A nickel-cobalt-sensing ArsR-SmtB family repressor. Contributions of cytosol and effector binding sites to metal selectivity.

NmtR from Mycobacterium tuberculosis is a new member of the ArsR-SmtB family of metal sensor transcriptional repressors. NmtR binds to the operator-promoter of a gene encoding a P(1) type ATPase (NmtA), repressing transcription in vivo except in medium supplemented with nickel or, to some extent, cobalt. In a cyanobacterial host, Synechococcus PCC 7942 strain R2-PIM8(smt), NmtR-mediated repression is alleviated by cobalt but not nickel or zinc addition, while the related sensor SmtB responds exclusively to zinc. Quantification of the number of atoms of nickel per cell shows that NmtR nickel sensitivity correlates with cytosolic nickel contents. Differential metal discrimination in a common cytosol by SmtB (zinc) and NmtR (cobalt) is not simply explained by affinities at equilibrium; although NmtR does bind nickel substantially more tightly than SmtB, it has a higher affinity for zinc than for cobalt and binds cobalt more weakly than SmtB. SmtB is known to bind and sense zinc at interhelical four-coordinate, tetrahedral sites across the C-terminal alpha 5 helices, while absorption spectroscopy of Co(II)- and Ni(II)-substituted NmtR reveals five- and six-coordinate metal complexes. Site-directed mutagenesis identifies six potential cobalt/nickel ligands that are obligatory for inducer recognition but not repression by NmtR, four of which (Asp(91), His(93), His(104), His(107)) align with alpha 5 ligands of SmtB with two additional His provided by a carboxyl-terminal "extension" (designated alpha 5C). Gel retardation assays reveal that zinc does not allosterically regulate NmtR-DNA binding at concentrations where lower affinity cobalt does. These data suggest that two additional ligands form hexacoordinate metal complexes and are crucial for driving allosteric regulation of DNA binding by NmtR, thereby allowing NmtR to preferentially sense metals that favor higher coordination numbers relative to SmtB.

Cells contain regulatory proteins to detect and respond to deficiency or excess of essential metals to maintain sufficient atoms to satisfy the requirements of metalloproteins while avoiding toxicity (1). The ArsR-SmtB family of transcriptional repressors associate with the promoters of genes encoding proteins involved in the efflux and/or sequestration of excess metal (2). De-repression occurs when the repressors bind metal effectors coincident with the number of atoms exceeding an optimal cell quota. SmtB-mediated repression is alleviated by Zn(II) (3), ZiaR by Zn(II) (4), ArsR by As(III), Sb(III), and Bi(III) (5), CadC by Cd(II), Pb(II) and Bi(III) (6 -8), and CzrA by Co(II) and Zn(II) (9,10). Clearly these sensors discriminate between different metals in vivo, but the factors dictating which inorganic elements elicit responses remain to be defined.
There is rich literature describing metal coordination by numerous small molecules in vitro and established theories cataloguing the factors likely to influence metal selectivity in vivo (11). The ligand environments of metal ions are also known in a vast array of metalloenzymes (12). The binding sites of enzymes are not only influenced by metal selectivity but also by catalytic constraints, different secondary and tertiary structures, and evolutionary histories. Metalloregulatory proteins have some advantages for exploring metal selectivity in vivo. First, selectivity will have been a dominant factor in the evolution from a common ancestor of structurally similar sensors that detect different metals. Second, by associating their target promoters with reporter genes it is possible to monitor metal occupancy in vivo. In some, if not all, cells there is an absence of free copper (13), and it is likely that this is also true of several other essential metals including zinc (14). Thus, factors that influence the probability of sensors encountering different labile metal ions are likely to influence metal specificity. For example, metallochaperones assist in the delivery of metals, including nickel (15) to some proteins or target compartments (16), promoting advantageous metal-protein partnerships while inhibiting others en route. In a two-hybrid assay, a copper metallochaperone from Synechocystis PCC 6803 was shown to interact with copper transporting P 1 -type AT-Pases but not with structurally related zinc or cobalt transporters (17), which illustrates how the specificity of metallochaperone-metalloprotein interactions could define which metals are acquired by which proteins in vivo.
