DNA Melting within a Binary ς54-Promoter DNA Complex*

The ς54 subunit of the bacterial RNA polymerase requires the action of specialized enhancer-binding activators to initiate transcription. Here we show that ς54 is able to melt promoter DNA when it is bound to a DNA structure representing the initial nucleation of DNA opening found in closed complexes. Melting occurs in response to activator in a nucleotide-hydrolyzing reaction and appears to spread downstream from the nucleation point toward the transcription start site. We show that ς54 contains some weak determinants for DNA melting that are masked by the Region I sequences and some strong ones that require Region I. It seems that ς54 binds to DNA in a self-inhibited state, and one function of the activator is therefore to promote a conformational change in ς54 to reveal its DNA-melting activity. Results with the holoenzyme bound to early melted DNA suggest an ordered series of events in which changes in core to ς54 interactions and ς54-DNA interactions occur in response to activator to allow ς54 isomerization and the holoenzyme to progress from the closed complex to the open complex.

The 54 subunit of the bacterial RNA polymerase requires the action of specialized enhancer-binding activators to initiate transcription. Here we show that 54 is able to melt promoter DNA when it is bound to a DNA structure representing the initial nucleation of DNA opening found in closed complexes. Melting occurs in response to activator in a nucleotide-hydrolyzing reaction and appears to spread downstream from the nucleation point toward the transcription start site. We show that 54 contains some weak determinants for DNA melting that are masked by the Region I sequences and some strong ones that require Region I. It seems that 54 binds to DNA in a self-inhibited state, and one function of the activator is therefore to promote a conformational change in 54 to reveal its DNA-melting activity. Results with the holoenzyme bound to early melted DNA suggest an ordered series of events in which changes in core to 54 interactions and 54 -DNA interactions occur in response to activator to allow 54 isomerization and the holoenzyme to progress from the closed complex to the open complex.
Accessing the information in DNA often relies upon the action of DNA-binding proteins that are able to generate noncanonical B-DNA structures. Recombination, replication, methylation, repair, and transcription are processes that proceed through intermediates in which DNA is distorted. The process of RNA transcript formation by all RNA polymerases must involve a DNA-melting event to reveal the template DNA strand (1)(2)(3). Distortion of the DNA leading to the nucleation of strand separation occurs within the closed complex formed between RNA polymerases and the promoter. Following isomerization of the polymerase, full stable DNA opening is evident, which is thought to have spread from the initial nucleation site. Single-stranded DNA-binding activities in the RNA polymerase are required for DNA opening, and in the case of the bacterial RNA polymerase, the subunit plays an important role (1,(3)(4)(5)(6)(7). For the 70 -type factor, binding to the core enzyme induces conformational changes in a single-stranded DNA-binding region of the protein. As a consequence of these conformational changes, 70 gains specificity for the non-template strand of the melted region in the open complex (8,9). For the 54 -type factor, unrelated by sequence to 70 , the determinants of single-stranded DNA binding are less well described. Single-stranded DNA binding by 54 is evident, however (4,10,11). The sequences that 54 recognizes as single-stranded DNA are between the Ϫ12 promoter element and the start site (4). Importantly, the activity of the 54 holoenzyme is tightly regulated at the DNA-melting step, but promoter binding to form the initial 54 holoenzyme closed complex is not highly regulated (12).
The closed complexes formed with the 54 holoenzyme are silent for transcription unless acted upon by an enhancerbinding activator protein (13)(14)(15). A network of protein and DNA interactions involving 54 function to maintain a stable holoenzyme conformation that rarely changes spontaneously to allow DNA melting and transcript initiation (11, 16 -22). The conformationally restricted closed promoter complex isomerizes to an open promoter complex (in which the DNA strands are melted out) in a reaction in which the activator consumes ATP or another nucleoside triphosphate (14). As a part of this reaction pathway, 54 contributes to the creation of a local structural distortion within the closed complex (23). 54 binds tightly to the distorted promoter DNA and can be shown to isomerize independently of the core RNA polymerase in a reaction that has all the remaining requirements for open complex formation (4,24). Isomerization is associated with an increased DNase I footprint of 54 on DNA, extending toward the transcription start site (24).
Here we use DNA footprinting to show that the DNA within the isomerized 54 -DNA complex has melted and that some melting is negatively regulated by Region I of 54 . However, extensive melting requires Region I. Additionally, changed interactions between 54 and the nucleated DNA are evident in complexes in which melting has occurred. We show that the presence of core RNA polymerase inhibits those changes in 54 -DNA interactions that occur in response to activator, consistent with the view that tight binding to the early melted DNA limits DNA opening (4). The results provide clear evidence in favor of an activation mechanism in which conformational changes in a basal 54 -DNA complex are brought about by the enhancer-binding activator. Activator-independent melting suggests that the activator does not function exclusively as a site-specific DNA helicase for DNA opening (11,18,22).

