Stimulation of Open Complex Formation by Nicks and Apurinic Sites Suggests a Role for Nucleation of DNA Melting in Escherichia coli Promoter Function*

We report the effects of depurination and prenicking at various positions of the phage λprmup-1Δ265 promoter DNA on the rate of open complex formation. We have found that depurination and prenicking at positions around the −10 region strongly stimulated the rate of open complex formation. Since nicking and depurination are known to destabilize DNA helical structure, our observations indicate that the instability of the −10 region is important for open complex formation. We further infer that (i) the nucleation of DNA melting, which occurs during the isomerization from the closed complex into the open complex, contributes to the rate of open complex formation; (ii) the nucleation of melting occurs around the −10 region; and (iii) the propagation of DNA melting from the nucleation region is not rate-limiting. In addition, we have found that depurination at several positions inhibited open complex formation. We used dimethyl sulfate modification protection studies to show that most of the guanine bases that are among these positions are in contact with RNA polymerase in the open complex.

Open complex formation is the major rate-determining step in the process of transcription initiation with E. coli RNA polymerase (E 70 ). Many of the previous studies have been based on a model containing two functional steps (1,2) that include the initial binding of RNA polymerase to the promoter DNA to form the closed complex followed by the isomerization of the closed complex into the open complex. Some studies have also suggested that the isomerization step in the two-step model can be further dissected into two discrete steps: the rate-limiting isomerization step per se and a DNA melting step (3)(4)(5)(6)(7)(8)(9)(10). In this extended model, the DNA melting step is argued to be very rapid under normal transcription conditions, and consequently it is normally not ratelimiting. It has been suggested that the isomerization step involves a major conformational change in RNA polymerase, which is thought to be the cause for the nucleation of DNA melting (10,11).
It is known that the Ϫ10 region of the promoters recognized by 70 holoenzyme is thermodynamically less stable than av-erage DNA. The conserved Ϫ10 hexamer sequence has a melting free energy that is close to the maximum (least stable) possible based on a calculation using the nearest-neighbor thermodynamic data of Breslauer et al. (12). Therefore, as expected, Margalit et al. (13) have shown that 80% of the up and down mutations in the Ϫ10 region correlated qualitatively with the change in the melting free energy. A closer inspection showed that most of the mutations among the exceptions are located outside the Ϫ10 (hexamer) region, and consequently are not bona fide exceptions if only the melting free energy of the Ϫ10 region is important. There were several exceptions within the Ϫ10 region. However, in such cases it can be argued that the effect of the mutation on the specific contact between RNA polymerase and promoter DNA is larger than the effect of the melting free energy change, and consequently obscured this effect.
We have tested whether the structural instability of the Ϫ10 region is important for promoter function (presumably in the nucleation of DNA melting) by carrying out depurination and prenicking studies. Both prenicking (14,15) and depurination (16,18) are known to destabilize DNA double-helical structure, and prenicking may also increase DNA structural flexibility (19). We found that both defects at positions around the Ϫ10 region had strong stimulatory effect on the rate of open complex formation on the prmup-1⌬265 promoter. This suggests that DNA structural instability in the Ϫ10 region is important for promoter function, and that DNA melting contributes to the rate of open complex formation. Interestingly, the region displaying the stimulatory effect is much smaller than the melted region detected in the open complex. This is consistent with the hypothesis that DNA melting can be divided into two steps: nucleation, and the subsequent propagation of DNA melting from the nucleation region, which is not rate-limiting. The nucleation region as suggested by these studies is located in a relatively small region around the Ϫ10 region. In addition, in both the depurination and prenicking analyses, the isomerization rate constant from the closed complex into the open complex was stimulated by at least 5-fold. This indicates that the nucleation of DNA melting occurs in the isomerization from the closed complex into the open complex.
The stimulatory effect of depurination around the Ϫ10 region on open complex formation is clearly due to the involvement of DNA melting this region. For most protein-DNA interactions, depurination is expected to decrease binding due to disruption of essential protein-DNA contacts. Based on this reasoning, depurination has been used in other studies to reveal protein-DNA contact (20). Interestingly, we also found that depurination at some positions had an inhibitory effect on open complex formation, suggesting that these positions are involved in contact with RNA polymerase. We have carried out DMS 1 modification protection studies of the open complex to confirm that most of the guanine bases among these positions are likely to be involved in contacting RNA polymerase.