To identify the regulatory metal binding sites of SmtB we previously generated mutants of Cys and His candidate ligands. One, or both, of a pair of His residues (105/106) was/ were required for metal recognition but not repression (18). Difference electron density maps obtained after soaking apo-SmtB crystals with mercuric acetate suggested two symmetryrelated pairs of metal binding sites per dimer (19). A pair of metal sites was located close to the ␣3 helix within the DNAbinding helix-turn-helix motif and a second pair was formed by four ligands, two from each monomer, bridging antiparallel carboxyl-terminal ␣5 helices. Fractional occupancies were low (Ͻ2%), and it was subsequently established that SmtB binds only one zinc per monomer with affinity K Zn in excess of 10 11 M Ϫ1 at equilibrium (20). Substitution of cobalt into the zinc sites of SmtB gave spectral features diagnostic of tetrahedral coordination environments with one or two Cys ligands (20). Zinc and cobalt x-ray absorption spectroscopy (21) coupled with 15 N-1 H NMR peturbation spectroscopy (21) implicate Cys 14 , His 18 , and Cys 61 in a site designated ␣3N (Fig. 1A). However, Cys variant proteins, deficient in ␣3N, retain inducer responsiveness in vivo (18) and zinc-dependent DNA dissociation in vitro (22). In contrast H106Q SmtB was refractory to zincinduced disassembly of SmtB-DNA complexes in vitro (22) consistent with loss of inducer recognition in vivo (18). Occupancy of ␣5 sites (which include His 106 , Fig. 1A) regulates DNA binding by SmtB even though ␣3N sites are occupied in dissociated SmtB (Fig. 1B). In contrast, trigonal thiolate sites adjacent to a predicted helix-turn-helix region are required for inducer recognition by ArsR (23), while a tetrathiolate ␣3N site modulates CadC DNA binding in vitro (8,24) and CadC inducer recognition in vivo (7). Different allosteric sites (␣3N or ␣5) with distinct ligand sets and geometries correlate with, and presumably contribute toward, the biological metal specificities of individual ArsR-SmtB family members. Analogous observations have been made for Escherichia coli Fur homologues, Fur and Zur (25).
To identify ArsR-SmtB sensors with new specificities, sequence databases were searched for smtB-related genes adjacent to genes predicted to contribute to homeoestasis of other metals. Ten genes were identified in the fully sequenced genome of Mycobacterium tuberculosis (26), and Rv3744 was selected due to its proximity to a divergently transcribed gene (Fig. 1C) encoding a deduced protein with similarity to CoaT (27). Genes encoding identical proteins are present in Mycobacterium bovis (www.sanger.ac.uk) strain BCG (28). An initial aim of this research was to establish whether the product of Rv3744 binds the adjacent operator-promoter to repress transcription, except in the presence of an excess of some metal. Our data support this, and the genes are designated nmtR and nmtA (Fig. 1C).
In this paper we show that NmtR represses expression from the nmt operator-promoter in vivo with repression specifically alleviated by elevated nickel or cobalt, the former being the more potent. Comparative analyses of NmtR and the related zinc sensor SmtB, in vitro and in vivo, lead to some surprising findings. While SmtB detects different metals relative to NmtR in the same cell, revealing intrinsic differences in these sensors, the host cytosol can also influence biological metal selectivity of these metal sensors. We also show that metal sensor selectivity does not simply correlate with metal affinity at equilibrium, but is instead consistent with a model in which distinct coordination geometries of inducing metal complexes of NmtR versus SmtB drive distinct allosteric pathways to effect regulation of DNA binding in each case.

EXPERIMENTAL PROCEDURES
Bacterial Strains and DNA Manipulations-Mycobacterium smegmatis mc 2 155 and M. bovis BCG (Pasteur) were used as mycobacterial hosts, and Synechococcus PCC 7942 strain R2-PIM8(smt) (29), lacking functional smtA and smtB genes, was used as a cyanobacterial host. The smtB-deficient status of the latter alleviates any influence of SmtB (with a similar recognition helix to NmtR) on expression from the nmtA operator-promoter in this host. Mycobacterial cells were grown with shaking at 37°C in LB medium (30) containing 0.05% (v/v) Tween 80, and cyanobacterial cells were grown at 30°C in Allen's medium using conditions as described (29). E. coli strains JM109 (Stratagene) and BL21(DE3) were used and grown in LB medium. Cells were transformed to antibiotic resistance as described (29 -31). Standard DNA manipulations were performed as described by Sambrook et al. (30). All generated plasmid constructs were checked by sequence analysis.
