Helicobacter pylori NikR Protein Exhibits Distinct Conformations When Bound to Different Promoters*

Helicobacter pylori NikR (HpNikR) is a ribbon-helix-helix (RHH) DNA-binding protein that binds to several different promoter regions. The binding site sequences are not absolutely conserved. The ability of HpNikR to discriminate specific DNA sites resides partly in its nine-amino acid N-terminal arm. Previously, indirect evidence indicated that the arm exists in different conformations when HpNikR is bound to the nixA and ureA promoters. Here, we directly examined HpNikR conformation when it was bound to nixA and ureA DNA fragments by tethering (S)-1{[bis(carboxymethyl)amino]methyl}-2-{4-[(2-bromoacetyl)amino]phenylethyl}(carboxymethyl)amino]acetic acid, iron(III) to different positions in the N-terminal arm and RHH DNA binding domain. Different cleavage patterns at each promoter directly demonstrated that both the RHH domain and the arm adopt different conformations on the nixA and ureA promoters. Additionally, the two RHH domain dimers of the HpNikR tetramer are in distinct conformations at ureA. Site-directed mutagenesis identified an interchain salt bridge (Lys48-Glu47′) in the RHH domain remote from the DNA binding interface that is required for high affinity binding to ureA but not nixA. Finally, DNA affinity measurements of wild-type HpNikR and a salt bridge mutant (K48A) to hybrid nixA-ureA promoters demonstrated that inverted repeat half-sites, spacers, and flanking DNA are all required for sequence-specific DNA binding by HpNikR. Notably, the spacer region made the largest contribution to DNA affinity. HpNikR exhibits a substantially expanded regulon compared with other NikR proteins. The results presented here provide a molecular basis for understanding regulatory network expansion by NikR as well as other prokaryotic regulatory proteins.

Helicobacter pylori requires nickel ions as cofactors for the urease and hydrogenase enzymes, which are important for the colonization of animal models of infection (1)(2)(3). This virulence requirement has motivated numerous studies of nickel-dependent gene regulation in H. pylori. The Ni 2ϩ -responsive tran-scription factor NikR from H. pylori (HpNikR) 3 regulates the expression of multiple genes by directly binding to their promoters in response to increasing nickel (4 -10). HpNikR is an exquisitely sensitive nickel sensor with a K D,Ni of ϳ2 pM (4,5); however, how that affinity translates to gene regulation in cells remains unclear (e.g. in rich media, changes in gene expression are observed in response to ϳ100 M) (6, 9 -11). From these studies, it appears that the recognition sequences in each promoter are defined by a series of poorly conserved 6-bp imperfect inverted repeat half-sites separated by a 15-bp spacer (5,7,12). HpNikR binds to these promoters with a range of affinities (5,12). The promiscuity of DNA binding by HpNikR contrasts with high sequence selectivity of the NikR proteins from Escherichia coli (EcNikR (13,14)) and Geobacter uraniireducens (GuNikR (15)), which recognize highly conserved inverted repeats found in only one or two promoter regions per genome. The NikR protein family represents an excellent system with which to understand the role of protein and DNA sequence in the expansion of prokaryotic regulatory networks.
The archetypical RHH fold (ϳ45-50 amino acids) consists of a single ␤-strand followed by two ␣-helices, with the two ␤-strands of each obligate RHH dimer forming an antiparallel ␤-sheet motif (23). Structural studies of the Arc and MetJ repressors first demonstrated that the RHH ␤-sheet motif sits in the major groove of DNA, with three solvent-exposed residues making base-specific hydrogen bonds via their side chains (23)(24)(25). Nonspecific DNA phosphate contacts were also observed in the structures, some by tandem turn regions N-terminal to the ␤-sheets and others by the N terminus of helix ␣2 of the RHH domain (23)(24)(25). These protein-phosphate interactions are hypothesized to attach the N terminus and helix ␣2 to the DNA backbone and link these structural elements to the ␤-sheet (23). The helix ␣2-phosphate backbone interactions are shared among all of the RHH protein-DNA co-crystal structures (16, 24 -30).
The EcNikR-operator structure revealed DNA interactions similar to those of the Arc and MetJ repressors (27). Additionally, the nickel binding C-terminal domain interacts with the phosphate backbone of the central base pairs that separate the inverted repeat sequences. No structures exist for HpNikR bound to DNA, so any differences between the protein-DNA interfaces for the two NikR proteins that might explain the broad DNA recognition properties of HpNikR remain unknown. The HpNikR and EcNikR DNA-contacting ␤-sheet residues are different (Arg-Ser-Ser and Arg-Thr-Thr, respectively), which probably changes the specific DNA sequence motif that is recognized.