MATERIALS AND METHODS
DNA Fragment and RNA Polymerase-The 564-base pair HindIII-EcoRI fragment containing the prmup-1⌬265 promoter was isolated and labeled as described previously (21). The end-labeled DNA fragment was digested with HinfI and run on a 5% polyacrylamide gel. The 189-base pair fragment containing the promoter and labeled on either strand at the EcoRI end was eluted from the gel using the Maxam and Gilbert (22) crush and soak procedure.
E. coli RNA polymerase holoenzyme (E 70 ) was isolated according to Burgess and Jendrisak (23) and Lowe et al. (24), and its activity was determined as described by Hawley and McClure (25). The RNA polymerase used in these studies had an activity of 65%. The RNA polymerase concentrations in the text are expressed as active concentrations.
DNA Prenicking and Depurination-Random phosphodiester bond cleavages (nicks) were introduced into the labeled 189-base pair fragment DNA at a frequency of about one nick per molecule by treating with DNase I. Following the treatment, the DNA was purified by phenol extraction and ethanol precipitations. The precipitated DNA was dissolved in TE (10 mM TrisCl, pH 8.0, 1 mM EDTA).
Partial depurination of the DNA was performed as described (22) with some modifications. Briefly, 3 l of 1.0 M formic acid (J.T. Baker) with pH adjusted to 2.0 using piperidine (Fisher) was added to 30 l of the end-labeled DNA in TE buffer. Following an incubation of 1 h at 25°C, the DNA was precipitated with ethanol. The precipitation steps resulted in ϳ95% renaturation of the DNA as judged by running samples on a 5% polyacrylamide gel. To assure complete renaturation, the DNA was dissolved in 180 mM NaCl, incubated at 90°C for 10 min to denature, and then cooled slowly to allow reannealing. The DNA was once again purified by ethanol precipitation, and then dissolved in TE buffer.
Assay for the Effects of Prenicking and Depurination on Open Complex Formation-Open complex formation on the prenicked or depurinated DNA was assayed using the gel retardation method as described (26,27) with some modifications. Open complexes were formed with 40 nM RNA polymerase and 1 nM prenicked or depurinated DNA at 19°C in standard reaction buffer (30 mM Hepes (adjusted to pH 7.5 with KOH), 200 mM potassium glutamate, 10 mM MgCl 2 , 1 mM dithiothreitol, and 100 g/ml bovine serum albumin) (21). At time zero, 4 l of RNA polymerase in reaction buffer was mixed with 16 l DNA in the same buffer. At various times ranging from 0.5 to 60 min, the reactions were stopped with the addition of 4 l of 180 g/ml heparin and 4 l of 40% glycerol. The samples were immediately loaded into a running 4% polyacrylamide gel (acrylamide to bis-acrylamide ratio of 59:1). The electrophoresis buffer was 10 mM TrisCl (pH 7.8) and 1 mM EDTA. After electrophoresis, the gel was exposed to Kodak X-Omat AR film for about DNA in the gel slices was isolated and prepared for electrophoresis on a sequencing gel following the procedure described by Ausubel et al. (28). The polyacrylamide gel slices were placed in an agarose gel with a piece of DEAE membrane inserted in front of each of them. Electrophoresis was carried out to transfer DNA from the gel slices onto the DEAE membranes (with ϳ70% recovery). To elute the DNA, each of the membranes was incubated with 200 l of elution buffer (1 M NaCl, 10 mM TrisCl, pH 8.0, and 1 mM EDTA) at 65°C for 30 min, and then rinsed with 200 l TE of buffer. The eluted open complex DNA and free DNA samples (with Ͼ80% recovery) were extracted with phenol, and precipitated with ethanol. After this step, the samples of depurinated DNA were dissolved in 1 M piperidine, incubated at 90°C for 30 min for strand scission at the depurinated positions (22), and then lyophilized. All of the prenicked or depurinated DNA samples were counted in a scintillation counter and were then dissolved in formamide sequencing sample buffer (22). The volume of the sample buffer added to each sample was adjusted so that the concentration of the DNA was proportional to the amount of DNA in the corresponding open complex or free DNA band in the retardation gel. Equal volume of each sample was then loaded onto a 9% polyacrylamide, 8 M urea sequencing gel. After electrophoresis, the sequencing gel was dried and exposed to Kodak X-Omat AR film.
