Predicting metals sensed by ArsR-SmtB repressors: allosteric interference by a non-effector metal


*E-mail; Tel. (+44) 191 222 7695; Fax (+44) 191 222 7424.


Many bacterial genomes encode multiple metal-sensing ArsR-SmtB transcriptional repressors. There is interest in understanding and predicting their metal specificities. Here we analyse two arsR-smtB genes, ydeT and yozA (now aseR and czrA) from Bacillus subtilis. Purified AseR and CzrA formed complexes in gel-retardation and fluorescence-anisotropy assays with fragments of promoters that were derepressed in ΔaseR and ΔczrA cells. Candidate (i) partly thiolate, α3-helix (for AseR) and (ii) tetrahedral, non-thiolate, α5-helix (for CzrA) metal binding sites were predicted then tested in vitro and/or in vivo. The precedents are for such sites to sense arsenite/antimonite (α3) and zinc (α5). This correlated with the respective metal inducers of AseR and CzrA repressed promoters in B. subtilis and matched the metals that impaired formation of protein–DNA complexes in vitro. The putative sensory sites of 1024 ArsR-SmtB homologues are reported. Although AseR did not sense zinc in vivo, it bound zinc in vitro exploiting α3 thiols, but AseR DNA binding was not impaired by zinc. If selectivity relies on discriminatory triggering of allostery not just selective metal binding, then tight non-effector metal complexes could theoretically inhibit metal sensing. AseR remained arsenite-sensitive in equimolar zinc, while CzrA remained zinc-sensitive in equimolar arsenite in vitro. However, cupric ions did not impair CzrA–DNA complex formation but did inhibit zinc-mediated allostery in vitro and prevent zinc binding. Access to copper must be controlled in vivo to avoid formation of cupric CzrA.


Cells detect metals to maintain supplies for metalloproteins while avoiding toxic excesses. The ArsR-SmtB family of repressors is one group of DNA-binding proteins that detect metal surplus to control the production of proteins that either export (Nies, 2003) or sequester (Blindauer et al., 2002) the respective ions. What determines which metals are sensed by these transcriptional regulators? Open reading frames (ORFs), encoding proteins with similarity to ArsR and SmtB, are commonly found in eubacterial and archaebacterial genomes with some organisms containing multiple representatives; for example, there are 10 in the genome (Cole et al., 1998) of Mycobacterium tuberculosis H37Rv. The metals that alleviate repression in vivo have been determined for at least eight family members: Zn(II) for SmtB from Synechococcus PCC 7942 (Huckle et al., 1993), As(III) and Sb(III) for ArsR (Wu and Rosen, 1991), Cd(II), Pb(II) and Zn(II) for CadC (Endo and Silver, 1995; Sun et al., 2001; Busenlehner et al., 2002), Zn(II) for ZiaR from Synechocystis PCC 6803 (Thelwell et al., 1998), Zn(II) and Co(II) for CzrA from Staphylococcus aureus (Kuroda et al., 1999; Singh et al., 1999), Ni(II) and Co(II) for NmtR from M. tuberculosis (Cavet et al., 2002), Cd(II) and Pb(II) for CmtR from M. tuberculosis (Cavet et al., 2003; Wang et al., 2005), Cu(I), Ag(I), Zn(II) and Cd(II) for BxmR from Oscillatoria brevis (Liu et al., 2004). The diversity of metal binding sites and allosteric mechanisms has been described as a ‘theme and variations model’ (Busenlehner et al., 2003). An objective is to be able to identify which subset of the more than 1000 sequenced but uncharacterized ArsR-SmtB homologues do detect metals, and ideally to correctly predict which metals each one senses.

Metal-sensing transcriptional regulators are also useful for probing the disposition of metals in cells, and exploring the kinetic and thermodynamic factors that allocate metals to the correct proteins (Outten and O’Halloran, 2001; O’Halloran and Finney, 2003) because: (i) they can reveal metal occupancy in vivo via the association of reporter genes with their target promoters and (ii) in comparison to the binding sites of many other metalloproteins, metal selectivity is likely to have been a dominant factor driving the divergence from common ancestors of the metal binding sites of related sensors that detect different elements (Tottey et al., 2005). Using pair-wise comparisons of ArsR-SmtB sensors of differing metal selectivities, contributions to specificity of: (i) metal partitioning based on metal-affinity, (ii) metal-specific allostery and (iii) differential access to metals in vivo have been documented (Cavet et al., 2002; 2003). Analogous pair-wise comparisons within other families of metal sensors have similarly provided molecular details of differences in metal binding sites that mediate selective allostery and/or affinity, for example, copper-CueR versus zinc-ZntR, and manganese-MntR versus iron-DtxR (Changela et al., 2003; Guedon and Helmann, 2003). Finding that metal specificity is not always solely based on selective metal binding and partitioning has intriguing implications that must now be investigated. For example, the cobalt and nickel sensor, NmtR, has a tighter affinity for zinc than for nickel or cobalt, but zinc is not efficiently sensed because its preferred co-ordination geometry is unlike that of nickel and cobalt so zinc is less effective at correctly re-positioning NmtR ligands to trigger allostery (Cavet et al., 2002; Pennella et al., 2003). NmtR binds zinc 1000-fold tighter than cobalt, and therefore a simple untested prediction is that this poor effector will prevent cobalt binding and inhibit inducer recognition. Evidence of such inhibition in vitro, but not in vivo, will reveal where access to different metals might be kinetically controlled in cells. Here we begin to explore antagonism between non-effector metals and inducer recognition in vitro.

Derepression of ArsR-SmtB-regulated promoters occurs when the proteins bind the respective metal(loid) resulting in a weaker affinity for DNA. Predicted helix–turn–helix DNA binding regions of SmtB, CadC and ArsR are conspicuously associated with Cys residues, which are obvious candidate metal ligands (Morby et al., 1993). Substitution of these Cys residues generated ArsR repressors that were not responsive to their cognate inducers (Shi et al., 1994). In contrast, analogous thiols in SmtB are not required for inducer recognition but carboxyl-terminal His are essential for zinc detection by SmtB (Turner et al., 1996). ArsR-SmtB family members have been modelled on the known winged helix homodimeric structures of SmtB (Cook et al., 1998) (Fig. 1A), and S. aureus CzrA (Eicken et al., 2003), plus most recently CadC (Ye et al., 2005). Spectral studies id

Figure 1.

A. Two of the types of sensory sites in ArsR-SmtB transcriptional repressors indicated on the known structure of SmtB. B. subtilis CzrA has candidate metal ligands at predicted α5-helices (in parenthesis) analogous to the sensory ligands of SmtB (not in parenthesis). B. subtilis AseR has candidate metal(loid) ligands at α3-helices (in parenthesis) analogous to the sensory sites of ArsR, and the non-sensory α3 ligands in SmtB (not in parenthesis).
B. Candidate promoter binding sites for CzrA aligned with the core consensus binding motif for SmtB-related sensors (Busenlehner et al., 2003).
C. Gene architecture around aseR.ydfA showing proposed promoter binding sites for AseR.
D. Gene architecture around cadA, czrA and czcD.trkA.

\entified two pairs of zinc binding sites in SmtB, but only the tetrahedral sites formed by antiparallel carboxyl-terminal α5-helices are required for inducer recognition (VanZile et al., 2002a). A comparison of the crystal structures of apo- and zinc-SmtB dimers uncovered a hydrogen-bond network connecting the imidazole ring of a zinc-liganded His to the carboxyl-terminal region of helix α4 of the opposing monomer's helix–turn–helix, explaining how metal binding at α5-helices can reposition the remote DNA binding regions at helices α3 and α4 to become suboptimal for DNA interaction (Eicken et al., 2003). In contrast, trigonal oxyanion binding at helix α3 of the helix–turn–helix in ArsR (Shi et al., 1994; 1996) might be expected to directly interfere with DNA binding, require no elaborate hydrogen-bonding network and be responsive to any metal that can bind. However, in this report we find that zinc can bind to α3 sensory-site residues of an arsenite-responsive regulator, but fail to alter DNA binding.