In previous work, alanine scanning and truncation mutagenesis indicated that the nine-residue N-terminal arm of HpNikR exhibits conformational differences when the protein is bound to the nixA and ureA promoters, based on changes in DNA binding affinity (5). Here, we have further examined this con-formational difference using chemically modified and new sitedirected mutants of HpNikR, in combination with engineered promoter sequences, to demonstrate that protein conformational differences exist beyond the N-terminal arm when the protein is bound to the nixA and ureA promoters. The data also show significant contributions to binding affinity from DNA regions flanking the inverted repeat sequences in the two promoter regions. These observations will help to understand NikR regulon expansion in H. pylori and, more generally, the interplay between protein and DNA sequence in prokaryotic regulon expansion.

EXPERIMENTAL PROCEDURES
Mutagenesis of HpNikR-Site-directed mutagenesis of HpNikR was carried out using the QuikChange protocol (Stratagene, La Jolla, CA). Complementary oligonucleotides (Table  1) with the mutated codon were used with plasmid pEB116 (5) to generate C96A (pEB173) and each salt bridge mutant. The ͓AAGCGTTCATTAACTAATTAATAC͔TTAAAATGAATTA͓GTAATATATCTTTGAAGCGCTATA͔ a Underlined bases correspond to restriction sites, and capitalized bases correspond to regions complementary to genomic DNA sequence. b Previously constructed (5). c Kinased primers were used to generate labeled PCR products from various plasmid templates containing wild-type or hybrid promoters (see "Experimental Procedures"). d NA, not applicable; primers used to generate PCR products subsequently cloned into pBluescript (see "Experimental Procedures"). e Brackets indicate ureA sequence introduced into an otherwise wild-type nixA promoter. HS, half-site; SP, spacer; FL, flanking sequence.
T3C, N5C, and N20C mutants were generated using pEB173 as a template. The presence of each mutation was verified by DNA sequencing (SeqWright, Houston, TX). Protein Expression and Purification-All HpNikR proteins and mutant variants were expressed and purified as described previously (5). The T3C, N5C, and N20C mutant proteins were purified with 1 mM ␤-mercaptoethanol present throughout the purification protocol. Protein concentration was determined in 6 M guanidine hydrochloride using ⑀ 276 ϭ 9895 M Ϫ1 cm Ϫ1 (31).
Promoter Fragments; Cloning and Labeling-Hybrid nixA-ureA promoter fragments were generated by two sequential rounds of PCR using overlap extension with two oligonucleotides that contained the base mutations and flanking primers designed to amplify ϳ200 -400 bp of DNA spanning the nixA or ureA promoters and genes ( Table 1). The final PCR product was digested with EcoRI (New England Biolabs, Beverly, MA) and KpnI (New England Biolabs) and ligated into pBluescript (Stratagene). DNA sequences of the hybrid promoters were verified by sequencing (SeqWright).
Fe-BABE Modification and DNA Cleavage by Modified Cys Mutants-Purified HpNikR Cys mutants were desalted into 20 mM Tris-Cl (pH 8.0), 300 mM NaCl to remove the ␤-mercaptoethanol, and NiCl 2 was added back to apoprotein at a stoichiometry of 2:1 Ni 2ϩ /HpNikR monomer to ensure saturation of the high affinity binding site. 40 M protein was incubated with 15 mM Fe-BABE in a final reaction volume of 37 l and incubated at 37°C for 1 h according to established protocols (32). The reaction was quenched with an equal volume of 1 M Tris-Cl (pH 8.0) and desalted two times into 20 mM Tris-Cl (pH 8.0), 300 mM NaCl to eliminate unreacted Fe-BABE.
Fe-BABE cleavage reactions were performed in the same buffer used for DNase I footprinting and used a variation of a previously published protocol (32). The reactions were started with 5 l of a 10ϫ stock of freshly mixed sodium ascorbate (100 mM; stored in aliquots at Ϫ20°C) and H 2 O 2 (250 mM), incubated for 2 min at 22°C, and quenched with 36 l of a 2.5ϫ stock solution of thiourea (100 mM) and EDTA (75 mM). DNA fragments were analyzed by denaturing polyacrylamide gel electrophoresis.
