Influence of Mg2+ and temperature on formation of the transcription bubble.

The transcription bubble formed in the binding complex of T7A1 promoter upon Escherichia coli RNA polymerase was analyzed by chemical probes, namely by single-strand specific reagents, to map the unpaired bases in the bubble, and by FeEDTA, to analyze the accessibility of the DNA backbone. The latter probe could also be used as a local hydroxyl radical probe placed close to the Mg2+-binding site in the active center. The data show that the transcription bubble consists of two parts, an Mg2+-dependent part and an Mg2+-independent part, both having individual transition temperatures. The data further suggest that formation of a transcription active open complex is preceded by a transition state complex having enhanced affinity for those Mg2+ ions presumably participating in the formation of the catalytic site. Our data also suggests that the three catalytically active Mg2+ ions in RNA polymerase are functionally not equivalent. One/two of the three Mg2+ ions are responsible for the polymerization, the other two/one for enlargement of the transcription bubble.

The transcription bubble formed in the binding complex of T7A1 promoter upon Escherichia coli RNA polymerase was analyzed by chemical probes, namely by single-strand specific reagents, to map the unpaired bases in the bubble, and by FeEDTA, to analyze the accessibility of the DNA backbone. The latter probe could also be used as a local hydroxyl radical probe placed close to the Mg 2؉ -binding site in the active center. The data show that the transcription bubble consists of two parts, an Mg 2؉ -dependent part and an Mg 2؉independent part, both having individual transition temperatures. The data further suggest that formation of a transcription active open complex is preceded by a transition state complex having enhanced affinity for those Mg 2؉ ions presumably participating in the formation of the catalytic site. Our data also suggests that the three catalytically active Mg 2؉ ions in RNA polymerase are functionally not equivalent. One/two of the three Mg 2؉ ions are responsible for the polymerization, the other two/one for enlargement of the transcription bubble.
The transcription bubble is a characteristic attribute of the eubacterial transcription-competent complex. Although no three-dimensional structural information on the transcription bubble is yet available, one can visualize it as a region in which the DNA strands are separated in order to facilitate "reading" of the sequence information for RNA synthesis. The transcription bubble is initially formed upon polymerase binding upstream of the start point of RNA synthesis (1)(2)(3). It changes its size and position when RNA synthesis proceeds, as analysis in subsequent steps of RNA synthesis between registers 11 and 20 has revealed (4). The position of the upstream lagging end of the transcription bubble remains constant in the registers 11-18, while the position of the downstream lagging end moves downstream in accordance with progress of RNA synthesis. In this way the size of the transcription bubble increases from 11 base pairs in the 11-mer complex to 18 base pairs in the 18-mer complex. If RNA synthesis proceeds beyond register 18, the transcription bubble collapses in the upstream region. As a consequence, it shrinks to its initial size of 11 base pairs. While information about the transcription bubble in the RNA synthesizing complex is rather detailed, there is no correspondingly large body of information on the transcription bubble in the binary complex.
The concept of the transcription bubble (5,6) was developed from the finding that the transcription active complex, the open complex, is preceded by the closed complex in which the promoter DNA is assumed to be base paired (7). These two complexes are distinguished from each other by their characteristic temperature sensitivity toward heparin. The closed complex exists below 17°C and is heparin-sensitive (8), while the open complex exists above 17°C and is heparin-resistant. The validity of this concept was proven by showing that the temperature dependence of DNA strand separation and heparin sensitivity can be correlated (7,9). Isomerization of the binding complex is a reaction scheme which holds good for most factor-independent promoters. This scheme had to be modified by the finding that the RNA polymerase promoter complex undergoes at least two further isomerization steps before the final transcription active open complex is formed. These transition complexes are the intermediate complex (10 -12) and the Mg 2ϩ -independent and the Mg 2ϩ -dependent complexes (13).
Transition state complexes have been defined by kinetic studies in which concentrations of the reactants, temperature, and ionic strength were varied (10,13). It is plausible that by lowering the temperature, transition state complexes can be "frozen" and accumulated. However, correspondence between time-dependent and temperature-dependent isomerization steps has not been convincingly demonstrated. Having these reservations in mind, chemical probing studies at different temperatures can provide useful information on structural changes of transition state complexes. It has been shown that RNA polymerase has contact with the promoter in the closed complex from base position Ϫ53 to Ϫ4 (14 -16). In the intermediate complex the protected region is extended downstream to position ϩ21 (14,15).