Several metal sensors and their target genes have been characterized in Bacillus subtilis (Gaballa et al., 2002; 2003; Guedon et al., 2003; Moore and Helmann, 2005), including ArsR located on the skin element and required for arsenical/antimonial-responsive expression of arsenite/arsenate-resistance genes (Sato and Kobayashi, 1998). At the outset of this research we noted two further B. subtilis ORFs encoding ArsR-SmtB-like proteins of unknown function. One of these, ydeT (aseR), has the motif Cys–Xaa–Cys–Xaa3–Asp associated with a predicted helix–turn–helix DNA binding region and with some analogy to the trigonal α3 ligands of ArsR. The other, yozA (czrA), is devoid of thiols and contains the motifs Asp–Xaa–His and His–Xaa2–His at opposing ends of a predicted carboxyl-terminal α-helix, analogous to the tetrahedral α5 ligands of SmtB but with a ligand set more closely matching that of zinc/cobalt-sensing CzrA from S. aureus (Xiong and Jayaswal, 1998; Kuroda et al., 1999; Pennella et al., 2003). Based on these observations, we predicted AseR to have α3 sites that sense arsenite and antimonite, acknowledging that an ArsR with these specificities had already been characterized from this genome, and that CzrA has α5 sites that sense zinc and possibly cobalt. The predicted inducers are confirmed here and also in studies reported during the drafting of this article (Moore et al., 2005). Assays (in vivo or in vitro) of mutant strains, or mutant recombinant proteins, missing the proposed ligands were consistent with α5 (CzrA) and α3 (AseR) sensory sites, confirming that it is possible to correctly identify sensory sites of, at least a subset of, ArsR-SmtB proteins and thereby predict the most probable effectors. However, a hypothetical mechanism by which a similar CzrA site could become responsive to cupric ions is also described. A catalogue of the 1024 SmtB-ArsR homologues currently in the databases, annotated to show the distribution of predicted metal sensory sites, has been created.

Metal(loid)s identical to the in vivo effectors, zinc not arsenite for CzrA, and arsenite not zinc for AseR, also impaired DNA–protein binding in vitro, critically establishing that for these ions and sensors, selectivity is not dictated in vivo solely by the two proteins having different metal(loid) access; for example, if only AseR, not CzrA, interacted with an arsenite metallochaperone or arsenite importer. Having established that AseR only responds to arsenite not zinc in vitro we investigated whether this was because the protein does not bind zinc, or because the ‘wrong’ metal binds but fails to impair DNA binding. The latter was shown to be true, despite arsenite-sensory α3 thiols contributing to zinc binding, implying either that the correct binding geometry is crucial to trigger allostery even at α3 sites that are proximal to DNA-binding subdomains, or that zinc dissociates from the protein when in contact with DNA. Finally, we have tested the effects of allosterically ineffective arsenite on zinc sensing by CzrA and conversely of zinc on arsenite sensing by AseR. In both cases the non-effectors do not prevent inducer recognition. However, cupric ions render CzrA inducer non-responsive but competent to bind DNA supporting the notion that CzrA must not come into contact with exchangeable intracellular copper.


Candidate SmtB-ArsR repressor sites occur in promoters that are derepressed in ΔczrA and ΔaseR

We initially sought to identify the promoter(s) likely to be regulated by CzrA. A search of the B. subtilis genome with a consensus nucleotide binding site (5′-ATANNTGAN NANNTNNTCANNTAT-3′) for SmtB-like regulators (from PnmtA, PziaA, PczrA and both PsmtA sites) exclusively identified regions within 150 nt upstream of the start codons of cadA and czcD (Fig. 1B). An iterative manual search using the dyad symmetrical element in PcadA identified only a weakly similar core sequence upstream of czrA itself. A gene expression profile of ΔczrA showed a greater than 10-fold increase in abundance, relative to wild type, in descending order, of czcD (144-fold), trkA (56-fold) and cadA (24-fold) transcripts plus enhanced expression from PczrA (12-fold) (Table S1), and this is consistent with subsequent assays of reporter gene expression from pcadA and pczcD in ΔczrA. CzcD and CadA are metal exporters (Guffanti et al., 2002; Gaballa and Helmann, 2003).

A region of dyad symmetry with similarity to ArsR-SmtB DNA binding sites is located upstream of aseR (Fig. 1C), consistent with this being the AseR target. Expression from PaseR-ydfA was greatly (305-fold) enhanced in ΔaseR (Table S2) and this is confirmed in subsequent assays of reporter gene expression. YdfA is a deduced arsenite exporter. No other transcripts showed a greater than fivefold increase in abundance in ΔaseR. A threefold increase was detected for yheI but no binding of AseR to the yheI promoter was observed in gel-retardation assays (data not shown).

AseR and CzrA bind specifically to the promoters they repress

CzrA is predicted to bind PcadA and PczcD-trkA, while AseR is predicted to bind PaseR-ydfA. To test this AseR and CzrA were expressed in Escherichia coli, purified using a combination of Heparin-affinity and size-exclusion chromatography, then analysed for DNA binding. Purified CzrA retarded DNA fragments that included sequences from either PczcD or PcadA but did not retard otherwise identical, shorter, control DNA fragments without the promoter sequences (Fig. 2B and C). More than one retarded band was detected analogous to known higher-order SmtBn–DNA complexes (VanZile et al., 2002b). Fluorescently labelled oligonucleotides corresponding to 31 bp of PcadA centred on the predicted CzrA binding site were titrated with purified CzrA and a decrease in rotation confirms formation of CzrA–DNA complexes with Kapp in the region of 5 × 10−7 M−1 under these experimental conditions, in the presence of EDTA and DTT (Fig. 2A). The DNA-binding isotherm must be a complex function of more than one protein oligomerization event with more than one DNA affinity, as partly evidenced by the multiple bands in gel-retardation assays, and therefore curve fitting has not been attempted. Purified CzrA did not retard DNA fragments that included sequences from PczrA to any greater extent than a control DNA fragment without the promoter sequences (data not shown).

Figure 2.

CzrA binding to czcD.trkA and cadA promoters.
A. An aliquot of 6-hexachlorofluorescein-PcadA 31mer DNA (10 nM) titrated with apo-CzrA and anisotropy, robs, monitored.
B. Gel retardation assay using aliquots (0.2 µM) of either control DNA (N = 136 bp) or probe (F = 170 bp) with an extra 32 bp of PczcD DNA titrated (0, 0.18, 0.35, 0.7, 0.92, 1.4, 1.75, 3.5, 7 µM) with purified CzrA generating CzrA–PczcD complexes (C).
C. Aliquots (0.2 µM) of control DNA (N = 136 bp) and probe (F = 166 bp) containing the same 136 bp plus 30 bp of PcadA DNA titrated with CzrA (0, 0.15, 0.3, 0.6, 1.5, 3, 15 µM) generating CzrA–PcadA complexes (C).

Purified AseR retarded a DNA fragment that included sequences from PaseR-ydfA but did not retard an otherwise identical, but shorter, control DNA fragment that did not contain PaseR-ydfA sequences (Fig. 3B). Fluorescently labelled oligonucleotides corresponding to 30 bp of PaseR-ydfA centred on the predicted AseR binding site were titrated with purified AseR and a corresponding decrease in rotation was detected reflecting an increase in mass due to protein–DNA complex formation, with Kapp in the region of 5 × 10−7 M−1 in the presence of EDTA and DTT (Fig. 3A). It is noted that DTT and EDTA were excluded from all subsequent fluorescence anisotropy experiments to allow analyses of metal(loid) protein complexes.