DNase I Footprinting-DNase I footprinting was performed as described previously (13,33) in a binding buffer containing 10 mM Tris-Cl (pH 8.0), 100 mM KCl, 3 mM MgCl 2 , and 50 M NiCl 2 . Labeled DNA fragments were incubated with protein at 22°C for 30 min prior to the addition of DNase I (Sigma) (final concentration 300 ng/ml). Formic acid cleavage of labeled DNA was performed using the standard protocol for Maxam-Gilbert sequencing (2). DNA fragments were analyzed by denaturing gel electrophoresis.
Electrophoretic Mobility Shift Assays-Mobility shift assays were performed using 7% polyacrylamide gels and electrophoresis buffer containing 50 mM Tris (pH 8.8), 25 mM boric acid and with 50 M NiCl 2 . The binding buffer was identical to that used for DNase I footprinting except that 10 g/ml E. coli thioredoxin (34) and 4 ng/l salmon sperm DNA (Fisher) were added to prevent nonspecific HpNikR accumulation in the gel wells. Labeled DNA fragments were incubated with HpNikR or mutant proteins at 22°C for 30 min, and 20 l of the 25-l total volume was loaded directly onto a running gel (120 V). Apparent affinities measured by mobility shift and DNase I footprinting assays were calculated from binding curves determined by the ratio of bound (all shifted species) versus free counts as quantified using a GE Healthcare Typhoon Trio variable mode imager and ImageQuant version 5.1 software. The data were fit using MICROMATH SCIENTIST version 2.01 and the following equation, where y represents fraction of DNA bound (ratios described above), K d is protein concentration required for half-maximal DNA binding, x is protein concentration, and n is the Hill coefficient. All reported affinities are the average of at least two independent experiments using a dilution series of 16 protein concentrations, and the S.D. value is also reported.
Circular Dichroism and Fluorescence Spectroscopy-CD spectra were collected on a Jasco J-715 spectrapolarimeter using a 900-l sample volume in a cylindrical cuvette with a 1-cm path length. All spectra were collected at 22°C in a buffer containing 20 mM Tris-Cl (pH 7.5), 100 mM NaCl.
Urea melts were performed with 6 M protein that was incubated in the absence or presence of increasing concentrations of urea in 20 mM Tris-Cl (pH 7.5) 100 mM NaCl at 22°C for a minimum of 16 h prior to measurements. Fluorescence spectra were collected on a Cary Eclipse fluorescence spectrophotometer in a 150-l, 1-cm path length Hellmann quartz cuvette. Samples were excited at 295, and emission spectra were collected from 315 to 380 nm with excitation and emission slit widths set to 10 nm. Center of mass calculations were performed as described previously for the Arc repressor (35).

RESULTS
To test if HpNikR adopts distinct conformations when bound to DNA, we endowed the protein with nuclease activity by covalent modification with Fe-BABE. This approach was used previously with another RHH protein, the Mnt repressor (36). Fe-BABE-modified Cys residues hydrolyze DNA within ϳ22 Å due to hydroxyl radical production by the chelated Fe 2ϩ under the appropriate reaction conditions (37). The locations of the Cys substitutions in the N-terminal arm (T3C and N5C) and RHH domain (N20C) of HpNikR were chosen based on the EcNikR-DNA structure and the likelihood that the side chains would be oriented toward, but not in direct contact with, the DNA (Fig. 1). To introduce unique Cys residues for modification, we also mutated a native Cys residue (C96A, hereafter referred to as HpNikR*), which does not affect DNA binding activity (data not shown) (38). The essential Cys 107 required for nickel binding was protected from modification by the presence of Ni 2ϩ .

HpNikR RHH Domain Conformations Are Different on the nixA and ureA Promoters
Cys mutant proteins (T3C, N5C, and N20C) were purified under reducing conditions and modified with Fe-BABE according to published protocols (32). Fe-BABE-treated HpNikR* did not produce detectable cleavage for either nixA or ureA promoter DNA fragments (Fig. 2). The FeBABE-treated proteins (HpNikR* and the single Cys mutants) all produced the expected DNase I footprints, indicating that recognition of the HpNikR binding site was not perturbed by modification.