The Rate of Open Complex Formation on Prenicked or Depurinated DNA-The autoradiograms of the sequencing gels resulting from the prenicking and depurination analyses were quantified using scanning densitometry. The technical details of the densitometry analysis including densitometer scanning of the sequencing autoradiograms, background subtraction, and integration of peaks in the densitometer tracings have been described (21). The fraction of open complex formation (F RPo ) for each feature (band or group of bands) was calculated as the integration value of the feature in the lane for an open complex DNA sample divided by the sum of this integration value and the integration value of the corresponding feature in the lane for the free DNA sample at the same time point. The F RPo was normalized by the maximum fractional open complex formation, F max , and plotted as ln(1 Ϫ F RPo ) versus time; the rate of open complex formation, k obs , was calculated from the slope of the plot. The F max used for normalization was the F RPo obtained at 60 min in the gel retardation assay. Using a single value of F max for determining k obs for all of the features was based on the assumption that overall open complex formation was complete at 60 min. This assumption was confirmed by the observation that the maximum values of F RPo for the bands that showed rapid open complex formation ( obs Ͻ 2 min) were equal to F max . However, several exceptions to this rule were found as discussed in the text. In addition, for the bands that appeared with rapid kinetics, k obs was determined by fitting to only the first three time points.
DMS Modification Protection at Guanine Bases-Open complexes were formed by incubating 40 nM RNA polymerase and 0.25 nM labeled DNA for 60 min at 19°C in 185 l of standard reaction buffer. The DMS modification reaction was started with the addition of 1 l of DMS (50 mM), and was stopped 30 s later with the addition of stopping solution (3.5 l of 14 M (Ϫmercaptoethanol, 50 l of 1.5 M sodium acetate, 3 l of 3 mg/ml tRNA, and 700 l of 95% ethanol). The sample was immediately put into a dry ice-ethanol bath. The DNA was isolated by centrifugation, and precipitated again with ethanol. The precipitated DNA was treated with piperidine, lyophilized, and analyzed on a sequencing gel as described above for the depurinated DNA samples.

Analysis of Open Complex Formation by Gel Retardation
Assays-We analyzed open complex formation on the prenicked, depurinated DNA, as well as untreated DNA using the gel retardation method (27,28). The result from such a experiment using prenicked template is shown in Fig. 1. We found that on all three templates, a single band corresponding to open complex was observed, and the fraction of open complex formed following 60 min of incubation was about 0.8 (Ϯ 0.05). In addition, core enzyme and subunit each alone did not bind to the promoter DNA.
To reveal whether the pretreated DNA has an overall effect on the rate of open complex formation, we have analyzed the open complex formation following various incubation times. We found that the rate of open complex formation determined was similar on pretreated DNA as well as normal untreated DNA ( obs between 17 and 21 min), and was comparable to the rate determined from a parallel experiment using the abortive initiation assay ( obs ϭ 14 min, not shown). Thus, prenicking and depurination did not significantly alter the overall process of open complex formation. This is expected if open complex formation were affected by prenicking or depurination at only a limited number of positions.

Strategy for Analyzing the Effects of Prenicking and Depurination at Each Position of Promoter DNA on Open Complex
Formation-After showing that partial prenicking and depurination did not alter the overall rate of open complex formation, we carried out further studies to determine whether prenicking and depurination at specific positions would stimulate or inhibit open complex formation. For this purpose, we first separated the open complexes formed on the pretreated DNA at varying times from free DNA using gel retardation method as described above. The DNA samples from both the open complex and free DNA bands were isolated and analyzed by electro-phoresis on a sequencing gel. The resulting autoradiograms were quantified by densitometry scanning. In these autoradiograms, each band represented a population of DNA molecules that carried a defect (nick or apurinic site) at a certain position. To determine the magnitude of the stimulatory or inhibitory effect of prenicking at each position, we determined the fraction of open complex formed (F RPo ) and rate of open complex formation (k obs ) for each band or group of bands as described above. The average rate of open complex formation (k av ) was determined by quantifying the whole region from Ϫ45 to ϩ45. Fig.  3A shows the plots