Molecular Mechanism of Mot1, a TATA-binding Protein (TBP)-DNA Dissociating Enzyme*

The essential Saccharomyces cerevisiae ATPase Mot1 globally regulates transcription by impacting the genomic distribution and activity of the TATA-binding protein (TBP). In vitro, Mot1 forms a ternary complex with TBP and DNA and can use ATP hydrolysis to dissociate the TBP-DNA complex. Prior work suggested an interaction between the ATPase domain and a functionally important segment of DNA flanking the TATA sequence. However, how ATP hydrolysis facilitates removal of TBP from DNA is not well understood, and several models have been proposed. To gain insight into the Mot1 mechanism, we dissected the role of the flanking DNA segment by biochemical analysis of complexes formed using DNAs with short single-stranded gaps. In parallel, we used a DNA tethered cleavage approach to map regions of Mot1 in proximity to the DNA under different conditions. Our results define non-equivalent roles for bases within a broad segment of flanking DNA required for Mot1 action. Moreover, we present biochemical evidence for two distinct conformations of the Mot1 ATPase, the detection of which can be modulated by ATP analogs as well as DNA sequence flanking the TATA sequence. We also show using purified complexes that Mot1 dissociation of a stable, high affinity TBP-DNA interaction is surprisingly inefficient, suggesting how other transcription factors that bind to TBP may compete with Mot1. Taken together, these results suggest that TBP-DNA affinity as well as other aspects of promoter sequence influence Mot1 function in vivo.

Snf2/Swi2 enzymes play critical roles in regulation of essentially all DNA metabolic processes by utilizing the energy of ATP hydrolysis to alter protein-DNA interactions (1)(2)(3)(4). Mot1, 2 a Snf2/Swi2 protein, dissociates the TATA-binding protein (TBP) from promoters and thereby regulates TBP distribution in the cell (5)(6)(7)(8)(9). By exploiting the Mot1 experimental system, biochemical analyses of the TBP-DNA dissociation reaction have provided general insight into mechanisms by which Snf2/Swi2 ATPases catalyze rearrangements of stable protein-DNA complexes (4). The N terminus of Mot1 interacts with TBP through its flexible HEAT repeats and in this way recruits Mot1 to TBP-DNA (7, 10 -14). The C-terminal region of Mot1, which comprises the ATPase domain, interacts with DNA upstream of the TATA box (13,14). Given the sequence and structural similarity among Snf2/Swi2 ATPase domains (15,16), a central goal in the field has been to decipher the mechanisms by which these enzymes couple ATP hydrolysis to dissociation or rearrangement of protein-DNA complexes.
The Snf2/Swi2 ATPases comprise a group within the SF2 helicase superfamily (17). Thus, their mechanisms are proposed to be similar to those of helicases except that they do not catalyze DNA strand separation (15,18). Nonetheless, the Snf2/ Swi2 family has well over a thousand members, and the great majority of them have not been characterized biochemically (19). Biochemical studies of helicases have revealed diverse mechanistic properties (3), suggesting that Snf2/Swi2 enzymes may likewise utilize the conserved ATPase domain in diverse and unexpected ways. In the case of Mot1, several catalytic mechanisms have been proposed that may not involve translocation. These include potential Mot1-mediated changes in TBP conformation and a proposed "spring-loaded" mechanism that exploits the severely bent DNA in the TBP-DNA complex (4, 20 -22). Moreover, recent work on related enzymes has shown that rigid coupling of the motor domain to the target complex is not required for catalytic activity (23,24). These observations challenge prior models for Mot1 action in which it functions by pulling or pushing TBP along DNA (4). We have used numerous approaches in an attempt to detect ATP-driven DNA translocation by Mot1 (14,25). The inability to detect DNA translocation by Mot1 could be due to its use of another mechanism, or it could use short range translocation (i.e. 1 or just a few base pairs) to dissociate TBP, and in this way translocation may have escaped detection. This provides the motivation for novel approaches to probe the Mot1 catalytic mechanism.
The ATPase domains are composed of RecA-like subdomains (referred to as 1A and 2A), which together can form the nucleotide binding pocket as a cleft between them. The ATPase domains also possess Snf2/Swi2-specific subdomains referred to as 1B and 2B that are thought to play roles in coupling ATP hydrolysis to DNA translocation (26 -28). A flexible linker allows the two RecA-like domains to adopt different conformational states with resulting changes in the orientations of critical residues in the ATP binding pocket. Crystal structures of SF2 ATPases, as well as related SF1 superfamily members, have revealed the presence of two types of conformational states, "open" and "closed," whose interconversion appears to be mediated by the binding and hydrolysis of ATP (26, 28 -30). Biochemical evidence for nucleotide-mediated conformational changes inferred from the structural data was obtained in fluorescence resonance energy transfer (FRET) studies using the Snf2/Swi2-related enzyme SsoRad54 (31). Importantly, emerging evidence provides support for the idea that the conversion between such conformational states is important for ATPase regulation (27).
