A Cu^I-sensing ArsR family Metal Sensor Protein with a Relaxed Metal Selectivity Profile

Abstract

ArsR (or ArsR/SmtB) family metalloregulatory homodimeric repressors collectively respond to a wide range of metal ion inducers in regulating homeostasis and resistance of essential and nonessential metal ions in bacteria. BxmR from the cyanobacterium Osciliatoria brevis is the first characterized ArsR protein that senses both Cu^I/Ag^I and divalent metals Zn^II/Cd^II in cells by regulating the expression of a P-type ATPase efflux pump (Bxa1) and an intracellular metallothionein (BmtA). We show here that both pairs of predicted α3N and α5 sites bind metal ions, but with distinct physicochemical and functional metal specificities. Inactivation of the thiophilic α3N site via mutation (C77S) abolishes regulation by both Cd^II and Cu^I, while Zn^II remains a potent allosteric negative effector of operator/promoter binding (ΔG_c≥+3.2 kcal mol⁻¹). In contrast, α5 site mutant retains regulation by all four metal ions, albeit with a smaller coupling free energy (ΔG_c≈+1.7 (±0.1) kcal mol⁻¹). Unlike the other metals ions, the BxmR dimer binds four mol•equiv Cu^I to form an α3N binuclear Cu^I₂S₄ cluster by x-ray absorption spectroscopy. BxmR is thus distinguishable from other closely-related ArsR family sensors, in having evolved a metalloregulatory α3N site that can adopt an expanded range of coordination chemistries, while maintaining redundancy in the response to Zn^II. The evolutionary implications of these findings for the ArsR metal sensor family are discussed.

All cells must obtain their quota of biologically required transition metal ions for use as cofactors in metalloenzymes or for structural or regulatory roles through specific membrane-associated transporters or uptake systems (9). However, all metal ions are toxic in excess, and the intracellular availability of each is tightly controlled in such a way that metal homeostasis allows organisms to rapidly respond to changes in their microenvironments (10, 11). Likewise, metal ions that play no biological role, e.g., heavy metal pollutants and other environmental contaminants, must either be detoxified via biotransformation or efflux from the cytosol (12, 13).

This adaptive response is mediated by gene regulatory proteins, coined metalloregulatory proteins (10), or metal sensor (14) proteins. These specialized proteins have evolved metal coordination sites that "sense" specific metals ion(s) in the cytosol by forming specific coordination complexes. This, in turn, functions to activate or inhibit DNA binding to operator sites(s) in the promoter thereby regulating the expression of genes. These genes include transporters, intracellular chelators and detoxification enzymes, that mediate metal homeostasis or resistance in what is thought to be a selective adaptive response.

There are currently seven major families of metal-sensing transcriptional regulators that have thus far been structurally characterized in prokaryotes (see (14) for a review). These seven families span the detection of the six primary biologically essential first row transition elements Mn^II, Fe^II, Co^II, Ni^II, Cu^I and Zn^II, as well as nonbiological heavy metals Ag^I/Au^I and Cd^II/Hg^II, respectively, found in the environment. In addition, ArsR¹ or ArsR/SmtB (6) and MerR (15) family sensors have been identified that detect Pb^II, as well as As^III/Sb^III and Bi^III. Sensor families are typically named for the founding member(s) that gave rise to the family (14). Of interest here is the ArsR/SmtB family, named for E. coli R774 ArsR, which regulates expression of the ars operon in response to trivalent As^III and Sb^III and their oxyanions and organoarsenicals (16) and the Zn^II sensor Synechococcus SmtB (17), which regulates expression of the smt (Synechococcus metallothionein) operon largely as a result of Zn^II toxicity.

ArsR family sensors (COG0640) constitute a very large class of bacterial transcriptional regulators. A recent report identifies over 500 unique sequences that could be divided into eight major groups on the basis of global sequence similarity (7). A least four structurally distinct, subunit-bridging metal sensing sites have now been identified and characterized in ArsR family sensors, a remarkable collective feature that appears to distinguish the ArsR family from all other metalloregulatory protein classes (4, 14). In other metal sensor families, a single regulatory metal site often acquires distinct metal specificities by more subtle structural alterations in the first coordination shell (14, 18, 19), a feature that also characterizes ArsR sensors that harbor a single subtype of metal site, e.g., α5/α5C (see below) (20).

