Comparison of TATA-binding protein recognition of a variant and consensus DNA promoters.

Assembly of transcription pre-initiation complexes proceeds from the initial complex formed between "TATA" bearing promoter DNA and the TATA-binding protein (TBP). Our laboratory has been investigating the relationships among TATA sequence, TBP center dot TATA solution structure, recognition mechanisms, and transcription efficiency. TBP center dot TATA interactions have been modeled by global analysis of detailed kinetic and thermodynamic data obtained using fluorimetric and fluorometric techniques in conjunction with fluorescence resonance energy transfer. We have reported recently that TBP recognition of two consensus promoters, adenovirus major late (AdMLP: TATAAAAG) and E4 (TATATATA), is well described by a linear two-intermediate mechanism with simultaneous DNA binding and bending. Similar DNA geometries and high transcription efficiencies characterize these TBP x TATA complexes. Here we show that, in contrast to the consensus sequences, TBP recognition of a variant sequence (C7: TATAAACG) is described by a three-step model with two branching pathways. One pathway proceeds through an intermediate having severely bent DNA, reminiscent of the consensus interactions, with the other branch yielding a unique conformer with shallowly bent DNA. The resulting TBP x C7 complex has a dramatically different solution conformation than for TBP x DNA(CONSENSUS) and is correlated with diminished relative transcription activity. The temperature dependence of the TBP x C7 helical bend is postulated to derive from population shifts between the conformers with slightly and severely bent DNA.

Our laboratory has been studying the detailed recognition mechanism of DNA promoters by S. cerevisiae TBP (15,16,20) and the solution geometries of the resulting TBP⅐TATA complexes (34,35). Both lines of investigation utilize dye-labeled TATA-bearing oligomers and steady-state, stopped-flow, and time-resolved fluorescence techniques in conjunction with fluorescence resonance energy transfer (FRET). Global analysis of extensive real-time kinetic and thermodynamic data sets first revealed a linear three-step mechanism for the TBP⅐AdMLP reaction (16), with intermediate conformers having DNA bent to the same extent as in the final complex. These conformers are present at high mole fraction throughout the reaction and persist at equilibrium. The parallel investigation of TBP⅐DNA solution structures concurrently revealed a strong sequence dependence of the DNA helical bend in such complexes (34). Because the latter suggested TATA sequence-dependent recognition interactions, the mechanistic study was extended beyond AdMLP to include other sequences.
We first asked the question, "How do the detailed recognition processes of different consensus sequences by TBP compare?" The E4 sequence was chosen as the alternate consensus sequence, because functional differences between the TBP⅐AdMLP and TBP⅐E4 interactions had been identified previously using DNase I footprinting (18,19). Extensive data collection and model fitting using global analysis showed the linear two-intermediate model to be common to the reaction of TBP with both strong promoters (20). The energetics of the partial reactions for the two promoters was very similar for the initial step, formation of the first intermediate conformer. Beyond that initial binding and bending event, however, the reaction progression differed substantially, with the TBP⅐E4 interaction nearly complete after the second step but with the TBP⅐AdMLP reaction continuing to undergo large energetic changes in the final step.
Having thus established a detailed comparison between the interactions of TBP with two consensus sequences, we have now examined TBP recognition of a variant sequence. The C7 sequence (TATAAACG) is a naturally occurring single base variant of the AdMLP sequence. TBP-bound C7 has a helical bend dramatically different from that in bound AdMLP that correlates with significantly reduced relative transcription efficiency (34). This solution geometry is highly sensitive to the presence and concentration of osmolytes (35), in contrast to that for the complexes bearing consensus sequences. Detailed kinetic and energetic profiles for the TBP⅐C7 interaction have been determined using FRET-based measurements. A comprehensive comparison with the TBP⅐AdMLP and TBP⅐E4 profiles clearly distinguishes the recognition process by TBP of the variant sequence from those of the consensus sequences. The ensemble of TBP⅐C7 data is well described by a three-step model with two branching pathways. One of these pathways is unique to the variant sequence mechanism and yields a bound species with DNA bent only slightly. Along the other pathway, the variant reaction proceeds through an intermediate conformer bearing strongly bent DNA in a process remarkably reminiscent of the consensus reactions. The temperature dependence of the TBP-bound C7 solution bend angle is proposed to arise from population shifts between conformers with slightly and severely bent DNA within a two-state model.