There is some evidence that Mot1 exhibits conformational heterogeneity. Prior work suggests that at least two kinetically distinguishable types of Mot1-TBP-DNA complexes can be formed in solution in the absence of ATP (32). In addition, a point mutation in the linker connecting the ATPase subdomains affected the biochemical properties of the Mot1-TBP-DNA ternary complex (32). Formation of a closed conformation in response to nucleotide binding is also supported by recent cross-linking-mass spectrometry data (14). To gain more insight into the mechanism by which ATP hydrolysis is coupled to TBP-DNA dissociation, we used the Fe(III) (S)-1-(p-bromoacetamidobenzyl)ethylenediamine tetraacetic acid (FeBABE)-mediated hydroxyl radical cleavage assay to probe the ATPase conformational state during the reaction cycle. Helicases and translocases tend to primarily move along one of the two DNA strands, and for this reason ATP-mediated translocase activity can be blocked by discontinuities (single-stranded gaps) in the DNA (3,33). Therefore, in parallel, experiments were conducted using gapped DNA probes. These allowed us to determine which of the DNA bases in the required region were required for Mot1 binding and catalysis of TBP-DNA disruption. The combined results suggest that ATPdriven TBP-DNA dissociation by Mot1 involves at least two distinguishable steps: the conversion of an open ATPase state to the closed state induced by ATP binding and short range DNA translocation, which occurs subsequent to ATP hydrolysis and probably coincides with ADP or P i release. During the course of this work, we also found to our surprise that Mot1 is a catalytically inefficient enzyme, requiring multiple rounds of ATP hydrolysis to displace a single stably bound TBP molecule from the promoter. We discuss how this catalytic inefficiency relates to Mot1 function in vivo.

Results
Affinity of Mot1 for Gapped DNA Complexes-To investigate how the Mot1 ATPase utilizes DNA in the TBP-DNA dissociation reaction, functional studies focused on the DNA segment upstream of the TATA sequence, which had been shown in prior work to be important for Mot1 action (32,34). To determine the roles of DNA bases and backbone contacts in Mot1mediated catalysis, we tested a series of DNA templates with 3-base gaps in either strand (Fig. 1A). References to the "top" or "bottom" DNA strand that follow are with regard to the sequences as shown in Fig. 1A. To determine the contribution of these bases to Mot1 affinity, Mot1 was added in various concentrations to preformed TBP-DNA complexes, and after a 20-min incubation the reaction products were resolved on nondenaturing polyacrylamide gels (electrophoretic mobility shift assays (EMSAs)) such as the one shown in Fig. 1B. Equilibrium binding constants for Mot1 interaction with TBP-DNA were then measured by quantifying the Mot1-TBP-DNA complex formed (Fig. 1C). The measured affinity of Mot1 for TBP-DNA complexes formed using the fully duplex ("WT") probe is in good agreement with the affinity reported previously using TATA-containing probes with somewhat different sequences (32,35). Although the upstream DNA segment is required for stable association of Mot1 with TBP-DNA (32,34), surprisingly, the affinity of Mot1 for TBP-DNA was not much affected by 3-base gaps within this segment (Fig. 1D). This is consistent with nonspecific interactions between Mot1 and DNA being mediated by a flexible N-terminal Mot1 domain (13), which would allow for alternative interactions with DNA to compensate for loss of particular short sequences.
Differential Roles for DNA Bases in Mot1 Action-Next, we measured by EMSA the ability of Mot1 to displace TBP from the gapped DNAs using ATP. Representative results using the fully duplex WT probe are shown in Fig. 2A. Addition of Mot1 shifted TBP-DNA to the more slowly migrating Mot1-TBP-DNA complex position, and ATP resulted in Mot1-mediated depletion of both TBP-DNA and Mot1-TBP-DNA ternary complexes as has been reported previously (32,34). As shown in the EMSA, a small proportion of TBP-DNA complexes remained in the presence of Mot1 and ATP. Essentially all detectable TBP-DNA complexes can be cleared by Mot1 when using a competitor DNA to measure the Mot1-mediated TBP-DNA dissociation rate (32), but under the conditions used in Fig. 2A (addition of ATP for 5 min prior to loading onto the gel) cycles of Mot1-catalyzed TBP-DNA dissociation are antagonized by rebinding of TBP to DNA. On the WT probe, ϳ75-80% of the TBP-DNA complexes detected in the gel were disrupted in reactions with Mot1 and ATP; in contrast, many of the gapped probes supported less efficient TBP-DNA dissociation (Fig. 2B, pink bars). Gapped probes 1, 2, 3, 4, 5, 6, 9, and 10 were all at least partially defective for Mot1-induced TBP-DNA dissociation (p Ͻ 0.005). Probe 6 was unique in supporting a lower overall extent of Mot1 interaction in the absence of ATP compared with the other probes. This is consistent with a modest apparent decrease in the affinity of Mot1 for TBP-DNA complexes formed using this probe (Fig. 1D). An unanticipated observation is that probe 7 supported complexes that were significantly more efficiently cleared by Mot1 than those formed on the WT probe (Fig. 2B, green bar). That the gap in probe 7 is immediately opposite the gap in probe 1, which was defective for TBP clearance by Mot1, emphasizes the DNA strand specificity of Mot1 action. When the ternary complex levels were quantified, it was observed that for most of the DNA probes only about 15% of the Mot1-TBP-DNA complexes remained in reactions to which ATP had been added (Fig. 2C). However, the residual ternary complex levels in ATP-containing reactions were increased compared with the WT probe for complexes formed on gapped probe 1 (Fig. 2C, red bar). Ternary complex levels on probes 2 and 5 were apparently increased; the increase on probe 2 was not statistically significant (p ϭ 0.18), but the difference obtained using probe 5 (p ϭ 0.06) fell just short of reaching the customary level of statistical significance using p Ͻ 0.05 as the cutoff (Fig. 2C). For reasons discussed below, the possible defect in probe 5 compared with WT is potentially interesting. Probes 7 and 8 appeared to support ternary complex levels even lower than the WT probe in reactions with ATP, but these differences were not statistically different. The results in Fig. 2, B and C, suggest that gaps at different positions confer different biochemical defects. On most catalytically defective probes (probes 3, 4, 6, 9, and 10), ATP caused Mot1 to dissociate from the ternary complex with less efficient dissociation of TBP from DNA occurring as a result. This suggests that these gaps uncouple or partially uncouple the ATPase cycle from TBP displacement. On probe 1, however, significant proportions of the ternary complex appeared stable in the presence of ATP, suggesting that a gap at this position blocks a catalytic step that destabilizes both Mot1 and TBP association with DNA. Taken together, the results with the gapped templates indicate that bases along an 18-base stretch of the top strand are important for Mot1 function. In contrast, bases on the bottom strand required for Mot1 function fall within a much more discrete 6-base region.
