Cooperativity and interaction energy threshold effects in recognition of the −10 promoter element by bacterial RNA polymerase

RNA polymerase (RNAP) melts promoter DNA to form transcription-competent open promoter complex (RPo). Interaction of the RNAP σ subunit with non-template strand bases of a conserved −10 element (consensus sequence T−12A−11T−10A−9A−8T−7) is an important source of energy-driving localized promoter melting. Here, we used an RNAP molecular beacon assay to investigate interdependencies of RNAP interactions with −10 element nucleotides. The results reveal a strong cooperation between RNAP interactions with individual −10 element non-template strand nucleotides and indicate that recognition of the −10 element bases occurs only when free energy of the overall RNAP −10 element binding reaches a certain threshold level. The threshold-like mode of the −10 element recognition may be related to the energetic cost of attaining a conformation of the −10 element that is recognizable by RNAP. The RNAP interaction with T/A−12 base pair was found to be strongly stimulated by RNAP interactions with other −10 element bases and with promoter spacer between the −10 and −35 promoter elements. The data also indicate that unmelted −10 promoter element can impair RNAP interactions with promoter DNA upstream of the −11 position. We suggest that cooperativity and threshold effects are important factors guiding the dynamics and selectivity of RPo formation.


INTRODUCTION
Formation of the transcription-competent open promoter complex (RP o ) by bacterial DNA-dependent RNA polymerase (RNAP) is a critical checkpoint on the pathway of gene expression. In RP o , the DNA duplex is melted over a stretch of 12-15 bp, which makes the transcription start point (position +1) accessible to the RNAP catalytic center. RNAP initiates transcription in the form of a holoenzyme (subunit composition aIaIIbb'os). The dissociable specificity subunit s is required for both promoter recognition and melting (1,2). Specific interactions of the Escherichia coli primary s subunit (s 70 ) with non-template strand (nt-strand) nucleotides of conserved À10 promoter element (consensus sequence T À12 A À11 T À10 A À9 A À8 T À7 ) are an important source of energy-driving localized melting of s 70 -dependent promoters (3)(4)(5)(6)(7). The strand separation usually is initiated at the À11A/T base pair and propagates in the downstream direction (2). The À12 bp likely remains in the double-stranded (ds) form in most promoters (8,9). At physiological conditions, the RP o formation is a highly cooperative process (2,(10)(11)(12). Yet, intermediate promoter complexes with transcription bubbles not extended to include the transcription start point have been detected at several promoters (12)(13)(14) as well as when studying RNAP mutants (15,16). In contrast, promoter complexes bearing partially melted À10 promoter element have not been revealed at physiological temperatures, implying a particularly high degree of cooperativity in unpairing of the À10 element bases.
Oligonucleotides and fork junction promoter fragments containing single-stranded (ss) extensions corresponding to the nt-strand of the À10 promoter element have been used as model substrates to study RNAP interactions with melted DNA in promoter complexes (3,4). Binding studies using these DNA probes confirmed that in the context of the RNAP holoenzyme, s subunit recognizes the À10-ntstrand sequence in the single-stranded form (3)(4)(5)(6)9,17). Recently reported structures of s domain 2 and RNAP holoenzyme complexes with model promoter fragments reveal that the nt-strand bases of consensus À10 element interact with numerous residues from s conserved regions 2 and 1.2 with multiple s residues simultaneously contacting more than one nucleotide (18,19). In the structures, the nt-strand segment of the À10 element adopts a conformation that is incompatible with the ds DNA conformation, particularly because the A-11 and T-7 bases are flipped out of the DNA base stack (18,19). These results indicate that the recognition of the À10 element must be coupled with its unwinding and melting.
The molecular details of events that trigger the À10 promoter element recognition and strand separation remain unclear. In the light of the proposed mechanism of the À10 element recognition, we reasoned that studying interdependences between RNAP interactions with individual À10 element nucleotides may help clarify fine details of these processes. Although non-additive effects of multiple substitutions in À10 element bases on transcription (20) and RNAP binding (5) have been observed, experimental data on interdependences between partial s interactions with the À10 element bases are lacking. Here, we systematically studied mutual effects of partial RNAP interactions with À10 element bases in the context of RNAP complexes with model promoter fragments by using a highly sensitive and quantitative fluorometric RNAP molecular beacon assay. The data reveal a strong degree of cooperation between specific RNAP contacts with individual À10 element nucleotides and show that the recognition of the À10 element bases occurs only when the overall interaction acquires a free energy below a $À3 kcal/mol threshold. We suggest that the threshold effect contributes to the selectivity of open promoter complex formation by hindering RNAP binding to non-promoter DNA.