Probing of the A1 promoter using FeEDTA-generated hydroxyl radicals has revealed that closed and intermediate complexes differ not only in size but also in the kind of interaction. In the closed complex the polymerase faces one side of the DNA (14,15), a result which is in line with neutron solution studies (17). In the intermediate complex, domains of RNA polymerase wrap around the DNA in the region between base position Ϫ15 and ϩ18 covering both DNA strands and including the starting point of RNA synthesis (14,15), a finding which is in line with information obtained from electron microscope studies (18). The footprint of the open complex visible above 30°C does not differ from that of the intermediate complex except for a region of the template strand at base position Ϫ1 which shows enhanced reactivity toward FeEDTA. It has been suggested by Schickor (14) that this hyper-reactivity reflects enhanced accessibility of the template strand due to DNA strand separation. Record and colleagues (16) found a similar effect in the Mg 2ϩ containing complex and explained it by a charge effect. We show here that the enhanced cleavage is due to interaction of the FeEDTA probe with the RNA polymerase promoter complex. FeEDTA acts as local hydroxyl radical source, presumably by binding at or near the site occupied by the catalytically active Mg 2ϩ , similar to the recent observation with Fe 2ϩ (19).
A previous study concerned Mg 2ϩ -dependence of the transcription bubble of P r , a promoter which in order to form a transcription active complex, requires supercoiling of the promoter DNA. Our study focuses an the Mg 2ϩ effect on the linear promoter A1. Using this promoter, Escherichia coli RNA polymerase forms a transcription active complex at 37°C with promoter DNA in the relaxed state. In this study three different probes for analysis of unpaired thymidines, adenines, and cytidines were applied, allowing us to determine the size of the transcription bubble precisely within 1 base pair.

EXPERIMENTAL PROCEDURES
RNA polymerase was isolated from E. coli as described in Ref. 17. Promoter fragment A1-130 containing the sequence of the T7A1 promoter between Ϫ69 and ϩ61 was prepared as described (20). The 3Ј-end labeling and 5Ј-end labeling was performed as described in Ref. 14. The radioactivity of the preparations was typically 1 Ci/pmol.
For some experiments promoter fragment A1-220 (representing A1 promoter sequence between Ϫ96 and ϩ123) was used. End labeling was performed by filling the BamHI termini with deoxy-[␣-32 P]NTP, subsequent removal of the short fragments near one of the termini with either HaeIII (non-template strand labeling) or AluI (template strand labeling), and final purification of the end-labeled promoter fragment with QIAGEN PCR purification kit. Sequencing ladders were obtained by AϩG cleavage with formic acid and AϾC cleavage with NaOH (21).

Formation of Binary Complexes
A1 promoter fragment (1 g of A1-130 or 2 g of A1-220 containing typically 10 5 cpm of the corresponding radioactive fragment) and 12 g of RNA polymerase were incubated for 5 min at 37°C in 50 l of 8 mM Hepes, pH 8, containing 6 mM MgCl 2 . The reaction mixture was then dialyzed for 1-2 h against proper buffer using the floating membrane filter technique (Millipore, VS 0.025 m).

Hydroxyl Radical Footprinting using FeEDTA
Hydroxyl radical cleavage was conducted in a similar way as described in Ref. 9. Three drops, 1.2 l each, of 2 mM (NH 4 ) 2 Fe(SO 4 ) 2 , 4 mM EDTA, 3% H 2 O 2 , and 0.1 mM dithiothreitol were separately placed onto the inner wall of a Eppendorf tube. The cleavage was started by simultaneous mixing of them with 10 l of binary complex formed as described above. Cleavage was performed for 4 min at 37°C. The reaction mixture was then quickly passed through a nitrocellulose filter (13-mm Sartorius) and washed with 200 l of 8 mM Hepes, 50 mM NaCl in order to get rid of free DNA. The DNA was eluted from the filter using a 100-l solution containing 1% SDS, 0.3 M sodium acetate, and 0.1 mg/ml carrier DNA (15 min at 37°C) and precipitated with 300 l of ethanol. The pellet was dissolved in 80% formamide containing 0.02% bromphenol and xylene cyanol. The solution was heated for 2 min at 95°C and applied on a 8% sequencing gel. As a control, free DNA was cleaved. For that purpose the polymerase promoter complex was destroyed by addition of SDS, final concentration 0.5%, and subsequently subjected to all cleavage reactions, as described above. The reaction was stopped by addition of 15 l of 2% glycerol, 0.6 M NaAc, 0.1 mg/ml carrier DNA and by precipitation with 90 l of ethanol.

Hydroxyl Radical Footprinting using Potassium Peroxonitrite (KOONO)
The cleavage with potassium peroxonitrite was performed as described in Ref. 22. Binary complex was dialyzed against 0.05 M sodium cacodylate buffer, pH 7.2. One l of 60 mM KOONO in 0.3 M KOH was added to 20 l of a solution containing the binary complex (final pH of the solution was pH 8). After 30 s of the incubation (the reaction is finished actually within a few seconds) the carrier DNA was added and the mixture precipitated with ethanol.

Single-strand Probing
The reaction was performed in two steps, first the single strand specific reagent was added for modification of the corresponding bases and subsequently the DNA was cleaved at the modified base by applying piperidine.

Modifications
Thymidines Using Osmium Tetroxide (OsO 4 )-One l of a freshly prepared solution of 10 mM OsO 4 , 15 mM bipyridine was added to 10 l of binary complex. The reaction mixture was incubated at 37°C for 2 min. The cleavage reaction was stopped by addition of 10 l of 0.1 M dithiothreitol, 0.6 M NaAc, containing 0.1 mg/ml carrier DNA and precipitation with 60 l of ethanol.