To determine whether the impaired catalytic activity of gapped probes depended on the size of the gap, probes with a single base deletion or simply a strand nick were also tested. As shown in Fig. 2D, removal of the 5th base upstream of the TATA sequence or placement of a nick between the 5th and 6th bases upstream of the TATA sequence generated probes that supported Mot1 catalytic activity as well as the WT probe. These strand discontinuities fall within the 3-base gaps of the defective probes 1 and 2, arguing that disruption of Mot1 activity requires a single-stranded region of a minimum particular length as opposed to a simple break in one strand. The results from Fig. 2 are summarized in Fig. 2E. Nucleotides highlighted in pink show TBP-DNA dissociation defects, whereas red indicates defects in both TBP-DNA dissociation and Mot1 dissociation. The results with probe 5, similar to probe 1, are included in purple rather than red because of the ambiguous statistical significance mentioned above. Gap 7 bases are shown in green to highlight the improved TBP-DNA dissociation observed on this DNA. The most important result from these studies is that the DNA strands upstream of the TATA sequence are not equivalent in terms of their roles in the Mot1 catalytic cycle. Although a discrete region of the bottom strand is important, essentially the entire length of the top strand segment occupied by the ATPase is important for Mot1 catalytic function.
Mot1 Catalytic Efficiency-Mot1 destabilizes stable TBPpromoter interactions and thereby regulates transcription (11,32,34,35). However, recent work also showed that Mot1 pref- erentially increases the dynamics of less stably bound TBP molecules at diverse sites (36). Combined with the observation that perhaps a dozen or more ATP molecules could be hydrolyzed in the time it takes for dissociation of one high affinity TBP-DNA complex (32), these observations led us to investigate the efficiency of Mot1-catalyzed TBP-DNA dissociation using the immobilized template system. In standard assays in solution, it is difficult to assess Mot1 catalytic efficiency because the reactions contain a mixture of free and DNA-bound species, and thus the change in the proportion of TBP-DNA complexes in response to Mot1 could be due to potentially multiple cycles of Mot1 binding and ATP hydrolysis. To test Mot1 catalytic effi-ciency in preformed, purified complexes, TATA-containing DNA molecules were tethered to beads via a biotin linkage (Fig.  3A), and TBP and Mot1 were assembled on them. Fig. 3, B and C, show that, as expected, TBP binding to the beads was DNAdependent, and this association could be blocked by the addition of competitor DNA prior to the addition of TBP. Also as expected, Mot1 binding to the beads was largely dependent on DNA and TBP. This was true regardless of which DNA end was attached to the beads.
Next, we investigated the fate of TBP after the addition of Mot1 and ATP to immobilized TBP-DNA complexes. In contrast to what was routinely observed in solution, addition of  Fig. 1A). The level of TBP-DNA complex formed on each probe in the absence of Mot1 was set to 1, and the relative levels of TBP-DNA present in reactions containing Mot1 Ϯ ATP were normalized to those levels. Note that addition of Mot1 without ATP reduced the TBP-WT DNA complex level to about 30% of the level present in the absence of Mot1 coincident with the formation of the Mot1-TBP-DNA complex. In contrast, the extent of TBP-DNA complex formation was reduced to a comparable level when ATP was added because Mot1 catalyzed TBP-DNA complex disassembly. Gapped DNA probes impaired for ATP-dependent TBP-DNA dissociation are indicated by the pink bars (p Ͻ 0.005; those probes not significantly affected had p values Ͼ0.05). Complexes formed on probe 7 (green bar) were cleared to significantly greater extent than on the WT probe. For one probe (probe 6), the level of TBP-DNA complex formation was less affected by addition of Mot1 than for the other probes (asterisk; see text). C, relative Mot1-TBP-DNA complex formation measured in reactions containing the indicated gapped DNA probes, TBP, Mot1, and ATP compared with results obtained using the WT probe. Relative Mot1-TBP-DNA complex levels remaining in reactions containing ATP were higher using probes 1, 2, and 5 than when using WT DNA. The difference obtained using probe 1 was significant (p ϭ 0.04) using p Ͻ 0.05 as the cutoff. However, the difference obtained using probe 5 was just outside the 0.05 cutoff (p ϭ 0.065) and quite possibly also significant. Comparisons of results from all other probes with WT had p values Ͼ0.1. p values were determined using a two-tailed Student's t test. D, relative TBP-DNA complex levels present in reactions formed using fully duplex WT probe or probes with a 1-base gap or a single strand nick as indicated in the inset diagram. The strand discontinuity is located 5 bases upstream from the TATA sequence. E, summary of the main findings from EMSA analysis using the gapped DNA probes. Sequences highlighted in pink indicate gap positions that interfere with Mot1-mediated, ATP-dependent dissociation of TBP-DNA complexes. The sequence in red indicates the gap at position 1 that supports ternary complex assembly but is less affected by ATP addition than the other probes. Error bars represent S.D.