Construction of nmt-lacZ Fusions, Site-directed Mutagenesis, and ␤-Galactosidase Assays-M. tuberculosis H37Rv genomic DNA was used as template for PCR with primers I (5Ј-GAAGGATCCGGCCAA-CATATCAG-3Ј) and II (5Ј-GAAGAATTCTGGGGTCTGTAAAGCTCG-3Ј) and the amplification product (497 bp) containing the nmtA operator-promoter and nmtR (Fig. 1C) ligated to pGEM-T prior to subcloning into the SalI/BamH1 site of pLACPB2 (32) or the ScaI/BamH1 site of pJEM15 (31) to create transcriptional fusions with lacZ. "QuikChange" (Stratagene) site-directed mutagenesis was subsequently used, according to the manufacturer's protocols, to generate derivatives with codon substitutions within nmtR: Gln41 to a UAG stop codon; Asp 51 , Asp 91 , Asp 99 , and Asp 114 to Ala; and His 93 , His 104 , His 107 , His 109 , and His 116 to Arg. The pLACPB2-and pJEM15-based constructs were introduced into cyanobacterial and mycobacterial hosts, respectively. R2-PIM8(smt) containing smtB and the smtA operator-promoter in pLACPB2 (3) was used to examine expression from the smtA operatorpromoter. ␤-Galactosidase assays were performed as described previously (33), in triplicate on at least three separate occasions. The media were supplemented with various [metal] (described in individual experiments) for approximately 20 h immediately prior to assays. These assays therefore differed from previous reported assays (3,29,33), to examine expression from the smtA operator-promoter, which used much shorter ( Gel Retardation Assays-For these experiments recombinant NmtR was generated as a fusion to glutathione S-transferase by subcloning the nmtR coding region, generated by PCR using primers II and III (5Ј-GAAGGATCCATGGGGCACGGGGTCGAAG-3Ј) with M. tuberculosis H37Rv DNA as template, into the BamHI/EcoRI site of plasmid pGEX-6P2 (Amersham Biosciences). Recombinant fusion protein was expressed in E. coli JM109, cleaved using precision protease, purified according to manufacturer's protocols, and dialyzed against 10 mM Tris-HCl (pH 7.5), 1 mM EDTA, followed by 10 mM Tris-HCl (pH 7.5). A single prominent band corresponding to the predicted size of NmtR and five residues of glutathione S-transferase (13.238-kDa) was detected by PAGE. Crude cyanobacterial cell extracts were prepared as described (33) and protein concentrations determined using Coomassie Blue R-250 (using bovine serum albumin standards). Gel retardation assays were performed (18) with EDTA omitted from buffers unless otherwise stated. The probe in all cases was 71-bp XbaI/BamHI fragment from pGEM-T containing the nmt operator-promoter region generated by PCR using primers III and IV (5Ј-GAATCTAGATGGTTAGGCAGCC-3Ј) with M. tuberculosis H37Rv DNA as template. Examination of metalinduced DNA dissociation involved adding increasing [Co(II)] or [Zn(II)] to binding reactions containing recombinant NmtR; to check for reversibility 1 mM EDTA was also added to some reactions after 50 min.