EXPERIMENTAL PROCEDURES
DNA, Protein, and Solution Conditions-Fourteen base fluorescently labeled DNA oligonucleotide probes and the unlabeled complementary strands were synthesized and purified by Sigma-Genosys (The Woodlands, TX) as described previously (16,20,35). TAMRA and fluorescein were linked covalently to the 5Ј-and 3Ј-ends, respectively, via six carbon linkers to form the double-labeled C7 top strand, TAMRA-5Ј GGGCTATAAACGGG 3Ј -fluorescein (duplex denoted as T*C7 dpx *F) with the single-labeled oligonucleotide having only the 3Јfluorescein (duplex denoted as C7 dpx *F). All duplexes were formed using a 2ϫ excess of complement. Full-length S. cerevisiae TBP was prepared as described previously (15,18). Studies were conducted in 10 mM Tris-HCl (pH 7.4), 100 mM KCl, 2.5 mM MgCl 2 , 1 mM CaCl 2 and 1 mM dithiothreitol at the temperatures indicated. Determination of equilibrium isotherms at 15, 20, 25, and 30°C is described in the legend of Fig. 1.
Theory and Inter-dye Distance Distributions of Free and TBP-bound C7-The interaction of TBP with T*C7 dpx *F was monitored using fluorescence resonance energy transfer (FRET) together with various fluorescence techniques. Detailed discussions of FRET and its applications to the present study have been published (15, 16, 34, 36 -42). Briefly, the rate constant for the non-radiative transfer of excited state energy from a donor to an acceptor fluorophore is highly dependent upon the distance separating the two dyes. Because the binding of TBP to a TATA-bearing promoter induces a bend in the DNA helical axis, binding to T*C7 dpx *F results in a decrease in the 5ЈTAMRA-3Ј fluorescein distance. The corresponding change in the efficiency of energy transfer is reflected in the fluorescence emission of the dyes and may be used to monitor these interactions with TBP in real time.
For non-rigid molecules such as the fluorescently labeled oligomeric probes used in these studies, the 5Ј-end to 3Ј-end distance is described by a probability distribution, P(R), rather than by a single fixed value. Such distributions were extracted from time-resolved fluorescence emission measurements of free and TBP-bound T*C7 dpx *F at 15, 20, 25, and 30°C as described (34). Model-dependent bend angles for the bound duplex were then determined as described (34,35).
TBP⅐C7 Association Binding Kinetics-The association binding of TBP with T*C7 dpx *F has been investigated as described for TBP binding the adenovirus major late (AdMLP) and E4 promoters (16,20) using stopped-flow fluorimetry. Briefly, solutions of 40 nM T*C7 dpx *F were flowed against TBP solutions of 400, 800, or 1600 nM (prior to mixing), using a 1:1 mixing ratio, at 15, 20, 25, and 30°C. Averaged replicate curves were fit using exponential decay models. The resulting set of eight curves, each deriving from a unique combination of temperature and TBP concentration, was used in global analyses.
TBP⅐C7 Dissociation Kinetic Measurements and Analysis-TBP⅐T*C7 dpx *F complexes were challenged with a large excess of unlabeled C7 duplex (C7 dpx ), and dissociation was monitored using steady-state FRET as described (16,20). The initial complex incorporating labeled DNA was formed and equilibrated using 19.2-20.3 nM T*C7 dpx *F and 992.3-1004 nM TBP, the latter sufficient to ensure Ն94% duplex saturation at all temperatures. Dissociation kinetic data were obtained as a function of unlabeled DNA concentration, with C7 dpx added to final concentrations of 17.9, 24.8, or 35.5 M. Data obtained at 15, 20, and 25°C were all well fit using a biexponential decay model as described (16,20).