In ArsR family proteins, metal sites are named for the secondary structural elements that are known or thought to provide ligand donor atoms to the individual metal ions (20, 21). These sites have been designated α3N/α3 (21) (or metal site 1 in S. aureus pI258 CadC (1)), α4C (22), α5/α5C (3, 21, 23) (site 2 in S. aureus pI258 CadC), and most recently α5-3 (7).¹ Further evidence for an even greater structural diversity of sensing sites in ArsR family proteins comes from the recent characterization of two As^III-sensing ArsRs shown to harbor tris-thiolato As^III binding sites distinct from the S₃ α3 site originally characterized in E. coli R773 ArsR (4, 5). Similarly, our physical characterization of a predicted Cd^II/Pb^II-sensing CmtR from Streptomyces coelicoler (SCO0875) reveals a second S₃ Cd^II/Pb^II binding site involving a pair of adjacent Cys residues at the extreme C-terminus (Y. Wang and D. Giedroc, unpublished observations) in addition to the canonical α4C Cd^II/Pb^II site structurally characterized (24, 25).

While the metal selectivity of individual metal site subtypes is rather well understood and sometimes predictable on the basis of the amino acid sequence (20, 26), approximately half of the projected ArsR family members do not conform to any of these four metal-binding motifs. For example, one poorly characterized subgroup contains a conserved Met and two conserved Cys, the latter derived from the projected α2 and α5 helices that are predicted to be quite close in the structure but may not bind metal ions directly (7, 27). These include Paracoccus pantotrophus SoxR, the thiosulfate-induced sulfur-oxidizing (sox) operon regulator (28), V. cholerae HyU (7) and Xylella fastidiosa BigR, a regulator implicated biofilm formation (29). Thus, it remains a considerable challenge to unambiguously identify the mode of induction of ArsR sensors that do not precisely conform to one of the four known structural classes on the basis of amino acid sequence alone. In addition, since some ArsR family sensors contain two metal sites, e.g., α3N and α5 (6, 30), a further complication is that one site might can be metalloregulatory while the other site may bind metal ions but in a way that does not mediate allosteric regulation of DNA binding, and is thus nonfunctional (30, 31).

Initial biological studies established that an ArsR family repressor, termed BxmR¹, regulated the expression of genes encoding both a heavy metal transporting ATPase, Bxa1¹, as well as an intracellular metallothionein, BmtA¹, in the freshwater cyanobacterium Oscillatoria brevis (32, 33). Oscillatoria spp. produce geosmin and 2-methylisoborneaol which give drinking water a musty or earthy odor and has a major negative economic impact on the catfish aquaculture industry due to absorption in the flesh of the fish (34). Copper sulfate has been extensively used in water purification systems as a copper algaecide (34), and it was subsequently demonstrated that bxa1 expression was inducible by both monovalent (Cu^I/Ag^I) and divalent (Zn^II/Cd^II) ions in O. brevis, a novel property (33). The thiol oxidant diamide is also an inducer of bmtA expression (35). In vitro DNA binding and coupled transcription-translation experiments suggested that metal and diamide stress-induced disassembly of BxmR-operator/promoter complexes was directly responsible for regulation (35, 36).

In this report, we present a comprehensive characterization of the metal binding properties and metalloregulation of bxa1 operator/promoter binding by BxmR (33). We show that both pairs of predicted α3N and α5 sites bind metal ions, but with distinct physicochemical and functional metal specificities. BxmR is thus distinguished from other closely-related ArsR family sensors, in having evolved a metalloregulatory α3N site that can adopt an expanded range of coordination chemistries to include Cu^I/Ag^I, while maintaining redundancy in the response to Zn^II. These studies provide a glimpse as to how an ArsR family metal sensor with a relaxed metal specificity profile could have evolved to effect resistance against an expanded range of heavy metal ions.

EXPERIMENTAL PROCEDURES

Construction of E. coli overexpression plasmids for BxmRs and protein purification

The wild-type bxmR-coding region was amplified by PCR from O. brevis genomic DNA and cloned into pET3a (Novagen) between NdeI and EcoRI restriction sites to construct pET3a-BxmR. In order to obtain overexpression clones for α3NΔ and α5Δ BxmRs, a PCR-based quick-change mutagenesis method was employed to replace Cys77 to Ser in the α3N site, or ¹³²HLDEE¹³⁶ to ¹³²NLTYA¹³⁶ (H132N/D134T/E135Y/E136A) in the α5 site, using pET3a-BxmR as a template to create pET3a-α3NΔ BxmR and pET3a-α5Δ BxmR, respectively. C31S BxmR was prepared in exactly the same way. The integrity of all plasmid constructions was confirmed by complete DNA sequencing.