Cleavage of nixA-DNA cleavage by the Fe-BABE-modified N20C HpNikR* spanned and extended just past the outer boundaries of the characteristic DNase I footprint (Fig. 2a) (5). There were seven symmetrically arranged cleavage regions of roughly equal intensity that were grouped into two pairs of outer cleavage regions (rOC1 to -4 (for RHH outer cleavage; Fig. 2, a and e) and one set of three inner regions (rIC1 to -3 (for RHH inner cleavage). The rOC2-rIC1 and rOC3-rIC3 pairs straddle the two NikR half-sites (Fig. 2e, gray boxes), consistent with the location of Asn 20 at the end of helix ␣1 and near the two edges of the antiparallel ␤-sheet (Fig. 1b) (27). Because HpNikR binds to DNA as a tetramer, the presence of more than four cleavage regions suggests that each modified Cys adopted more than one conformation. The data do not distinguish between differences in side chain conformers or larger changes in protein conformation.
Cleavage of ureA-Fe-BABE-modified N20C HpNikR* produced eight cleavage regions on the ureA promoter that were grouped into two pairs of cleavage regions outside and overlapping the inverted repeat half-sites (rOC1 to -3, rIC5; Fig. 2, b and f) and four cleavage regions inside the half-sites (rIC1 to -4). Whereas rOC1 and -2 and rIC5 and rOC3 are each separated by 6 bp, the smaller spacing between the inner cleavage regions and the offset center of cleavage relative to the DNase I footprint result in an asymmetric positioning of rOC1 and rOC3. Similar to the nixA cleavage results, the eight cleavage regions at ureA indicate that modified Cys residues were present in at least two different conformations. The band intensities for the outermost cleavage regions were noticeably greater than those of the inner regions.
The clear differences in the patterns and intensities of cleavage of nixA and ureA by Fe-BABE N20C HpNikR* provide the first direct evidence of distinct conformations in HpNikR binding. Furthermore, the asymmetric cleavage pattern of ureA suggests that the two RHH domains of the HpNikR tetramer are in distinct conformations when bound to this promoter.

The HpNikR N-terminal Arms Are in Different Conformations at the nixA and ureA Promoters
Cleavage of nixA-Fe-BABE-modified T3C or N5C HpNikR* also produced eight distinct cleavage regions on nixA that spanned the DNase I footprint (Fig. 2, c and e). Three regions were located outside (aOC1 to -3, for arm outer cleavage) and five inside (aIC1 to -5, for arm inner cleavage) of the inverted repeat half-sites, with all but one pair of the eight regions being separated by 2 bases and symmetrically positioned relative to the footprint. The cleavage observed for T3C and N5C often overlapped with that seen for N20C (Fig. 2, compare a, c, and e), with an offset of 1-2 bases/region. This suggests that the Fe-BABE moieties of Thr 3 , Asn 5 , and Asn 20 are similarly but not identically positioned in space relative to the DNA backbone.
Cleavage of ureA-Fe-BABE-modified T3C or N5C HpNikR* cleaved ureA at six regions (Fig. 2, d and f) that can be grouped as three outer regions (aOC1 to -3) and three inner regions (aIC1 to -3). aIC1 and aIC2 were distinct from all other HpNikR cleavages because of their size (6 bases/region), spanning much of the DNA spacer between the two inverted repeat half-sites. Notably, there was larger spacing between the upstream cleavage regions compared with the downstream regions (aOC1-aOC2 versus aIC3-aOC3; aOC2-aIC1 versus aIC2-aIC3), resulting in an overall asymmetric cleavage pattern for ureA. This cleavage asymmetry was similar to that observed for N20C cleavage of ureA and is consistent with the two arms and one RHH domain dimer from each half of the HpNikR tetramer  (40)). B-form DNA is also shown. The structure and DNA were modified using the Chimera package from the Resource of Biocomputing, Visualization, and Informatics at the University of California, San Francisco (supported by National Institute of Health Grant P41 RR-01081 (55)).
being in two different conformations when bound to the ureA promoter region. This result contrasts with the symmetric cleavage patterns observed for the three modified proteins when bound to nixA promoter DNA. The distinguishable cleavage patterns observed for nixA and ureA with modifications of either the ␣1 helix or the arm provide strong support for different HpNikR conformations when bound to the promoter regions.

Mutation of an RHH Residue That Does Not Contact DNA Selectively Impairs ureA Binding
The RHH domain of HpNikR is less stable to chemical denaturation than that of EcNikR (5). This reduced stability may correlate with conformational flexibility and expanded DNA sequence recognition. To probe residues that might contribute to this differing stability, we generated a K48A HpNikR mutant. This residue position is poorly conserved across NikR orthologs (39) and is often a small residue in RHH proteins. The K48A mutant was not affected in stability (supplemental Fig. S1) but showed profound differences in DNA binding affinity between the nixA and ureA promoters (Fig. 3 and Table 2). Based on examinations of different RHH-DNA co-complex structures, it is very unlikely that Lys 48 of HpNikR directly contacts DNA. Although the mutation did not account for the difference in stability between the HpNikR and EcNikR RHH domains, this result provides the first evidence for HpNikR conformational  differences beyond the RHH domain-DNA interface when the protein is bound to different promoter sequences.