Figure 3.

AseR binding to aseR.ydfA promoter.
A. An aliquot of 6-hexachlorofluorescein-paseR 30mer DNA (10 nM) was titrated with AseR and anisotropy, robs, monitored.
B. Gel retardation assay performed using aliquots (0.2 µM) of control DNA (N = 136 bp) and probe (F = 260 bp) containing the same 136 bp plus 124 bp of PaseR DNA titrated with increasing concentrations (0, 0.1, 0.2, 0.4, 1, 2, 10 µM) of purified AseR generating AseR–PaseR complexes (C).

Expression from CzrA- and AseR-responsive promoters is consistent with predicted metal specificities

What, if any, metals are sensed by AseR and CzrA in vivo? Assays of β-galactosidase activity in cells exposed to maximum permissive concentrations of a broad range of different metals (data not shown) revealed maximally enhanced expression from PaseR in response to arsenite and antimonite (Fig. 4A) and from PcadA in response to zinc (Fig. 4B). Changes in expression from PcadA and PczcD in response to a range of concentrations of cadmium, cobalt and copper were small relative to those observed for zinc (Fig. 4C). Loss of repression of PcadA and PczcD in ΔczrA cells, and of PaseR in ΔaseR cells, was also confirmed (Fig. 4D and E). In view of the lack of in vitro interaction between CzrA and PczrA, it is suggested that the small increase in expression from PczrA in ΔczrA cells is an indirect effect. Expression from PcadA was also examined in ΔcadA mutants containing lacZ integrated within the cadA coding region (strain BFA1117). Increased β-galactosidase activity was detected in response to zinc and to a minor extent copper, but not cadmium or cobalt (data not shown). During the drafting of this article, an article was published describing extensive assays of the responses of B. subtilis genes to metals with arsenite similarly being the potent elicitor of PaseR-ydfA expression (Moore et al., 2005). In both studies CzrA-responsive genes show only small responses to copper. Furthermore, RNA dot blots also showed zinc to be the most potent (and/or rapid) inducer of czcD transcripts and zinc to be the most potent inducer of PczcD–lacZ expression, again consistent with the metal selectivity reported here for PcadA–bgaB and PczcD–bgaB. A minor difference between the two studies relates to the expression of PczcD–lacZ (Moore et al., 2005) in response to cadmium and cobalt where we see only negligible effects (Fig. 4C). However, we do find that ΔcadA is hypersensitive to cadmium but not zinc (data not shown) consistent with published data (Guffanti et al., 2002; Gaballa and Helmann, 2003). The differences may relate to the use of more prolonged metal exposures in the present work and therefore it remains possible that cadmium also mediates allostery but that this is only observed in vivo after short exposure times.

Figure 4.

Bacillus subtilis AseR and CzrA respond to opposing metal(loid)s.
A and B. β-Galactosidase activity in cells containing bgaB fused to PaseR(A) or PcadA (B) in response to maximum permissive concentrations of zinc (175 µM), arsenite (20 µM) or antimonite (20 µM).
C. Expression from PcadA and PczcD up to maximum permissive concentrations of zinc (squares; 0–175 µM), cadmium (diamonds; 0–2 µM), cobalt (triangles; 0–50 µM) and copper (circles; 0–2 mM).
D. Expression from PcadA, PczcD and PczrA in either the absence or the presence of maximum permissive concentrations of zinc in wild type (black) or ΔcadA (grey).
E. Expression from PaseR in wild type (black) or ΔaseR (grey) with no additional metal.

Arsenite and not zinc destabilizes AseR–DNA complexes

The complementary sensitivities of AseR and CzrA observed in vivo could be due to differing access to metal(loid)s in vivo perhaps via distinct interactions with metallochaperones, differing metal(loid) affinities and/or differing allosteric mechanisms that demand use of ligands and geometries peculiar to distinct inorganic elements. These options were investigated via fluorescence anisotropy. DNA (fluorescent PaseR 30mer) was titrated with AseR in the presence of arsenite (2 µM) and a decrease in rotation, increase in polarization and robs, was detected with Kapp approximately 15 × 10−7 M−1, weaker than that observed for apo-AseR (Fig. 5A). It is noted that the apo-titration in Fig. 5A implies a Kapp approximately 5 × 10−7 M−1, but nonetheless binding appears slightly weaker than in Fig. 3, and this phenomenon reproducibly correlated with the presence (Fig. 3) or absence (Fig. 5) of EDTA and DTT. In contrast, an equivalent concentration of zinc did not inhibit formation of AseR–DNA complexes with Kapp approximately 5 × 10−7 M−1, similar to that observed for apo-AseR (Fig. 5B). The effect of 2 µM arsenite on AseR DNA binding appears to be modest, but it is worth noting that at 7 × 10−7 M AseR, there is negligible formation of DNA complex in the presence of arsenite but near saturation binding for apo-protein. We do not know the amounts of the components in vivo and a relatively small change in DNA affinity of arsenite-AseR relative to apo-AseR may be sufficient to mediate regulation in this system. Using pre-formed AseR–DNA complexes, zinc did not cause a reduction in robs, but addition of 2 µM arsenite did cause robs to drop (data not shown). Thus, selectivity in AseR involves either (i) exclusive binding of arsenite but not zinc or (ii) binding of both but (a) only arsenite promoting the necessary allosteric change or (b) displacement of zinc upon protein contact with DNA.

Figure 5.

Arsenite not zinc destabilizes AseR–DNA complexes.
A. An aliquot of 6-hexachlorofluorescein-PaseR 30mer DNA (10 nM) was titrated with AseR in the absence (circles) or presence (triangles) of arsenite (2 µM) under anaerobic conditions to give an increase in anisotropy, robs.
B. Analogous pair of titrations to (A) but both in the presence of zinc (2 µM).

Zinc does not inhibit effector-mediated allostery at the α3 site of AseR

Where selectivity operates at the level of allosteric triggering rather than solely metal binding, it becomes feasible that the formation of tight complexes with allosterically ineffective metals might exclude the correct metals. Does zinc prevent arsenite detection by AseR? Titration of fluorescent PaseR 30mer DNA with AseR showed that formation of AseR–DNA complexes were inhibited by arsenite even in the presence of equimolar zinc (Fig. 5B). Furthermore, arsenite promoted a decline in robs indicative of the dissociation of pre-formed AseR–DNA complexes even in the presence of equimolar zinc (data not shown). Thus, there is no a priori reason why AseR must be prevented from gaining access to zinc.

Is the arsenite-sensing site in AseR formed by α3 ligands? The proposed α3 ligands are Cys33, Cys35 and Asp39. An AseR mutant, missing the two thiols via serine substitution, was purified and titrated against fluorescent PaseR 30mer DNA in the presence of 2 µM arsenite (Fig. 6). An increase in robs confirmed that the mutant protein remained competent to bind to DNA, despite the location of the substitutions within the predicted helix–turn–helix DNA binding region. A Kapp in the region of 5 × 10−7 M−1 implies that the formation of these complexes was unaffected by the presence of arsenite consistent with loss of arsenite detection (Fig. 6). As expected addition of arsenite to pre-formed AseRC33/35S–DNA complexes also did not promote any decline in robs (data not shown). Thus, Cys33, Cys35 thiols at α3 are obligatory for arsenite-mediated allostery in AseR. Analogous mutants of ArsR have previously been found to be DNA binding-competent and hence repression-competent but inducer non-responsive (Shi et al., 1994), although differences between AseR and ArsR include the presence of a carboxyl-terminal di-Cys motif and one further additional Cys in AseR.

Figure 6.

AseR α3 ligands are required for arsenite-mediated allostery. An aliquot of 6-hexachlorofluorescein-PaseR 30mer DNA (10 nM) was titrated with AseRC33/35S in the presence of arsenite (2 µM) under anaerobic conditions.