Proteins
Escherichia coli RNAP core was purchased from Epicenter. RNA polymerase holoenzyme containing the s 70 derivative labeled at position 211 with fluorescent label 5-tetramethylrhodamine (RNAP beacon) was prepared as previously described (6).

DNA probes
DNA oligonucleotides were synthesized by Integrated DNA Technologies. Fork junction and double-stranded DNA probes were prepared as described previously (6).

Fluorometric assays
Fluorescence measurements were performed using a QuantaMaster QM4 spectrofluorometer (PTI) in transcription buffer [40 mM Tris-HCl (pH 8.0), 100 mM NaCl, 5% glycerol, 1 mM DTT and 10 mM MgCl 2 ] containing 0.02% Tween 20 at 25 C. Final assay mixtures (800 ml) contained 1 nM labeled RNAP holoenzyme and DNA probes at various concentrations. The fluorescence intensities were recorded with an excitation wavelength of 550 nm and an emission wavelength of 578 nm.
To obtain equilibrium dissociation constants (K d ), the experimental dependence of the fluorescent signal amplitude (F) on DNA probe concentration was fit to Equation (1), unless otherwise noted (6,21).
where X = (F À F 0 )/(F max À F 0 ), F 0 is the initial value of the amplitude, and F max is the limiting value of the amplitude at [DNA] = 1. The data were analyzed using SigmaPlot software (SPSS, Inc.  Figure S2A) producing negligible signal on binding to the RNAP beacon was used as a reference competitor, as described previously (6). Time-dependent fluorescence changes were monitored after manual mixing of RNAP beacon (800 ml) and a DNA probe (<20 ml) in a cuvette; the mixing deadtime was 15 s.
In line with previous works, we used values of the free energy gain/loss resulted from substitution a consensus base Y for a non-consensus base Z at position N in the À10 element ÁÁG(Y N Z) to characterize strengths of the specific interactions in RNAP complexes with studied templates. The changes in ÁG were calculated using Equation (2): Free energy gain resulted from RNAP binding to the single-stranded segments of fork junctions 2-5 (the structures are shown in Figure 1C and Supplementary Figure  S1) was calculated using Equation (3) where K d (M) and K d (1) are dissociation constants for assayed probes and probe 1 that corresponds to the double-stranded fragment of the probes. As K d (6) could not be directly measured (see later in the text), ÁÁG 6 was evaluated using Equation (4) assuming that improvements of fork junction affinities because of extension of their ss segments from À8 to À7 are equal in the context of probes 5, 6 and 7, 8.

