The nitrogen fixation protein NifA is a member of the protein family activating transcription by the alternative eubacterial [sigma]N ([sigma]54) RNA polymerase holoenzyme. Binding sites for NifA, upstream activator sequences (UASs), are remotely located. Interaction between holoenzyme bound in a closed promoter complex and NiFA is facilitated by bending of the intervening DNA by integration host factor (IHF). We have examined NifA contact with the Klebsiella pneumoniaenifH promoter UAS in the presence and absence of holoenzyme and IHF. Footprints with UV light were made on 5-BrdU-substituted DNA and DNase I and laser UV footprints on conventional DNA templates. Results establish that the consensus thymidine residues of the UAS motif 5'-TGT are in close proximity to NifA. Reactivity suggests that each UAS thymidine is not structurally equivalent. Titration of NifA binding to the UAS in the presence or absence of the closed promoter complex indicates that the interaction of NifA with the UAS is not strongly co-operative with holoenzyme or IHF, a result supportive of an activation mechanism not reliant upon simple recruitment of factors to the promoter. Laser footprints demonstrated that holoenzyme suppressed reactivity of promoter consensus -14, -15 and -16 T residues , indicating close contact. Binding of holoenzyme resulted in a specific increase in 5-BrdU reactivity at -9 within the holoenzyme binding site, likely reflecting DNA distortion. Enhanced -9 reactivity required [sigma]N N-terminal sequences that are necessary for activation. Since T-9 is melted in open complexes the closed complex appears poised for melting. Open promoter complex formation was accompanied by a distinct change in laser footprint signal at -11, consistent with the view that nucleation of strand separation occurs within or close to the -12 promoter element.
Gene activation by one class of regulatory bacterial proteins requires the alternative RNA polymerase [sigma]N holoenzyme (1 ,2 ). The [sigma]N holoenzyme binds promoters in a transcriptionally inactive form and is the target for enhancer binding proteins that are responsible for activating transcription of gene sets needed for diverse processes (3 ,4 ). The NifA protein specifically activates nitrogen fixation (nif) genes by catalysing isomerization of the closed recognition complex between E[sigma]N and nif promoters to the open promoter complex, which can then initiate transcription (5 -7 ). As with other activators of the [sigma]N holoenzyme, NifA must hydrolyse a nucleoside triphosphate to catalyse open promoter complex formation (6 ,8 -10 ). A change in configuration of the [sigma]N holoenzyme likely accompanies open promoter complex formation and it appears that one function of NifA is to stimulate a change within the closed promoter complex to facilitate strand separation (3 ).
Until recently study of the NifA protein from a variety of diazotrophs was restricted by the difficulty of preparing purified active material (6 ). In contrast to the extensively studied NifA protein from Klebsiella pneumoniae, NifA from Azotobacter vinelandii can be purified in a soluble and active form (6 ). Using a derivative of the A.vinelandii NifA protein with a hexahistidine N-terminal tag to facilitate purification, we have examined the contacts that the NifA protein makes in vitro with its upstream activator sequences (UASs; 11 ). Such contacts are important for establishing the fidelity of activation through tethering NifA near to its target [sigma]N holoenzyme closed promoter complex (5 ). Whether the binding of NifA to UASs is conditioned by binding of [sigma]N holoenzyme in the closed promoter complex was also examined. The binding of NifA and [sigma]N holoenzyme was examined by DNase I footprinting and by UV light and laser light (12 ) footprints on 5-BrdU-substituted and thymidine-containing DNA respectively. Results show that NifA makes DNA major groove contacts to non-equivalent consensus T residues of the UAS, providing a direct demonstration of their contribution towards NifA binding. Interaction of NifA with the UAS appeared largely independent of [sigma]N holoenzyme binding, providing supporting evidence for an activation mechanism in which only transient contacts between NifA and [sigma]N holoenzyme are needed (13 ). Recruitment of factors to the promoter appears not to be a major mechanistic aspect of activation of the [sigma]N holoenzyme. Footprints also showed that: (i) binding of [sigma]N holoenzyme results in a DNA distortion around -9, likely nucleation of strand separation (14 ), which requires N-terminal sequences of [sigma]N that are associated with activation; (ii) open promoter complex formation is accompanied by an altered protein-DNA interaction at -11, results consistent with nucleation of and subsequent DNA melting events involving [sigma]N (14 ,15 ). It appears that the holoenzyme in the closed complex is poised for melting the DNA.