A dependence of the TBP⅐T*C7 dpx *F dissociation kinetics on C7 dpx concentration was observed, indicating two contributions to the dissociation process: first-order replacement of labeled by unlabeled duplex and active removal, or second-order displacement, of labeled duplex. Similar complexity has been reported for TBP⅐T*E4 dpx *F dissociation kinetics (20). Extraction of the pure replacement curve at each temperature was necessary, because only that process is considered in subsequent global analysis. Two model-independent methods (20) were used for this procedure. The recovered replacement kinetic curves at 15, 20, and 25°C were subsequently used in the global analyses.
Global Analyses-The eight stopped-flow binding and three replacement kinetic curves and four equilibrium binding isotherms characterizing the TBP⅐C7 interaction were analyzed globally to explore their correspondence with various kinetic models. Analyses based on twoand three-step models, including both linear and branching schemes, are described below. To ensure meaningful comparisons with the previously determined results for TBP binding to consensus sequences, analyses were conducted in a manner analogous to those for TBP⅐AdMLP and TBP⅐E4 (16,20).
Smoothed representations of the association and replacement curves and equilibrium binding isotherms were constructed as described (16,20). The latter were scaled from 0 to 0.127 (15°C), 0.216 (20°C), 0.301 (25°C), or 0.335 (30°C), the temperature-dependent amplitude changes observed upon TBP⅐T*C7 dpx *F binding in steady-state FRET measurements. The theoretical response functions were determined as described (16,20). The overall average squared residual, global 2 , derived from the weighted variance between the observed and theoretical points describing the shapes of the association ( SF 2 ), replacement ( R 2 ), and equilibrium binding ( K 2 ) curves, with the coefficients reflecting the relative information content of each term. A fit with each term within its experimental error thus yields a value for global 2 Յ 1. Error estimates for the optimal parameter values were obtained exactly as described (16).
The correspondence of the ensemble of TBP⅐C7 data was first tested against linear (Equation 2) and branching (Equation 3) two-step models, where the subscript "INT" denotes an intermediate conformer.
Global analysis for either model yielded values for the four rate constants at 30°C as well as the corresponding activation enthalpies. The quantum yields of the donor fluorescein in the two TBP-bound duplexes, relative to that of free T*C7 dpx *F, were also determined. These quantum yield values indicate the relative extent of DNA bending in each binary complex, as they reflect the 5Јdye-3Јdye distance (15,16,34).
The data were further tested against various three-step mechanisms. First explored was a linear two-intermediate ki- shown previously to describe both the TBP⅐AdMLP and TBP⅐E4 reactions (16,20). Determined in the analysis were values for the six rate constants at 30°C, the corresponding activation energies, and the quantum yield for each TBP-bound conformer relative to that of the free duplex. To reduce the number of fitting parameters, additional analyses were conducted with quantum yields constrained to be equivalent for either conformers A and B or conformers B and C, consistent with a two-state model for this interaction (34,35).
Two additional three-step models were explored, with two and three branching pathways, with the rate constants k 2 and k 3 in Equation 5 and k 2 , k 3 , and k 5 in Equation 6 being second-order. Relative quantum yield values were determined for each TBP-bound conformer, and additional analyses conducted with constraints on the quantum yields as for Equation 4. Finally, Equation 5 was extended to allow direct conversion between TBP⅐DNA A and TBP⅐DNA B, INT via rate constants k 7 and k 8 .