DNA probes
To discern interdependencies of partial RNAP interactions with individual nucleotides of the À10 promoter element, we measured RNAP affinity to a large set of model promoter fragments. The structures of DNA probes used are presented in Supplementary Figure S1 and are also schematically depicted in main figures. The majority of probes are based on T5 N25, a strong promoter containing consensus À10 element. Some experiments were performed with probes based on the sequence of a weak Pr promoter of Pseudomonas putida bearing a suboptimal À10 element (22). The affinities of DNA probes to RNAP were characterized by dissociation constants of their complexes with RNAP as determined by the fluorometric RNAP molecular beacon assay (6). The K d values varied widely-over a 4-log range. With its high sensitivity and low intensity of non-specific background signal, the RNAP beacon assay is ideally suited for performing such measurements, as it allows to quantitatively measure both weak and strong interactions.
Binding of fork junction promoter fragments reveals strong cooperativity of individual À10 element nt-strand nucleotides interactions with RNAP An adenine at the À11 position and a thymine at the À7 position are the most conserved bases of the À10 element (23). Introduction of non-consensus bases at the À11 position generally strongly decreases promoter activity
-38 -12-11   (24,25). To test the effects of specific RNAP contact with À11A on RNAP interactions with other À10 element nt-strand nucleotides, we compared RNAP affinities with a series of fork junction DNA probes bearing consecutive one-nucleotide extensions of the nt-strand from the À11 to À7 positions ( Figure 1A and C) and to a set of similar probes in which the À11 position was occupied by a non-consensus T. Probes 1-6 bear consensus À35 element sequence TTGACA. The K d of RNAP complexes with each probe, the ratios of K d values for matching probes bearing either an A or a T at the À11 position and corresponding free energy changes caused by the A-11T substitution are shown in Figure 1C; Figure 1B shows representative experimental data. K d for RNAP complex with fork junction 6 (bears consensus À10 element) could not be calculated from data shown in Figure 1B, as RNAP binding to this probe was too strong and fluorescence intensity reached saturation level at a minimal probe concentration used. Determination of the K d value by equilibrium competition-binding assay also could not be carried out, as reactions did not reach equilibrium even after a 20-h incubation (data not shown), which can be explained by slow dissociation of RNAP complex with this probe (26). Therefore, we determined a change in K d caused by the extension from À8 to À7 in the context of fork junctions 7 and 8 ( Figure 1C and Supplementary Figure S2), whose affinity to RNAP is weakened because of a non-consensus À35 element sequence TTG CTT (a À35 element of the T5 N25 promoter) and a junction point at position À13 rather than at À12 as in probes 1-6 (4). The calculated free energies of RNAP binding to consensus single-stranded segments of fork junctions are shown in Figure 1D. Overall, the data presented in Figure 1C and 1D show that for the most part, consecutive one-nucleotide extensions considerably improve affinities of À11A containing probes. The only exception is probe 3, where the introduction of À10T causes a drop in affinity compared with the shorter probe 2. A similar inhibitory effect of an extra top strand À10 nucleotide on formation of heparin-resistant RNAP-fork junction complexes was reported by Guo and Gralla (4). In contrast to the À11A containing probes, for the À11T series of probes only the extension from À8 to À7 resulted in a large increase in affinity, whereas other one-nucleotide extensions had at most a slight effect on binding. The AÀ11T substitution caused an $5-fold drop in affinities of probes with 3 0 termini at À11 and À10. This ratio increased to $200 for probes extending to À8, whereas a nearly 1000-fold difference was observed in the context of probes extended to À7. The AÀ11T substitution also caused a large $300-fold effect on RNAP binding to fork junction 9 that does not contain the À35 element but bears an extended À10 element and includes the À6 to À4 bases that interact with the s conserved region 1.2 (19,27,28). Clearly, these results indicate that efficiency of À11A recognition is strongly stimulated by RNAP contacts with other À10 element bases.