EXPERIMENTAL PROCEDURES
DNA and Proteins-The promoter fragments used in this work were Escherichia coli glnHp2 from Ϫ60 to ϩ28 in which T at Ϫ13 was replaced by G and the Ϫ11/Ϫ10 sequence was replaced with a CA/TG mismatch to create a short unpaired DNA element next to a consensus GC (Ϫ13/Ϫ12) promoter element or a mismatched sequence between Ϫ11 and Ϫ6 (see Fig. 1). Sinorhizobium meliloti nifH promoter fragments from Ϫ60 to ϩ28 either containing a CA/TG mismatch immediately adjacent to the consensus GC element or with the A-12 top strand base missing (gapped duplex) were also used (see Fig. 1 DNA strands from Ϫ60 to ϩ28 were annealed to create the duplex, with either strand 5Ј-32 P-end-labeled. The unlabeled strand was at 2-fold molar excess. The Klebsiella pneumoniae 54 protein, its Region I-deleted derivative lacking the first 56 amino acids (⌬I 54 ), and Region I (amino acids 1-56) were prepared as described previously (11,25). The activator was E. coli PspF lacking a functional C-terminal DNA-binding domain (PspF⌬HTH) (26). E. coli core RNA polymerase was from Epicentre Technologies Corp.
DNA Binding Assays-End-labeled DNA (16 -100 nM) and 1 M 54 or ⌬I 54 in a 10-l reaction in buffer containing 25 mM Tris acetate (pH 8.0), 8 mM magnesium acetate, 10 mM KCl, 1 mM dithiothreitol, and 3.5% (w/v) polyethylene glycol 8000 were incubated for 5 min at 30°C. Activator PspF⌬HTH (0.5-4 M) and dGTP, GTP, or GTP␥S 1 (4 mM) were added for a further 10 min. Region I was at 0.5 M. Where indicated, heparin (100 g/ml) was added for 5 min prior to gel loading. Free DNA was separated from -bound DNA on 4.5% native polyacrylamide gels run in 25 mM Tris and 200 mM glycine at room temperature.
DNA Footprints-Binding reactions were conducted as described above; footprinting reagents were added; reactions were terminated; and bound and unbound DNAs were separated on native gels as described above. DNA was then excised, processed, and analyzed on a denaturing 10% polyacrylamide gel. For DNase I footprints, 1.75 ϫ 10 Ϫ3 units of enzyme (Amersham Pharmacia Biotech) was added to a 10-l binding reaction for 1 min, followed by addition of 10 mM EDTA to stop cutting. For KMnO 4 footprinting, 4 mM fresh KMnO 4 was added for 30 s, followed by 50 mM ␤-mercaptoethanol to quench DNA oxidation.
Gel-isolated DNA was eluted into 0.1 mM EDTA (pH 8.0) (DNase I footprints) or H 2 O (KMnO 4 footprints) overnight at 37°C. KMnO 4oxidized DNA was cleaved with 10% (v/v) piperidine at 90°C for 20 min. Recoveries of isolated DNA were determined by dry Cerenkov counting, and equal numbers of counts were loaded onto gels.

RESULTS
Previously, we showed that purified 54 bound to the S. meliloti nifH promoter was able to isomerize if the DNA template had an unpaired sequence downstream of the GC element of the promoter (24). The isomerization also required activator and nucleoside triphosphate hydrolysis and was characterized as an extended 54 -DNA interaction toward the transcription start site. The unpaired DNA downstream of the GC promoter element was suggested to mimic the nucleation of DNA melting (early melted DNA) seen in 54 closed complexes, which normally requires 54 and core RNA polymerase (23). Here we have used variants of the 54 -dependent E. coli glnHp2 promoter (27) (Fig. 1) to explore DNA melting by 54 . We chose to generate a nucleated glnHp2 promoter because the high AT content of the sequence should facilitate the detection of unstacked T residues using KMnO 4 as a DNA footprinting reagent (28). Base unstacking occurs when DNA melts, and the associated increased reactivity to KMnO 4 is readily detected.