Thymidines Using Potassium Permanganate (KMnO 4 )-One l of freshly prepared 0.1 M KMnO 4 solution was added to 10 l of binary complex. The reaction mixture was incubated at 37°C for 1 min. The cleavage reaction was stopped by addition of 10 l of 0.1 M dithiothreitol, 0.6 M NaAc, containing 0.1 mg/ml carrier DNA and precipitation with 60 l of ethanol.
Adenines Using Diethyl Pyrocarbonate-One l of freshly prepared 3% diethyl pyrocarbonate was added to 10 l of binary complex. The reaction mixture was incubated for 4 min at 37°C. The cleavage reaction was stopped by addition of 10 l of 50 mM imidazole, 0.6 M NaAc, pH 7, 0.1 mg/ml DNA and precipitation with 60 l of ethanol.
Cytidines Using Dimethyl Sulfate and Hydrazine (Hz)-The cleavage was conducted according to Ref. 23. At first, DNA in the binary complex was methylated by addition of 1 l of 100 mM dimethyl sulfate to 10 l of the binary complex. The reaction mixture was incubated for 2 min at 37°C. The methylation was stopped by addition of an equal volume of 0.1 M imidazole, 0.6 M NaAc, containing 0.1 mg/ml carrier DNA. The DNA was precipitated with 60 l of ethanol and dissolved in 20 l of water. Subsequently 10 l of 1-butanol and 20 l of hydrazine were added in order to cleave the methylated cytosines. The reaction mixture was incubated at 0°C for 5 min. After precipitation with ethanol, the DNA was subjected to piperidine treatment.

Piperidine Treatment
The pellet was dissolved in 90 l of 10% piperidine and incubated at 90°C for 20 min. Ten l of 5 M LiCl was then added and DNA precipitated with 300 l of ethanol. The precipitate was dissolved in 5-10 l of 80% formamide containing 0.02% of bromphenol and xylene cyanol.

Gel Electrophoresis
The samples were heated at 90°C for 2 min, chilled on ice, and then applied on 22 ϫ 50-cm (wedge thickness from 0.2 to 0.4-mm) slab of 6 or 8% polyacrylamide gel containing 8 M urea and 50 mM TBE buffer. Gels were run at 55 W for 1-1.5 h using heating plates (55°C).
The gels were incubated for 15 min in 10% acetic acid, then washed twice with water (total washing time: 1 h), dried, and exposed to x-ray film (3 M). For quantification, gels were exposed to Fuji Imager plate BAS IIIS, which was scanned with Bas-1000 PhosphorImager. Scans were processed with MacBas software.

Temperature Profile of the Activity
Binary complexes were prepared at different temperatures in the presence of MgCl 2 , as described above. One l of "initiation mixture" (1 mM ApUpC ϩ 1 mM GTP ϩ 1 mM CTP ϩ 0.1 mM ATP ϩ Ϸ100 nCi [␣-32 P]ATP) was added to 10 l of the binary complex and incubated for 5 min at the same temperature (initial experiments have shown that the kinetics is linear for at least 10 min). The reaction was stopped by addition of 10 l of 20 mM EDTA in formamide, and 5 l was analyzed on 20% sequencing gel. The gel was visualized by PhosphorImager, and the bands corresponding to 20-mer product were quantified.

RESULTS
E. coli RNA polymerase was incubated at 37°C with DNA containing the sequence of the strong promoter A1 of the phage T7, as described under "Experimental Procedures." The complex was subjected to analysis with two kinds of probes, namely single strand-specific reagents in order to map the transcription bubble, and FeEDTA as a hydroxyl radical source in order to map the RNA polymerase on the DNA. The DNA was subsequently analyzed on a sequencing gel to determine the cleavage sites.
Mg 2ϩ -dependent Changes of the Transcription Bubble-Three types of single strand-specific reagents were applied, namely (a) dimethyl sulfate, which methylates cytosines at the N-3 position, thus activating them toward hydrazine (Hz) (23); (b) OsO 4 , which oxidizes thymidines at the C5-C6 double bond (24), (c) diethyl pyrocarbonate, which attacks purines with strong preference for adenines at the N-7 position (25). KMnO 4 was also used as an alternative to OsO 4 .
Two sets of experiments were performed with the radioactive label either in the template or in the non-template strand, each consisting of three samples, namely DNA (a) without RNA polymerase as reference, (b) in the complex with RNA polymerase without Mg 2ϩ , and (c) in the complex with RNA polymerase and Mg 2ϩ . Each of these samples was treated with OsO 4 , diethyl pyrocarbonate, and dimethyl sulfate/Hz, then cleaved at the modified nucleotides with piperidine and analyzed on a sequencing gel, as described under "Experimental Procedures." Since dimethyl sulfate modifies guanines with high yield, probing by dimethyl sulfate or Hz alone was included as a control experiment. Fig. 1 shows the results obtained at the template strand ( Fig.  1A) and the non-template strand (Fig. 1B). The accessible bases indicated in Fig. 1, A and B, are shown in a schematic representation in Fig. 2. Accessibility is defined here as enhanced reactivity of a base upon polymerase binding. A base pair is considered as being part of the transcription bubble if one of the two complementary bases is accessible.