Direct Metal Binding of NmtR and SmtB-Titrations of apo-NmtR (Յ0.05 mol eq of Zn(II) as determined by atomic absorption spectroscopy; ⑀ 280 ϭ 4470 M Ϫ1 cm Ϫ1 ) with Ni(II) and Co(II) were monitored by UV-visible optical absorption spectroscopy (600 M NmtR) or by steadystate tyrosine fluorescence (for Ni(II), Co(II), and Zn(II); 5.0 M NmtR) in 10 mM HEPES (pH 7.0), 450 mM NaCl, 22°C essentially as described previously for SmtB (20,21). To monitor Zn(II) binding in competition between SmtB and NmtR, Co(II)-SmtB was titrated with Zn(II) in the presence and absence of stoichiometric apo-NmtR and the optical spectra recorded as described and used to monitor bleaching of Co(II)-SmtB complexes. To prepare Co(II)-SmtB, recombinant apo-SmtB was ex-pressed from pET29a, prepared as described for NmtR, and preincubated in an anaerobic chamber with a 4-fold molar excess of Co(II). Studies of competitive metal binding to NmtR (Zn(II) versus Ni(II) versus Co(II)) involved titrating apo-NmtR in 20 mM MES (pH 6.0), 150 mM NaCl, with combinations of Co(II), Ni(II), and Zn(II), and measuring [metal] bound using atomic absorption spectrophotometry, following separation on Sephadex G-25.
Determination of Metal Quotas-Cells, R2-PIM8(smt) and M. smegmatis mc 2 155, were harvested during exponential growth and washed three times with 10 mM Tris-HCl (pH 7.5), 1 mM EDTA and once with Milli-Q H 2 0. Pelleted cells were dried overnight at 80°C, dissolved in 70% (v/v) nitric acid, and the metal content measured by atomic absorption spectrophotometry. Metal contents were determined as atoms mg Ϫ1 cellular protein and atoms cell Ϫ1 (from hemocytometer counts) or cyanobacterial cell volume equivalent for the mycobacteria as determined from packed cell volumes. Assays were performed in triplicate on at least three separate occasions, and parallel control experiments eliminated any metal contamination from the materials used.

NmtR Binds to the nmt Operator-Promoter Region and Is a
Nickel-and Cobalt-responsive Repressor-Similarity of the deduced product of open reading frame Rv3744 (NmtR) to the zinc-responsive repressor SmtB (3), and of the divergently transcribed gene (nmtA) to metal-transporting P 1 -type ATPases, suggests that the former might bind to the intervening operator-promoter region to regulate the latter. To test whether or not NmtR binds to the operator-promoter region separating nmtR and nmtA (Fig. 1C), NmtR was expressed in E. coli, purified, and used in gel retardation assays. A single retarded complex was detected ( Fig. 2A), which was retained in the presence of nonspecific (poly(dI-dC)⅐poly(dI-dC)), but not specific (nmt operator promoter region), competitor DNA (data not shown).
NmtA shows greater similarity to the cobalt exporting P 1type ATPase CoaT (27) than to related zinc exporters ZntA (34,35) or ZiaA (4). It is probable that the same metals will be sensed by NmtR and transported by NmtA and therefore speculated that metal specificities of at least one of the proteins differ from those naively inferred from homologies. To establish which (if any) metals induce transcription from the nmtA operator-promoter, a 497-bp region including the operator-promoter region separating the two genes plus the entire nmtR coding region was fused to a promoterless lacZ in plasmid pJEM15 and introduced into M. smegmatis mc 2 155. Elevated ␤-galactosidase activity was detected in response to exposure (20 h) to maximum permissive concentrations of cobalt or nickel but no other metals (Fig. 2B). Elevated ␤-galactosidase activity in the absence of added metal ions was detected from an analogous construct in which codon Gln 41 within the nmtR coding region had been converted to a stop codon, confirming that NmtR acts negatively toward expression from the nmtA operator-promoter. Exposure to a range of concentrations of zinc, cobalt, and nickel ( Fig. 2C) revealed that no viable concentration (insets in Fig. 2C) of zinc induced expression from the nmtA operator-promoter, while nickel was the most potent inducer. Equivalent trends were observed in M. bovis BCG (not shown). In a Common Cell, R2-PIM8(smt), NmtR Responds to Cobalt and SmtB to Zinc-Differences in the metals that alleviate repression of transcription from the nmtA (Fig. 2) and smtA operator promoters (3) suggest that NmtR and SmtB can bind and/or allosterically respond to different metals. However, the promoters have been analyzed in different cell types, mycobacteria or cyanobacteria, where the abundance and/or chemical forms of the labile pools of these metal ions may differ. To investigate whether or not metal specificity is an intrinsic property of the sensors and/or influenced by the host cytosol, nmt reporter constructs were introduced into a cyanobacterium, R2-PIM8(smt), for direct comparison with SmtB in the same cell. Extracts from these cells formed complexes with the nmt operator-promoter in gel retardation assays; the complexes were absent when extracts were used from cells containing an equivalent construct with a stop codon in the nmtR open reading frame, confirming expression of NmtR in the heterologous host (Fig. 3A). In the absence of metal supplements, ␤-galactosidase activity was lower in cells with functional nmtR than nmtR containing a stop codon confirming NmtRmediated repression (Fig. 3B). Exposure to maximum permissive concentrations of cobalt, but not zinc or any other tested metal, alleviated NmtR-mediated repression, while only zinc alleviated SmtB-mediated repression of expression from the smtA operator promoter under the same assay conditions (Fig.  3B). The two sensors show converse discrimination between cobalt and zinc in the same cytosol.