Lys 48 Participates in a Salt Bridge That Is Required for High Affinity ureA Binding
To better understand the role of Lys 48 in HpNikR-DNA interactions, we examined different HpNikR crystal structures (40). The negatively charged side chains of two residues, Glu 47Ј from the other polypeptide chain in the RHH dimer and Asp 52 from the same chain (i to i ϩ 4), are both close to Lys 48 (Յ3.8 Å). To determine if either residue interacts with Lys 48 when HpNikR is bound to ureA, E47A and D52A mutants were constructed and tested for DNA binding in mobility shift assays. Both mutants bound to the nixA and ureA DNA fragments with only slightly weakened affinities (Table 2) compared with the ϳ60-fold reduction of K48A affinity for ureA. The E47A and D52A mutants had 2-and 4-fold decreases in binding affinity for nixA, whereas both mutants showed a similar (6 -7-fold) decrease in affinity for the ureA promoter region.
The absence of a significant decrease in affinity for the two mutants suggests that neither residue interacts with Lys 48 , nor are they important by themselves for DNA binding. Alternatively, the remaining acidic side chain may partially compensate for the mutated residue and preserve the interaction with Lys 48 . This possibility was explored with the E47A/D52A double mutant, which had lower affinity for the nixA and ureA promoters relative to the single mutants (12-and 37-fold, respectively; Table 2). The greater reduction in affinity for binding to ureA is similar to the effect of the K48A mutation. These data support the idea that at least one negatively charged residue interacts with Lys 48 when HpNikR is bound to ureA, although the double mutant is not as impaired as K48A.

Inverting Candidate Salt Bridge Residues Identifies a Critical Glu 47 -Lys 48 Interaction
As a further probe of potential Lys 48 interactions, two residue pair swap mutants were constructed (E47K/K48E and K48D/ D52K), along with two single residue mutants (E47K and K48E). Intriguingly, the E47K/K48E double mutant displayed only minor decreases in affinity for nixA and ureA of 1.5-and 2-fold, respectively (Table 2). In contrast, the K48D/D52K mutant was severely impaired for DNA binding to both promoter fragments (164-fold lower affinity for nixA and no detectable binding for ureA up to 5 M protein). The similar behavior of the K48A and K48D/D52K mutants suggests that the double mutant cannot make the Lys 48 -dependent interaction required for ureA binding. The E47K and K48E single mutants had reduced affinity for ureA (E47K) or both nixA and ureA (K48E) compared with the E47K/K48E charge swap double mutant.
These data are most consistent with the formation of a specific intersubunit interaction between the Glu 47Ј and Lys 48 side chains when HpNikR is bound to ureA. The various deleterious effects of the single mutants and the K48D/D52K on DNA binding affinity are also consistent with an important structural interaction involving Lys 48 .

The Effects of RHH Domain Mutations on DNA Binding Do Not Correlate with Protein Stability
Urea melts were performed in the presence of stoichiometric Ni 2ϩ to determine if the charge mutations affect DNA binding affinity by perturbing RHH domain stability. Domain unfolding was monitored by following the red shift and quenching of Trp 54 fluorescence upon unfolding (41). Trp 54 is in the linker between the RHH and C-terminal domains of HpNikR. Its fluorescence reports solely on RHH domain folding because comparison of fluorescence shift (as determined by center of mass calculations for each spectrum; see "Experimental Procedures") with loss of CD signal at 230 nm from HpNikR urea melts indicates only a single fluorescence transition that coincides with the RHH domain transition in the CD data (supplemental Fig.  S1).