Based on the above definition, the pattern obtained by each reagent provided an upper and a lower limit of the transcription bubble at both ends. By combining the results from all three reagents, the bubble could be located accurately within 1 base pair. Fig. 2 shows that the upstream lagging end of the bubble is situated at base position Ϫ12 and the downstream lagging end at base position Ϫ1, if the incubation of the complex was performed in the absence of Mg 2ϩ .
Comparison of the accessibility patterns with and without Mg 2ϩ (Fig. 2) shows that Mg 2ϩ has two effects. In the presence of Mg 2ϩ the transcription bubble is enlarged downstream to base position ϩ2 and the interaction between RNA polymerase and DNA is enhanced, as indicated by a reduced modification of some bases. Thus A(Ϫ4) and A(Ϫ6) in the non-template strand ( Fig. 1B) are protected in the Mg 2ϩ containing complex and modified in the Mg 2ϩ free complex. The thymidine at base position Ϫ8 in the non-template strand shows a similar, although less pronounced, effect to that observed with the adenines mentioned above. T(Ϫ8) is moderately accessible for OsO 4 in the Mg 2ϩ free complex and not accessible in the Mg 2ϩ containing complex. There is an exception to this scheme. Cytidine at position Ϫ3 in the template strand is inaccessible without Mg 2ϩ and moderately accessible if Mg 2ϩ is added.
Guanine is the only base for which no probe has been suggested which would allow a differentiation between paired or unpaired state. Dimethyl sulfate is assumed to methylate guanines irrespective of whether they are base paired or not. Our probing studies using dimethyl sulfate alone show that this is not correct. Dimethyl sulfate has a preference for guanines in the single stranded region, as indicated by the improved cleavage of G(Ϫ3) in the non-template strand and G(Ϫ1) and G(Ϫ2) in the template strand, as the patterns in Fig. 1, A and B, show.
G(Ϫ9) in the template strand deviates from the pattern described above in that it is protected, while the complementary C(Ϫ9) is accessible. This might be due to a similar effect to that observed with A(Ϫ4) and A(Ϫ6). These bases are fully or partly protected, although the complementary base is accessible, reflecting perhaps improved interaction with RNA polymerase. G(Ϫ5) at the template strand is not affected at all by RNA polymerase binding, either in the presence or absence of Mg 2ϩ .
It is interesting to note that some of the guanines located outside of the transcription bubble show differences in the accessibility depending on whether Mg 2ϩ is present or not. G(Ϫ13) and G(Ϫ14) in the non-template strand are protected, indicating strong interaction of these bases with RNA polymerase. It is surprising that G(ϩ12) in the template strand is enhanced, although this base is not part of the transcription bubble according to the criteria developed above. We assume that this is due to distortion, perhaps bending of the DNA.
Analysis of the Transcription Bubble at Different Temperatures-The mapping studies on the transcription bubble in the previous paragraph allowed differentiation between two parts of the transcription bubble, namely an Mg 2ϩ -independent part reaching from base position Ϫ12 to Ϫ1 and an Mg 2ϩ -dependent part reaching from base position ϩ1 to ϩ2. In order to analyze whether the two parts show different melting behaviors, the Mg 2ϩ containing complex was analyzed in the temperature range from 4 to 37°C. For the purpose of this analysis, it was sufficient to probe only the accessibility of thymidines, since the Mg 2ϩ dependent part of the transcription bubble comprising the bases at position ϩ1 and ϩ2 (see Fig. 2) consists of AT pairs only.
The binding complex was formed in the presence of Mg 2ϩ , as described above, and subjected to treatment with KMnO 4 or OsO 4 at the temperatures indicated. The resulting accessibility patterns obtained at the non-template strand and the template strand are shown in Fig. 3A and Fig. 4A, respectively. The thymidine patterns obtained at different temperatures were quantified and the resulting intensities were plotted, in Fig. 3B for the template strand and in Fig. 4B for the nontemplate strand. The plot revealed that the thymidines have different transition temperatures (T m ) which are compiled in Table I along with the corresponding enthalpies (⌬H). Two groups of thymidines could be identified according to their transition temperatures, namely thymidines belonging to the Mg 2ϩ -dependent and those belonging to the Mg 2ϩ -independent part of the transcription bubble. The transition temperatures of the thymidines in the Mg 2ϩ -independent part are between T m ϭ 10 -15°C, while those of the Mg 2ϩ -dependent part are shifted to higher temperature by about 15°C, as shown in Table I. The positive enthalpy values for the transition indicate that the formation of the transcription bubble is enthalpy driven, in line with a suggestion from Suh et al. (13).