Isomerization of the 54 -glnHp2 Complex-Initially, we used a gel shift assay to show that 54 bound to the modified glnHp2 promoter with unpaired DNA at Ϫ11/Ϫ10 and, in a reaction  (27), heteroduplex promoter fragments contained either unpaired DNA from Ϫ11 to Ϫ10 with A-10 (bottom strand) replaced with G to form a pre-melted structure comparable to that for the early melted S. meliloti nifH DNA (24) or from Ϫ11 to Ϫ6 (highlighted). In glnHp2 m12, the wild-type T:A base pair at Ϫ13 is replaced by G:C to increase binding. S. meliloti nifH promoter fragments are as described previously (24). requiring hydrolyzable nucleoside triphosphate (dGTP) and activator, produced a supershifted heparin-resistant complex (ss-DNA, Fig. 2A, lane 4). As seen before, the same mobilitysupershifted complex formed with activators of different molecular weights, providing evidence that the activator was not stably associated with the isomerized complex (Ref. 24 and data not shown). Formation of this complex required 54 Region I ( Fig. 2A, compare lanes 4 and 8). Addition of Region I in trans to ⌬I 54 resulted in a new species with similar mobility to that of the supershifted complex, independent of activator and nucleotide ( Fig. 2A, lanes 9 and 10). Using DNase I footprinting, we showed that the DNA within the isomerized 54 complex was protected more than in the complex formed with 54 in the absence of activator and nucleotide (Fig. 2B). The downstream edge of the 54 footprint extended to about Ϫ5 (Fig. 2B,  lane 3). In the isomerized complex, the footprint extended clearly to ϩ2, but a partial footprint to ϩ5 was also detected (Fig. 2B, lane 4). The upstream edge of the footprint was not easily discernible due to background DNA fragments in the undigested sample (Fig. 2B, lanes 1 and 5). Binding of ⌬I 54 did not lead to the extended DNase I footprint of 54 seen with the activator-dependent isomerized complex, but weakly footprinted to about Ϫ7 ( The overall results showed that 54 -DNA interactions at glnHp2 are changed by the action of activator in a nucleotidedependent manner. Non-hydrolyzable nucleotide GTP␥S did not substitute for dGTP to produce an extended footprint (data not shown). Qualitatively, the results are similar to those obtained with the S. meliloti nifH promoter (24) and clearly indicate that activator-dependent 54 isomerization may be readily demonstrated with a number of different promoters.
Activator-dependent DNA Melting by 54 -We used KMnO 4 to probe for DNA melting within isomerized complexes. KMnO 4 footprints at 30°C using the S. meliloti nifH early melted DNA (Ϫ12/Ϫ11) in the isomerized complex did not convincingly show extra DNA melting, but the unpaired T residue at Ϫ12 of the heteroduplex region was much less reactive in both 54 -DNA complexes (data not shown). These KMnO 4 footprints were repeated at 37°C and with a promoter derivative having a single base pair of heteroduplex at Ϫ12 (top strand A replaced with C) (24). Results with the Ϫ12/Ϫ11 heteroduplex showed no extra DNA melting at the elevated temperature, but the unpaired T residue at Ϫ12 in the isomerized complex was more reactive to KMnO 4 (data not shown). However, footprints using the C-12 heteroduplex DNA in the isomerized complex showed that, at 37°C, the template strand T at Ϫ9 had a 2-fold increase in KMnO 4 reactivity, indicating some extra DNA melting (data not shown). This contrasts with results obtained with glnHp2 derivatives (see below and Fig. 3), where considerable extra DNA melting in the isomerized complex was seen at 30°C. It seems that melting of the nifH promoter within the isomerized complex occurs less frequently than with glnHp2 and may relate to differences in the ease with which the DNA strands of the two promoters can separate (see "Discussion").
Footprints of the glnHp2 Ϫ11/Ϫ10 promoter DNA show convincingly that the isomerized complex has extra DNA melting. As shown in Fig. 3A, 54 strongly protected the template strand unpaired T residue at Ϫ11 from KMnO 4 attack (compare lanes 2 and 3). In the isomerized complex, this protection was lost, and KMnO 4 reactivity was evident at the unpaired T residue at Ϫ11 as well as at the new positions Ϫ9 and Ϫ7 (Fig. 3A,  compare lanes 3 and 4). The bases at Ϫ11, Ϫ9, and Ϫ7 must be within an altered DNA structure compared with the free DNA -DNA complexes were challenged with heparin (100 g/ml for 5 min) and loaded onto a native polyacrylamide gel, where bound and unbound DNAs were separated. B, the isomerized 54 -DNA complex has an extended DNase I footprint. Comparison of 54 and isomerized complexes shows that the footprint is extended by the action of activator. Removal of Region I prevents formation of the extended footprint. Reactions (10 l) were as described for A, except that glnHp2 Ϫ11/Ϫ10 DNA was at 50 nM, and PspF⌬HTH was at 4 M. Following exposure to DNase I, bound and unbound DNAs were separated on a native gel, eluted, and analyzed on a sequencing gel. Additions to each binding reaction are shown above each lane. Lanes 1 and 5, untreated DNA; lanes 2 and 6, DNase I-cut DNA. and the non-isomerized 54 -DNA complex. It seems that activator brings about some extra DNA melting as well as changing the interaction of 54 with the T residue at Ϫ11. The KMnO 4 reactivity of the bands was quantified to enhance the reliability of the interpretation (Table I). The same patterns of enhanced reactivity and protection were seen in three independent experiments. Controls in which either activator or hydrolyzable NTP was omitted or a non-hydrolyzable NTP (GTP␥S) was used showed that the extra DNA melting at Ϫ9 and Ϫ7 and the changed footprint at Ϫ11 required activator plus hydrolyzable nucleotide (data not shown). In the absence of these components, the 54 -DNA footprint remained unchanged, and the only complex evident in the gel shift assay was the fast running 54 -DNA complex (data not shown).