The Trp fluorescence of several of the charge mutants was perturbed in the native protein prior to the addition of urea (supplemental Fig. S2). Given the spatial proximity of Trp 54 to the positions of ␣2 helix mutations, these changes probably represent local perturbations of structure that influence the quenching of Trp fluorescence. Most of the charge mutants showed little to no change in the RHH domain unfolding transition (supplemental Fig. S1) with two exceptions. The K48D/ D52K double mutant displayed an obvious loss in the cooperativity of unfolding. This decreased stability may explain the poor DNA binding affinities of this mutant ( Table 2). The K48E mutant displayed reduced stability with a shift in the midpoint of the unfolding curve to significantly lower urea concentrations. This decrease in stability was almost fully alleviated in the E47K/K48E double mutant. These data indicate that the ability of HpNikR variants to bind nixA and ureA did not correlate with changes in RHH domain stability except for the destabilized K48D/D52K mutant. In this case, the severe disruption in DNA binding is probably the result of a more general structural perturbation rather than a specific salt bridge disruption. The absence of a correlation between protein stability and DNA binding affinity among the other mutants is consistent with a specific requirement for a Glu 47 -Lys 48 salt bridge.

Half-sites, Spacers, and Flanking DNA Sequences Contribute to HpNikR High Affinity Binding to nixA and ureA
The stark difference in affinities of K48A HpNikR for the nixA and ureA DNA fragments lead us to probe the relative contributions of different DNA sequence elements to the affinity change. Typically, the inverted repeat sequences present in RHH DNA recognition sites are thought to provide the majority of the interactions necessary for specific protein-DNA complex formation (16,23), although this has not been probed extensively. Thus, changes in half-site sequences represent likely candidates for influencing the conformation of HpNikR when bound to the nixA and ureA promoters. Comparison of the inverted repeats in the nixA and ureA promoter fragments, which are based on DNase I, Fe-EDTA, and DMS footprinting (5), indicated that 4 of the 12 bp of the nixA and ureA 6-bp half-sites are different (Figs. 2 and 4).
To identify the DNA sequence features that foster a "nixAlike" conformation of K48A HpNikR, several nixA-ureA hybrid promoters were constructed in which the two half-sites, the flanking sequences, and the intervening spacer regions of the two promoters were arranged in different combinations (Fig.  4). These constructs represent all of the possible transitions between the nixA and ureA promoter sequences, or vice versa.
Wild-type HpNikR exhibited only modest affinity reductions for all six of the hybrid promoters (ϳ4 -6-fold; Fig. 4). Substitution of the nixA or ureA half-sites in the context of the other promoter (ureA-HS and ureA-FLϩSP in Fig. 4) had similar effects on affinity (3-4-fold decrease). Substitution of the spacer or flanking regions in either context had a slightly greater effect (5-6-fold). The limited decrease in affinity with the hybrid promoters indicated that the sequence changes did not substantially affect any key protein-DNA contacts. Nonetheless, the affinity changes in the hybrid DNA fragments suggest some conformational distortion of native HpNikR along the transition between native binding site sequences. However, the effects on affinity were neither additive nor synergistic.
In contrast to wild-type HpNikR, the K48A mutant was very sensitive to any substitution that increased the ureA sequence content of the promoter (Fig. 4), showing 9-and 10-fold decreases in affinity when the ureA half-sites (ureA-HS) or flanking regions (ureA-FL) were inserted into the nixA sequence. Surprisingly, the largest -fold decrease (59-fold) in response to exchange of a single sequence element occurred upon swapping the ureA spacer into the nixA sequence. A single sequence swap of this type was previously shown not to affect the EcNikR-DNA interaction (14). These data suggest that the ureA spacer destabilizes K48A binding more than either the ureA half-sites or flanking sequences. Similar to the results for wild-type HpNikR, these effects were neither additive nor synergistic. In particular, the ureA spacer was the main determinant of binding affinity when at least one nixA promoter element remained.
The affinity changes observed for the native and mutant HpNikR proteins support two key aspects of HpNikR DNA recognition: (i) sequence elements throughout individual promoter sequences are important for DNA binding, and (ii) protein conformational flexibility is required for HpNikR to recognize the two promoters. This flexibility relies at least in part on interactions involving Lys 48 in the context of binding to ureA.

DISCUSSION
The HpNikR regulon is more extensive than those of most RHH proteins (42), which generally bind to only one recognition site in their respective genomes. Here, three complementary lines of evidence indicate that the expansion of the HpNikR regulon includes differential recognition of various promoter regions through changes in protein-DNA interactions that affect protein conformation. HpNikR appears to have at least two conformations when bound to DNA, and protein conformation is sensitive to DNA sequence or structure beyond the inverted repeat sequences in the binding sites tested here. These results suggest that interpretation of HpNikR-DNA interactions in the context of the HpNikR regulon and nickeldependent gene regulation in H. pylori should extend beyond previously inferred changes in DNA contacts afforded by the flexible N-terminal arm. The different protein conformations could be differentially sensitive to changes in intracellular phys-  Table 2, respectively. Asterisks denote affinities that are lower limit estimates.
iology other than nickel concentration and allow for selective regulation of a subset of NikR-controlled genes (e.g. in response to decreased pH) (43).