The non-effector zinc binds to AseR

A simplistic expectation was that selectivity in metal sensors would be driven by metal-affinities with each repressor binding tightly to sensed metals while binding to other metals weakly or not at all. However, in a previous paper we reported that the cobalt and nickel sensor, NmtR, had a higher affinity for zinc than for nickel or cobalt, but zinc is a poor mediator of allostery. NmtR spectra revealed that the sensory nickel-cobalt site was six co-ordinate octahedral (Cavet et al., 2002) while zinc-bound in a four co-ordinate tetrahedral geometry (Pennella et al., 2003). Unlike in NmtR, any metal(loid) associating with α3 sites located within the AseR DNA-binding domain might be expected to impair DNA binding. Does AseR bind zinc via α3 thiols?

The metallochromic indicator, 4-(2-pyridylazo)-resorcinol (PAR), forms a PAR2–zinc complex which absorbs at 500 nm (Δɛ = 66 000 M−1 cm−1) with an overall conditional stability constant documented to be 3.85 × 1012 M−2, and under conditions of surplus PAR, as used here, the metallochromic indicator has been used to detect nanomolar to picomolar zinc–protein dissociation equilibrium constants (VanZile et al., 2000). Zinc-dependent PAR absorbance is reduced by the addition of AseR, but this is reversed by the coincident addition of the thiol-modifying reagent p-mercuriphenylsulphonic acid (PMPS) (Fig. 7A). This places the zinc-AseR dissociation equilibrium constant somewhere within the stated range and furthermore the displacement of protein-associated zinc by PMPS confirms that the metal is thiol-co-ordinated. Separate binding isotherms (data reflect two replicate experiments) were performed in small volume multiwell microtitre plates, zeroed against apo-PAR, to monitor the association of zinc (up to 2 µM) with PAR (400 µM) in either the absence or presence of 2 µM AseR (Fig. 7B). This matches the concentrations of protein and zinc used in fluorescence anisotropy experiments. Again, competition between protein and indicator implies tighter than nM binding for Zn-AseR. Thus, zinc at a concentration of 2 µM would be predicted to bind AseR in the fluorescence anisotropy experiments and this is empirically observed in these AseR titrations.

Figure 7.

AseR binds zinc.
A. Spectra of the metallochromic indicator 4-(2-pyridylazo)-resorcinol (PAR) (200 µM) in the absence (solid line) or presence (all other spectra) of 2 µM zinc, with 2 µM AseR (dashed lines) and also p-mercuriphenylsulphonic acid (PMPS) (upper dashed line), expressed M−1 metal–PAR2 complex equivalent.
B. Microtitre plate assays (open and closed symbols represent replicate experiments) of zinc-PAR formation (400 µM PAR), measured at 492 nm zeroed against apo-PAR, as a function of [zinc] in the absence (triangles) or presence (circles) of 2 µM AseR. PMPS was added to the AseR and 2 µM zinc containing samples, and the absorbance re-read (indicated by arrows).

Addition of PMPS displaced the anticipated amount of bound zinc from AseR implying thiol binding (Fig. 7). Purified AseR was further incubated with two molar equivalents of zinc, bound and free metal resolved by size-exclusion chromatography on Sephadex G-25 (Fig. 8). Zinc and AseR co-migrated as a complex. The experiment was repeated with AseRC33/35S but zinc and protein now migrated separately (Fig. 8). Thus, α3 thiols are required for zinc ions to bind AseR sufficiently tightly to remain associated during size-exclusion chromatography implying that Cys33 and Cys35 contribute towards the aberrant, allosterically non-effective, zinc site in addition to the allosterically effective arsenite site. A crucial caviat is that conformation change in AseR upon contact with DNA could weaken the affinity for zinc but not for metalloids.

Figure 8.

AseR α3 ligands are required for zinc binding. AseR (triangles) or AseRC33/35S (squares) were incubated with zinc (2:1 zinc:protein) and fractionated on Sephadex G-25. Fractions were analysed for protein (closed symbols) or zinc (open symbols) by atomic absorption spectrophotometry.

CzrA has a pseudotetrahedral non-thiol metal site, and its α5 region is essential for Zn sensing

Is the zinc-sensing site in CzrA formed by α5 ligands? A nucleotide deletion arose in a subcloning routine creating a gene encoding a CzrA mutant with six carboxyl-terminal α5-residues changed, including the substitution of three His and one Asp that are predicted to contribute two α5 ligands, plus adding an extra six codons (Fig. 9A and B). This mutant was introduced into ΔczrA containing a PcadA–bgaB fusion, causing an approximately threefold reduction in zinc tolerance, and an approximately 1.2-fold reduction in cadmium-tolerance, relative to an otherwise identical strain without the CzrA variant (data not shown). This is presumed to arise due to repression of genes encoding proteins involved in metal efflux that is not alleviated by high zinc. In the presence of elevated zinc, β-galactosidase activity was elevated in cells containing, or devoid of, wild-type CzrA but negligible in cells containing the CzrA α5 variant, consistent with retention of DNA binding and repression, but loss of zinc perception (Fig. 9C). Identical trends were obtained using PczcD–bgaB rather than PcadA–bgaB (data not shown).

Figure 9.

CzrA α5 ligands are required for metal sensing and contribute to metal binding.
A. Sequence of the carboxyl-terminal region of CzrA including two predicted α5 ligands (underlined).
B. Sequence of a variant of the carboxyl-terminal region of CzrA due to a nucleotide deletion (boxed in A).
C. β-Galactosidase assays showing expression from PcadA in B. subtilis cells containing CzrA, devoid of CzrA or containing the frameshift mutant in (B), in the presence of [maximum permissive] zinc (200 µM).
D. Apo-subtracted spectra of either 188 µM CzrA (left) or 190 µM CzrAH92Q (right) in cobalt (0.5:1 metal:protein) expressed M−1 metal-protein.

Cobalt is often used to probe spectrally silent zinc binding sites. Titration of purified CzrA with Co2+ revealed cobalt-dependent molar absorptivity in the visible region with a maximal intensity ɛ568 = 278 M−1 cm−1 (Fig. 9D), corresponding to the d–d electronic transition envelope (VanZile et al., 2000) and consistent with cobalt bound in either a tetrahedral or distorted pseudo-tetrahedral environment. CzrA is devoid of Cys and therefore Co2+-CzrA did not show intense Co2+-dependent absorptivity in the near UV which would otherwise characterize cobalt-thiolate bonding [ɛ318 ≈ 900–1200 M−1 cm−1 per S2-Co(II) bond; VanZile et al., 2000] and the lack of thiolate ligands also decreases both the wavelength and intensity of features due to d–d transitions. His92, along with Asp90, is predicted to contribute ligands of the proposed α5 site in conjunction with two ligands supplied by the other monomer (Fig. 9A) and previously mutated (Fig. 9C). A purified CzrAH92Q variant showed greatly reduced molar absorptivity in the visible region (ɛ568 < 60 M−1 cm−1) relative to wild-type protein upon titration with equivalent amounts of cobalt and reflecting impaired metal binding (Fig. 9D). Addition of greater amounts of cobalt caused an increase in the spectral baseline due to light scatter coincident with protein aggregation, preventing titration to saturation to obtain an estimate of the weaker KCo for CzrAH92Q and presumably reflecting the propensity for cobalt to form additional intermolecular bonds when its ligand sphere is otherwise not fully satisfied. Taken together, these data support the proposed role for α5 ligands in metal binding in vitro and in metal sensing in vivo.