KMnO 4 was used to probe the non-template strand in 54 -DNA and isomerized 54 -DNA complexes forming with the modified glnHp2 promoter. As shown in Fig. 3B (see also Table  I), T residues at Ϫ8 and Ϫ6 were KMnO 4 -reactive in isomerized complexes that formed in response to activator and hydrolyzable nucleotide (Fig. 3B, lane 4). In the absence of activator and hydrolyzable nucleotide, the same sequences were no more reactive to KMnO 4 than the naked DNA (Fig. 3B, compare lanes 2 and 3; see also Table I). As seen for the template strand (Fig. 3A), isomerization of the 54 -DNA complex was accompanied by some local DNA melting. In both cases, melting seems to extend from the Ϫ11/Ϫ10 heteroduplex region by at least (interpretation is limited by the placement of a potentially reactive T residue) an extra 4 base pairs toward the transcription start site (summarized in Fig. 7). Melting at these locations is seen in natural open promoter complexes forming with the 54 holoenzyme (13)(14)(15)29). The lack of reactivity of nontemplate T at Ϫ4 suggests either that the transcription start site sequence is not stably opened in the isomerized complex or that T at Ϫ4 is protected by 54 from KMnO 4 attack.
Weak Deregulated DNA Melting by 54 Lacking Region I-The amino-terminal 50 amino acids of 54 (Region I) function to inhibit isomerization of the RNA polymerase holoenzyme as well as to promote enhancer responsiveness (30 -33).
Activities of Region I include contributions to the DNA-binding function of 54 , particularly the recognition and creation of the nucleated DNA near Ϫ12, and an interaction with core RNA polymerase (4,16,23,31,34). Removal of Region I results in activator-independent transcription if the DNA is transiently opened and allows the holoenzyme to engage with pre-melted DNA (11,18,20,25). Using KMnO 4 to probe the template strand of the glnHp2-⌬I 54 complex, we found some evidence for weak activator-independent melting. As shown in Fig. 3A (lane 8), the T residue at Ϫ9 showed some increased KMnO 4 reactivity in the ⌬I 54 complex compared with the zero protein control (lane 2). This region of extra DNA melting appears to be a subset of that found within the activator-dependent complex forming with full-length 54 (Fig. 3A, compare lanes 4 and 8).
The restricted pattern of melting at Ϫ9 seen with ⌬I 54 was independent of nucleotide and activator (Fig. 3A, compare  lanes 7 and 8). Quantitative treatment of the KMnO 4 reactivity (Table I) showed a constant increase in the reactivity of T at Ϫ9 when ⌬I 54 was bound. The increase was seen in three independent experiments. The T reside at Ϫ11 in the ⌬I 54 complex was slightly more protected compared with the isomerized 54 complex (Fig. 3A), but significantly more reactive than 54 alone to KMnO 4 attack (compare lanes 3, 4, and 8; also see Table I). It seems that removal of Region I partially deregulates 54 melting activity. Region I supplied in trans did not result in a significant change in KMnO 4 reactivity to suggest a shift in the footprint toward that of the non-isomerized 54 -DNA complex (Fig. 3A, compare lanes 3 and 5; and Table I). It seems that, although Region I binds to the Ϫ12/Ϫ11-⌬I 54 complex, this does not lead to large changes in KMnO 4 reactivity (24).
KMnO 4 was used to probe the non-template strand in the ⌬I 54 -DNA complexes forming with the modified glnHp2 promoter. As shown in Fig. 3B (lanes 5-8) and Table I, no sequences were significantly more KMnO 4 -reactive than in the unbound DNA (Fig. 3B, lane 2). For the ⌬I 54 -DNA complex, no further changes in KMnO 4 reactivity were detected in response to activator and hydrolyzable nucleotide (Fig. 3B, lane 7). Region I supplied in trans to ⌬I 54 -DNA binding assays did not change non-template strand KMnO 4 reactivity (Fig. 3B, lanes 5  and 6).