End-to-End Distance Distributions, P(R), and TBP-bound C7 Solution Bend Angles-The parameters describing the probability distribution of the 5Јdye-3Јdye distance for free T*C7 dpx *F were R free ϭ 54.7 Ϯ 0.1 Å and free ϭ 6.8 Ϯ 0.2 Å. These values were invariant over the experimental temperature range, consistent with the observed temperature independence of the T*C7 dpx *F steady-state spectrum. In contrast, the values of R bound and bound were temperature-sensitive. These values are shown in Table I together with the corresponding bend angles. Successive decreases in R bound with increasing temperature indicate a temperature-dependent TBP⅐C7 conformational change. The change in the inter-dye distance from 53.1 Å at 15°C to 50.4 Å at 30°C corresponds to an increase of ϳ1.7-fold in the DNA bend angle. As the bend angle for TBP⅐C7 increases with temperature, a corresponding decrease in bound is observed, consistent with similar results from sequence-and osmolyte concentration-dependent studies. This inverse correlation is attributed to increasing compatibility of duplex and protein structures at the interface, restricting the DNA helical mobility (34,35).
TBP⅐C7 Association Kinetics-The binding and bending of T*C7 dpx *F by TBP was monitored in real time as a function of temperature and protein concentration using fluorescence stopped-flow and FRET. A total of 40 curves were collected, yielding a set of eight averaged curves, each a unique combination of temperature (15-30°C) and TBP concentration (200 -800 nM). All curves were very well represented by biexponential decay (15,16). The overall change in amplitude at each temperature was consistent with the corresponding steady-state change obtained for the binding isotherms. The association of TBP with C7 is notably slower and more biphasic than with either consensus promoter ( Fig. 2 (16, 20)).
TBP⅐C7 Dissociation Kinetics-A solution of TBP⅐T*C7 dpx *F complex was challenged with high concentrations of unlabeled C7 duplex and dissociation from all bound species monitored via changes in steady-state emission. The dependence of the TBP⅐C7 dissociation kinetics on the unlabeled DNA concentra-FIG. 1. Equilibrium binding isotherm for C7 with TBP at 25°C. The fraction of bound TBP at each titration point is calculated directly from the change in the steady-state emission spectrum (40). The van't Hoff plots shown in the inset are constructed from the experimentally derived K a values for C7 (q), AdMLP (f), and E4 (OE) and yield values for ⌬H 0 of 17.1 Ϯ 3.9, 16.9 Ϯ 2.2, and 25.1 Ϯ 1.7 kcal mol Ϫ1 , respectively. The steady-state spectrum of TBP-bound T*C7 dpx *F changes significantly with temperature from 15 to 30°C, whereas the free T*C7 dpx *F spectrum does not. The donor-only spectrum changes slightly upon addition of TBP, independent of temperature. That the change observed in the donor-only fluorescence upon addition of TBP is independent of temperature suggests that those changes observed for TBP-bound T*C7 dpx *F with temperature derive not from trivial changes in static quenching but, rather, from energy transfer effects, reflecting a temperature-dependent DNA conformational change. a A total of 48 decay curves, each one an average of three curves, were collected and analyzed. All curves were well described by biexponential decay, with a mean value for 2 of 1.00 Ϯ 0.07. Subsequent analysis was performed to determine P(R) modeled as a shifted Gaussian (40). P(R) values were refit to P(R) bound ϩ P(R) free , weighed using K a , to correct for the small amount (Ͻ5%) of free duplex.
(Eq. 5) (Eq. 6) tion revealed both passive replacement and facilitated displacement processes, as previously observed for TBP⅐E4 but not TBP⅐AdMLP (20). The pure replacement curve was extracted from the set of decays collected at each temperature using two procedures. A Taylor series expansion about an intermediate C7 dpx concentration yielded results that were essentially identical to those obtained from a second method, in which linear plots of experimental values of ␣ fast , fast , and slow versus [C7 dpx ] were extrapolated to [C7 dpx ] ϭ 0.
The nature of the biphasicity characterizing the kinetics of C7 replacement from the TBP-bound complex differs markedly from that of the consensus sequences (Fig. 3). The slow phase eigenvalue, 0.0010 Ϯ 0.0002 s Ϫ1 , and its essential invariance with temperature are nearly identical with slow for AdMLP and E4 replacement. However, this phase ranges in amplitude from only 11 to 17%, in sharp contrast to the dominance of the slow phases for AdMLP and E4, which range from 82 to 88% and from 66 to 80%, respectively.