NmtR Does Not Respond to Nickel in R2-PIM8(smt) where the Nickel Content Is Lower and Less Variable than the Mycobacterium-Exposure to maximum permissive concentrations of nickel alleviated NmtR-mediated repression in the mycobacterium (Fig. 2B) but not in the cyanobacterium (Fig. 3B). To further investigate whether or not NmtR responds to nickel in the heterologous host, ␤-galactosidase activity was measured in response to a range of metal concentrations up to and including inhibitory doses (insets in Fig. 4). No increase was detected in response to any concentration of nickel, indicating a difference in the labile nickel pools of the two bacteria. NmtR only responded to cobalt, and SmtB to zinc, in the cyanobacterium (Fig. 4).
The nickel contents of both organisms were determined in cultures grown in normal medium or following supplementation with maximum permissive (labeled H on Figs. 2 and 4) or half-maximum permissive (L) nickel concentrations. Values were expressed as (i) number of atoms per cell (for the cyanobacterium) or per cyanobacterial cell volume equivalent for the mycobacterial cells and (ii) relative to protein content. These values increase by 37-and 38-fold in the mycobacterium but only 3.5-and 3-fold in the cyanobacterium following nickel supplementation. No increase was detected between half-maximum permissive and maximum permissive concentrations in the cyanobacterium. It is inferred that NmtR does not respond to nickel in the cyanobacterium, because the metal is excluded from the cell, relative to the mycobacterium.
SmtB Outcompetes NmtR for Zinc- Table I shows no significant increase in the zinc content of zinc-supplemented mycobacterial cells, but shows a greater than 10-fold increase in cyanobacteria. Exclusion of zinc from the mycobacterium could theoretically have accounted for the lack of a response of NmtR to zinc. Only by having established that NmtR also fails to respond to zinc in the cyanobacterium, where zinc is clearly available for detection by SmtB (Fig. 4), can it be concluded that this difference in selectivity reflects intrinsic differences between SmtB and NmtR. A simplistic explanation is that SmtB has a high affinity for zinc, low affinity for cobalt, and NmtR has a low affinity for zinc but high affinity for cobalt. The heirarchy of affinities for SmtB is Zn(II) Ͼ Ͼ Co(II) Ͼ Ͼ Ni(II) (20). To establish the heirarchy of metal binding to NmtR, competitive binding experiments and direct titrations (see be- low) were performed. Apo-NmtR was incubated with an excess (with respect to protein) of equimolar amounts of each paired combination of these ions, fractionated on Sephadex G-25, bound and free metal quantified by atomic absorption spectrophotometry to determine which ion was predominantly bound and indicating the order Zn(II) Ͼ Ni(II) Ͼ Ͼ Co(II). It is possible that while both proteins bind zinc with highest affinity only SmtB can compete with endogenous ligands, implying that NmtR binds zinc with lower affinity than SmtB. The two ␣5 sites of SmtB, which mediate zinc-dependent DNA dissociation, have K Zn Ϸ 7.8 ϫ 10 11 M Ϫ1 . The ␣3N sites that are occupied in free solution have K Zn of at least 10 13 M Ϫ1 (21). The zincbinding site(s) of NmtR remain to be defined, and K Zn may also exceed 10 13 M Ϫ1 ; competitive binding was therefore used to directly establish which protein preferentially acquires zinc.