used to determine k av based on data from both strands. We found that the value of k av derived is comparable to the k obs determined in the gel retardation assay. This indicates that prenicking in the region from Ϫ45 to ϩ45 did not significantly change the overall rate of open complex formation. Fig. 3B shows several plots used to determine the k obs for single bands or small groups of bands. The plots are representative of three types of bands each displaying a k obs very different (larger or smaller) or similar to k av . Three fitted lines representing each type of the bands are shown: the fitted line in the middle was derived from the data for the whole region from Ϫ45 to ϩ45 on the top strand, and corresponded to a k obs (defined as k av ) value of 7.8 Ϯ 0.5 ϫ 10 Ϫ4 s Ϫ1 ; the lower line was fitted to the data for a small group of bands, Ϫ7 to Ϫ9, and corresponds to a k obs of 1.4 Ϯ 0.3 ϫ 10 Ϫ2 s Ϫ1 , which is approximately 20-fold higher than k av; the upper line was fitted to the data for the Ϫ35 band and corresponds to a k obs of 3.4 Ϯ 0.4 ϫ 10 Ϫ4 s Ϫ1 , which is less than half k av . For the bands that displayed rapid kinetics, the rate constants could not be accurately determined; the values reported here are lower estimates. On the other hand, for the bands with slow kinetics, open complex formation might not have reached a final value at 60 min. Consequently, we are not certain whether open complex formation for these bands would reach the same F max that we have used to normalize F RPo in determining the k obs . In our analysis, the effects on both the rate and F max would contribute to the apparent changes in k obs . Consequently, the magnitudes of inhibition on k obs would be slightly overestimated in those cases where smaller F max would be reached.
The effects of prenicking at different positions on the rate of open complex formation are summarized in Fig. 4. The magnitudes of stimulation and inhibition are expressed as R (ϭ k obs /k av ) and Ϫ1/R, respectively, i.e. in terms of how many fold the rate of open complex formation was stimulated or inhibited. It is clear from the figure that prenicking at most positions within the promoter did not significantly affect the rate of open complex formation. However, prenicking at positions around the Ϫ10 region (Ϫ12 to Ϫ1 on the top strand, and Ϫ12 to Ϫ4 on the bottom strand) strongly stimulated open complex formation. The magnitude of stimulation is up to 20-fold. The k obs values for the these positions were up to 8 times larger than the isomerization rate constant k f (1.8 ϫ 10 Ϫ3 s Ϫ1 ) measured on the unmodified template using abortive initiation assays (15), i.e. prenicking at these positions stimulated the isomerization rate constant by at least 8-fold (see "Discussion"). Significant stimulation was not observed outside this region. Position-dependent Effects of Depurination on Open Complex Formation-Representative portions of the sequencing autoradiograms resulting from the depurination studies are shown in Fig. 5. The average rate of open complex formation, k av , was determined by quantifying the region from Ϫ45 to ϩ30 as a whole, and corresponded to a value of k av ϭ 7.0 Ϯ 0.6 ϫ 10 Ϫ4 s Ϫ1 . Again, this value is very similar to that determined in gel retardation assay, suggesting that partial depurination in this region does not have an overall effect on the rate of open complex formation.
As shown for prenicking, even though depurination at most positions did not have a strong effect on open complex formation, at some positions it had a strong stimulatory effect, and at yet some other positions an inhibitory effect was observed. For example, the Ϫ33 band had a k obs of 3.1 Ϯ 0.2 ϫ 10 Ϫ4 s Ϫ1 , which is less than half k av ; the Ϫ9 band had a k obs of 7.0 Ϯ 4.7 ϫ 10 Ϫ3 s Ϫ1 , which is about 10 times larger than k av . As discussed for prenicking studies, the rate of open complex formation may be overestimated for the bands with slow kinetics, while for bands showing rapid kinetics, the rate constants determined are probably lower estimates. The k obs values for these positions were up to 6 times larger than the isomerization rate constant k f , i.e. depurination at these positions stimulated the isomerization rate constant by at least 6-fold as discussed for prenicking. Depurination at positions Ϫ2 and Ϫ3 resulted in some stimulation, but it was of much lower magnitude (2-3-fold). In addition, 2-4-fold stimulation was also observed at positions ϩ19, ϩ20, and ϩ26 on the bottom strand.
Those positions where depurination inhibited open complex formation can also be identified in Fig. 6. The positions showing more than 50% inhibition on k obs were ϩ3 to ϩ5, Ϫ13, Ϫ14, Ϫ16, Ϫ17, Ϫ30, Ϫ32, and Ϫ33 on the top strand, and Ϫ28, Ϫ35, and Ϫ36 on the bottom strand. These positions might be in contact with RNA polymerase during open complex formation (see "Discussion").
