding and/or hydrolysis drives a conformational change in the thumb that moves TBP-DNA into the groove of the globular carboxyl-terminal domain. This may allow the ATPase domain to interact with the TBP-DNA complex, thus forming a clamp that sandwiches DNA between TBP and the ATPase domain. DNA bending and/or TBP-TATA interactions could then be modified. Presumably ATP hydrolysis would release this interaction and open the clamp. Consistent with this hypothesis, domain-swapping experiments suggest that in Mot1p, ATP-driven dissociation requires the interaction of the ATPase domain with the TBP-DNA complex (43). Also, TBP and DNA stimulate the ATPase activity of BTAF1 (7). Furthermore, experiments altering functional groups in the major groove of TATA box DNA suggest that Mot1p contacts TATA DNA on the undersurface of the TBP-DNA complex (43). It is also conceivable that conformational changes are transmitted directly from the ATPase domain through the thumb to the amino terminus of BTAF1 and TBP-DNA. In light of these hypotheses it will be important to establish whether there is a link between an ATPase catalytic cycle and conformational changes by investigating structural differences in B-TFIID in various nucleotide states. Our results provide an important framework for this analysis. Insight into this would also be relevant for an understanding of the molecular mechanism of Mot1p and the more complex SNF2-like ATPase family members.