The Entirety of the TBP⅐C7 Data Is Well Described by a Three-step Model with Two Branching Pathways-The correspondence of the TBP⅐C7 thermodynamic and kinetic data to numerous kinetic models was explored. The multiphasic character of the association and dissociation kinetics and the inequality between the measured equilibrium constants and the ratio of the forward and reverse rate constants rule out a simple single-step reaction. Global analyses demonstrated a lack of correspondence of the data with linear and branching two-step models, in Equations 2 and 3, yielding average weighted residuals ϳ2-5ϫ the experimental error.
Various three-step mechanisms were then examined. The linear two-intermediate model (Equation 4) and the model with three branching pathways (Equation 6) both failed to accommodate the entire data set. Although good correspondence to both models was obtained using only the eight association and four equilibrium binding curves, inclusion of the three replacement curves resulted in weighted residuals ϳ2ϫ the experimental error. Reduction in the number of parameters via additional constraints on the quantum yields did not improve the quality of these fits.
The simplest model accommodating the entirety of the TBP⅐C7 data was that given by Equation 5, a three-step branching model incorporating one intermediate species. Glo-bal analysis to this model yielded a global ϭ 0.96 and weighted residuals within experimental error for the association, replacement, and equilibrium binding curves (Fig. 4). The optimal kinetic and thermodynamic values corresponding to this model are shown in Table II. The relative quantum yield values for the bound species were obtained first using the constrained analysis and did not change in the subsequent unconstrained analysis (see "Experimental Procedures"). Notably, two distinct relative quantum yields were obtained, 0.985 for species A in having strongly bent DNA. Adding an additional step to Equation 7 for direct conversion between TBP⅐DNA A, SHALLOW and TBP⅐DNA B, SEVERE yielded values for the two additional rate constants that were very small relative to those for k 1 through k 6 . The values for the latter and the corresponding activation enthalpies remained essentially unchanged, as did the quality of the fit. This more complex model was therefore eliminated, because a good fit could be achieved with a negligibly slow conversion process.
The relative fraction of each species along the reaction pathway and at equilibrium for Equation 7, determined using the microscopic rate constants and a saturating TBP concentration, are shown over the experimental temperature range in Fig. 5. TBP⅐DNA A, SHALLOW is the dominant complex at 15°C and persists in high mole fraction at equilibrium. Nearly equivalent equilibrium populations of TBP⅐DNA A, SHALLOW and TBP⅐DNA B, SEVERE are present at 20°C, whereas TBP⅐ DNA B, SEVERE is the dominant reaction species at the higher temperatures. TBP⅐DNA C, SEVERE is present at relatively low concentrations with a maximum mole fraction at equilibrium of 18% at 30°C.
Thermodynamic Pathway for the TBP⅐C7 Reaction Pathway-The thermodynamic profile for TBP⅐C7 association is shown in Fig. 6

DISCUSSION
Extensive kinetic and thermodynamic data sets describing the interaction of TBP and C7 have been shown to correspond within experimental error to a three-step branching model. The consequent energetic and kinetic profiles reveal the partial reactions that constitute the recognition mechanism of TBP for this variant promoter sequence. This work makes possible a detailed comparison of TBP interactions with the variant C7 sequence and two strong consensus promoter sequences, AdMLP and E4, characterized previously.
The interactions of S. cerevisiae TBP with the consensus AdMLP and E4 promoters have been shown previously to be accommodated by a three-step linear model (Equation 4) (16,20). This model cannot accommodate the TBP⅐C7 interaction. Features common to both consensus reactions include a linear progression through two intermediates having fully bent DNA, the first intermediate species present at high mole fraction throughout the reaction and at equilibrium, and the second intermediate present at low mole fraction. The reactions proceed through very similar entropic and enthalpic changes during the first step but differ significantly for the second and third steps. These changes lead ultimately to very similar TBP⅐DNA structures in solution and co-crystals, both with the DNA helix severely bent (23-28, 34, 35).