DMS Modification Protection at Guanine Bases-We carried out DMS modification protection studies to reveal the RNA polymerase-promoter contacts on the guanine bases in the open complex. As shown in Fig. 7, the guanines at positions Ϫ14 and Ϫ16 on the top strand, and at positions Ϫ3 and Ϫ31 on the bottom strand were the only bases protected from DMS modification. The guanines at positions Ϫ2, Ϫ10, Ϫ17, and Ϫ33 on the top strand were enhanced. These results will be compared with those from the depurination analysis in terms of RNA polymerase-promoter interaction (see "Discussion"). DISCUSSION We have found that both prenicking and depurination around the Ϫ10 region strongly stimulated the rate of open complex formation. This indicates that: (i) the structural instability of the Ϫ10 region is important for promoter function; (ii) the process of destabilizing a small region of promoter DNA around the Ϫ10 region, which we refer to as the nucleation of Instability of the Ϫ10 Region Is Important for Open Complex Formation-Based on the results from promoter mutations, it is clear that base pair identity in the Ϫ10 region is important for RNA polymerase-promoter DNA recognition during open complex formation. Our observation that destabilization of the Ϫ10 region by prenicking and depurination strongly stimulated open complex formation provides strong evidence that the instability of DNA double-helix structure in this region is also important for promoter function. The Ϫ10 region of E. coli promoters is less stable than average DNA (13). In addition, it has been shown that all promoters have at least two of the three highly conserved base pairs, Ϫ12T, Ϫ11A, and Ϫ7T (1). Therefore, the Ϫ10 region of the promoter has two functions. First, the base pair identities in this region are important for RNA polymerase recognition; second, the sequence of this region seems to be optimized for DNA melting (presumably at the step of nucleation) during open complex formation.
That depurination and prenicking can destabilize DNA double-helix structure has been shown in several studies. The effects of an abasic site include significant reduction of T m of oligonucleotides (16), and a free energy loss of 6.5 kcal/mol (17,18). NMR studies have shown that a nick in an oligonucleotide destabilized the DNA with changes in enthalpy and entropy that roughly corresponded to the loss of a single base pair (14). Much stronger effects were observed when a nick was introduced into a dumbbell-shaped, double-hairpin molecule based on thermodynamic studies (15).
Mechanism of DNA Melting-The observation that the rate of open complex formation can be significantly stimulated by destabilizing a specific region of promoter DNA indicates that destabilization of this region may normally contribute to the rate of open complex formation. We refer to this destabilization of a small region in promoter DNA during open complex formation as the nucleation of DNA melting. Consequently, the region where stimulation was observed is taken as a rough estimate of the nucleation region. Based on our results, the nucleation process has the following characteristics: First, the nucleation is at least part of the isomerization step from the closed complex to the open complex. This suggestion is supported by our observation that depurination and prenicking stimulated the rate constant, k f , for the isomerization step by as much as 8-fold. The actual stimulatory effect on k f must be many times larger than what was observed because helical defects in the Ϫ10 region are expected to have two opposite effects: one is the stimulatory effect caused by destabilization of the DNA double-helix structure, and the other is the inhibitory effect caused by the elimination of the interactions between RNA polymerase and the promoter DNA at the depurinated positions.
Roberts and Roberts (29) have identified the non-template strand as being responsible for the sequence-specific interaction of RNA polymerase in the Ϫ10 region of P R Ј. The lack of stimulation by depurination at the Ϫ11 position observed here is consistent with their proposal. Part of the stimulation they observed on heteroduplex templates with mismatched base pairs is likely due to the effect of helical defects described here.