Mot1 and ATP did not cause any detectable dissociation of TBP from DNA (Fig. 3D, compare lanes 3, 5, and 7). EMSAs conducted in parallel showed that the Mot1 used was fully capable of displacing TBP from TBP-DNA complexes in solution as expected (32). These observations suggest that under these conditions Mot1 has poor TBP-DNA dissociation activity when associated with purified bead-bound complexes. One concern was that the high density of DNA on the beads might mask the TBP-DNA dissociation activity of Mot1 by facilitating rebinding of dissociated TBP molecules to other DNA molecules on the bead rather than allowing displaced TBPs to diffuse into the bulk solution.
To test the possibility that displaced TBP molecules rebind the bead-bound DNA, the same reaction was performed in the presence of TATA-containing competitor DNA not bound to the beads. As shown in Fig. 3D, addition of competitor DNA increased the efficiency of catalysis, possibly by preventing the rebinding of TBP after displacement, although the effect was not very robust (Fig. 3D, lane 6). Addition of competitor DNA along with additional Mot1 increased the proportion of TBP released from DNA still further (Fig. 3D, lane 8). Importantly, the competitor DNA was fully capable of blocking binding of TBP to the beads as shown in order-of-addition experiments (Fig. 3C), and the competitor DNA had no effect on either TBP-DNA or Mot1-TBP-DNA complexes in the absence of ATP (Fig. 3D, lanes 2 and 4 versus lane 6). When the fate of Mot1 was investigated, about 50% of the bound Mot1 dissociated from the beads upon ATP addition, whereas 50% remained bound to the complexes (Fig. 3E). Overall, under "optimal" conditions with both Mot1 and competitor DNA in solution and in the presence of ATP, we observed only ϳ40% of the DNA-bound TBP molecules displaced from the beads.
To better test the possible influence of the bead on the dissociation reaction, we added a 20-nucleotide single-stranded tail to the 5Ј end of the DNA duplex (5Ј tail; Fig. 3A). Fig. 3, F and G, show that the complexes formed on the tailed duplex behaved similarly to complexes formed on DNAs without the tail, arguing that the ability of Mot1 to dissociate TBP-DNA was not determined by proximity to the bead. Notably, TBP dissociation from tailed DNA was only detected in the presence of competitor DNA (Fig. 3F, lanes 6 and 8), and the addition of Mot1 in solution did not increase the removal of TBP as was observed on the complexes formed without the tail (Fig. 3D).
This difference suggests that Mot1 in solution facilitates dissociation of complexes formed on the shorter DNA segment compared with the 5Ј tail duplex. However, even when well removed from the bead by the 20-nucleotide tail, Mot1 remains a relatively inefficient enzyme, indicating that proximity of the complex to the bead cannot explain these results. Taken together, these data show that although ATP destabilizes the association of Mot1 with TBP-DNA, a single molecule of Mot1 bound to TBP-DNA has limited ability to dissociate high affinity TBP-DNA complexes without the aid of competitor DNA in solution and in some cases additional Mot1 molecules.
Structural Organization of the ATPase Domain on Upstream DNA-The FeBABE-mediated hydroxyl radical cleavage assay has been used previously to map the general position of the Mot1 ATPase domain along DNA (13). Using the immobilized template system, we extended this assay to obtain more precise localization information and to obtain deeper insight into the conformation of the Mot1 ATPase during steps in the catalytic cycle. FeBABE molecules were covalently coupled to specific sites in duplex DNA centrally located within the DNA segment that gap analysis showed was critical for Mot1 catalytic activity (Fig. 4A). Two FeBABE molecules were coupled near each other to increase the sensitivity of the assay (37). As in the prior experiments, biotinylated DNA probes were linked to streptavidin-conjugated agarose beads for rapid and quantitative separation of DNA-associated constituents from unbound FeBABE, TBP, or Mot1. In separate experiments, the functionality of the modified DNA probes for TBP and Mot1 binding and ATP-dependent TBP-DNA dissociation was confirmed by EMSA (data not shown).