How do the partial recognition reactions compare for the variant and consensus sequences? Two striking differences present themselves. Most importantly, TBP binding with the C7 variant yields complexes having DNA with different bends, one with DNA bent to only approximately half the extent of that in the other complexes. This result contrasts with the consensus reactions in which TBP induces exclusively the final, severe DNA bend in all bound species. In addition, the variant interaction proceeds via two branching pathways, one in which TBP⅐DNA A, SHALLOW is formed directly rather than through an intermediate species.
On the other hand, the interaction with the variant sequence along the other branch is remarkably reminiscent overall of the reaction with the consensus sequences. Binding and DNA bending are simultaneous, proceed through an intermediate species, and yield conformers with strongly bent DNA. The initial binding/bending event for both variant and consensus promoters (TBP ϩ C7 3 TBP⅐C7 B, SEVERE and TBP ϩ DNA 3 I 1 , respectively) occurs via a large energetic barrier together with the largest stepwise increase in entropy. Likewise, the thermodynamic changes characterizing the second step of this pathway for C7 (TBP⅐C7 B, SEVERE 3 TBP⅐C7 C, SEVERE ) follow a pattern very similar to those for the combined second and third steps of the TBP⅐AdMLP and TBP⅐E4 interactions (I 1 3 I 2 3

FIG. 4. Average fits (solid line) from the TBP⅐C7 global analysis obtained using the three-step model with two branching pathways (Equation 7) for (A) stopped-flow curves, 25°C with [TBP]
‫؍‬ 200 nM, (B) replacement curves, 20°C, and (C) equilibrium binding isotherms, 20°C. Overall 2 SF , 2 R , and 2 K values of 0.95, 1.1, and 0.68, respectively, reflect theoretical fits to observed curves within experimental noise and were associated with globally randomized residuals. Because additional stringency is introduced using noisefree experimental curves and absolute TBP activities in the global analysis, the close correspondence between the theoretical and experimental data is notable. The average fit to the replacement curves obtained using the linear two-intermediate model (Equation 4) is shown in B for 20°C (dashed line). Although good correspondence to the association and equilibrium curves was obtained using Equation 4, a 2 R value of 1.79 signified a clear lack of fit to the replacement curves.
TBP⅐DNA CONSENSUS ) (Fig. 6). The TBP⅐C7 reaction has only one intermediate along the path with a severe bend, unlike that for AdMLP and E4. This result may not reflect a real difference between the recognition processes for the two types of promoters but, rather, that meaningful determination of two additional rate constants and one additional quantum yield is not possible from the data. The dominance of TBP⅐C7 B, SEVERE over the course of the reaction and its persistence at equilibrium, much like I 1 for the strong promoters (16,20), reflects kinetic similarities between the variant and consensus partial reactions. A prevalent TBP⅐C7 intermediate, corresponding closely both structurally and kinetically to I 1 , further implicates this conformer as a biologically relevant species to which transcription factors may bind in subsequent steps of pre-initiation complex assembly (16,20).
The atomic resolution TBP⅐C7 co-crystal structure revealed a C:G Hoogsteen base pair at position 7 (28). This adaptation was unique among twelve such sequences investigated and apparently allowed the C7 helix to conform to the binding site to yield a complex with strongly bent DNA. The branching pathways of the TBP⅐C7 interaction may correlate with this potential to form an alternate base pairing. In this view, a steric incompatibility between the exocyclic NH 2 of guanine and Leu-72 of TBP associated with Watson-Crick base pairing at position 7 interferes with successful intercalation of the 3Ј-phenylalanine pair, yielding a bound conformer with only partially bent DNA. This scenario is consistent with the negative enthalpic change and very slight entropic increase leading to formation of TBP⅐DNA A, SHALLOW . On the other hand, the "right" branch may reflect formation of TBP⅐C7 B, SEVERE via the Hoogsteen base  [20.6, 20.9] a Numbers in brackets represent the upper and lower error bounds corresponding to the 68% confidence region, determined as described previously (16). pairing, in a process more similar to that of the consensus sequences.