The 1.6 kb NdeI-BamHI fragment containing A.vinelandii NifA coding sequence from pDB737(6 ) was cloned into NdeI-BamHI-restricted pET28b+ vector (Novagen), to generate pMB737/28b+, and subsequently transformed into BL21(DE3)/plysS to direct synthesis of an N-terminal hexahistidine tagged NifA protein. For hexahistidine tagged NifA protein preparation cells were grown at 25oC in 1 l batches of 2* YT medium, from a 20% inoculation previously grown to saturation at 37oC, and induced after 4-6 h by addition of 1 mM IPTG. Cells were harvested 3-4 h after induction, re-suspended in cold 0.5 M NaCl, 25 mM sodium phosphate, pH 7.0, 5% (v/v) glycerol and lysozyme added to 200 [mu]g/ml. After 30 min incubation on ice, cells were sonicated and then centrifuged at 39 000 g for 30 min. Between 25 and 50% of the overproduced NifA protein was in the supernatant following the temperature downshift induction procedure. The soluble form of the tagged NifA protein was purified immediately by Ni affinity chromatography at 4oC and eluted with a gradient of imidazole. Pooled peak fractions were dialysed into 50 mM Tris-HCl, pH 8.0, 50 mM NaCl, 0.1 mM EDTA, 1 mM DDT, 50% glycerol and aliquoted for storage at -80oC. Assuming NifA protein to be dimeric the final stored concentration was 7 [mu]M. Where necessary, centrifugal membrane concentrators (Filtron `nanosep') were used to achieve higher molarity.
An in vivo estimate of tagged NifA activity was made by transforming pMB1 (11 ) (a K.pneumoniae nifH-lacZ fusion) into BL21(DE3)/plysS containing pMB737/28b+. [beta]-Galactosidase activity was measured in exponential phase cells without the addition of IPTG.
These were conducted using the Klebsiella pneumoniae nifH promoter or the nifH049 variant (which has a high affinity RNA polymerase binding site) as the target DNA sequence (5 ) and core RNA polymerase, [sigma] and IHF proteins prepared as described previously (6 ). Briefly, single-stranded promoter DNA from an M13 clone was extended with Klenow DNA polymerase using a 5'-32P-labelled universal primer (in the case of 5-BrdU footprinting the primer was extended with 5-BrdUTP instead of dTTP) to generate end-labelled double-stranded DNA for footprinting (16 ,17 ). Dimethyl sulphate (DMS), exonuclease III, 5-BrdU or DNase I footprinting was then conducted to detect binding of NifA and/or other proteins to promoter DNA. Binding reactions (25-50 [mu]l, 30oC) containing 1.2-2.4 nM template DNA were conducted in 25 mM Tris-acetate, pH 8.0, 8 mM Mg acetate, 10 mM KCl, 1 mM dithiothreitol, 3.5% (w/v) PEG 6000 for DMS and exonuclease III assays and in 50 mM Tris-acetate, pH 8.0, 0.1 M potassium acetate, 8 mM magnesium acetate, 27 mM ammonium acetate, 3.5% PEG6000, 0.4 mg/ml bovine serum albumin, 0.4 mg/ml salmon sperm DNA for DNase I and 5-BrdU assays. Protein concentrations were as indicated in the figure legends. Protein solutions contributed up to 10% of the final assay volume. Proteins and DNA were incubated for 15-20 min prior to addition of one of the following footprinting reagents: 175 U exonuclease III (Pharmacia) for 3 min; 5 [mu]l diluted DMS (1.5 [mu]l DMS freshly diluted in 100 [mu]l H2O) for 2 min; 0.0070 U freshly diluted DNase I (Amersham) for 20 or 60 s. In the case of the 5-BrdU nucleoprotein complex this was transferred to a Petri dish on ice before illumination with UV light (254 nm) for 1-2 min, followed by digestion with proteinase K for 1 h at 37oC in the presence of 0.1% (w/v) SDS. Enzymatic footprint reactions were terminated by rapid phenol extraction and the DMS methylation protection assay by quenching with mercaptoethanol followed by base cleavage using procedures employed for making the chemical cleavage G ladder (18 ). Nucleic acids were recovered by ethanol precipitation and then electrophoresed through 6% denaturing polyacrylamide gels which were dried and autoradiographed. DNA size markers were generated by chemical cleavage of DNA with piperidine following partial methylation with DMS (18 ).