A Mot1 Conformational Intermediate-As described above, Snf2/Swi2 ATPases have been crystalized in two different canonical conformational forms that differ dramatically in the relative orientations of the dual RecA folds that comprise the ATP binding pocket (26,28). In the closed form, the RecA subdomains are oriented appropriately for ATP binding in the cleft between them. In the alternative open conformation, one RecA subdomain is swiveled ϳ180°, and there is no ATP binding pocket. We hypothesized that Mot1 can adopt two similar conformations, that they are interconverted by nucleotide binding and release, and that they are intermediates on the catalytic path (32). This interconversion of RecA subdomains can alter the contacts between the ATPase and DNA, suggesting a molecular explanation for how ATP hydrolysis leads to DNA translocation (14,28) (38). To test this idea, we used the FeBABE cleavage assay to define the conformations of the Mot1 RecA subdomains vis-à-vis the upstream DNA. We compared the FeBABE cleavage pattern of Mot1 when bound to probes 6Fe, 4Fe, and 10Fe (Fig. 4A) in the presence and absence of ADP-AlF 4 , a transition state analog of ATP hydrolysis (39). As shown in Fig. 4B, a new cleavage product appeared in reactions containing ADP-AlF 4 and when using probe 6Fe (arrow). There were no significant differences in the cleavage patterns obtained with probes 10Fe and 4Fe. The appearance of this new cleavage site suggests that Mot1 undergoes a nucleotide-mediated conformational change that juxtaposes the cleaved region of the protein in proximity to the FeBABE molecules on DNA.
Top Strand DNA Is Required Subsequent to Reorganization of the ATPase Domain-Although bases close to the TATA box were important for the ATP-dependent displacement of TBP (e.g. gap probe 1), there was no difference in the Mot1 cleavage pattern with and without ADP-AlF 4 using probe 10Fe in which FeBABE molecules were positioned within this same region of DNA. To test the relationship between the inferred nucleotidemediated Mot1 conformational change and TBP displacement, FeBABE molecules were localized as in probe 6Fe, and a gap was introduced at the position of the gap in probe 1 (Fig. 4A). Remarkably, the new cleavage site induced by ADP-AlF 4 was present in complexes formed on probe 6gapFe as well as probe 6Fe (Fig. 4C, lane 2 versus lane 1, arrow). This result suggests that the top strand DNA specified by the gap is required at a catalytic step subsequent to the nucleotide-induced conformational change in the Mot1 ATPase.
Probe-specific Difference in ATPase Conformation-To determine whether DNA sequence can influence ATPase conformation, we tested a DNA probe called Moyle6Fe, which was used previously (40) and contains the same TATA sequence as probe 6Fe, but the upstream DNA sequence is different. We also compared the effect of ADP-AlF 4 with ADP-BeF 3 , a "ground state" ATP analog (41,42), using each of these two DNA probes. As shown in Fig. 5, A and B, using the 6Fe probe, the novel cleavage product (arrow) that was observed with ADP-AlF 4 was also observed using ADP-BeF 3 albeit to a lesser extent. In comparing the results obtained with the two DNA probes, although the overall cleavage patterns are similar, the cleavage product that was nucleotide-dependent on the 6Fe probe was present in complexes formed using the Moyle6Fe probe even in the absence of nucleotide. We conclude that Mot1 exists in two distinguishable conformational forms in the Mot1-TBP-DNA complex and that both nucleotide and DNA sequence can influence which conformational form predominates. Interconversion between conformational states was mediated by nucleotide binding when Mot1 was bound to the 6Fe probe, whereas on the Moyle6Fe probe the alternative conformational form was detectable even in the absence of an ATP analog.
As shown in Fig. 5C, all of the FeBABE-induced cleavage sites mapped to the Mot1 ATPase domain, and the site that differed depending on the DNA sequence or ATP analog mapped to domain 2B. As discussed in more detail below, we interpret the differential cleavage in domain 2B as indicative of its juxtaposition near DNA as is predicted in the closed state (Fig. 5D). In contrast, complexes formed on the Moyle6Fe probe appear to adopt a closed state conformation even in the absence of ATP analog.  Fig. 6, A and B. The interconversion between these conformational forms would involve the rotation of domains 2A and 2B by ϳ180° (26, 28 -30). As shown in the figures, formation of the closed conformation appropriately positions previously identified Mot1 catalytic residues (6,7,28) that are far apart in the open form. In the open form, the primary interactions between the ATPase and DNA are mediated by side chains in domain 1A (28), so when bound to DNA the closed complex formation most likely occurs by rigid body rotation of domain 2B (see supplemental Movies 1-3). To visualize the conformational change in the context of DNA, the N-terminal RecA domain structures in the open and closed models were aligned with the analogous structure in the SsoRad54-DNA co-complex (28), and the nucleotide-or DNA sequenceinduced cleavage site is highlighted in cyan (Fig. 6, C and D; and see supplemental Movies 1-3). The induced cleavage site mapped to a solvent-exposed ␣ helical hairpin and loop at the tip of domain 2B, which is expected to undergo large scale movement during the transition between the two forms. Notably, the cleavage site points away from DNA in the open form but projects into the DNA minor groove in the closed form, offering an explanation for why the site was detected on probe 6Fe in the presence of nucleotide analogs. The results further support the suggestion that DNA sequence variation of the Moyle6Fe probe somehow facilitates closed complex formation even in the absence of nucleotide.