The different patterns of Fe-BABE-dependent cleavage of the nixA and ureA promoter fragments as well as the clear asymmetry in the cleavage pattern about the center of the DNase I footprint for ureA but not nixA (Fig. 2) argue for distinct conformations of HpNikR in the two protein-DNA complexes. These data are consistent with previous findings that showed a cation requirement for the arm truncation mutant nt9-HpNikR only for the ureA promoter in mobility shift assays (5). More detailed studies are required to understand the structural basis for these differences; however, it is likely that the N-terminal arm makes DNA contacts that are important for DNA binding. The two most likely possibilities are an encircling of the DNA by HpNikR, as has been demonstrated for repressor (44), and/or accommodation of spacer sequence asymmetry by altering DNA contacts via arm flexibility. We note that previous DMS footprinting results were identical for native and nt9-HpNikR and did not show protection of the few G residues in the spacer regions of either DNA fragment.
The effect of the K48A mutation on the HpNikR-ureA interaction was unexpected. The identification of a putative Glu 47Ј -Lys 48 salt bridge between the ␣2 helices of the RHH dimer as a requirement for high affinity binding to ureA but not nixA adds a new perspective for the identification and understanding of differences in protein conformation that exist between HpNikR tetramers bound to different promoters. In particular, these data suggest that conformational changes are not restricted to the protein-DNA interface. The available crystal structures of HpNikR reveal that the Lys 48 -Glu 47Ј interaction (Lys-N⑀ to Glu-O␦ or -O⑀ distance Ͻ3.2 Å) is present in several instances (38,40). In many cases, only one of the two Lys residues in the RHH domain is visible. None of these structures includes DNA, although SO 4 2Ϫ ions from the crystallization buffer are visible in some cases. There is no correlation between the occurrence of the salt bridge interaction and the presence of nickel ions in the structure. A Lys 48 -Asp 52 interaction (Ͻ3.2 Å) can be seen in two of the structures (40), and although intrahelical salt bridges (i to i ϩ 4) are known to influence helical stability based on model peptide studies (45,46), this type of interaction does not appear to be important for the HpNikR-DNA interactions studied here.
Only a few NikR sequences have the Glu 47 -Lys 48 residue pair (39), so this interaction does not appear to be widespread among NikR orthologs. It is possible that different residue pairs are used in a similar fashion by other NikR orthologs. Alternatively, residue 48 may interact with a different partner, such as residue 52, in some NikR proteins. The low sequence conservation of this protein region and its proximity to the interface of the Ni 2ϩ -and DNA-binding domains suggest that it may be particularly amenable to changes that influence function.
The location of the Glu 47 and Lys 48 in the ␣2 helix suggests that their interaction may affect the HpNikR DNA-contacting ␤-sheet differently at nixA and ureA. In every RHH protein-DNA co-crystal structure, there are interactions between residues at the N terminus of the ␣2 helix and the phosphate backbone flanking each end of the repeat sequences (23-25, 47).
These protein-DNA contacts are thought to help anchor the ␤-sheet motif to DNA (23). The structural mechanism by which Ni 2ϩ binding controls NikR DNA binding is not yet fully understood but probably involves an allosteric network of residues that affects non-covalent interactions at the N-and C-domain interface (39,48). The Lys 48 -Glu 47Ј interaction may provide a connection between Ni 2ϩ and DNA binding, as can be observed in one HpNikR structure (Protein Data Bank entry 2WVE) where a Lys 48 -Glu 47Ј -Lys 137 linkage is present. However, this structure probably does not represent the ureA-bound conformation of HpNikR because the K48E/E47ЈK swap mutant is unlikely to interact with Lys 137 but is still competent for DNA binding.