A
The fact that the introduction of À7T confers a $120fold improvement in affinity of À11T probes [compare K d values for probes 5(À11T) and 6(À11T) in Figure 1C] demonstrates that specific interaction of À11A with RNAP is not strictly required for the recognition of À7T in the context of fork junctions. We evaluated the specificity of RNAP interaction with À7T in the context of several fork junctions with non-consensus bases at positions À11 to À8 (Figure 2). Introduction of A at position À7 caused, respectively, 210-and 710-fold drops in affinity of probes 8 and 9 with consensus À10 element. A 110-fold drop was observed in the context of a probe with a nonconsensus -11T [probe 6(À11T)]. A much lower, $8-fold, effect was observed with probe 10 bearing a G À10 T À9 T À8 non-consensus base stretch. Finally, the TÀ7A substitution caused only a 2.4-fold effect in the context of fork junction 11 with non-consensus bases at positions À11 to À8. Thus, the efficiency of À7T recognition is clearly modulated by the strength of RNAP contacts with other À10 element bases.
Overall, the results show that specific interactions between individual À10 element bases and RNAP are highly interdependent. Further, a strong specific interaction between RNAP and À10 promoter element bases occurs only when free energy of the overall RNAP interaction with the À10 element reaches a certain critical level. The specific RNAP interaction with À11A is much weaker in fork junctions with short ss extensions (probes 2 and 3) than in fork junctions 8 and 9 bearing ss extensions spanning the entire À10 element [ÁÁG(AÀ11T) $1 and $4 kcal/mol, respectively]. The threshold effect is also pronounced for less conserved À10T, À9A and À8A bases. Indeed, the T À10 A À9 A À8 segment improved the binding of probe 5 as compared with probe 2 $50-fold, whereas in the context of À11T substituted probes, this segment increased the binding only $2-fold ( Figure 1C). A similar effect is observed for À7T base recognition in fork junctions 8, 9 and 11 bearing all-consensus (probes 8 and 9) or non-consensus (probe 11) bases within the À11 to À8 segment ( Figure 2). These results may be explained by individual interactions between RNAP and À10 element nt-strand nucleotides cooperatively contributing to retention of a conformation of the À11 to À7 segment backbone (18,19) that favors recognition of the À10 element bases. We suggest that the $3 kcal/mol difference in ÁÁG(AÀ11T) values for probes 2, 3 and 8, 9 approximately corresponds to the energetic cost required to retain such conformation. The threshold energy value likely depends on promoter sequence, in particular on identity of the À12 bp (see later in the text).
Inhibition of fork junction DNA binding by a terminal nt-strand À10 base We further examined inhibition of RNAP binding observed on the introduction of consensus À10 T (probe 3) into probe 2 ( Figure 1C). We determined K d values for RNAP complexes with derivatives of probe 3 bearing an A, a G, a C or an abasic site at the À10 position (shown in Figure 3A) and found that only the abasic probe [probe 3(À10Ab)] bound slightly better than probe 2. The K d values for other probes were higher than that of probe 3(À10Ab) by 1.7-to 4.6-fold ( Figure 3A and B). This result correlates with reported negligible effect of the introduction of an abasic site at the À10 position on heparin resistance of an RNAP-fork junction complex (29). It seems that the need to unstack the À11A base, which is a prerequisite for its specific binding by s (18,19), may, at least in part, account for the slight energetically unfavorable effect of the presence of a base at position À10 observed in fork junction 3 and its derivatives. Disruption of stacking interaction between the À11 and À10 bases should consume a part of the binding energy, whereas the contribution of the À10 position interaction with RNAP to the overall binding energy is likely to be insignificant, as RNAP interacts only with the backbone of the À10 nt in reported structures of RNAP complexes with promoter fragments (18,19). Consistent with the latter suggestion, substitutions of À10T have a rather slight effect on the affinity of an oligonucleotide probe 30 corresponding to positions À12 to +2 of the nt-strand segment of the T5 N25 promoter (Supplementary Table S1). It is noteworthy also that purines at À10 cause somewhat higher inhibitory effects than pyrimidines ( Figure 3B). This is consistent with the expectation that stacking interactions between neighboring purines should be stronger than between a purine and a pyrimidine because of higher surface area of the contact (30).

Recognition of À7T base in the context of fork junctions based on the sequence of Pr promoter
The s 70 -dependent Pr promoter controls catabolism of phenolic compounds by P. putida CF600 (22). The Pr promoter bears a poor À10 element C À12 TGGCT À7 containing only one consensus base À7T (22). Consequently, the Pr promoter is intrinsically weak and requires ppGpp and DksA for optimal activity (31). Substitutions of À7T for any other base abolish activity (31). We wondered whether the critical importance of À7T for Pr activity is displayed in the context of RNAP interaction with fork junction probes. We measured RNAP binding to Pr-based fork junction probe 12 and to control probes that either bore an A at the À7 position [probe 12(À7A)] or lacked the À7 nucleotide altogether (probe 13) ( Figure 4A). The K d for probe 12 was 11 nM, whereas the K d values for control 12(À7A) or 13 probes were found to be $100 nM ( Figure 4B). This result demonstrates that RNAP clearly recognizes À7T in fork junctions based on the Pr promoter, whereas the RNAP interaction with À7T in probe 12 is much weaker than in probe 8 bearing consensus À10 element bases [ÁÁG(T-7A) values of 1.3 and 3.2 kcal/mol, respectively]. Further extending the nt-strand from À7 to À4 in probe 14 considerably increased the binding affinity compared with probe 12 ( Figure 4B), indicating that the RNAP interaction with the Pr discriminator segment is not affected in the probe.