Cobalt-SmtB is blue with distinctive spectral features around 550 nm, and between 300 and 400 nm (20). In contrast, cobalt-NmtR is nearly colorless with molar absorptivities in the visible region of Յ100 M Ϫ1 cm Ϫ1 (see below). Addition of zinc displaces cobalt and thereby bleaches cobalt-SmtB (20). Zincmediated bleaching of cobalt-SmtB was unaffected by the presence of an equimolar amount of NmtR (Fig. 5), establishing that zinc binds to SmtB in preference to NmtR and implying a difference in affinity of at least one order of magnitude.
At Equivalent Concentrations Cobalt, but Not Zinc, Cause Dissociation of NmtR from DNA-Of the three ions tested, zinc binds to NmtR with the highest affinity, cobalt the lowest, and yet the latter alleviates NmtR-mediated repression in vivo but the former does not (Fig. 4). Perhaps cobalt-NmtR undergoes an allosteric change to impair DNA binding, while zinc-NmtR does not. This was tested in gel retardation assays. Purified NmtR remains associated with nmt operator-promoter DNA in the presence of zinc at concentrations where cobalt inhibits complexes (Fig. 6). Higher zinc concentrations inhibited the formation of NmtR DNA complexes, but this requires further investigation due to evidence of protein precipitation. Preliminary titration of DNA with preformed complexes of apo-, cobalt-, nickel-, or zinc-NmtR, monitored via fluorescence anisotropy, was also consistent with nickel and cobalt similarly regulating complex formation with both metals more effective than zinc (data not shown).
Identification of Six ␣5C Residues That Are Essential for Nickel and Cobalt Sensing by NmtR-Some difference in the effector recognition site of NmtR compared with SmtB might (i) favor binding of nickel over cobalt, (ii) disfavor binding of zinc in competition with SmtB (Fig. 5B), and/or (iii) favor allosteric regulation by nickel and cobalt in preference to zinc (Fig. 6). The next challenge was to identify the inducer recognition site of NmtR. Asp 51 in NmtR aligns with conserved Asp residues in other family members (8) and, at least in SmtB, is thought to contribute one ␣3N ligand (21). Ala substitution of Asp 51 did not impair either NmtR-mediated repression or nickel/cobalt recognition (Fig. 7). Inducer recognition by NmtR either does not involve ␣3N sites, or Asp 51 is not an essential ␣3N ligand.
While inducer recognition by many ArsR-SmtB family members requires ␣3N sites (8,23), in SmtB ligands from antiparallel ␣5 helices at the carboxyl-terminal dimer interface are obligatory for metal-mediated DNA dissociation (21). The ␣5 helices of NmtR were predicted based upon the coordinates for SmtB, and Fig. 7 shows a hypothetical dimer interface at the carboxyl-terminal region of NmtR. Four candidate nickel/cobalt ligands in NmtR, Asp 91 , His 93 , His 104 , and His 107 (Fig. 7) correspond to the ␣5 residues Asp 104 , His 106 , His 117 , and Glu 120 of SmtB (Fig. 1A). Substitution of any one of these residues in NmtR created functional repressors that mediated low expression of lacZ from the nmtA operator-promoter in mycobacterial cells grown in the absence of metal supplements. In contrast ␤-galactosidase activity was constitutively elevated in cells containing a non-functional mutant in which the codon for Gln 41 had been substituted with a stop codon. Most importantly, substitution of NmtR residues aligning with the ␣5 ligands of SmtB caused loss of inducer recognition with ␤-galactosidase activity remaining low in the presence of nickel and cobalt concentrations that cause loss of repression by wild-type NmtR.    6. Cobalt inhibits NmtR binding to nmt operator-promoter DNA more effectively than zinc in vitro. Gel retardation assays were performed using 1 ϫ 10 Ϫ9 M nmt operator-promoter DNA as probe (FP, free probe) and 1 ϫ 10 Ϫ8 M NmtR with (left to right) 0, 1 ϫ 10 Ϫ5 , 2.5 ϫ 10 Ϫ5 , 5 ϫ 10 Ϫ5 , or 7.5 ϫ 10 Ϫ5 M Co(II) or Zn(II) added to binding reactions. EDTA (1 mM) was subsequently added to duplicate reactions containing 5 ϫ 10 Ϫ5 or 7.5 ϫ 10 Ϫ5 Co(II) or Zn(II) as indicated.