Discussion
Roles for DNA Bases in Mot1 Catalysis-The results obtained with the gapped DNA probes indicate that probes with 3-base gaps on the top DNA strand were defective for Mot1-mediated catalysis (Fig. 2E). In contrast, most probes with gaps on the bottom strand were fully functional or, in the case of one gapped template, functioned even better than fully duplex DNA. The gapped template that functioned better than the fully duplex probe is notable as the position of the gap is imme- diately opposite a gap on the other strand that resulted in a template that was defective for Mot1 action. We suggest that the non-equivalent roles of DNA bases at the same position along the DNA length but on opposite strands argue for strand specificity in Mot1 action. Prior work provided evidence for Mot1-induced DNA unbending (14,40), which was suggested to "prime" the TBP-DNA complex for dissociation (Fig. 7). The strand specificity in gap effects indicates that the catalytic defects at most positions are not simply due to conformational flexibility in the upstream DNA imparted by the gap as a gap in either strand would accomplish this. However, the improved activity of Mot1 using a template with a gap at position 7 suggests that there are positions where DNA flexibility can facilitate DNA unbending without compromising the ability of Mot1 to dissociate TBP-DNA. In contrast, the gap immediately opposite it (probe 1) would be expected to increase flexibility, but in this case we suggest that the integrity of the top strand is required for Mot1 action so the probe shows a defect. The combined results suggest that the continuity of the top DNA strand is required for translocation along it. The gaps did not substantially affect the affinity of Mot1 for TBP-DNA, which is consistent with Mot1 binding being dictated mainly by a network of interactions formed between the Mot1 N-terminal domain and the convex surface of TBP observed in co-crystal structures as well as the flexibility of the Mot1 N-terminal domain, which would allow the ATPase to potentially adopt different orientations when docking DNA (4,13,43). Instead, the DNA gaps that impair activity apparently affect a catalytic step subsequent to Mot1 binding to TBP-DNA. By analogy with other ATPases in the SF2 ATPase/helicase family (3,33,44,45), the simplest interpretation of these results is that Mot1 uses ATP hydrolysis to translocate primarily along one DNA strand and this translocase activity can be impeded by single-stranded gaps pro-vided that the gaps are sufficiently long (i.e. 3 bases but not 1 base). The different effects of gaps at different positions along the DNA suggest that different DNA segments may be contacted by the ATPase at different steps of the catalytic cycle. To gain an understanding of how these DNA segments are used catalytically, we displayed them on a structural model of the TBP-DNA complex (Fig. 6E). The model shows in red the position of bases at the gap in probe 1, which was responsible for impaired Mot1 and TBP dissociation in the presence of ATP. Top strand gaps that conferred reduced TBP-DNA dissociation span the upstream DNA (i.e. distal to TBP) and are shown in pink. The probe 5 gap, which led to a defect in TBP-DNA dissociation and nearly reached statistical significance for a defect in ternary complex disassembly, is shown in purple. It is apparent from the model that these gaps define a broad surface associated with the top surface and one side of the upstream DNA. The red and purple gapped regions are well aligned on the top surface of DNA and separated by one turn of the helix.
This surface defined by analysis of gapped templates likely defines the approximate localization of the ATPase domain as well as the "footprint" of where Mot1 translocates during ATP hydrolysis. Three-dimensional reconstructions of the Mot1-TBP-DNA ternary complex have been obtained by analysis of single particles in electron micrographs. Models for the ternary complex obtained by docking an ATPase-DNA co-complex within these EM envelopes locate the ATPase in a roughly similar position (14,28). The FeBABE results presented here localize the ATPase to this same region, but a more refined placement cannot be determined based on the results presented here because FeBABE-mediated cleavage can reach a distance of ϳ12 Å (46 -48). By analogy with SF2 helicases, domains 1B and 2B in Swi/Snf enzymes are thought to couple ATP hydrolysis to rearrangements of their substrate protein-DNA complexes, including changes in DNA conformation (27). In this regard, the proximity to TBP and DNA of the Mot1-specific domain 2B in the closed state is notable as domain 2B is then well positioned to transmit ATP hydrolysis to an interaction with TBP-DNA, resulting in dissociation.
Biochemical studies of other Snf2/Swi2 ATPases showed that ATP hydrolysis can induce translocation along one DNA strand in the 3Ј to 5Ј direction (33,45). This is consistent with the translocation direction of SsoRad54 along DNA proposed based on the co-crystal structure (28). When combined with the Mot1-TBP-DNA ternary complex model described previously, the presumed 3Ј to 5Ј translocation direction would suggest that the Mot1 ATPase translocates toward the TBP-DNA complex to displace it (14). Although our results do not provide clear evidence for translocation along one strand, as discussed above, the integrity of the bottom strand is not essential for Mot1 action, whereas the integrity of the entire top strand appears critical. It has been pointed out that ATPase motor movement and directionality are governed by chemical properties of the ATP hydrolysis cycle and interactions of the macromolecules, not by mechanical properties of the motors as commonly assumed (49). Regardless of the behavior of related ATPases, it is therefore conceptually possible that the Mot1 ATPase moves along one strand in the 5Ј to 3Ј direction (i.e. toward TBP) or even both directions, and in fact the relative inefficiency of Mot1-mediated TBP-DNA dissociation observed here may be due to thermodynamic accessibility of other "unwanted" reactions or processes that do not lead to destabilization of the TBP-DNA complex. An obvious explanation for why this occurs in our in vitro system is that in vivo Mot1 mainly targets TBP-DNA complexes that are less stable than those formed on high affinity TATA sequences. This suggestion is supported by recent evidence showing that Mot1 targets less stably bound TBP molecules associated with intergenic and transcribed regions (36). The extent to which TBP molecules bound to non-promoter sites in mot1 mutant cells contribute to inter-or intragenic transcription has not been clearly established, although the spectrum of altered RNA lengths in mot1 cells suggests that Mot1 plays a role in preventing aberrant initiation from such sites (50). However, to date, we have not defined a biochemical system for studying the effects of Mot1 on DNA sequenceselective interactions between TBP and such weak DNA sites in vitro. Alternatively or in addition, Mot1 activity may be more efficient when it operates in conjunction with other factors. This could explain why mutation of Mot1 leads to increased TBP occupancy at presumed high affinity sites (i.e. TATA sequences) in vivo (51). NC2 is a good candidate for a factor that impacts Mot1 function in vivo. Mot1 and NC2 regulate many of the same genes and co-occupy many of the same promoters (52)(53)(54)(55)(56). It has been proposed that the NC2containing TBP complex is a physiologically relevant Mot1 target (57), although a significant difference in Mot1 biochemical behavior in the presence of NC2 has not been detected thus far (10,14).