Zinc and not arsenite destabilizes CzrA–DNA complexes

DNA (fluorescent PcadA 31mer) was titrated with CzrA in the presence of arsenite (2 µM) and a decrease in rotation and an increase in polarization and robs were detected with Kapp approximately 5 × 10−7 M−1(Fig. 10A) similar to that observed for apo-CzrA (Fig. 2). In contrast, an equivalent concentration of zinc inhibited formation of CzrA–DNA complexes (Fig. 10B) and using pre-formed CzrA–DNA complexes, arsenite did not reduce robs, while 2 µM zinc did (data not shown). Thus, selectivity in CzrA involves either exclusive binding of zinc but not arsenite, or binding of both but only zinc promoting the necessary allosteric change. Can arsenite prevent zinc detection by CzrA? Titration of fluorescent PcadA 31mer DNA with CzrA showed that formation of CzrA–DNA complexes were inhibited by zinc even in the presence of equimolar arsenite (Fig. 10C). Furthermore, zinc promoted a decline in robs indicative of the dissociation of pre-formed CzrA–DNA complexes even in the presence of equimolar arsenite (data not shown). Thus, there is no a priori reason why CzrA must be prevented from gaining access to arsenite.

Figure 10.

Zinc not arsenite destabilizes CzrA–DNA complexes.
A. An aliquot of 6-hexachlorofluorescein-PcadA 31mer DNA (10 nM) was titrated with CzrA in the presence of arsenite (2 µM) under anaerobic conditions to give an increase in anisotropy, robs.
B. Analogous titration to (A) but in the presence of zinc (2 µM) rather than arsenite.
C. Analogous titration to (A) and (B) but in the presence of arsenite (2 µM) and zinc (2 µM).

Cupric ions do not inhibit CzrA–DNA binding but do inhibit zinc-mediated allostery

The E. coli copper sensor CueR has zeptomolar (10−21 M) affinity for copper, suggesting that there is no freely available copper in the cytosol; one atom per cell volume formally equates to nM (Changela et al., 2003). Why might it be necessary to maintain a bacterial cytosol that is devoid of freely available copper? We recently found that a critical cytosolic zinc site in a zinc-exporting P1-type ATPase of Synechococystis PCC 6803 bound cuprous ions more tightly than zinc, raising the possibility that free copper might mispopulate cytosolic zinc sites and requiring that any cytosolic copper is trafficked by ligand exchange (Borrelly et al., 2004a). Does the zinc site in CzrA bind copper more tightly than zinc and, if so, what impact does this have on activity?

The formation of CzrA–DNA (PcadA fluorescent 31mer) complexes was not affected by the presence of copper with Kapp in the region of 5 × 10−7 M (Fig. 11A), analogous to apo-CzrA (Fig. 2). As before (Fig. 10) zinc inhibited formation of these complexes (Fig. 11B) but in the presence of both copper and zinc the protein behaved essentially like apo-CzrA with Kapp again in the region of 5 × 10−7 M (Fig. 11C). Analogous effects of these metals were detected on complex dissociation using pre-formed apo-CzrA–DNA complexes, observing that zinc and not copper promoted a decline in robs due to the liberation of free DNA but robs was unaltered by zinc in the presence of copper (data not shown). Thus, CzrA does not ‘sense’ cupric ions in vitro but, importantly, zinc becomes allosterically uneffective if added in the presence of cupric ions. Equivalent experiments were also performed under a N2 atmosphere in an anaerobic chamber, sealed into a gas-tight cuvette, then titrated with cuprous ions using a gas-tight syringe, but CzrA precipitation was repeatedly observed.

Figure 11.

Cupric ions inhibit zinc-mediated allostery in CzrA.
A. An aliquot of 6-hexachlorofluorescein-PcadA 31mer DNA (10 nM) was titrated with CzrA in the presence of cupric ions (2 µM) under anaerobic conditions to give an increase in anisotropy, robs.
B. Analogous titration to (A) but in the presence of zinc (2 µM) rather than cupric ions.
C. Analogous titration to (A) and (B) but in the presence of cupric ions (2 µM) and zinc (2 µM).
D. Predicted α5 ligands of B. subtilis CzrA and hydrogen bonds connecting zinc-His102 to helix α4 modelled on zinc-CzrA from S. aureus (Eicken et al., 2003).
E. Tetrahedral co-ordination geometry preferred by zinc.
F. Tetragonal co-ordination geometry preferred by cupric ions (Gray et al., 2000).

After incubation with either zinc, or cupric, ions the metals co-migrate with CzrA via gel filtration on Sephadex G-25 (Fig. 12). Most importantly, in the presence of equimolar cupric and zinc ions, cupric ions co-migrate with CzrA but zinc ions migrate separately confirming that CzrA preferentially binds cupric ions rather than zinc.

Figure 12.

CzrA binds cupric ions in preference to zinc.
A. CzrA was incubated with copper (1:1 copper:protein), fractionated on Sephadex G-25 and fractions were analysed for protein (closed circles) and copper (closed triangles). The column was calibrated with free copper (closed squares).
B. CzrA was incubated with zinc (1:1 zinc:protein), fractionated on Sephadex G-25 and fractions were analysed for zinc (open triangles) and protein (closed circles). The column was calibrated with free zinc (open squares).
C. CzrA was incubated with copper and zinc (1:1:1 copper:zinc:protein), fractionated on Sephadex G-25 and fractions were analysed for zinc (open triangles), protein (closed circles) and copper (closed triangles).

Sensory motifs in SmtB-ArsR homologues

Candidate sensory motifs associated with predicted elements of secondary structure have been identified in 1024 ArsR-SmtB-related protein sequences (Table S3). There are 179 representatives that contain the motifs found in AseR or CzrA and a further 143 containing the other types of known sensory motifs previously described in CadC, NmtR and CmtR (Table 1). Its unclear what proportion of the remaining 702 representatives that have no currently identifiable sensory motif are nonetheless metal sensors but with novel metal binding sites. There are 41 sequences that have both the α3N and α5 motif, some of which may exploit both sites in sensing, as in ZiaR, while one site may be vestigial in other representatives, analogous to α3N of SmtB and α5 of some CadC’s. Within the current 1024 representatives, the CXCXXC ArsR-like α3 motif is never associated with amino-terminal ligands and is most likely to correspond with oxyanion sensing rather than α3N CadC-like cadmium sensing. The α4 motif found in cadmium sensing CmtR is never found with another motif and is restricted to just six species with eight representatives in Corynebacterium glutamicum. There are only nine sequences with the NmtR-like α5C motif, associated with nickel and cobalt sensing, and no precedent for co-occurrence of α5C with another motif such as α3N.

Table 1.  Sensory sites in ArsR-SmtB homologues.
Sensory motifCharacterized precedentsNumber of representatives
  • a.

    In ZiaR both sites are obligatory for inducer recognition while SmtB and some CadC proteins that have both sites only require one for metal-mediated allostery.

  • Seven characterized representative ArsR-SmtB metal sensors were used to produce an alignment of reference upon which secondary structure elements were mapped based on the known structure of SmtB. Each of the 1024 related sequences in the Pfam database was aligned to the reference and parsed for the occurrence of the metal-binding patterns defined in the methods. The full list of proteins and their predicted metal motifs is given in Table S3.