Binding reactions (20 [mu]l) containing 5 nM double-stranded pWVC88049 (mutant K.pneumoniae nifH promoter with the -15 to -17 C residues changed to T residues) plasmid DNA were conducted in 50 mM Tris-acetate, pH 8.0, 0.1 M potassium acetate, 8 mM magnesium acetate, 27 mM ammonium acetate, 3.5% PEG6000 at 30oC for 10 min, using the protein concentrations indicated in the figure legends. High intensity laser light was generated by a neodymium ytrium aluminium garnet (NdYAG) laser machine producing a beam of polarized coherent light at a wavelength of 1064 nm. An optical system consisting of two frequency doubling crystals and a series of dichroic mirrors allowed an incident beam of 266 nm to be directed onto the sample. Under optimal conditions a single pulse of radiation at an intensity of ~0.5 * 1011 W/m2 of 5 ns duration was used. The estimated energy dose per experiment was 100 J/m2. Samples in 0.5 ml microfuge tubes were placed vertically in a heating block and irradiated with a high energy laser pulse at 266 nm (12 ). After dividing the binding reaction into two capped microfuge tubes and heat denaturing the nucleoprotein complex (95oC for 5 min, then chilled on ice for 2 min), 0.5 pmol 5'-32P-labelled oligonucleotide (primer upstream of UAS, 5'-CAAGCTGTTGAACAGGC-3'; primer downstream of UAS, 5'-GATACCGCCTTTACCGT-3') were annealed (48oC for 15 min) to the DNA separately and extended in a reaction with 1 U Klenow polymerase plus 0.5 mM dNTPs at 48oC for 10 min. Nucleic acids were recovered by ethanol precipitation and then loaded onto 6% sequencing gels. Data were collected by phosphorimager analysis.
A labelled Rhizobium meliloti nifH promoter DNA duplex was made from pMB210.1 (5 ) by a thermal cycle amplification reaction using a pair of primers (5'-TATTTCGGTTGTTCGGACACATGAA-3', 5'-AAAAATAGGCGTATCACGAGGCCCT-3') to generate a 610 bp DNA fragment containing the NifA UAS and the RNA polymerase binding site. The amplified fragment was 5'-32P-labelled by kinasing. Binding reactions (15 [mu]l) containing 2 nM end-labelled DNA were conducted in 50 mM Tris-acetate, pH 8.0, 0.1 M potassium acetate, 8 mM magnesium acetate, 27 mM ammonium acetate, 3.5% PEG6000 at 30oC for 10 min, with 400 nM NifA, 100 nM core RNA polymerase, 200 nM [sigma]N, 50 nM IHF (19 ) plus GTP (1 mM) if needed (6 ). Prior to electrophoresis 3 [mu]l loading buffer (binding buffer plus 20% glycerol, xylene cyanol and bromophenol blue) plus 100 [mu]g/ml heparin were added to each sample and samples were immediately loaded and run at room temperature on a 4% native acrylamide gel at 120 V for 2 h in 25 mM Tris-HCl, 0.2 M glycine, pH 8.6. After drying, the gel was autoradiographed at -80oC.
To facilitate rapid purification of A.vinelandii NifA protein a construct directing synthesis of an N-terminal hexahistidine tagged NifA protein was made. The tagged protein retained the ability to efficiently activate transcription at the K.pneumoniae nifH promoter in vivo (data not shown). We examined a variety of growth conditions for protein overproduction. Maximum yields of soluble tagged NifA were obtained from induced cultures grown in 2* YT medium at 25oC but inoculated with cells grown at 37oC. Other growth conditions examined included growth in 2* YT throughout at 25, 30 and 37oC and minimal medium at 30 and 37oC. These conditions all led to reduced levels of soluble tagged NifA compared with the temperature shift-down procedure (see Fig. 1 ). To purify tagged NifA, induced cells were lysed and the soluble cell fraction was directly applied to a metal chelate-Sepharose column. Following Ni chelate chromatography at 4oC the tagged NifA protein was judged to be >95% pure on gels (see Fig. 1 ).