DNA Sequence Effects on Mot1
Conformation-An unexpected finding reported here is that the conformation of the Mot1 ATPase can be influenced by DNA sequence. Our observation of the open form of Mot1 in 6Fe DNA complexes is consistent with the open form of the ATPase in the SsoRad54-DNA co-complex (28). In contrast, in solution, interconversion between conformational forms of Swi/Snf ATPases has been observed (31,58), and the closed form predominated when bound to DNA in solution (31), consistent with results using the Moyle6Fe template. We favor a model in which the Mot1 open and closed forms both exist in solution in the absence of ATP. ATP binding then shifts the equilibrium to the closed form. In support of this, kinetic analysis uncovered evidence for two types of Mot1-TBP-DNA complexes in solution in the absence of ATP; one was notably less stable than TBP-DNA, and one was more stable (32). Because DNA sequence can influence Mot1 conformation, the combined results suggest that Mot1-TBP-DNA complexes can have differential stability dictated by the promoter sequence. Although the function of Mot1 in gene activation is not fully understood, the presence of Mot1 in complexes at active promoters (53,55,59) suggests that its dissociation activity may be regulated. We suggest that the influence of DNA sequence on Mot1 conformation may be relevant for control of Mot1 activity in vivo, possibly helping it to distinguish active versus repressed promoters.
Mot1 Function in Isolated Complexes-Given the wealth of biochemical data on the ATP-dependent catalytic activity of Mot1, it was surprising to find that when ATP was added to preformed, immobilized Mot1-TBP-DNA ternary complexes Mot1 was unable to efficiently dissociate them. The requirement for nanomolar concentrations of Mot1 in solution and/or competitor DNA in addition to ATP suggests that either more than one catalytic event or more than one molecule of Mot1 was required to effectively displace TBP from DNA. This apparent inefficiency would not have been observable in our ensemble biochemical reactions in solution in which multiple rounds of binding and dissociation can occur over the reaction time course. A requirement for multiple ATP hydrolysis events to dissociate one high affinity TBP-DNA complex would be consistent with the stoichiometry of ATP hydrolysis to dissociation suggested previously (32). As ATP induced some Mot1 to dissociate from TBP-DNA without dissociating TBP from DNA, it is possible that having Mot1 in solution is the key to better catalysis. Alternatively, although complexes with more than one molecule of Mot1 have not been identified, an additional molecule(s) of Mot1 may transiently participate in the dissociation reaction. Some members of the helicase superfamily function as monomers, and others function as multimers. Mot1 is an abundant protein in vivo with at least one-third the number of TBP molecules in yeast and human cells (60,61). Taken at face value, this suggests that in vivo either Mot1 is not a very efficient catalyst or that it has other roles that require widespread and stoichiometric interaction with other components of the transcription apparatus.

Experimental Procedures
TBP and Mot1 Purification-Full-length recombinant Saccharomyces cerevisiae TBP was obtained by purification from a

Mechanism of Mot1
bacterial overexpression strain as described previously (10,62) and quantified as described previously (32). Full-length recombinant S. cerevisiae Mot1 was isolated by affinity purification using a yeast overexpression strain as described previously (10,32).
DNA Probes-The synthetic deoxyoligonucleotides (oligos) used in this study are listed in supplemental Table S1. The gapped DNA probes were prepared by annealing combinations of the oligos corresponding to the sequences shown in Fig. 1A. For each gapped probe, a full-length top or bottom strand was annealed to two complementary shorter oligos to leave missing sequences as indicated. In each case, the shorter non-TATAcontaining oligo was labeled, and the labeled duplex was gelpurified, ensuring that any complexes observed by EMSAs contained all three oligonucleotides.
The oligos used for FeBABE assays (obtained from Life Technologies) were biotinylated on the 5Ј end of the top strand and hybridized to unmodified bottom strand oligos. The duplexes were formed in 25 mM Tris-Cl, pH 8, 10 mM EDTA, 0.1 M NaCl by heating to 100°C for 5 min and then slowly cooling overnight to room temperature. A 2-fold excess of the bottom strand was used during annealing to increase the yield of the duplex product.
For testing the catalytic activity of Mot1, oligonucleotides were biotinylated at the 5Ј end (upstream) or 3Ј end (downstream) of the TATA box. A 2-fold excess of the unmodified strand was added, and annealing conditions were the same as those for the FeBABE DNAs. A 20-nucleotide stretch of singlestranded DNA was added to the 5Ј end of the 5Ј Bio duplex to create the 5Ј tail duplex.