Discriminatory Metal-sensing by NmtR
NmtR has 11 additional carboxyl-terminal residues relative to SmtB (drawn as unstructured ribbon on Fig. 7), including three additional potential nickel/cobalt ligands, His 109 , Asp 114 , and His 116 . Substitution of Asp 114 had no detectable effect on repression or inducer recognition, but substitution of either His created inducer non-responsive functional repressors, thereby identifying a total of six residues, all of which are obligatory for either cobalt or nickel recognition (Fig. 7).
UV-visible Absorption Spectroscopy of Ni(II)-and Co(II)-substituted NmtR-Site-directed mutagenesis of NmtR (Fig. 7) suggests that inducer recognition could involve hexadentate Ni(II) or Co(II) coordination complexes formed by extended carboxyl-terminal ␣5C sites in a way in which six ligands (rather than four for SmtB) are required for allosteric regulation of DNA binding, and thus sensing. Consistent with this, the saturated UV-visible absorption spectrum of Ni(II)-NmtR (Fig. 8A) recorded at a 1:1 Ni(II)-NmtR monomer ratio reveals three very weak (⑀ Յ80 M Ϫ1 cm Ϫ1 ) and very broad ligand field absorption transitions diagnostic (36) of six-coordinate d 8 Ni(II). It is noted that the gradual upward slope in the cor-rected spectrum of Ni(II)-NmtR is not due to light scattering, but rather to the non-resolved nature of low intensity absorption bands characteristic of Ni(II) in coordination complexes deviating from perfect trigonal bipyramidal or octahedral coordinate symmetry (37). The inset (Fig. 8A) reveals that the stoichiometry of Ni(II) binding to NmtR is 1 Ni(II) per monomer or 2 per dimer (NmtR is fully dimeric under these conditions) 2 with a lower limit of the affinity for Ni(II), K Ni Ն2 ϫ hedral Co(II) complexes described in the literature (39). Consistent with the competitive metal binding experiments described above, Co(II) binds to NmtR with an affinity Ն40-fold weaker than Ni(II) and at least 500-fold weaker than Zn(II) (inset, Fig. 8B) again implying that for NmtR, K Zn Ն K Ni Ͼ K Co . Remarkably, K Co for NmtR is Ϸ3000-fold smaller for NmtR relative to SmtB under similar solution conditions (20,21) despite the finding that NmtR senses Co(II), while SmtB does not (Fig. 3). DISCUSSION Several lines of evidence demonstrate that NmtR is a nickel/ cobalt-responsive DNA-binding repressor of transcription from the divergent nmtA operator-promoter. (i) Purified NmtR forms specific complexes with the nmtA operator promoter in vitro ( Fig. 2A); (ii) equivalent complexes are detected using crude lysates of R2-PIM8(smt) containing nmtR but not cells containing an internal stop codon in nmtR (Fig. 3A); (iii) expression of ␤-galactosidase activity from the nmtA operator-promoter is elevated in mycobacterial (Fig. 7) and cyanobacterial (Fig. 3) cells devoid of functional nmtR compared with cells containing NmtR; (iv) expression of ␤-galactosidase activity via the nmtA operator-promoter is elevated in both bacterial cell types in response to elevated cobalt (Figs. 2 and 3), and nickel is the most potent inducer at viable concentrations in the mycobacterium (Fig. 2). Factors that contribute to metal selectivity in the context of a cell have been identified, i.e. which metals are detected by NmtR or SmtB.