Initially we examined whether the purified tagged NifA could support open complex formation. We used an end-labelled linear Rhizobium meliloti nifH promoter DNA fragment replete with its UASs in a gel shift assay designed to detect promoter complexes stable in the presence of heparin. Such complexes are open or initiated complexes and are more stable at the R.meliloti nifH promoter than at the K.pneumoniae nifH promoter, therefore facilitating their identification by gel shift assay (6 ,20 ). As shown in Figure 2 , a retarded species was detected in the presence of NifA, holoenzyme, IHF and GTP, the latter serving as both the initiating nucleotide and hydrolysable nucleoside triphosphate for NifA. The extent of heparin-stable complex formation detected at 400 nM NifA was comparable with that observed in similar assays with non-tagged NifA (6 ). Thus we conclude that the tagged NifA is competent as a transcriptional activator.
To examine potential interactions between NifA and [sigma]N holoenzyme a series of DNase I footprints were conducted (Fig. 4 ). The promoter template used was K.pneumoniae nifH049, to allow efficient closed promoter complex formation. A promoter with a high occupancy in the closed complex was chosen to maximize detection of any potential interaction between [sigma]N holoenzyme and NifA that might increase NifA binding and provide evidence of an activator-[sigma]N holoenzyme contact. Other than three transitions from -17 to -15 to generate a high affinity [sigma]N binding site, the nifH049 promoter sequence is unaltered compared with the wild-type. Brief 20 s exposures to DNase I were made to maximize the detection of transient interactions of proteins with the promoter. Initially, footprints corresponding to the binding of NifA, IHF and [sigma]N holoenzyme were obtained (see Fig. 4 ). When all three proteins were present together there was no significant change in the strengths of the NifA or [sigma]N holoenzyme footprints compared with those of the single proteins. Similar results were repeatedly obtained at the wild-type nifH promoter and with [sigma]N holoenzyme or NifA concentrations that individually gave weak footprints (data not shown). At the nifH049 promoter a small difference at the 3'-end of the [sigma]N holoenzyme footprint was evident; in the presence of NifA the footprint appeared slightly stronger at the 3'-end (Fig. 4 , lanes 6 and 7 compared with 8 and 9). Overall the binding of NifA and [sigma]N holoenzyme appeared independent. The IHF footprint was apparently stronger in the presence of NifA and [sigma]N holoenzyme (Fig. 4 , lanes 4 and 5 compared with 9), indicating some interaction between proteins (possibly IHF and holoenzyme) but not necessarily directly between NifA and [sigma]N holoenzyme. An increased IHF footprint was not observed with NifA in the absence of holoenzyme (data not shown).
Promoters bound by [sigma]N holoenzyme are protected from around -31 to -5 in DNase I and hydroxyl radical footprints (16 ). When [sigma]N holoenzyme was included in the 5-BrdU footprints at the K.pneumoniae nifH promoter an increase in reactivity at -9 was detected as the binding signal (Fig. 5 , lane 4). This likely reflects a distortion of the DNA by [sigma]N holoenzyme binding in the closed promoter complex (14 ). Distortions by [sigma]N holoenzyme around the -12 promoter element have been detected with o-copper phenanthroline, KMnO4 and diethylpyrocarbonate footprinting reagents (10 ,14 ). Holoenzyme also reduced reactivity of the consensus -26 T base. When [sigma]N holoenzyme and NifA were included together in the 5-BrdU footprint assay, no significant increase in the strength or quality of the signal due to NifA binding or [sigma]N holoenzyme binding was evident (Fig. 5 , lanes 2 and 4 compared with 5 and 6). No increase in IHF footprint (as observed with DNase I; Fig. 4 , lane 9) was observed in the comparable 5-BrdU footprint (Fig. 5 , lane 6), likely reflecting the different reactivities of the footprinting agents. Within the sensitivity of the assay it appears that binding of [sigma]N holoenzyme and NifA are largely independent of each other and of IHF at the K.pneumoniae nifH049 promoter. Similar conclusions were drawn from experiments with BrdU-substituted wild-type nifH promoter (data not shown). As shown in Figure 5 , the [sigma]N holoenzyme partly suppressed the NifA-enhanced reactivity at -44. Binding of IHF fully diminished -44 reactivity (Fig. 5 , compare lanes 5 and 6 with 2). Probably [sigma]N holoenzyme sterically blocks NifA access to -44 and IHF preferentially occupies the sequence across -44. Reduced reactivity around base pair -66 (marked 0 in Figs 5 and 6 ) was associated predominantly with addition of IHF or holoenzyme, but to a lesser extent NifA. The significance of this protection is unclear, but suggests that all three proteins might be able to transiently interact around -66.