FeBABE-mediated Hydroxyl Radical Cleavage Assay-Six millimolar FeBABE (Dojindo, F279-10) was coupled to 0.7 M oligonucleotide hybrid (via specific phosphorothioate modifications) in 20 mM MOPS, pH 7.9, for 16 h at 50°C. The FeBABE-coupled oligonucleotide hybrid was then bound to streptavidin beads (Dynabeads M-280 Streptavidin, 112.05D, Invitrogen) using the biotin linker on the 5Ј end of one strand. The beads were then washed twice with 20 mM MOPS, pH 7.9, and once with the reaction buffer (4% glycerol, 4 mM Tris-Cl, pH 8, 60 mM KCl, 5 mM MgCl 2 , 100 mg/ml BSA). Mot1-TBP-DNA complexes were formed using buffer conditions identical to those used for EMSA described below. TBP was added to 20 nM final concentration and incubated on a rotator at room temperature for 20 min. The beads were then pulled down using a magnet, and the supernatant (containing unbound TBP) was removed. The beads were then resuspended in reaction buffer, and Mot1 was added to 8 nM final concentration followed by incubation at room temperature for 30 min. After this incubation, the beads were pulled down again and washed with reaction buffer without DTT. The beads were then resuspended in 7.5 l of reaction buffer without DTT, and 2.5 l of 50% glycerol was added as described by Chen and Hahn (63). To the reactions that contained ADP-AlF 4 , 1 l of 2 mM ADP, 0.5 l of 50 mM NaF, and 0.1 l of 9 mM AlCl 3 were premixed, and the mixture was added. To the reactions that contained ADP-BeF 3 , 1 l of 2 mM ADP, 0.1 l of 50 mM NaF, and 0.12 l of 9 mM BeCl 2 were premixed, and the mixture was added. The addition of either analog mixture was followed by a 30-s incubation before initiating the hydroxyl radical cleavage reaction. Hydroxyl radical cleavage was initiated by the addition of 1.25 l of 50 mM ascorbate to reduce the Fe 3ϩ to Fe 2ϩ followed by the addition of 1.25 l of 50 mM H 2 O 2 , 10 mM EDTA. After incubating for 5 min at 37°C, the reaction was stopped using 6 l of 4ϫ Laemmli protein gel sample buffer and 1 l of 1 M DTT. The products of the cleavage reaction were then resolved on 14% denaturing gels and analyzed by Western blotting using an antibody specific for the C terminus of Mot1 (7,32,34).
Mapping FeBABE Cleavage Sites-The molecular weights of the bands on the Western blots were calculated by measuring their migration distance from the well of the gel relative to molecular weight standards. The average molecular weight of a band was estimated from at least three independent blots, and the cleavage sites were mapped to the primary sequence using these molecular weight estimates.
EMSAs-Detection of TBP and Mot1 binding to DNA by native gel electrophoresis was conducted as described (34) using a radiolabeled fragment of the adenovirus major late promoter (Ͻ1 nM) and the concentrations of TBP and Mot1 as indicated in the figure legends. Where indicated, ATP was added to 50 M. Band intensities were quantified using a Typhoon phosphorimaging system and ImageQuant software (GE Healthcare). The data presented were obtained by averaging the results of at least two independent experiments.
Immobilized Template Assays-Oligonucleotide duplexes were coupled to streptavidin magnetic beads and washed as described above for the FeBABE experiments. TBP-DNA and Mot1-TBP-DNA complexes were formed under the same reaction conditions as described above. The complexes were washed once with reaction buffer and split into separate reaction tubes. Each reaction was then resuspended in 20 l of reaction buffer (the same conditions as described for EMSAs). For reactions in which additional Mot1 was added in solution, 2 l of 382 nM Mot1 was added. ATP was added to the indicated reactions to a final concentration of 100 M for 5 min at room temperature. Where indicated, a 31-bp competitor DNA fragment containing a TATA box sequence was added to a final concentration of 7.5 ng/l. Then the reactions were again pulled down using a magnet, and the supernatant was carefully removed. The beads were resuspended in 20 l of reaction buffer, and half the volume was used for the gel. The beads and supernatant were then added to 2ϫ sample buffer, boiled, loaded onto 12 or 10% SDS gels, and analyzed by Western blotting for TBP and Mot1 using ␣-TBP (64) and ␣-Py (7) antibodies, respectively. The same procedure was used for DNA probes biotinylated on the downstream end to test the effect of beads coupled to the end downstream of the TATA box.
Molecular Modeling-To visualize the biochemical results in the context of the TBP-DNA complex, several molecular models were developed. Models of the Mot1 ATPase in the open and closed forms were obtained by threading the Mot1 ATPase sequence through published structures of Snf2/Swi2 ATPase domains. The open form model was obtained using the Sso-Rad54 structure (Protein Data Bank code 1Z6A) and CPHmodels (65). The closed form model was obtained using the Danio rerio Rad54 structure (Protein Data Bank code 1Z3I) and PROTEUS2 (66). A model of the TBP-DNA complex was adapted from the S. cerevisiae TBP-DNA co-crystal structure (67) in which standard B-form duplex DNA was appended to TATA-containing DNA in the structure. The image in Fig. 6E was obtained by mapping the biochemical data onto the TBP-DNA structure using Chimera (68). Images shown in Fig. 6, A-D, were obtained using MacPyMOL 2 (Schrödinger, LLC).
Author Contributions-R. V., J. D. T., and D. T. A. conducted the experiments and wrote the manuscript.