Interactions of 54 with DNA Pre-opened from Ϫ11 to Ϫ6 and a Gapped Structure-The extra DNA opening from Ϫ9 to Ϫ6 seen in the activator-dependent isomerized complex (Fig. 3) could arise from activator functioning as a DNA helicase. To explore this issue, we wished to learn if pre-opening the DNA from Ϫ11 to Ϫ6 would allow the 54 to bind promoter DNA in an isomerized state without activator. The interaction of 54 with glnHp2 promoter DNA mismatched from Ϫ11 to Ϫ6 (see Fig. 1) to mimic the DNA opening seen in the activator-dependent isomerized complexes was examined in the presence and absence of activator. Gel shift assays showed that 54 bound the Ϫ11/Ϫ6 opened DNA to give an initial complex (ss*-DNA) with reduced mobility compared with the Ϫ11/Ϫ10 opened DNA complex (Fig. 4, compare lanes 2 and 5). The reduced mobility was similar to that of the activator-dependent isomerized complex (ss-DNA) forming on the Ϫ11/Ϫ10 opened DNA (Fig. 4, compare lanes 3 and 5). The mobility of the 54 complex with DNA opened from Ϫ11 to Ϫ6 was unchanged by activator and hydrolyzable nucleotide (Fig. 4, compare lanes 4 and 5). To more fully understand the properties of these slow running complexes, we probed them by DNase I and KMnO 4 footprinting. Using glnHp2 promoter DNA mismatched from Ϫ11 to Ϫ6, 54 gave a short DNase I footprint to Ϫ5 and no extra KMnO 4 reactivity, and footprints were insensitive to activator and nucleotide (data not shown). We conclude that the reduced mobility of the 54 complex with DNA opened from Ϫ11 to Ϫ6 is due to the altered DNA conformation and that pre-opening the DNA does not drive the change in 54 needed for the extended downstream DNase I footprint to at least ϩ2 seen in activatordependent isomerized complexes (Fig. 2B, lane 4) (24). Instead, a change in 54 conformation driven by the activator seems to be necessary for the extended footprint to ϩ2. The insensitivity of the 54 complex on the Ϫ11/Ϫ6 opened DNA to activator suggests that the DNA from Ϫ9 to Ϫ6 should be in a doublestranded form for activator to act on 54 , consistent with prior work with the S. meliloti nifH promoter (24).
We previously observed with the S. meliloti nifH promoter that removal of the top strand A-12 residue resulted in a 54 -DNA complex that did not form a new slow running species when incubated with activator and hydrolyzable or non-hydrolyzable nucleoside triphosphate (Fig. 5A, compare lanes 2, 4, and 5 with lane 6) (24). ⌬I 54 formed a slower running complex when Region I was in trans (Fig. 5A, compare lanes 2, 6, and 7).
To characterize these complexes and to determine their relationship to the isomerized complex, we used DNase I and KMnO 4 footprinting. The S. meliloti nifH promoter Ϫ12 gap (top strand) probe (see Fig. 1) gave a DNase I footprint with a distinct region that was cut poorly compared with the intact probe (data not shown). This region centered over the Ϫ12 gap   Fig. 3 KMnO 4 reactivities of T residues at positions Ϫ11, Ϫ9, and Ϫ7 (template strand) residues and Ϫ8 and Ϫ6 (non-template strand) in -DNA complexes are expressed by their ratio to signals without protein (reactivity ϭ 1; Fig. 3, A and B, lanes 2) Shown are the results from the gel shift binding assay of 54 with glnHp2 DNA (16 nM) opened from Ϫ11 to Ϫ6 compared with the Ϫ11/Ϫ10 opening. With glnHp2 Ϫ11/Ϫ6 DNA, a 54 -DNA complex (ss*-DNA) was formed that has mobility similar to that of the activator-and nucleotide-dependent glnHp2 Ϫ11/Ϫ10 supershifted complex (ss-DNA), but reduced mobility compared with the -DNA complex. ϩ indicates the presence of 54 (1 M), PspF⌬HTH (4 M), and dGTP (4 mM). Heparin was added prior to gel loading. Results without heparin were similar. and extended ϳ4 bases on either side, suggesting a locally altered DNA structure refractory to DNase I cutting. When 54 bound, extra cutting from Ϫ10 to Ϫ1 was also evident, suggesting that 54 stabilized a double-stranded DNA structure otherwise absent from the unbound gapped DNA (data not shown). There was no difference in the DNase I footprint under activating conditions (data not shown). The effects of activating conditions were then gauged by KMnO 4 footprinting. When 54 bound the gapped duplex, the single-stranded T residue at Ϫ12 (template strand) was protected from attack by KMnO 4 , and a modestly increased reactivity to KMnO 4 was seen at T-9 indicative of some activator-and nucleotide-independent DNA melting (Fig. 5B, compare lanes 2 and 3). The same pattern was seen in the presence of activator and nucleotide (Fig. 5B, compare lane 3 with lanes 5-7) without evidence for extra activationdependent melting. This suggests that a mismatch at Ϫ12 is needed for activator response. The slow running ⌬I 54 complex with Region I in trans footprinted like wild-type 54 (Fig. 5B,  compare lanes 3 and 8), whereas the ⌬I 54 complex showed less KMnO 4 reactivity at TϪ9 (compare lanes 8 and 9). This suggests that Region I stabilizes the melted DNA at Ϫ9. The DNA melting seen with the gapped DNA when bound by 54 (Fig. 5B) is consistent with the proposed role of the Ϫ12 nucleotide in restricting melting prior to activation (4). The gapped DNA allows melting within the 54 -DNA complex that is not evident with homoduplex DNA (Fig. 5B and data not shown).