α3ArsR/AseR 106
α5SmtB/CzrA  73
α3NCadC 110
α5CNmtR   9
α4CmtR  24
Unknown  702
Total 1024
α3N+α5ZiaRa  41


AseR and CzrA are metal(loid)-responsive DNA-binding (Figs 2–4) repressors of transcription from one and at least two, respectively, identified B. subtilis promoters (Fig. 4). They have complementary specificities of metal(loid) sensing in vivo such that the former responds to arsenite not zinc, the latter to zinc not arsenite (Fig. 4) and this perfectly matches the ions that impair formation of the respective repressor–promoter complexes in vitro (Figs 5 and 10). Selectivity operates at the level of metal(loid)-specific allostery, or metal(loid) affinity, rather than solely different access to these effectors in vivo[due to kinetic factors such as sensor-specific interaction with metal(loid)-donating metallochaperones or metal transporters]. The metals sensed validate predictions from protein sequence analyses which proposed a trigonal, thiol-containing, α3 metalloid site in AseR and a thiol-free tetrahedral α5 zinc site in CzrA. The properties of mutants confirm respective α3 and α5 sites (Figs 6 and 9). AseR can bind zinc (Fig. 7) via thiols of the metalloid-responsive α3 site (Fig. 8) but without triggering allostery (Fig. 5) implying that selectivity in AseR either operates at the level of allosteric triggering, not merely binding, despite sensing metal(loid)s at the helix–turn–helix DNA-binding subdomain, or that zinc can not associate with the protein when in complex with DNA. Zinc does not impair arsenite-mediated conformational change in AseR (Fig. 5), and arsenite does not impair zinc-mediated conformational change in CzrA (Fig. 10); however, cupric ions do block zinc-mediated allostery (Fig. 11) and block zinc binding (Fig. 12). Implications of copper preventing zinc sensing by CzrA (Fig. 11) are discussed in relation to (i) a theoretical CzrA-like copper sensor and (ii) the necessity for copper exclusion and/or intracellular trafficking of copper via ligand exchange. Candidate sensory sites have been identified in all 1024 ArsR-SmtB-like sequences in databases (Table S3).

We, and others, initially anticipated that in vivo metal selectivities of metal-responsive transcriptional regulators would always be intrinsic properties of the metal binding sites of the sensor proteins. However, NmtR-mediated repression of PnmtA–lacZ is most effectively alleviated by nickel and then cobalt in M. tuberculosis but only cobalt, not nickel, when expressed in a cyanobacterium (Cavet et al., 2002). Similarly DtxR from Corynebacterium diptheriae solely responds to iron in the homologous host, but iron and manganese when transferred to B. subtilis (Guedon and Helmann, 2003). Access to metals can thus be influenced by the cytosol potentially due to specific interactions between sensors and metallochaperones or importers, cytosolic exclusion of some metals and/or displacement of a relatively weaker bound metal from cell metal sites. Importantly, the in vivo metal selectivities of AseR and CzrA from B. subtilis in favour of or against arsenite and zinc (Fig. 4) correspond exactly with the capacities of these ions to destabilize repressor–DNA complexes in vitro monitored by fluorescence anisotropy (Figs 5 and 10). Thus, for these two ions and these two regulators selectivity is indeed inherent to the sensory sites of the repressors.

Evidence that zinc binds to α3 thiols of AseR but does not trigger allostery implies that distortion of the sensory site into the correct geometry and/or the selection of the full ligand set may be crucial for metal sensing at α3 sites even though the α3-helix is thought to be part of the helix–turn–helix DNA binding region, making it plausible that any metal association in this region could have caused enough conformational change to impair DNA binding. Thus, for α3 sensory sites, in common with α5 sites (Cavet et al., 2002), it is possible that metal specificity can reside at the level of selective triggering of allostery not solely selective metal binding and partitioning. However, it is also formally possible that AseR undergoes a conformation change when in contact with DNA such that only arsenite, and not zinc, is able to associate with α3 thiols under this condition. Where discrimination does occur at the level of allostery rather than simply metal binding, it provides an additional checkpoint for selectivity but also creates the potential for non-effectors to occlude binding of inducers. However, for AseR and CzrA we have established that the opposing non-effectors, zinc or arsenite, do not impair effector-mediated allostery, indicating that there is no requirement for these sensors to avoid access to the wrong metal(loid)s in vivo (Figs 6 and 7).

While cupric ions do not destabilize CzrA–PcadA complexes they do ‘block’ the detection of zinc, implying that these ions associate with at least a subset of the sensory-site ligands more tightly than zinc (Fig. 11). This was confirmed when CzrA was exposed simultaneously to both ions, cupric ions bound and zinc remained unbound (Fig. 12). Cuprous ions caused CzrA precipitation in vitro but potentially they may mediate (relatively small) in vivo responses to copper. In aerobic non-reducing media, copper will disproportionate to predominantly cupric ions while in the reducing cytosol it is generally expected that cuprous ions predominate. However, the chemical form of any intracellular copper has never been directly measured in B. subtilis and notably this organism lacks glutathione. In common with all bacteria other than some cyanobacteria, where a copper metallochaperone Atx1 supplies copper to plastocyanin and thylakoidal cytochrome c oxidase (Tottey et al., 2002), cytosolic copper-requiring enzymes remain to be discovered in B. subtilis. Nonetheless, B. subtilis does contain a copper metallochaperone similar to Enterococcus hirae CopZ and cyanobacterial Atx1 (Odermatt and Solioz, 1995; Banci et al., 2003; 2004; Radford et al., 2003). Metallochaperones allow copper to cross the cytosol via sequential ligand exchange, without release, and in Synechocystis PCC 6803 this may restrain copper from zinc sites, such as the amino-terminal domain of the zinc exporter ZiaA, which have failed to evolve a higher affinity for zinc than for copper (Borrelly et al., 2004a,b). The notion that freely available cytosolic copper ions must be avoided if some intracellular metalloproteins are to be occupied by less competitive metals is supported by the observation that copper ions form a non-productive but stable complex with CzrA.

The accuracy of the predicted locations and properties of the metal(loid) binding sites of B. subtilis CzrA and AseR, coupled with the accuracy of their predicted effector specificities, suggests that identifying these sites in related protein sequences may be useful for predicting metal specificities in at least a subset of ArsR-SmtB repressors (Table S3). The lack of Cys in the sensory site of CzrA will be unfavourable to thiophilic arsenite and antimonite and furthermore these oxyanions are unlikely to re-position ligands into the tetrahedral geometry preferred by zinc and probably required for allostery in this protein. In contrast, EXAFS revealed trigonal arsenite binding in ArsR (Shi et al., 1996) and the analogous sensory site of AseR is expected to exploit trigonal geometry.

Many metalloproteins contain pre-formed metal binding sites but association of metals with ArsR-SmtB sensors must, at least partly, reorganize the site to trigger conformation change. Metal association at α5 in SmtB and CzrA re-positions the imidazole group of a critical histidine to trigger formation of a hydrogen-bond network (Eicken et al., 2003). In B. subtilis CzrA the critical imidazole is predicted to be derived from zinc-His102 which forms a bond to Arg73 which in turn binds to Leu69 of helix α4 of the DNA binding region (Fig. 11D). In contrast to the preferred tetrahedral geometry of zinc (Fig. 11E), the preferred geometry for cupric ions is tetragonal, square planer in the absence of axial ligands (Fig. 11F) (Fraústo da Silva and Williams, 2001). It is therefore plausible that cupric CzrA-His102 is re-positioned according to tetragonal, rather than tetrahedral, geometry which must be unsuitable for initiating the hydrogen-bond network and so cupric CzrA fails to adjust the location of helix α4 and fails to impair DNA binding. Theoretically, in some currently uncharacterized ArsR-SmtB sensor, residues forming the hydrogen-bond connections could be re-positioned such that metal-dependent relocation of helix α4 demanded relocation of His102, or some other residue of the sensory site, into a tetragonal geometry, or some other geometry preferred by some other metal. This raises the intriguing possibility that an analogous sensor with an analogous α5 ligand set may not respond to zinc and might sense some other metal(loid), but this also sounds a cautionary note for predictions of metal specificities based on the current catalogue of sensory themes and variations.

Experimental procedures

Bacterial strains, plasmids and growth conditions

Escherichia coli and B. subtilis strains were routinely cultured in LB media at 37°C. Antibiotics were utilized at the following concentrations: 250 µg ml−1 (carbenicillin), 50 µg ml−1 (kanamycin), 7 µg ml−1 (neomycin), 35 µg ml−1 in E. coli, 3 µg ml−1 in B. subtilis (chloramphenicol).ΔczrA and ΔaseR strains of B. subtilis used in gene expression profile experiments were constructed by integration of pMUTIN4 (Vagner et al., 1998) into the ORFs (Kobayashi et al., 2003).