Second, the nucleation occurs in a relatively small region that overlaps the Ϫ10 region. It is known that upon open complex formation, the region from the Ϫ10 hexamer sequence to the transcription start site becomes single stranded (7, 30 -32). Based on KMnO 4 analysis, we have estimated that the minimum size of the DNA melting region on the prmup-1⌬265 promoter is 13 base pairs extending from Ϫ11 to ϩ2. 2 21.3 min). B, the rate of open complex formation of single bands or small groups of bands. The results shown (selected from 37 such plots) are representative of the three major classes of bands whose k obs were significantly larger than, smaller than, or similar to k av . The bands corresponding to the symbols are: Ⅺ, Ϫ18 and Ϫ19; ‚, Ϫ35; q, ϩ5 to ϩ8; OE, ϩ37 to ϩ39; f, Ϫ31 and Ϫ32; E, Ϫ7 to Ϫ9; छ, Ϫ10 to Ϫ12. nucleation region (where stimulation was observed) is smaller than the DNA melting region. Immediately downstream from the assigned nucleation region is a d(G-C) dinucleotide, which is thermodynamically very stable. The finding that depurination at these positions did not have a strong effect on open complex formation strengthens the idea that the nucleation occurs in a discrete region that does not extend to this position. Additional evidence for a stepwise process in promoter DNA melting comes from the protection studies of Chen and Helman (3), and from the characterization of an RNA polymerase mutant that melted promoter DNA in discrete steps (6).
The results and interpretation of Werel et al. (33) appear to argue for a larger region involved in the putative nucleation function. Their use of the T7 A1 promoter and pretreatment with hydroxyl radical do not allow a detailed or direct comparison with our results. Moreover, Werel et al. did not measure rates of association to the gapped templates they prepared. Instead, overall "affinities" were scored after a long dialysis step against TE buffer, and consequently they might have followed an effect on dissociation of preformed open complex. Probably for this reason, the stimulation observed for gaps in the melting region was modest ranging from about 2-fold to 5-fold at 22°C and up to 10-fold at 4°C. Nevertheless, it appears likely that a similar effect of DNA helical defects is responsible for our results and those reported by Werel et al.
Although we have shown here that the nucleation region includes the whole Ϫ10 RNA polymerase-recognition region and several additional base pairs as well, in other cases the recognition region and the nucleation region may be separable. For example, the Ϫ10 region of the promoters recognized by 32 holoenzyme can be divided into two segments, CCCC and ATt(Ϫ10)Aa (lowercase letters indicate weak conservation). It has been shown that the guanine residues of the first segment are all in contact with RNA polymerase (34). By analogy to the promoters recognized by 70 holoenzymes, the second segment but not the first would be melted in the open complex. Therefore, it seems that the first thermodynamically stable segment would be responsible for RNA polymerase recognition, while the second segment may have to do with the nucleation of DNA melting although it may also be involved in RNA polymerase recognition and binding. Thus, in this case the nucleation region may not overlap the recognition region completely. Similarly, the consensus sequence of the promoters recognized by the T7 RNA polymerase is about 20 base pairs long, but less than half of the sequence is melted in the open complex. The DNA melting region is centered around a TATA sequence. This sequence is similar to the Ϫ10 region of the promoters recognized by 70 holoenzyme, and consequently may be the nucleation site of DNA melting. Interestingly, Jorgensen et al. (35) have found that depurination in the DNA melting region of the T7 f10 promoter also enhanced the binding efficiency of T7 RNA polymerase. Although the sequences recognized by different factors have diverged during evolution, they may have maintained small segments that are suitable for DNA melting.
Third, the observation that only prenicking and depurination in a small region stimulated open complex formation suggests that propagation of DNA melting is not rate-limiting. If DNA melting in the region downstream from the nucleation region were rate-limiting, destabilization of the downstream region by prenicking or depurination should also significantly increase the rate of open complex formation.
It has been suggested that the isomerization from the closed complex into the open complex involves two discrete steps: the isomerization per se and DNA melting (4, 8 -10). DNA melting is not rate-limiting at higher temperatures, but may become rate-limiting at lower temperatures. We have argued above that the propagation of DNA melting from the nucleation region is not rate-limiting even though our studies were carried out at 19°C. This is consistent with the following findings. 2 The intermediate complex preceding DNA melting did not accumulate, even at a lower temperature (15°C); the apparent activation energy for open complex formation in the temperature range of 15°C to 25°C is only about 20 kcal/mol, which is similar to the activation energy observed on the P R promoter  at higher temperatures (10). Therefore, the stimulation by depurination and prenicking on the rate of open complex formation reported here was at the isomerization-nucleation step rather than the DNA unstacking step. Our suggestion that nucleation contributes to the rate of isomerization and that the properties of structural stability in the Ϫ10 region is important in this process is complementary to the suggestion of Roe et al. (10) that a major conformational change in RNA polymerase may occur in the isomerization step.