Interactions of the 54 Holoenzyme with Early Melted DNA-The early melted DNA structure just downstream of the GC promoter that enables 54 isomerization is believed to exist in closed promoter complexes, but is apparently absent in the activator-dependent open complex (23,24). The chemical reactivities at Ϫ12/Ϫ11 seen in closed complexes are not evident in open complexes; rather, new melting is evident nearer the transcription start site (23). To explore the activator responsiveness of the 54 holoenzyme on the S. meliloti nifH and E. coli glnHp2 early melted DNAs, we conducted gel shift and footprinting assays. As noted above, the 54 holoenzyme bound to the early melted DNA assumes a complex with greater resistance to heparin than does the holoenzyme complex on homoduplex DNA (31). This has been suggested to involve a changed interaction between 54 and core RNA polymerase since the binding of 54 to core RNA polymerase in the absence of early melted DNA is heparin-sensitive (31,34). Gel shift assays with either the S. meliloti nifH (Fig. 6A) 6 and 7). A reduction in 54 holoenzyme concentration led to increased formation of the supershifted 54 -DNA complex (ss-DNA, Fig. 6A, compare lanes 2-6) in the presence of activator and nucleotide, reflecting free 54 .
We next used DNase I and KMnO 4 footprinting to characterize the 54 holoenzyme-DNA complexes on the early melted DNA and to learn if they had isomerized. Results showed that, under non-activating conditions, the isolated 54 holoenzyme complex gave a short DNase I footprint to Ϫ5 and no extra KMnO 4 reactivity (data not shown). The footprints were essentially as for 54 , except that some extra protection from DNase I by the 54 holoenzyme upstream of Ϫ34 was observed with the E. coli glnHp2 promoter (data not shown). We next footprinted the 54 holoenzyme-DNA complexes under activating conditions. The isolated holoenzyme complexes gave footprints indistinguishable from those obtained under non-activating conditions, suggesting that core-bound 54 was not able to isomerize efficiently (data not shown). Further experiments with the use of initiating nucleotide (GTP) to potentially sta-bilize complexes on opened DNA through allowing initiation or with the omission of the heparin challenge to help preserve unstable complexes failed to produce 54 holoenzyme footprints in which activator-dependent changes were evident (data not shown). We conclude that 54 does not isomerize efficiently when bound to core RNA polymerase in assays using early melted DNA probes. We confirmed (data not shown), using S. meliloti nifH Ϫ60 to ϩ28 homoduplex DNA, that the heparin-resistant open promoter complexes formed by the action of activator and GTP had extended DNase I footprints to at least ϩ13, definition of an exact end point being limited by the resolution of the gel and fragment size to ϩ28 (14,20,35). However, the efficiency of activator-dependent stable complex formation was low in this assay. To address the issue that the 54 holoenzyme bound to the early melted DNA might form new activator-dependent complexes (not distinguished from the activator-independent complexes because of the common property of heparin resistance) but with low efficiency, we also used KMnO 4 as a probe of isomerization events. Here isomerization would be evident as increased reactivity to KMnO 4 rather than protection from DNase I (see Fig. 3A, lane 4). Unlike the activator-and nucleotidedependent complexes forming with 54 and the early melted glnHp2 promoter DNA (Fig. 3A), the 54 holoenzyme did not respond to activator to yield detectable extra melting or any associated loss of KMnO 4 reactivity of the glnHp2 heteroduplex sequence (data not shown). The overall results show that the holoenzyme complex, in contrast to 54 , is poorly (if at all) responsive to activation conditions when bound to the early melted DNA. As discussed below, this may relate to unusually stable complex formation between the early melted DNA and the 54 holoenzyme.