TBP⅐C7 and TBP⅐E4 display an additional correspondence: dissociation of these binary complexes ensues from both replacement and displacement processes (20). Moreover, the active displacement process does not occur primarily from the final bound TBP⅐DNA complex for either C7 or E4. Rather, bound DNA is displaced predominantly from TBP⅐C7 B, SEVERE (Fig. 7) and, for the consensus sequence, from the first TBP⅐E4 intermediate (20). Very similar temperature-independent displacement rate constants of ϳ2 ϫ 10 Ϫ3 M Ϫ1 s Ϫ1 were obtained for both sequences. These analyses point to the dominant role of the intermediate conformers in the displacement reaction and further support the proposal that TBP displacement by regulatory proteins may occur from these species (20).
TBP binding to C7 in solution induces a bend in the DNA helical axis with an average apparent angle significantly less than for AdMLP and E4 (34). The results of model fitting reported herein show that the C7 bend derives from a weighted average of the bends in three TBP-bound species present at equilibrium: one conformer with DNA bent only slightly, TBP⅐C7 A, SHALLOW , and two conformers with DNA bent severely, TBP⅐C7 B, SEVERE and TBP⅐C7 C, SEVERE . The enthalpy change corresponding to the transition from a shallow bend to a severe DNA bent in the complex is 37.1 kcal mol Ϫ1 , with ⌬S 0 ϭ 123.4 cal K Ϫ1 mol Ϫ1 . TBP⅐C7 A, SHALLOW and TBP⅐C7 B, SEVERE dominate the reaction even at equilibrium, with the mole fraction ratios of the three species dependent upon both temperature and TBP concentration.
The solution bend angle for TBP-bound C7 is temperaturedependent (Table I), in contrast to those of bound AdMLP and E4 (20,34). This variability was attributed to temperature-dependent population shifts between complexes with slightly and strongly bent DNA. The corresponding probability distributions were analyzed using a two-state model with the value of R SHALLOW fixed at 53.3 Å as described (34,35). The data were well accommodated and yielded a value for R SEVERE of 50.1 Å. These mean distances correspond to DNA bend angles for the two bound conformers of ϳ30°and ϳ60°. This two-state model has been shown previously to accommodate the sequence and  7. TBP⅐T*C7 dpx *F dissociation kinetic data obtained at 15°C using 35.5 M unlabeled DNA (q) and the corresponding pure replacement curve (؋). The fast phase of the replacement process for bound C7 occurs from two species, TBP⅐C7 A, SHALLOW and TBP⅐C7 B, SEVERE . The fast phase replacement rate is approximately a mole fraction-weighted average of k 1 and k 4 with an amplitude reflecting the sum of the fractional populations of the two species at equilibrium. The slow replacement rate derives from TBP⅐C7 C, SEVERE 3 TBP ϩ DNA with a rate Ϸ k 6 . To determine the predominant species from which displacement occurs, theoretical dissociation curves were generated as described (20), including displacement from either TBP⅐C7 A, SHALLOW (dashed line) or TBP⅐DNA B, SEVERE (solid line). Due both to the very low equilibrium population of TBP⅐C7 C, SEVERE and a slow rate constant of formation (k 5 ), this conformer is not a significant participant in the displacement process. Accommodation of the dissociation data to within error at all temperatures can result from displacement only from TBP⅐DNA B, SEVERE (solid line). In contrast, displacement only from TBP⅐C7 A, SHALLOW yields large errors (residuals 1.5-4ϫ error) and systematic deviation of the calculated from the observed curves. A rapid rate of displacement from TBP⅐C7 A, SHALLOW alone is required to be added to the pure replacement curve for the sum to mimic the fast phase of the observed dissociation. However, rapid removal of this predominant reaction species at 15°C precludes significant contributions of TBP⅐C7 C, SEVERE to overall dissociation, thus eliminating the slow replacement phase and producing a severe lack of fit (dashed line). The first 150 s of the TBP⅐C7 dissociation reaction at 15°C and the simulated results are shown in the inset.
osmolyte concentration dependence of TBP-bound DNA bend angles. The model is now extended to include the temperature dependence of the bound C7 bend, with the bend angle for the severely bent DNA somewhat smaller than that obtained previously.