β-Galactosidase measurements

Gene expression levels were monitored by creating a transcriptional fusion to bgaB (encoding a heat-stable β-galactosidase) in plasmid pBgaB (Pragai and Harwood, 2002). The operator-promoter region of cadA was amplified by PCR from B. subtilis genomic DNA using primers 5′-CCAGGATCCCTT CGTCGTCAAAAAGTA-3′ and 5′-CGCGAATTCTTTAGGGTT CCCTCTTCA-3′, ligated into pGEM-T, before subcloning into the BamHI/EcoRI site of pBgaB to generate a transcriptional fusion to bgaB. The resulting construct was linearized with ScaI and used to transform B. subtilis 168 to neomycin resistance. Integration at the amyE locus via a double-cross-over event was confirmed by PCR and loss of amylase activity (strain 168 amyE::PcadA–bgaB). The aseR.ydfA promoter was amplified with primer sequences 5′-TTTTGGATCCGG CTCTATGACCAC-3′ and 5′-TTTTGAATTCGTCAATCTAAT GC-3′, and the bgaB fusion created as described above. For gene expression experiments, cells were grown to mid-exponential phase in the presence of metal (6 h), cells were lysed as previously described (Pragai and Harwood, 2002) and β-galactosidase activity was determined (Morby et al., 1993).

Bacillus subtilis RNA was isolated and probes for microarray analyses were prepared using a two-step procedure as previously described (Kobayashi et al., 2001). B. subtilis DNA microarrays contained 4055 predicted ORFs with 45 not spotted due to problems with DNA amplification. The array also contained 39 calf thymus DNA spots which were used as negative controls. Hybridization and microarray analyses were performed as previously described with expression ratios calculated after normalization of Cy3/Cy5 intensities and subtraction of the background intensity. Expression of autoregulated genes in the knockout strains was observed by measuring the expression of the lacZ gene carried on the pMUTIN4 knockout vector.

Disruption of czrA in B. subtilis containing PcadAbgaB

Bacillus subtilis 168 genomic DNA was used as template for PCR with primers 5′-GTGTTGCAACCTGCAAAG-3′ and 5′-GCAAACAGCTCAGCAAGCTCTGC-3′ and the amplification product, 2025 bp, containing czrA with upstream and downstream flanking regions was ligated to pGEM-T (Promega, Madison, USA) to generate pGEM-T-czrA. The chloramphenicol acetyltransferase gene was then released from pCAT1 using EcoRV and ligated into the Eco47III site of pGEM-T-czrA. The resulting construct, pΔczrA, was linearized using SacI and used to transform B. subtilis IHA01 (Härtl et al., 2001) to chloramphenicol resistance (strain IHA01ΔczrA), with disruption of czrA confirmed by PCR. Chromosomal DNA from IHA01ΔczrA was subsequently used to transform strain 168 amyE::PcadA–bgaB to chloramphenicol resistance and czrA disruption again confirmed by PCR (strain 168ΔczrA amyE::PcadA–bgaB).

Examination of expression from PcadA in cells containing CzrA mutated at α5

Bacillus subtilis 168 genomic DNA was used as template for PCR with primers 5′-GTGTTTCACCCTCAATGTTTGG-3′ and 5′-GAACGGATCCTCTGACAATTGG-3′ and the amplification product, 660 bp, containing czrA and upstream operator-promoter sequences ligated to pGEM-T before subcloning into the BamHI/SacII site of pAXO1 (Härtl et al., 2001), generating pAXO1-czrA. The latter was used to transform B. subtilis IHA01 to erythromycin resistance and spectinomycin sensitivity and integration of czrA into lacA by a double-cross-over event confirmed by PCR (strain IHA01 lacA::czrA). Subsequent sequencing of the reintroduced czrA revealed a point mutation causing disruption of the deduced α5 metal binding site. Chromosomal DNA from IHA01–lacA::czrA was subsequently used to transform 168ΔczrA amyE::PcadA–bgaB to erythromycin resistance.

Cloning, expression and purification of CzrA

The CzrA ORF was amplified from B. subtilis genomic DNA using primers 5′-TGTTTCACCCTCAATGTTTGG-3′ and 5′-GAACGGATCCTCTGACAATTGG-3′, ligated to pGEM-T before subcloning into the NdeI/BamHI sites of pET29b (Novagen, Darmstadt, Germany) generating pET-CzrA. Constructs were confirmed to be free of nucleotide changes. E. coli BL21(DE3) cells harbouring this construct were cultured to OD595 = 0.5, expression induced with 1 mM IPTG for 5 h at 37°C, cells harvested and suspended in buffer A (10 mM Hepes, pH 7.4, 1 mM EDTA, 50 mM NaCl, 10 mM DTT) with 0.1 mM PMSF and lysed by sonication. The soluble fraction was loaded onto a Heparin-affinity column equilibrated in buffer A. Bound proteins were eluted using a linear gradient of 50 mM to 1 M NaCl in buffer A. Fractions containing CzrA were further purified using a Superdex 75 size-exclusion column equilibrated in buffer A plus 200 mM NaCl. Eluted CzrA was diluted and concentrated on a Heparin column as described above. Protein purity was judged to be > 90% by SDS-PAGE.

In order to obtain sufficient concentrations of CzrA for UV-visible spectroscopy, protein was solubilized from inclusion bodies containing the vast majority of overexpressed protein. Insoluble protein was isolated by centrifugation and washed three times in buffer A. Inclusion bodies were solubilized in 6 M guanidine hydrochloride and dialysed against buffer B (10 mM Hepes, pH 7.4, 1 mM EDTA, 1 mM DTT) overnight. Soluble CzrA was then purified as described above. The CzrA H92Q mutant was produced by Quikchange (Stratagene, La Jolla, USA) mutagenesis using oligonucleotide sequences 5′-ACTCTCCGGAGGATGAACAAGTTTTAGAT GTTCTTCAG-3′ and 5′-CTGAAGAACATCTAAAACTTGT TCATCCGCCGGACAGT-3′ with pET-CzrA plasmid. Protein was expressed and purified as described for the wild type.

Cloning, expression and purification of AseR and AseRC33/35S

The AseR ORF was amplified from B. subtilis genomic DNA using oligonucleotides designed to incorporate NdeI and BamHI restriction endonuclease sites. The purified PCR product was digested with NdeI and BamHI and ligated into pET-29a digested with the same enzymes generating pET-AseR. The aseRC33/35S ORF (AseR with C33S and C35S mutations) was produced by two rounds of Quikchange mutagenesis using pET-AseR as template. The C35S mutation was constructed using primers 5′-GAATATTGTGTCTCC CAATTGGTTGAT-3′ and 5′-ATCAACCAATTGGGAGACAC AATATTC-3′ and the C33S mutation using primer sequences 5′-GAACAAGAATATTCCGTCTCCCAATTG-3′ and 5′-CAAT TGGGAGACGGAATATTCTTGTTC-3′. Protein expression and purification for AseR and AseRC33/35S was carried out as described for CzrA.