There is also a key difference between the ␣2 helix N-terminal residues of HpNikR and EcNikR. This position is considered an N-Cap residue, and makes a side chain to main chain hydrogen bond interaction (i to i ϩ 4) that is known to stabilize the N terminus of an ␣-helix (49). This residue is Ser in HpNikR and Asn in EcNikR. In EcNikR, the Asn side chain also contacts the DNA phosphate backbone due to bifurcation of the polar amide group. This type of nonspecific DNA interaction seems less likely to occur in DNA-bound HpNikR because of the more limited hydrogen bond capability of the Ser-OH side chain, which probably only forms the N-cap hydrogen bond. Alternatively, if the Ser residue can interact with DNA, its position may be influenced by the relative orientation of helix ␣2 in a way that depends upon the interaction of Lys 48 -Glu 47Ј . DNA contacts made by the N-terminal arm of HpNikR may help to compensate for the loss of the N-cap residue DNA contact and provide flexibility in DNA recognition.
The Fe-BABE cleavage data are consistent with DNA interactions outside of the inverted repeat sequences. Such interactions have been observed in other RHH-DNA complexes, both experimentally and structurally (23)(24)(25)47). The hybrid promoter fragments provide complementary evidence in support of these interactions as well as differences between HpNikR-DNA interactions with ureA and nixA. Although the affinity changes of wild-type HpNikR for different promoters was modest, interactions of the K48A mutant revealed highly sensitive responses to subsets of the ureA promoter sequence swapped into the nixA fragment. These data show that a combination of half-site, spacer, and flanking sequences is required for high affinity DNA binding. Together with the evidence that the N-terminal arm impacts DNA binding specificity, this suggests that the structure of the DNA recognition sites outside of the HpNikR half-sites may be critical for specific DNA binding.
The role of the spacer DNA sequence in HpNikR DNA recognition is reminiscent of DNA binding by the 434 repressor protein (50). The central four bases of a 14-bp recognition motif are critical for 434 repressor binding to six distinct operator sites and, interestingly, are not all contacted by the repressor protein. DNA affinity is intimately linked to the A-T content of this 4-bp spacer, presumably due to the extent to which the spacer can overwind, compress, and properly orient the repressor-binding half-sites. The nixA and ureA spacers are A-T-rich, with the run of six As in nixA replaced by TTTTAA in ureA. The dramatic reduction in HpNikR K48A affinity after substitution of the nixA spacer with that of ureA supports the idea that the spacer structure is an integral determinant of HpNikR DNA binding. Spacers from the eight promoters to which HpNikR directly binds exhibit highly variable A-T content, with no two spacers containing the same length runs of As. One possibility is that distinct HpNikR conformations recognize variable half-site spacing, which result from differences in spacer sequence content. In this case, the spacer would constitute a significant determinant of DNA binding affinity, as was observed for the nixA and ureA spacers. A half-site sequence may bind preferentially to a particular HpNikR conformation that enables stable complex formation. Stepwise selectivity via cooperative protein interactions after half-site binding has been observed for the RHH Mnt repressor binding to DNA (51).
The NikR tetramer binds to the largest operator site of any RHH protein. The greater spacing between half-sites is a consequence of the C-terminal nickel-binding domain, which is physically interposed between the RHH domains in the DNA complex. The DNA duplexes of many of the RHH-DNA cocrystal structures are bent to varying degrees (Arc, 50°(25); MetJ, 50°(24); CopG, 60°(29); FitA, 44°(52); EcNikR, 22° (27); ParR, 46°(53)), which further supports the idea that RHH proteins with different ␤-sheet motifs require variations in flanking DNA structure for optimal DNA binding. It is likely, then, that models of RHH protein-DNA interactions need to be expanded to include the analysis of DNA structural specificity outside of the minimal binding sites often predicted from lower resolution experiments.
HpNikR directly binds to at least eight different promoters (4 -10, 12, 54), although detailed mapping of HpNikR-DNA interactions has only been performed for two of these (5). The interaction of HpNikR at the ureA promoter has been interpreted as transcriptional activation because of the enhancement in ureA expression upon the addition of nickel, which requires a functional nikR gene. However, transcriptional activation has not been demonstrated in vitro, so the simple conclusion that HpNikR-DNA interactions at nixA and ureA reflect different functional roles is probably premature. Preliminary binding experiments with K48A HpNikR and the HpNikR-repressed frpB4 promoter showed a ϳ50-fold reduction in affinity. 4 Because the sequences of the half-sites and spacer of frpB4 are different from both nixA and ureA, this result suggests that HpNikR does not adopt conformations that are specific to the binding site position relative to the start of transcription but instead adapts conformation to sequence. Future studies that take a comprehensive approach to defining the molecular details of HpNikR-DNA interactions with several promoters will help to identify how flexibility and specificity in RHH domain-DNA interactions are achieved. The K48A mutant will be a useful tool for these studies.