It is interesting to note that the base at the starting point of RNA synthesis, T(ϩ1), has a transition temperature of T m ϭ 29°C, whereas T(ϩ2) located further downstream has a lower transition temperature of T m ϭ 22°C. In this context it is worth noting that the temperature dependent accessibility of T(ϩ1) correlates with the temperature dependence of the initiation reaction for RNA synthesis, as shown in Fig. 4B.
In order to analyze whether charge or shielding effects, especially in the presence of Mg 2ϩ , influence the cleavage pattern, we compared the patterns obtained with KMnO 4 and OsO 4 . Both reagents attack the same positions of thymidines, but have different charges and sizes. Probing of the thymidines of the non-template strand using both probes (Fig. 3A) revealed that the pattern is qualitatively the same but differs quantitatively. When KMnO 4 is used, the yield of cleavage is generally higher, and moreover cleavage of T(Ϫ12) in the template strand is more pronounced. Otherwise no difference was observed between the two reagents with respect to temperature or Mg 2ϩ dependence of the transcription bubble.
For comparison purposes, the transition temperatures of the Mg 2ϩ -dependent and the Mg 2ϩ -independent part of the transcription bubble of another eubacterial RNA polymerase, Thermotoga maritima, a thermophilic organism, was included in Fig. 4C. The data were taken from a previously published study (26) on the Mg 2ϩ containing and a Mg 2ϩ -free binding complex of T. maritima RNA polymerase and the A1 promoter. The size of the transcription bubble of the thermophilic RNA polymerase differs from that of E. coli RNA polymerase, but the finding that the transcription bubble can be subdivided in an Mg 2ϩdependent and Mg 2ϩ -independent part having characteristic transition temperatures is the same in both systems (26). Of course, according to the thermophilic nature of T. maritima the transition temperatures are shifted to higher values, as shown in Table I.
Probing of the Binding Complex using FeEDTA-generated Hydroxyl Radicals-The binding complex of RNA polymerase and the A1 promoter is formed as described in the previous paragraph with and without Mg 2ϩ and subjected to footprint- The complex was formed in the presence of Mg 2ϩ , as described under "Experimental Procedures." It is visualized by labeling the 3Ј terminus of the non-template strand. A shows the pattern obtained after electrophoresis, the plot in B shows quantitatively the modification of the thymidines by KMnO 4 , as indicated, depending on temperature. The intensities obtained by PhosphorImaging were corrected for background intensity and normalized with respect to the applied total radioactivity on each lane. The curves were fit by applying van't Hoff equation, as described previously (36).
The probing pattern of the Mg 2ϩ -free binding complex shows, besides the footprint of RNA polymerase, an enhanced cleavage at position Ϫ1 in the template strand, previously discovered by Schickor et al. (14). This hyper-reactive region is located within a stretch of fully protected DNA reaching from base position Ϫ13 to ϩ13. This enhanced cleavage is also observed, although less pronounced, in the Mg 2ϩ -containing complex. We were interested in finding out the origin of this hyperreactivity, especially since its temperature behavior (14) is similar to that observed with T(ϩ1) in the above described Mg 2ϩ -containing complex. In order to understand the reason for the enhanced cleavage of the template strand around position Ϫ1, it is worth recalling how the hydroxyl radicals were generated, namely according to the Fenton reaction by reduction of H 2 O 2 with Fe 2ϩ . The resulting Fe 3ϩ is reduced by dithiothreitol back to Fe 2ϩ , starting a new cycle. EDTA is necessary in order to prevent Fe 2ϩ binding to DNA (27). A prerequisite for obtaining a footprint which reflects the contact sites between RNA polymerase and DNA is spatial independence of the hydroxyl radical source and the probed DNA molecule (14,27). This is achieved if the hydroxyl radicals stem from molecules freely diffusing in solution. An enhancement of the cleavage beyond that expected for free DNA could indicate activation of sugars by interaction with the enzyme or by a distortion of DNA (e.g. due to bending). An alternative explanation could be an increase of the local concentration of hydroxyl radicals due to binding of the hydroxyl radical source, similar to what was reported for, e.g. OpCuphenanthrolin (29). In order to determine which of these possibilities is correct, other OH radicals generating systems, namely KOONO and Fe 2ϩ , were applied and compared with the FeEDTA footprints.
(i) Hypercleavage Around the Start Point Is due to Bound FIG. 4. Temperature dependence of the thymidines accessibility in the template strand. The lanes show the accessibility patterns of thymidines in the temperature interval between 0 and 45°C, as indicated, by applying KMnO 4 . All other conditions are the same, as described in the legend to Fig. 3 for the non-template strand. The plot in C shows a quantification of the accessibility studies on the template strand using the same promoter A1 but upon binding of polymerase from T. maritima, a thermophilic organism. The data were taken from Ref. 26.  FeEDTA-It has previously been reported that potassium peroxonitrite (KOONO) produces hydroxyl radicals as freely diffusing molecules (22,30) and can thus be used as an alternative to FeEDTA. We used KOONO in order to decide whether a component of the Fenton reaction is responsible for the observed hyper-reactivity. We subjected the binding complex formed with and without Mg 2ϩ to KOONO treatment (Fig. 6, A,  lane 5, and B, lanes 3 and 4). The patterns obtained by using both hydroxyl radical generating reagents is essentially the same with exception of the hyper-sensitive spot. We conclude from this finding that a component of the Fenton reaction, namely FeEDTA, acts as local hydroxyl radical source.