Recognition of the T/A À12 base pair depends on RNAP interactions with other À10 element bases and with promoter spacer
A T at the À12 position is highly conserved among bacterial s 70 -dependent promoters (23) and substitutions of À12T decrease transcription from many promoters (24,25). Substitution of a T/A À12 base pair for an A/T base pair considerably decreases heparin resistance of RNAP complexes with fork junctions based on the lacUV5 promoter (32). In agreement with these data, we found that TÀ12A substitution decreased affinities of fork junction probes 5, 6(À11T) and 9 by 170-to 260-fold ( Figure 5). Heparin resistance assay data and structural modeling indicate that both nt-strand T and t-strand A of the T/A À12 base pair are recognized by the s subunit (4,18). Consistently, we found that affinity of fork junction probe 8 bearing an unpaired T at À12 was less affected by the T-12A substitution than affinities of probes 5, 6(À11T) and 9 ( Figure 5). In agreement with this result, a derivative of fork junction probe 5 lacking the template strand nucleotide at position À12 (probe 31) bound RNAP $10-fold weaker than probe 5 (Supplementary Figure S3). The affinity of oligonucleotide probe 30 depends very slightly on the identity of the base at À12, whereas substitutions of À11A and À7T greatly affected the binding (Supplementary Table S1). Modest $5-fold effects of substitutions of consensus base at the À12 position on the affinity have been observed with similar oligonucleotides that contained additional bases upstream of the À12 position (3,6,17). We considered a possibility that RNAP can effectively interact with the T/A À12 base pair in a ds probe truncated downstream at the À12 position. To strengthen specific binding of downstream DNA end, a TG motif of extended À10 element was incorporated in the probe ( Figure 5). However, RNAP recognized the T/ A À12 base pair in resulting probe 15 poorly, as probe 15(À12A) bound RNAP only $4-fold weaker ( Figure 5).
Next, we determined affinities of fork junction derivatives of probes 15 and 15(À12A) bearing A À11 , A À11 TAA À8 or A À11 TAAT À7 stretches of the consensus À10 element bases. The data presented in Figure 5 show that these extensions considerably improve affinities of TÀ12 containing probes, similarly to what was observed with consensus probes 1, 2, 5 and 6 ( Figure 1C). However, in the context of À12A probes, the extension from À12 to À11 results in a negligible change in affinity, whereas the extension to À8 only moderately improves the binding by $6-fold. In contrast, extension from À8 to À7 increases affinity by $300-fold in the context of 17(À12A) and 18(À12A) probes. Accordingly, the TÀ12A substitution strongly affects the binding of fork junction 17 ( Figure 5).
The data show that RNAP efficiently recognizes the T/A À12 base pair in fork junctions bearing long stretches of nt-strand À10 element bases, but the recognition is less effective in fork junction bearing the minimal A À11 extension and is poor in ds and ss probes truncated at À12. In principle, recognition of the T/A À12 base pair in probes 15 and 16 may be affected by fraying of probe termini (33). However, this effect cannot explain the large difference in the efficiency of recognition of unpaired À12T in the context of oligo 30 and fork junction 8. We propose that the position of the À12 bp in RNAP complex with probe 15 is not compatible with strong specific interaction of T/ A À12 with s. However, tight RNAP binding both to the À10 element bases and to promoter spacer segment located between the À10 and À35 elements may be sufficient to change spacer or/and s conformation and bring the À12 bp to a position that is optimal for the recognition of T/A À12 and adjacent nt-strand À10 element bases. This   implies that the recognition of T/A À12 should be coupled with initiation of promoter melting.