Global analysis of the collective TBP⅐C7 kinetic and thermodynamic data was conducted with the assumption that the microscopic ⌬C P 0 term associated with each intrinsic ⌬H 0 is negligible over the temperature range of the FRET measurements. The finding of multiple TBP⅐C7 species with different structural and energetic properties led us to investigate the contributions of temperature-dependent shifts in the equilibrium populations of these conformers to an overall heat capacity change for the binding reaction (46). The experimental equilibrium constants determined directly using steady-state FRET are shown in Fig. 8 together with the best least-squares fit to a straight line. The theoretical equilibrium constants corresponding to Equation 7, K a apparent , were generated from the values of the rate constants and activation energies (solid line in Fig. 8). Both lines accommodate the data over the experimental temperature range, although the kinetic model assumed neither heat capacity changes for the individual reaction steps nor a van't Hoff correspondence. ⌬Cp°R XN , the heat capacity change for the reaction corresponding to K a apparent , is not constant with temperature from 0 to 40°C (Fig. 9), with a maximum value of 1.58 kcal K Ϫ1 mol Ϫ1 at 18°C. This ⌬Cp°R XN derives solely from shifts in the populations of conformers with shallowly and severely bent DNA that have differing values of ⌬H°R XN , referenced to free TBP ϩ DNA.
Conclusions-This work further defines the relationships among TATA box sequence, TBP⅐TATA solution structures, and TBP⅐TATA recognition mechanisms. A single base pair substitution in the consensus AdMLP TATA sequence has dramatic functional consequences, yielding kinetics for association and dissociation that differ significantly from those of the parent promoter and another strong promoter, E4. Global analysis reveals a branching mechanism for the TBP⅐C7 interaction having a partial reaction that is absent from the consensus interactions. The solution structure of the TBP⅐C7 complex likewise differs from those of the consensus sequences and is temperature-dependent. A two-state model has now been shown to accommodate the dependence of TBP⅐TATA bend angles on sequence, osmolyte concentration, and temperature.
These functional and structural differences in the TBP⅐C7 variant complex correspond to diminished transcription efficiency relative to consensus binary complexes. Studies are currently underway to determine whether TBP recognizes variant TATA sequences similarly or uniquely.
FIG. 8. Analysis of heat capacity changes deriving from temperature-dependent shifts in TBP⅐DNA SHALLOW º TBP⅐DNA SEVERE . The K a values measured using steady-state FRET (q) are shown with the corresponding van't Hoff plot (dashed line). The apparent equilibrium constant corresponding to Equation 7, K a apparent , is equal to K 1 ϩ K 2 ϩ K 2 K 3 , where K 1 ϭ k 2 /k 1 , K 2 ϭ k 3 /k 4 , and K 3 ϭ k 5 /k 6 . The value of K a apparent was determined as a function of temperature (15-30°C; solid line) using the fitted values of k 1 through k 6 and corresponding activation energies. Extension of K a apparent from 0 to 40°C is shown in the inset with (ϩ) indicating the experimental temperature range. The non-linearity of K a apparent from 0 to 40°C reveals a ⌬Cp°R XN due to multiple conformations at equilibrium (46). However, the presence or absence of such a heat capacity change cannot be distinguished from the equilibrium data through the experimental temperature range.
FIG. 9. The predicted heat capacity change for the reaction modeled as in Equation 7 with ⌬C P°‫؍‬ 0 for each step in the mechanism. The variations in the value of [GRAPHIC2] among the TBP⅐C7 conformers with differently bent DNA give rise to values of ⌬H RXN that are strongly temperature dependent, in contrast to the usual calorimetric assumption.