In order to determine which part of the pattern can be attributed to hydroxyl radicals stemming from freely diffusing FeEDTA and which part to those stemming from bound FeEDTA, the Mg 2ϩ -free binding complex was subjected to FeEDTA treatment in the presence of glycerol. Glycerol is a hydroxyl radical scavenger which can absorb diffusing hydroxyl radicals (28). Lane 3 of Fig. 6A shows the pattern obtained. As expected, the typical RNA polymerase footprint is blurred, because the hydroxyl radicals generated in solution are captured by glycerol. On the other hand, the hyper-sensitive spot remains unchanged, as comparison with the glycerol free complex in Fig. 6A, lane 2, indicates. From this finding we conclude that the hyper-reactive spot generated by a hydroxyl radical source specifically bound to the binding complex is shielded against access by glycerol.
(ii) FeEDTA Competes with Mg 2ϩ for a Binding Site in the Transcription Complex near Base Position Ϫ1-As seen at Fig.  5, Mg 2ϩ ions specifically inhibit the hyper-cleavage of the template strand centered at base position Ϫ1, while the footprint outside the hyper-sensitive part remains unchanged. Fig. 7 shows the dependence of the hyper-reactivity on the Mg 2ϩ concentration in the range of 0 -10 mM. In the absence of Mg 2ϩ the cleavage intensity exceeds that of free DNA by a factor of 10. The hyper-reactivity decreases with increasing Mg 2ϩ concentration, indicating displacement of FeEDTA by Mg 2ϩ . But whether the displacement is competitive or noncompetitive is not clear from this curve. The cleavage intensity decreases to 50% at a Mg 2ϩ concentration of about 0.5 mM comparable with the applied FeEDTA concentration, indicating that both Mg 2ϩ and FeEDTA have an affinity constant of the same order of magnitude.
It appears that hyper-cleavage near the Ϫ1 position cannot be suppressed fully even at the highest MgCl 2 concentration. This fact along with the finding that the KOONO-generated hydroxyl radical pattern shows no hyper-cleavage either in the absence or presence of Mg 2ϩ (Fig. 6B, lanes 3 and 4) suggests that residual cleavage in the presence of Mg 2ϩ is also caused by FeEDTA, indicating that FeEDTA can bind but with reduced affinity even in excess of Mg 2ϩ . This result is in line with a finding by Craig et al. (16) on the P r , who showed slightly enhanced cleavage near the Ϫ1 position in the pre- sence of Mg 2ϩ .
(iii) FeEDTA Binds Adjacent to the Binding Site of Fe 2ϩ -It has previously been shown that Fe 2ϩ binds to the RNA polymerase active site by replacing Mg 2ϩ and produces hydroxyl radicals which cleave specifically the peptide chains of RNA polymerase as well as the template DNA of the promoter around position Ϫ1 (19). The authors conclude that Fe 2ϩ is bound by chelate formation with aspartates of the NADFDGD sequence of the ␤Ј subunit. This conclusion is based on the finding that a mutant RNA polymerase (DDD mutant), in which the aspartates are replaced, does not bind Fe 2ϩ . Support for this view that FeEDTA binds at the same or a similar site is provided by a comparison of the cleavage patterns obtained with Fe 2ϩ , shown in lane 1 of Fig. 6A, and with FeEDTA, shown in lane 2 of Fig. 6A. As expected, Fe 2ϩ , in contrast to FeEDTA, produces no footprint of RNA polymerase, since Fe 2ϩ does not generate hydroxyl radicals in solution. However, the intensive cleavage around base position Ϫ1 is visible with both reagents. Quantification of the patterns in Fig. 6C shows that the intensity distribution of the hyper-reactive spot differs slightly for both reagents. The intensity distribution is broader with FeEDTA, indicating that FeEDTA and Fe 2ϩ have slightly different binding sites.
Further support for the view that Fe 2ϩ and FeEDTA bind to RNA polymerase in different modes comes from experiments with the DDD mutant which is incapable of binding catalytically active Mg 2ϩ (19). The data in Fig. 8 show that while the wild type RNA polymerase is able to bind both Fe 2ϩ and FeEDTA (lanes 3 and 4), the DDD mutant binds only FeEDTA (lanes 6 and 7). This suggests that the three Asp residues mutagenized in the DDD mutant are not absolutely necessary for FeEDTA binding.

DISCUSSION
Formation of the Transcription Bubble-To map the transcription bubble, we have applied three different probes which modify unpaired thymidines, adenines, and cytidines. Until now there has been no adequate procedure for probing unpaired guanines. We have filled this gap by showing that dimethyl sulfate cleaves guanines located within the transcription bubble more efficiently than those located outside. The observation that dimethyl sulfate has a preference toward guanines in single stranded regions is in line with previous dimethyl sulfate methylation studies on the transcription bubble in the ternary complex (4). Analysis of results obtained from methylation studies on the binary complex using the P r promoter by Cowing et al. (15) also coincides with our view.