Core RNA Polymerase Binding to Isomerized 54 -DNA-Having shown that 54 bound to early melted DNA forms an isomerized complex (Ref. 24 and this work), but that 54 holoenzyme does so inefficiently if at all (see above), we conducted an experiment to determine whether the conformation of the isomerized 54 -DNA complex allowed core RNA polymerase binding. In this assay, the isomerized complex was formed using an end-labeled Ϫ35 to ϩ6 S. meliloti nifH Ϫ12/Ϫ11 DNA fragment (24) in excess of 54 to diminish the amount of free 54 and to ensure that core RNA polymerase interactions were potentially largely with DNA-bound 54 . Isomerization reactions were carried out and stopped by adding GDP (Fig. 6B, (24). Increasing amounts of core RNA polymerase (E) were then added to bind 54 -DNA complexes. As shown in Fig.  6B, addition of increasing amounts of core RNA polymerase (lanes 3-8) depleted the 54 -DNA complex; and in parallel, an increasing amount of the 54 holoenzyme bound to DNA (E-DNA) was detected. The amount of isomerized 54 -DNA complex (ss-DNA) remained relatively constant throughout the titration with core RNA polymerase. It seems that core RNA polymerase preferentially binds the non-isomerized 54 -DNA complex. At high core RNA polymerase concentrations (in excess of 0.6 M) (Fig. 6B, lanes 7 and 8), some core bound to DNA was detected, and this contributed to the apparent increased amount of 54 holoenzyme bound (graphed in Fig. 6C) since core-DNA and holoenzyme-DNA complexes were not fully resolved. We infer that weak binding of the isomerized 54 -DNA complex by core RNA polymerase is because the interface between core and 54 and DNA has changed upon isomerization. This leads to the suggestion that movements in 54 and DNA are concerted with some in core RNA polymerase for forming the natural open promoter complex and is discussed below.

DISCUSSION
DNA Melting by 54 -Activator-dependent isomerization of 54 results in the spreading of DNA melting away from the nucleated DNA structure and toward the transcription start site (Fig. 7). A structure melted over at least 6 base pairs is generated. Interactions with the nucleated DNA located at Ϫ11/Ϫ10 are lessened in the isomerized complex compared with the initial 54 -DNA complex. Although removal of Region I sequences allows some weak DNA melting by , this is not as extensive as activator-driven isomerization. A comparison of results obtained with ⌬I 54 and intact 54 shows that the full DNA-melting activity of 54 has some determinants outside of regulatory Region I (for weak melting at Ϫ9) and some that depend upon Region I (for extra melting from Ϫ8 to Ϫ6).
Isomerized 54 -nifH promoter complexes failed to show KMnO 4 reactivity as extensively as glnHp2 (Fig. 3 and data not shown). If this reflects less melting (as opposed to shielding from KMnO 4 ), differences in intrinsic DNA opening rates between the two promoters in combination with the sequenceindependent single-stranded DNA-binding activity in 54 (36) could contribute. DNA opening at Ϫ9 seen in natural nifH open complexes but weakly detected in isomerized complexes with 54 might therefore reflect a stabilizing contribution from the core enzyme (37). Clearly, changes in the 54 -DNA relationship seen in DNase I footprints of isomerized complexes in which melting may not have occurred is consistent with the idea that activator changes 54 structure and that pre-opening the DNA does not drive this change (22,24).
Comparisons of the closed and activator-dependent open 54 holoenzyme-promoter complexes using KMnO 4 and ortho-copper phenanthroline footprinting suggest that the structure of the DNA immediately downstream of the GC element differs between these complexes and that contact with G-13 is also altered (14,15,23). These observations are fully consistent with the lessened interaction at Ϫ11 detected by the KMnO 4 footprints of isomerized complexes reported here (Fig. 3). Based on the known interaction between Region I and core polymerase (34,38), we suggested that the silencing interactions associated with the Ϫ13 to Ϫ11 54 -DNA contact (4) can contribute significantly to inhibiting RNA polymerase isomerization through restricting conformational changes within the core subunits (16). Changes in promoter DNA and 54 holoenzyme conformation are probably coupled through 54 Region I to maintain either the closed or open state of the promoter. Open promoter complexes that form in deregulated transcription by the 54 holoenzyme are unstable compared with activator-dependent open complexes (18,19,25). Their instability may be related to incomplete DNA melting beyond Ϫ9 as a consequence of the mutations in Region I of 54 and a need for Region I to stabilize melted DNA (Fig. 5B).