RNAP binding to promoter fragments bearing À10 element template strand bases
Based on the aforementioned results, we created a set of ds and fork junction probes (probes 19-25, 27-29; Figure 6 and Supplementary Figure S1) bearing t-strand nucleotides downstream from the À12 position and measured affinities of these probes to RNAP. In the context of progressively extended ds probes 19-23, the introduction of the À10T/A bp resulted in inhibition of RNAP binding ( Figure 6A), similarly to what was observed with fork junction probes. As expected, the AÀ11T substitution strongly affected affinity of ds probes. Probe 23(À11T) extended to À7 binds RNAP only 4-fold stronger than probe 1 bearing no nucleotides downstream from À12, whereas probe 22(À11T) with downstream end at À8 binds RNAP even weaker than probe 1 (K d values are 16, 62 and 120 nM, respectively, Figures 1C and 6A). We further examined the effect of a ds segment bearing non-consensus À10 element bases in the context of probes 24 and 25 ( Figure 6B) containing a sequence upstream of the À35 element which interacts effectively with the RNAP a subunit C-terminal domain (34) and a TG motif of extended À10 element. The data show that introduction of four non-consensus base pairs downstream from the À12 position in probe 24 leads to a $50-fold decrease in the affinity ( Figure 6B). In contrast, fork junction derivatives of probe 24 containing either t-strand or nt-strand non-consensus bases bind RNAP stronger than the parent probe 24 (Supplementary Figure S4). Feklistov and Darst (18) proposed that melting of the À11 and downstream positions allows the À12 position to move closer to a s region 2.4 a helix, and that this movement is required for recognition of the À12 bp. This effect may at least in part account for the observed inhibition of the binding by mutations in positions À11/ À8. Indeed, the approach of the À12 bp to s region 2 should be constrained in RNAP complexes with ds probes 22(À11T) and 25 bearing non-consensus À10 element bases that likely remain paired. The large difference in affinities of probes 24 and 25 suggests that introduction of the non-consensus À10 element segment might affect not only the RNAP interaction with À12 position but also some other RNAP-promoter interactions in probe 25.
Previous investigations have revealed that RNAP binds to the t-strand segment of the transcription bubble considerably weaker than to the non-template segment (4,25). Consistently, we found that extensions of t-strand of probe 19 to the À8 and À7 position (respectively, fork junctions 28 and 29) conferred moderate ($5-and 13-fold) improvements in affinities ( Figure 6A).

DISCUSSION
High-resolution structures of s and RNAP bound to model promoter fragments have revealed that recognition of the À10 promoter consensus element is achieved through network of interactions between s residues and nt-strand À10 element nucleotides extruded from the DNA double helix (18,19). In this work, we investigated whether these interactions are interdependent and, if so, whether such interdependency is essential for promoter binding. Accordingly, we measured how changing the identity of one À10 element nucleotide affects RNAP affinity to other À10 element nt-strand nucleotides in the context of promoter fragment DNA probes. The study

20.
AT TA -38 -12 -9  Figure 6. RNAP binding to promoter fragments bearing À10 elementtemplate strand bases. (A) Calculated K d values. The sequence of À38 to À12 segment of the probes corresponds to that of probe 1. (B) Inhibitory effect of ds segment bearing non-consensus À10 element bases (shown in italic) on RNAP binding to promoter fragment. The sequence of probe 24 is shown on the top of the panel.