Using these different probes, the transcription bubble was located between base position Ϫ12 to Ϫ1 in the Mg 2ϩ free complex. In the Mg 2ϩ containing complex the transcription bubble is enlarged further downstream to base position ϩ2 encompassing the starting point of RNA synthesis (Fig. 2).
The necessity for opening the ϩ1 position is obvious, because the sequence information required for initiation of RNA synthesis would otherwise not be readable. Mg 2ϩ -dependent enlargement of the transcription bubble including the ϩ1 position seems to be a requirement for open complex formation of all eubacterial promoters. This enlargement was observed previously in the P r promoter, but only if the promoter DNA was supercoiled (31), and also in a RNA polymerase promoter complex of a thermophilic eubacteria, T. maritima (26). The Mg 2ϩ dependent enlargement was observed in this system at the physiological temperature of this organism which is 80°C.
Mapping of the transcription bubble was complicated due to the finding that there are several base pairs of which only one of both complementary bases was modified. Such a base pair was considered as open. But this finding also shows that other effects, such as shielding due to close interaction with the protein can prevent modification and can obscure the disruption of base pairs. Differentiation between the two effects was facilitated by using the surplus information obtained by probing the single strandedness of all four bases.
In order to judge the conclusiveness of results from single stranded probing studies, it is worth recalling the different modifications due to treatment by the single strand-specific reagents. Only dimethyl sulfate/Hz probes cytidines directly involved in Watson-Crick base pairing. All other single strandspecific reagents modify positions of the corresponding bases which do not directly participate in base pairing. The suggested single strand specificity of these reagents is a phenomenological finding for which there is no obvious explanation. Analysis of a DNA model suggests that for example, the C5-C6 bond of thymidines becomes accessible for OsO 4 only if the stacking interaction is disrupted. Such rather drastic conformational change can be induced by, e.g. breakage of the hydrogen bonding between the base pairs.
Among the bases participating in formation of the transcription bubble most adenines are protected despite accessibility of the complementary thymidine, indicating close contact between protein and adenines. This effect is even more pronounced in the Mg 2ϩ containing complex, suggesting that the adenines are the target sites for opening of the DNA strands.
Additional information about protein-DNA contacts are provided by footprinting studies using FeEDTA-generated hydroxyl radicals. This reagent cleaves the sugar moiety (28) of the bases. Comparison of the results from footprinting studies and single strand-specific probing studies show that the transcription bubble is located at the upstream end of a DNA stretch which is fully protected. This region reaches from base position Ϫ12 to ϩ16 at the non-template strand and from base position Ϫ13 to ϩ14 at the template strand. The fully protected region is interrupted by a window of accessibility at base positions Ϫ9, Ϫ10 in the non-template strand and by a window of enhanced cleavage around position Ϫ1 in the template strand. We speculate that the contacts with the sugar phosphate moiety hold the DNA in the proper position, while the contacts with the base moiety, especially of adenines, are required to turn the two strands against each other leading to base pair disruption.
Analysis of the Hyper-reactive Spot at Position Ϫ1 in the FeEDTA Footprint-A striking difference between the FeEDTA footprint of the Mg 2ϩ -free and the Mg 2ϩ containing complex is the hyper-sensitive spot in the template strand at position Ϫ1. While Schickor et al. (14) suggested that the observed effect reflects an enhanced accessibility of the DNA due to a conformational change of the DNA, our results rule out this possibility and suggest that the enhancement is due to hydroxyl radicals generated by a FeEDTA molecule bound to polymerase and thus acting as a local hydroxyl radical source. Analysis of the origin of the hyper-reactivity was complicated by the fact that FeEDTA can act as a hydroxyl radical source in solution which generates the footprinting pattern of RNA polymerase and can as well act as bound hydroxyl radical source which generates the hyper-sensitive spot. It was possible to differentiate between the two effects by using glycerol as scavenger of FeEDTA-generated hydroxyl radicals in solution (28) and using peroxonitrite (KOONO) as an alternative hydroxyl radical source instead of FeEDTA.
KOONO is assumed to produce hydroxyl radicals directly upon protonation of the OONO Ϫ anion as freely diffusing reagent. The footprint with KOONO has the same appearance as that obtained with FeEDTA, except for the hyper-sensitive spot. This observation supports the view that the hyper-cleavage observed with FeEDTA is due to binding of FeEDTA acting as local hydroxyl radical source. Glycerol has the opposite effect. It suppresses the RNA polymerase footprint but leaves unchanged the hyper-sensitive spot, suggesting that hydroxyl radical generated by bound FeEDTA are not accessible for glycerol, probably due to sterical hindrance.
It was previously shown using locally generated OpCu-hydroxyl radicals that, provided there are no shielding effects, hydroxyl radicals have a range of about 15 Å (32). Our data show that the bases ranging from base position Ϫ5 to ϩ3 are affected, which is in line with the proposed range of locally generated hydroxyl radicals.