Separate experiments using N-terminal deletion variants of [sigma]N (23 ) to form altered holoenzymes demonstrated that the -9 hyper-reactivity associated with [sigma]N holoenzyme binding required region I of [sigma]N (see Fig. 6 ). Region I appears to be a domain required for activation (4 ,14 ,15 ,23 ). N-Terminal deleted forms of [sigma]N did not form holoenzymes capable of producing -9 hyper- reactivity (Fig. 6 , lane 4 compared with lanes 7 and 10), although such holoenzyme forms show good promoter-specific DNA binding (23 ). Inclusion of [sigma]N rather than its holoenzyme in 5-BrdU footprints did not detectably enhance -9 reactivity, demonstrating that the reactivity was a property of the holoenzyme (Fig. 6 , lanes 3, 6 and 9). The enhanced -44 reactivity shown by NifA (Fig. 6 , lane 2) was suppressed somewhat in the presence of holoenzyme, irrespective of [sigma]N N-terminal sequences (Fig. 6 , compare lane 2 with lanes 5, 8 and 10). As indicated above, protection around -66 was afforded by holoenzyme (marked 0).
To further examine the interactions of NifA and holoenzyme with the nifH049 promoter, laser light footprinting experiments were conducted (12 ). In the laser assay binding reactions are exposed to a pulse of laser light. Reaction of the DNA can result in a DNA crosslink to a bound protein or in pyrimidine dimer formation. The latter reaction may be suppressed by protein binding to the DNA. By copying the irradiated DNA in a primer extension reaction the sites at which bound proteins modify the DNA reactivity can be determined. Proteins crosslinked to the DNA and pyrimidine dimers block primer extension. When laser assays were conducted the [sigma]N holoenzyme strongly suppressed DNA reactivity from -16 to -14 (primer extension stops at -15 to -13 were diminished; Fig. 7 A and D) at sequences corresponding to T residues required for strong [sigma]N binding (24 ). Likely the suppression reflects [sigma]N promoter contact across the T tract (16 ). The IHF protein produced a signal at -60 (Fig. 7 B) and a weaker signal at -51 (data not shown). NifA binding generated a primer extension stop signal around -133 within the nifH UAS (Fig. 7 C). When NifA was titrated in the presence of [sigma]N holoenzyme no significant change in the intensity of the NifA footprint compared with NifA alone was detected, a finding unchanged by IHF binding (data not shown). Similarly, the footprint of the [sigma]N holoenzyme in the closed complex was unchanged qualitatively or quantitatively by NifA in the presence or absence of IHF. We conclude that binding of NifA and [sigma]N holoenzyme in the closed complex are largely independent events. To confirm that the complexes assayed in laser footprinting were on route to open complex formation we added GTP, which serves as a hydrolysable substrate for NifA and as the initiating nucleotide (6 ). In the presence of IHF, which stimulates interaction between NifA and holoenzyme (28 ), the [sigma]N holoenzyme footprint changed to produce a strong primer extension stop signal at -11 (Fig. 7 D) and the NifA footprint appeared to increase slightly in intensity (Fig. 7 C). The bands at 22 and 28 mm did not reproducibly increase in reactivity with IHF (Fig. 7 C and data not shown). It is likely that the change at -11 under activating conditions is associated with strand opening, since nucleation of melting appears to occur at -12 and a changed [sigma]N holoenzyme-DNA interaction is predicted in the -12 promoter region upon activation (14 ). The more intense NifA footprint may reflect a NifA interaction with holoenzyme in the open promoter complex.