A T A TAT
required quantitative characterization of RNAP-DNA complexes with widely different stabilities, which is a technically challenging task because of RNAP propensity for non-specific DNA binding. Gralla and coworkers (4,29,32) studied the effects of substitutions in the À10 element on RNAP binding to DNA probes similar to those used in our work. However, the in-gel mobility retardation method they used allowed quantitative comparisons of probe affinities only within one order of magnitude range, which is insufficient for detection of effects related to cooperativity of RNAP interactions with the À10 element. We here relied on a highly sensitive molecular beacon RNAP assay that allowed quantitative characterization of RNAP-DNA complexes whose stabilities differed by as much as four orders of magnitude.
The main finding of our work is that specific RNAP interactions with nt-strand nucleotides of the À10 element are highly cooperative. The data also indicates that attaining a recognizable À10 element conformation is energetically costly. Accordingly, strong specific interaction with functionally most important and evolutionarily most conserved À10 promoter element bases occurs only when free energy gain resulting from the overall RNAP interaction with the À10 element exceeds a certain critical level. Such threshold-like behavior is a characteristic feature of various types of highly cooperative interactions (35,36). The threshold effect may help avoid unproductive RNAP interactions with occasional promoter-like DNA sequences, in particular when DNA melting is facilitated by negative DNA supercoiling, and thus improve the overall selectivity of promoter recognition.
RNAP interactions with the nt-strand À10 element nucleotides in progressively extended fork junctions ( Figure  1) should reflect interactions that arise in RNAP-promoter complex on gradual expansion of the transcription bubble. An adenine at the À11 position is of special importance for nucleation of promoter melting (25,(37)(38)(39). The data presented in Figure 1C demonstrate that specific interactions of À11A with RNAP in fork junctions with short A À11 and A À11 T À12 ss segments are considerably weaker than in fork junction bearing a single-stranded segment corresponding to complete À10 element. Accordingly, binding of the short ss fork junction segments to RNAP is also relatively weak ( Figure 1C and D). The average free energy of a base pair breakage within an A/T rich À10 element-like sequence is $1 kcal/ mol per bp (40), whereas initial nucleation of promoter melting can be much more energetically costly (40)(41)(42). This evaluation suggests that short bubbles around the À11A base, which may form because of thermal fluctuations (43,44), are unlikely to be stabilized by RNAP. In contrast, the ÁG gain resulting from RNAP interaction with ss stretch bearing À11 to À7 consensus À10 element bases (À7.1 kcal/mol, Figure 1D) should be sufficient to stabilize local melting. Thus, a first significantly stable melted intermediate promoter complex likely comprises unpaired À11 to À7 segment, at least in linear DNA templates. Overall, the results imply that cooperativity of promoter melting may be to a large degree accounted by the cooperativity of partial RNAPÀ10 element interactions. Our data also provide an explanation for the inhibitory effect of the À10 position (4), which is consistent with the proposed model of the À10 element recognition (18).
Our binding assays show that RNAP interacts with the T/A À12 base pair in fork junctions much stronger than in a double-stranded promoter fragment 15 bearing no bases downstream of the À12 position ( Figure 5). RNAP interaction with À12T is also weak in the context of singlestranded oligonucleotide probes. To explain these results, we propose that simultaneous RNAP binding to nt-strand bases of the À10 element and to double-stranded promoter spacer lead to a conformational change in promoter complex favoring specific recognition of the T/ A À12 base pair. Effective recognition of À12T in an ss DNA aptamer (28) suggests that some RNAP-aptamer interactions mimic RNAP contacts with promoter spacer. A sharp bend in DNA at the À16 position observed in the 6.5 Å resolution crystal structure of RNAP complex with a fork-junction promoter DNA (9) might be related to the hypothetical rearrangement improving the À12 bp recognition. Elucidation of this question may have to await high-resolution structures of various intermediates long the RP o formation pathway.
The results obtained here also show that short ds segments bearing non-consensus À10 element bases decrease affinity of promoter fragments ( Figure 6A and B). We note that ss stretches of non-consensus À10 element bases do not decrease the binding of fork junction probes (Supplementary Figure S4). This suggests us that duplex conformation of the À10 element can intrinsically impair promoter binding by constraining formation of RNAP-promoter contacts upstream from the À11 position. This explanation seems consistent with structural considerations indicating that RNAP interaction with unmelted À10 element segment should be weak and that unpairing of the À11 position and downstream À10 element bases should lead to strengthening of s region 2 interactions with the À12T base (18). This effect should shift the equilibrium between duplex and melted conformations of the À10 element in the direction of melting, and thus may play a role in the DNA opening step.

SUPPLEMENTARY DATA
Supplementary Data are available at NAR Online: Supplementary Table 1 and Supplementary Figures 1-4.