Electrophoretic mobility shift assays

The operator-promoter region of aseR.ydfA was amplified from B. subtilis genomic DNA using primers 5′-GATATACG GAGCGC-3′ and 5′-ACAACGAGTCATCGC-3′. Putative DNA binding sites for CzrA from the cadA and czcD promoter regions were created by annealing the following pairs of oligonucleotides: cadA 5′-CTTATATATGAGTATATGCTCATA TATATAA-3′ and 5′-TATATATATGAGCATATACTCATATATAA GA-3′; czcD 5′-TTTATATATGAACACATGCTCATATATAAAA-3′ and 5′-TTTATATATGAGCATGTGTTCATATATAAAA-3′. Purified PCR products were ligated to pGEM-T and re-amplified using primers designed to anneal to the plasmid backbone on either side of the cloning site. Competitor DNA was amplified by using the same primers with re-circularized plasmid as template. Equal quantities of target and competitor DNA were diluted in gel shift buffer, 20 mM Tris-HCl, pH 7.8, 1 mM DTT, 1 mM EDTA, 3% v/v glycerol, 0.05 mM spermidine. Increasing concentrations of protein were added and reactions were incubated at 20°C for 10 min. Products were separated on native 6% w/v polyacrylamide gels run in TAE buffer at 4°C and visualized with UV light after staining with ethidium bromide.

Fluorescence anisotropy analysis of DNA–protein interaction

Complimentary oligonucleotides were produced corresponding to the predicted DNA binding sequences for each protein: aseR.ydfA 5′-CGTGTATATAACGATTTGCTTATATATTGA-3′; cadA 5′-CTTATATATGAGTATATGCACATATATATAA-3′. One oligonucleotide of each pair was 5′-labelled using 6-hexachlorofluorescein. Oligonucleotides were annealed by drying down, resuspending in deionized water, heating to 95°C and cooling slowly to 20°C. DNA was subjected to ultracentrifugation using an airfuge (92 000 r.p.m. for 10 min) to remove any particulate matter before use. Double-stranded DNA was added at a final concentration of 10 nM to a 1 ml anaerobic quartz cuvette (10 mm path length) containing buffer 10 mM Hepes, pH 7.4, 200 mM KCl, 50 mM EDTA, 1 mM DTT. Protein was desalted into the same buffer and titrated into the cuvette using a Hamilton gas-tight syringe. The anisotropy of the solution was measured using a 8100 fluorometer (SLM-Aminco, Urbana, USA). The excitation wavelength was set to 530 nm and emission was detected through a 3 mm thick 570 nm cut-off filter. A vertical excitation polarizer was used with the emission polarizer alternated between the vertical and horizontal positions to measure IVH and IVV respectively. Anisotropy is defined by:

Anisotropy (robs)  =  (IVV  −  IVH)/(IVV  +  2IVH)

Differences in the detector response to vertical and horizontal polarized light (G-factor) were corrected by the fluorometer. Each measurement was averaged over a 99 s reading time. Associations in the presence of metal were carried out in an identical fashion except EDTA and DTT were omitted from all buffers and the required metal was added to a final concentration of 2 µM. Dissociation experiments were carried out by pre-forming the DNA–protein complex by adding 2 µM protein to 10 nM DNA and measuring the anisotropy. Samples were then titrated with anaerobic metal solutions using a Hamilton gas-tight syringe, except where non-reducing conditions were required.

In vitro analyses of zinc and copper binding

Purified protein (10 µM) was desalted into buffer (10 mM Hepes, pH 7.4, 200 mM NaCl, 1 mM DTT) and mixed with 2 Molar equivalents of Zn2+. Protein/metal mixtures were equilibrated for 10 min and then separated by gel filtration through a PD-10 column filled with Sephadex G-25 resin (10 ml bed volume). Fractions (1 ml) were collected and analysed for protein content by measuring the absorbance at 280 nm. Zinc content was analysed using a flame atomic absorption spectrophotometer AA-240 (Varian, Oxford, UK).

Spectral analysis was performed in buffer (10 mM Hepes, pH 7.4, 400 mM KCl and 1 mM DTT) using 200 µM PAR, with or without zinc (2 µM) or AseR (2 µM) pre-incubated with zinc (2 µM). To release thiol bound metal, an excess of the thiol-modifying agent, PMPS, was added and spectra measured.

Microtitre plate competition analysis was performed using purified AseR (2 µM) incubated with zinc (up to 2 µM) in buffer (10 mM Hepes, pH 7.4, 400 mM KCl, 1 mM DTT), followed by addition of PAR (400 µM). A control experiment was performed in the same way, except AseR was not added to the incubation mixture. Zn-PAR absorbance change was monitored at 492 nm. Saturating amounts of PMPS were added to the 2 µM zinc samples and Zn-PAR absorbance changes measured.

Purified CzrA was diluted to a concentration of 5 µM in 10 mM Hepes, pH 7.4, 200 mM NaCl. Protein (1 ml) was incubated with either 5 µM copper, 5 µM zinc or 5 µM of both metals, at room temperature for 10 min. Bound and free metals were resolved by size-exclusion chromatography on a Sephadex G-25 column equilibrated in the same buffer, with 1 ml fractions collected. Each fraction was analysed for protein content by measuring the A280 and using the theoretical extinction coefficient of ɛ = 2560 M−1 cm−1. Copper concentration was determined using a graphite furnace and zinc concentration using an air-acetylene flame on an atomic absorption spectrophotometer.

UV-visible spectroscopy

All experiments used a 1 ml quartz cuvette at 25°C and a Cary 4E spectrophotometer (Varian, Oxford, UK). Purified protein was desalted into buffer (10 mM Hepes, pH 7.4, 200 mM NaCl) and spectra collected from 200 to 800 nm. Metals were added as 0.5 Molar equivalents, allowed to equilibrate and spectra measured.

Computer-based identification of sensory motifs in ArsR-SmtB homologues

Representative sequences of different classes of known ArsR-SmtB sensors were selected to produce a clustalw (Thompson et al., 1994) alignment of reference using: (i) ArsR (Wu and Rosen, 1991) from E. coli (Swiss-Prot code P15905), (ii) CadC (Yoon et al., 1991) from S. aureus (Swiss-Prot code P20047), (iii) CmtR (Cavet et al., 2003) from M. tuberculosis (Swiss-Prot code P67731), (iv) NmtR (Cavet et al., 2002) from M. tuberculosis (Swiss-Prot code O69711), (v) SmtB (Huckle et al., 1993) from Synechococcus (Swiss-Prot code P30340), (vi) ZiaR (Thelwell et al., 1998) from Synechocystis (Swiss-Prot code Q55940) and (vii) CzrA (Xiong and Jayaswal, 1998) from S. aureus (Swiss-Prot code O85142). Secondary structure elements were then mapped on the sequences included in the alignment, based on the available crystal structure of SmtB (Cook et al., 1998) (PDB code 1SMT). Subsequently, each of the 1024 protein sequences included in the HTH_5 family at the Pfam database (Bateman et al., 2004) was aligned with clustalw to the alignment of reference, keeping the latter fixed. This allowed identification of putative secondary structure elements, which were then extended by two residues at both the ends and parsed for the occurrence of potential metal-binding patterns. Metal-binding patterns which were searched for are: (i) the CXCX(2)C pattern in the α3-helix, (ii) the CXC pattern in the α3-helix, (iii) the CX(2)D pattern in the α3-helix, (iv) the CX(3)C pattern in the α4-helix and (v) the DXHX(10)HX(2)[E,H] pattern in the α5-helix. A more loosely defined pattern (vi), expressed as [C,D,E,H]X[C,D,E,H]X(10)[C,D,E,H]X(2)[C,D,E,H], was also searched for in the α5-helix to allow some flexibility in the nature of the metal ligands. Furthermore, potential metal ligands (His, Cys, Asp or Glu) were sought in the amino-terminal region (if present) of all the sequences matching at least one of patterns (i), (ii) and (iii), as well as in the carboxyl-terminal region (if present) of all the sequences matching pattern (v) (either strict or loose), to account for the occurrence of potential α3N and α5C sites respectively.


This work was supported by a research grant from the BBSRC plant and microbial sciences committee. We thank Chris Dennison for insights into co-ordination geometries and Margaret Knight for assistance with metal analyses.