Destabilizing the double-helix structure of the Ϫ10 region might facilitate open complex formation by decreasing the activation energy for the nucleation. If this is true, the observed magnitude of stimulation (up to 20-fold) resulting from prenicking and depurination would correspond to a change in activation energy of about 1.7 kcal/mol. This is only a small portion of the total activation energy for open complex formation, which is 20 kcal/mol for this promoter.
RNA Polymerase-Promoter Interactions-We have found that depurination at several positions had an inhibitory effect on the rate of open complex formation, presumably by disrupting protein-DNA interactions. Five of these positions are located in the Ϫ35 region of the promoter, and are expected, considering the importance of base identities of this region in promoter function. Interestingly, removal of Ϫ31G by depurination did not have much effect on open complex formation even though this base was protected in the open complex from DMS modification on this promoter, as well as several other promoters (36,37). A possible explanation is that the protection resulted from an indirect effect rather than a specific contact to the N7 group of Ϫ31G by RNA polymerase.
Two of the bases (Ϫ14G and Ϫ16G), whose removal by depurination showed inhibitory effects on open complex formation, were also protected from DMS modification in the open complex. This suggests that these two bases may have contacts with RNA polymerase during open complex formation, in agreement with the results of Michin and Busby (38) on the gal P1 promoter. The removal of Ϫ33G and Ϫ17G by depurination also significantly inhibited open complex formation, indicating that these bases are in direct interaction with RNA polymerase. In the DMS protection experiment, however, we observed enhancement at these positions. A possible explanation that is consistent with both results is that RNA polymerase makes contact with the O6 groups of these two bases so that their N7 groups are still available for DMS methylation in the open complex. This type of protein-DNA contact at the O6 group of a guanine residue has been shown to exist in the interactions of repressor (39) and phage 434 repressor (40) with their respective operator site. Interestingly, the positions Ϫ14, Ϫ15, and Ϫ17 showed weak conservation in the compilation of the sequences of known promoters; and mutations have been isolated at positions Ϫ14, Ϫ15, and Ϫ16 on several promoters (41). This evidence and our results suggest that the contacts of RNA polymerase at the positions in the region from Ϫ14 to Ϫ17 may be important for open complex formation on this and other promoters.
We have also observed an inhibitory effect upon depurination at two other guanine positions (ϩ3 and ϩ5) where DMS modification protection or enhancement in the open complex was not observed. It is possible that, as mentioned above, RNA polymerase might make contact with the O6 group but not the N7 group of these guanine residues so that DMS modification, which occurs at N7 of guanine bases, would not be blocked.
Prenicking Also Inhibited Open Complex Formation-We found that prenicking at several positions inhibited open complex formation slightly. A possible explanation is that prenick-ing at these positions altered DNA structural properties such as flexibility. It has been suggested (42)(43)(44)(45) that there might be a DNA rotational change, or the formation of other DNA structural stress, during open complex formation. The DNA structural stress was argued to facilitate DNA melting. Therefore, prenicking might have inhibited open complex formation by eliminating the DNA stress. An example showing that a nick can relieve DNA torsional stress and consequently alter protein-DNA binding affinity came from a study by Koudelka et al. (19). It was shown that a nick in the middle of the operator DNA increased the phage 434 repressor binding, presumably by relieving a structural stress resulting from DNA bending caused by the repressor binding. An alternative explanation for our observations is that prenicking might have caused disruption of RNA polymerase-promoter interactions by altering DNA structure around the nick. It has been shown that nicks can cause slight distortion in DNA structure (14). This explanation is consistent with the observation that the positions showing inhibitory effects are within or close to protected regions in the hydroxyl radical footprint of the open complex.
Conclusion-Our finding that depurination or prenicking in the Ϫ10 region greatly stimulated open complex formation suggests that the intrinsic instability of this region is important for promoter activity. Our results also suggest that the nucleation of DNA melting, i.e. the destabilization of a small region of DNA, contributes to the rate of open complex formation. Nucleation may occur around the Ϫ10 region. Moreover, we have also shown that depurination at some positions had an inhibitory effect on open complex formation, indicating that these positions are important for open complex formation on this promoter. About half of these positions were found to be in the Ϫ35 region, which is consistent with the importance of base pair identity in this region for promoter function. Therefore, our results support the following model: both regions of the promoter are important in the direct interactions with RNA polymerase, whereas, the DNA melting free energy around the Ϫ10 region but not that of the Ϫ35 is important for open complex formation.