It was reported recently that Fe 2ϩ can act as local hydroxyl radical source replacing Mg 2ϩ (19). We rule out the possibility that FeEDTA dissociates and that enzyme-bound Fe 2ϩ generates the hydroxyl radical, since the binding constant of the Fe 2ϩ -EDTA equilibrium is about 9 orders of magnitude higher than that of the Fe 2ϩ -RNA polymerase equilibrium. 1 While replacement of Mg 2ϩ by Fe 2ϩ is feasible due to the same positive charge of the two reagents, it is less feasible for FeEDTA which is negatively charged. Despite this charge difference, the binding site of the two reagents at the polymerase promoter complex must be adjacent but is not identical, as indicated by the similarity of the hydroxyl radical cleavage patterns obtained by the two reagents.
Role of Mg 2ϩ Ions at the Active Site-Crystallographic analysis of different template-dependent polymerases (33) indicates that two or three Mg 2ϩ ions are required to form the active site for nucleic acid polymerization. These Mg 2ϩ ions are bound by chelating with aspartates (or glutamates) belonging to a sequence motif which is conserved in all nucleic acid polymerases (34,35). Comparison of the different polymerase structures known today shows that the distance geometry of the Mg 2ϩ ions and the aspartate residues in the active site is the most conserved structural detail.
There are indications that the polymerization active site of E. coli RNA polymerase is assembled in a similar way as in the other polymerases: kinetic studies using E. coli RNA polymerase showed that uptake of three Mg 2ϩ ions is required for formation of a transcription active complex (13); site-specific hydroxyl radical probing by Fe 2ϩ shows cleavages around the suggested Mg 2ϩ pocket (19); the ␤Ј subunit of E. coli RNA polymerase contains the preserved sequence motive with three aspartates which is assumed to participate in formation of the binding pocket of Mg 2ϩ (19).
If the aspartates are replaced, the RNA polymerase (DDD mutant) is unable to synthesize RNA, but still retains promoter binding specificity and the capacity to form the Mg 2ϩ -dependent enlarged form of the transcription bubble (19). We conclude from these findings that the three Mg 2ϩ ions participating in formation of the active site fulfill different functions. That Mg 2ϩ ions which are coordinated by the aspartates facilitates the polymerization activity. The other Mg 2ϩ ions facilitate enlargement of the transcription bubble. These conclusions derived from functional studies using wild type and mutant RNA polymerase are supported by our site-specific hydroxyl radical cleavage studies of the binding complex using Fe 2ϩ and FeEDTA.
Using wild type RNA polymerase both probes generate similar patterns suggesting that they have binding sites which are adjacent but not identical. Both probes can be displaced by Mg 2ϩ indicating that the sites are identical with, or close to, at least one of the putative Mg 2ϩ -binding sites in wild type RNA polymerase. But they differ with respect of their affinity to the polymerase promoter complexes formed with mutant enzyme (DDD), where Mg 2ϩ -chelating aspartates are replaced by alanines. While Fe 2ϩ does not bind to the DDD mutant RNA polymerase (19), FeEDTA does. These findings support the view mentioned above, in that the two probes monitor different subsets of Mg 2ϩ ions.
Temperature-dependent Change of the Size of the Transcription Bubble-Temperature-dependent analysis of thymidines shows that size and position of the transcription bubble in the Mg 2ϩ free complex is the same as that in the Mg 2ϩ containing complex at low temperature reaching from base position Ϫ12 to Ϫ1. Increasing the temperature of the Mg 2ϩ containing complex leads to enlargement of the transcription bubble downstream by 2 base pairs. The two parts of the transcription bubble, namely the Mg 2ϩ -independent part between base position Ϫ12 and Ϫ1 and the Mg 2ϩ -dependent part encompassing base positions ϩ1 and ϩ2, melt at different transition temperatures, the Mg 2ϩ -independent part at T1 ⁄2 ϭ 10 -15°C and the Mg 2ϩ -dependent part at a temperature more than 10°C higher. If one accepts that lowering the temperature leads to freezing of transition state complexes, our finding show that formation of the transcription active open complex is preceded by formation of a complex which has the appearance of a Mg 2ϩ free complex. A similar suggestion was made previously by Record for the supercoiled P r promoter (13,31).
It is interesting to note that temperature course of the hypercleavage reaction in the FeEDTA footprint is the same as the modification reaction of T(ϩ1) by KMnO 4 having a transition temperature of 30°C. However, the first effect is observed in the Mg 2ϩ free complexes, while the second one in the Mg 2ϩ containing complex. We suggest from this finding that the transition at 30°C reflects a conformational change of the polymerase promoter complex which does not require Mg 2ϩ , but enhances the affinity for Mg 2ϩ . We suggest further that the bound Mg 2ϩ facilitates then the apparent Mg 2ϩ dependent effects, such as enlargement of the transcription bubble and enhancement of the contacts between bases of the transcription bubble and RNA polymerase.