Role of Activator-Results with DNA mismatched from Ϫ11 to Ϫ6 across the sequences melted in the isomerized 54 -DNA complex support the view that activator drives a conformational change in 54 rather than generating isomerization by solely creating an opened DNA structure for 54 binding. For both transcription and 54 isomerization, pre-opening of the DNA through heteroduplex formation does not by pass activator requirements (22), unless the structure recognized by 54 at Ϫ11 is destroyed (36). Rather, activator-dependent conformational changes in 54 and the holoenzyme seem necessary for open complex formation and 54 isomerization (24,36). DNase I footprinting shows that 54 clearly binds to double-stranded DNA ahead of the locally melted DNA or the fork junction that forms next to the GC promoter element in closed complexes (this work and Refs. 4 and 24). This suggests that the further melting of the double-stranded DNA within the 54 -DNA complex is an active process in the sense that it is not a primary result of a domain of 54 translocating along the duplex and trapping DNA strands at a fraying fork junction. Energy for duplex destabilization may come from the tight binding of 54 to the initially locally melted DNA. An activator-driven change in 54 -DNA interaction might release 54 from binding to the double-stranded DNA downstream of the Ϫ12 GC and switch 54 to a conformation that then allows it to bind singlestranded DNA. Combined with the single-stranded DNA-binding activities in 54 (4,36) and the core RNA polymerase (39), 54 isomerization would lead to formation of the stable open promoter complex.
Stable Holoenzyme Binding-When bound to core RNA polymerase, activator-dependent isomerization of 54 was not evident, in marked contrast to its efficient isomerization without core. It seems that the short sequence of heteroduplex downstream of GC inhibits isomerization of 54 within the holoenzyme since, on linear DNA, activator-dependent isomerization of the holoenzyme is evident (11, 23, 39). The activatordependent movement of 54 across the sequence opened down-stream of GC implied by the results of our KMnO 4 footprints may be inhibited when 54 is bound by core RNA polymerase. This suggests that isomerization of 54 within the holoenzyme may normally require that the local DNA opening next to GC does not strongly persist. In the homoduplex, the DNA can base pair again, but not in the heteroduplex. Consistent with this view is the observation that the local DNA distortions downstream of GC and present in the closed complex are apparently changed in the open complex, as judged by the reduced sensitivity of the DNA to two chemical probes of DNA structure (23). The spread of melting observed in the isomerization assays would then normally be associated with a changing of the DNA structure believed to be locally melted in the closed complex, achieved through a breaking of DNA contacts and a rebinding of 54 to DNA. These considerations suggest an ordered series of events in which changes in core RNA polymerase to 54 interactions and 54 -DNA interactions occur in response to activator to allow 54 isomerization and the holoenzyme to progress from the closed complex to the open complex. This view is consistent with activator interacting with both core RNA polymerase and 54 (24,40) to achieve isomerization of the holoenzyme and with the view that the promoter sequences around GC contribute to preventing the holoenzyme from isomerizing prior to activation and to set the target of the activator (4,24).
Core-Interactions and DNA Melting-Results of core RNA polymerase binding with the isomerized and non-isomerized 54 -DNA complexes strongly suggest that some points of interaction between 54 and core are changed upon isomerization of the 54 -DNA complex. The poorer core binding of the isomerized complex likely correlates with the changes in protease sensitivity of 54 in the isomerized complex (24). Changed DNA structure within the isomerized 54 -DNA complex may also contribute to poor binding by core RNA polymerase. The strong reduction in isomerization of 54 resulting from core RNA polymerase binding prior to exposure to activating conditions (Fig. 6A) and the weak binding of isomerized 54 to core RNA polymerase (Fig. 6, B and C) is striking. It seems that normally for efficient 54 holoenzyme isomerization and open complex formation, activator-dependent changes in 54 structure would occur in concert with a changed binding of parts of 54 to core RNA polymerase; but on the early melted DNA, this is not occurring properly. As indicated above, when using the early melted DNA as template, the failure to reconfigure interactions with DNA next to the GC promoter may simply strongly stabilize a 54 conformation that is unfavorable for core RNA polymerase binding. We therefore suggest that, in closed complexes, activator drives a conformational change in 54 that results in altered contacts with the early melted DNA (as detected in our KMnO 4 footprinting; see above), allowing binding of 54 and core RNA polymerase to permit full 54 isomerization and the associated isomerization of the closed complex to the open complex. It seems that Region I of 54 greatly contributes to these events through (i) its requirement for creating the early melted DNA when the holoenzyme binds homoduplex DNA, (ii) directing 54 binding to the early melted DNA in heteroduplexes and associated fork junction structures, (iii) a binding interaction with core RNA polymerase, and (iv) the changing Region I structure in isomerized 54 and the activated 54 holoenzyme (reviewed in Ref. 12). The recent demonstration that Region I sequences localize over the Ϫ12 promoter region is fully consistent with these observations and points to a central role of the protein and DNA elements that localize there in establishing new interactions that allow DNA melting and a changing binding relationship between core subunits and 54 (41). We also note that the refractory behavior of the 54 holoenzyme bound to the early melted DNA and the poor core binding of isomerized 54 are fully consistent with the view that interactions 54 makes with the Ϫ12 promoter region, in particular sequences just downstream of GC, are key in limiting spontaneous activator-independent open complex formation (4).