The most potent allosteric effector of NmtR (nickel) in mycobacteria (Fig. 2C) is totally ineffective when nmtR is introduced into a cyanobacterium (Fig. 4). Clearly NmtR is produced and accumulated in the cyanobacterium (Fig. 3A), is functional as a repressor (Fig. 3B), and is competent to detect metal (cobalt) but unable to acquire and detect nickel even at concentrations inhibitory to cell growth (Fig. 4). These observations demonstrate that different cytosolic metal pools can determine the metals that metallo-sensory proteins detect in different cells. By analogy, cadmium and copper are the most potent inducers of dMTF-1 in Drosophila, but in transfected mammalian cells dMTF-1 responds to zinc, similar to its mammalian counterpart (40). In theory, related sensors from different bacteria responding to alternative metals in vivo could have identical metal-binding sites, identical allosteric responses to metals in vitro, but "restricted access" to different available metal pools in vivo.
Loss of nickel induction in R2-PIM8(smt) could indicate that a nickel metallochaperone for NmtR is absent in the cyanobacterium, suggesting a technical basis for selective screens for metallochaperones. However, cell nickel quotas (Table I) establish that more effective exclusion of nickel from the cyanobacterial, compared with mycobacterial, cytosol can account for the cellular differences observed here. Strikingly, SmtB does not respond to cobalt in the cyanobacterium even though it is intrinsically capable of allosterically responding to cobalt in vitro (22), and cobalt is available for detection in the cell at least by NmtR (Fig. 4); furthermore, SmtB binds Co(II) with a far greater affinity than does NmtR (Fig. 8). Although the nature of the available cobalt pool is unknown (association with proteins of cobalamin biosynthesis seems likely), these data imply that NmtR acquires cobalt from this pool far more effectively than does SmtB.
Crucially, SmtB and NmtR detect different metals (zinc and cobalt) when analyzed in the same cell type (Fig. 4) establishing that intrinsic features of these two proteins discriminate in favor of different metals. Most importantly this is not solely based on absolute binding preferences, since SmtB and NmtR both have higher affinities for zinc than for cobalt or nickel (Fig. 6), and two contributory factors have been identified. (i) SmtB outcompetes NmtR for zinc (Fig. 5) and could therefore more readily acquire zinc from other ligands in the cytosol. (ii) Zinc mediates allosteric regulation of DNA binding ineffectively (Fig. 6).
Why is cobalt (and nickel) more effective than zinc at promoting DNA dissociation by NmtR? NmtR absorption spectra are indicative of distorted octahedral Ni(II) coordination and five-or six-coordinate Co(II) liganding (Fig. 8). In contrast, SmtB is characterized by four-coordinate, tetrahedral complexes of both Co(II) and Zn(II) (20,21). Although the coordination geometry of Zn(II)-NmtR is not yet known, at least one explanation is that the difference in the allosteric regulation of DNA binding by NmtR versus SmtB requires metal coordination bonds to six protein-derived ligands, rather than four. This is consistent with the identification of six potential ligands by site-directed mutagenesis, each obligatory for inducer recognition in vivo. Ni(II) and Co(II) show a greater propensity to form octahedral coordination complexes relative to Zn(II) (11), and this is the most common geometry for Ni(II) and Co(II) catalogued in protein structural databases (12); in contrast, tetrahedral coordination geometry predominates for Zn(II) (12). Consistent with these trends, Zn(II) complexes of E. coli glyoxalase are five-coordinate and catalytically inactive, whereas the Ni(II) and Co(II) complexes recruit an additional water molecule into the first coordination sphere creating a nearly perfect octahedral complex, which exhibits high catalytic activity (41). It is tempting to speculate that two additional ligands provided by the COOH-terminal extension of the ␣5 helices in NmtR (Fig. 7) adapt negative regulation of operator/promoter binding previously observed for SmtB (22) to require hexadentate metal ligation environments. Metal detection within a cell might therefore be achieved in the absence of strict discrimination at the level of protein-metal binding. It remains unknown why the higher affinity zinc does not associate with NmtR in vivo in such a way that detection of nickel and cobalt is inhibited. This work highlights the need to identify the chemical form(s) of the labile pool(s) of metals accessible by each metal sensor and indeed other metalloproteins.