Activation of transcription is a key point controlling gene expression. A variety of mechanisms operate to increase the rate of transcription initiation. These include: (i) changes in DNA structure by activators to improve binding of RNA polymerase; (ii) the recruitment of factors, including RNA polymerase, to promoters by protein-protein contacts. In the latter case the binding of RNA polymerase can be demonstrated to be co-operative with the activator factor. Our results argue that the NifA activator protein binds to its UASs largely independently of the target RNA polymerase [sigma]N holoenzyme closed promoter complex. It appears that any interaction between NifA and [sigma]N holoenzyme prior to open complex formation at the K.pneumoniae nifH promoter is transient and insufficiently strong to improve occupancy by [sigma]N holoenzyme and vice versa (20 ,25 ). The apparent absence of any strong co-operativity is entirely consistent with a well-demonstrated and characteristic property of the [sigma]N holoenzyme; its ability to bind as a stable closed promoter complex unable to isomerize to the open complex without the assistance of activator (4 ). Furthermore, increasing the affinity of the nifH [sigma]N RNA polymerase binding site for [sigma]N holoenzyme does not overcome the requirement for activator and does not stimulate promoter activity (26 ), indicating that the major role for NifA is after closed complex formation. However, it has been suggested that at the supercoiling-sensitive, NtrC-activated, [sigma]N-dependent K.pneumoniae nifL promoter closed complex formation is influenced by a mutant NtrC protein unable to stimulate open complex formation (27 ). Examples of promoters where the binding of activator and [sigma]N holoenzyme is co-operative may exist.
Under conditions allowing open complex formation laser footprinting detected a small increase in the interaction of NifA with its UAS, suggesting that the interaction NifA has with the open complex is different to the interaction that NifA may have with the closed complex. Related observations with the NtrC activator using electron microscopy have shown that NtrC is detectably present within a DNA looped complex and probably touching [sigma]N holoenzyme only after open complex formation has taken place (13 ). By analogy with the DctD activator, which has been shown to specifically crosslink to components of the [sigma]N holoenzyme, contacts between NifA and [sigma]N holoenzyme may be made with the [beta] subunit of the core RNA polymerase and with [sigma]N (28 ). N-Terminal region I of [sigma]N is a potential contact site for NifA (15 ). However, the phenotypes of holoenzymes containing region I-altered [sigma]Ns is complex, because region I appears to contribute to the nucleation of strand separation as well as inhibiting low level activator-independent isomerization to the open complex (14 ,15 ). Our 5-BrdU footprints detected a clear change in the environment of T-9 in the closed complex. The increase in -9 reactivity was dependent upon region I of [sigma]N and likely related to a local DNA distortion associated with nucleation of DNA strand separation required for subsequent open complex formation (14 ). Significantly, T-9 (as 5-BrdU) is distorted in the closed complex and melted in the open complex (6 ), suggesting that the closed complex is poised for melting (14 ). The laser footprint open complex signal at -11 is fully consistent with transition from the nucleated closed complex to the open complex involving a DNA-protein transaction across the site of nucleation. The distinctive changes in DNA structure proximal to the -12 element that are detected in closed complexes (10 ,14 ,23 ) suggest that the promoter complex that NifA acts upon is distinct from the initial recognition complex that forms between promoter and [sigma]N holoenzyme. The initial recognition complex is predicted to convert to the nucleated closed complex independently of NifA. Interestingly, NifA appeared to interact in some way with the -44 position of the nifH promoter, suggesting that it can communicate with components of the closed promoter nucleoprotein complex, possibly IHF. However, the significance for activation is not clear. Overall our results have demonstrated that the NifA-IHF-[sigma]N holoenzyme nucleoprotein complex at the nifH promoter assembles without strong co-operative interactions between heterologous components. Similar conclusions have been drawn for the activator NtrC at the [sigma]N-dependent glnAp2 promoter (20 ) and in experiments where NtrC replaced NifA at the K.pneumoniae nifH promoter (25 ).
X.-Y.W. was supported by a CEC funded China-UK collaborative grant. The work was supported by a Wellcome Trust grant to M.B. We thank Henri Buc and Malcolm Buckle for their interest and use of the laser and Ray Dixon for pDB737.
*To whom correspondence should be addressed. Tel: +44 171 59 45442; Fax: +44 171 584 2056; Email: m.buck@ic.ac.uk
REFERENCES
+Present address: Shanghai Institute of Plant Physiology Academia Sinica, 300 Fonglin Road, Shanghai, China