The resultant constructs were transformed into E. coli BL21(DE3) and grown in LB to mid-log phase, after which time 0.4 mM IPTG was added and cells allowed to grow for an additional 2.5 h. Freshly harvested cells were pelleted by low-speed centrifugations and suspended in 100 mL Buffer A (25 mM MES, 5 mM DTT, 1 mM EDTA, pH 6.0) and lysed by sonication. The cellular lysate was centrifuged and the supernatant pooled and subjected to precipitation by addition of polyethylenimine (PEI) to 0.15% (v/v) at pH 6.0. The PEI supernatant and precipitated fractions were collected separately. The PEI pellet was resuspended in Buffer A containing 0.50 M NaCl to extract bound proteins. Protein fractions derived from the NaCl-washed PEI pellet were precipitated by the addition of solid (NH₄)₂SO₄ to 50% saturation. Each ammonium sulfate pellet was gently resuspended in 100 mL Buffer A with 0.05 M NaCl and separately subjected to SP Fast Flow chromatography (20 mL bed volume) on an Äkta-10 purifier, with elution achieved with a linear NaCl gradient (0.05-0.75 M) in Buffer A. BxmR-containing fractions eluted at approximately 0.15 M NaCl and 0.35 M NaCl and were pooled separately. The resultant protein fractions were further purified by Superdex 75 size-exclusion chromatography, and anion-exchange HPLC using a procedure similar to that described previously (8). Individual fractions containing highly purified BxmR were pooled, concentrated and dialyzed against 3 L of Buffer B (10 mM HEPES, 0.40 M NaCl, pH 7.0) in an anaerobic Vacuum Atmospheres glovebox. The purity of the final products was estimated by visualization of Coomassie-stained 18% Tris-glycine SDS-PAGE gels to be ≥95%. The concentrations of all BxmR stock solutions were determined by quantitative amino acid analysis carried out by the Texas A&M University Protein Chemistry Laboratory. The number of reduced thiols in wild-type BxmR was determined by anaerobic DTNB colorimetric assay to be 4.1 (4.0 expected) and 2.8 (3 expected) in the WT and C31S BxmRs (8). The possible presence of copper in apo-BxmR was determined using a Perkin-Elmer AAnalyst 700 atomic absorption spectrophotometer operating in the flame mode using hollow cathode lamps specific for each metal (37). The purification of α3NΔ (C77S) and α5Δ BxmRs were purified in essentially the same fashion and characterized as indicated.

Copper Binding Experiments

All metal binding experiments were carried out anaerobically at ambient temperature (≈22 °C) using either a Hewlett-Packard Model 8452A Spectrophotometer (for UV-Vis electronic absorption spectra) or at 25 (±0.1) ºC with an ISS PC1 photon counting spectrofluorometer (for fluorescence and luminescence titrations, respectively) essentially as described previously (37, 38). Solution conditions were 10 mM MES, 0.1 M NaCl, pH 6.3 unless otherwise indicated. 10.5 and 11.2 µM wild-type and C31S BxmR monomer, respectively, was used for the UV-Vis absorption experiments, while 10.0 µM BxmR monomer was used for the luminescence experiments (λ_ex=300 nm with the emission spectrum scanned from 500-750 nm). For tyrosine fluorescence measurements, 5.0 μM BxmR monomer in Buffer B (10 mM HEPES, 0.40 M NaCl, pH 7.0) was titrated with 2-5 μL aliquots of either a 400 μM Cu^I stock in Buffer B or mock-titrated with Buffer B (λ_ex=280 nm; λ_em=305 nm). At each ith addition, the emission intensity was recorded and corrected for dilution, photobleaching (from the mock-titrated sample) and the inner filter effect exactly as described previously (21).

Other Metal Binding Experiments

Zn^II binding stoichiometries and affinities of wild-type, α3NΔ and α5Δ BxmRs were determined using chelator competition assays with either mag-fura-2¹(K_Zn=5.0 × 10⁷ M⁻¹ at pH 7.0) (39) or quin-2¹ (K_Zn=2.7 × 10¹¹ M⁻¹ at pH 7.0, 25 ºC) (40) using absorption spectroscopy essentially as described (23, 41). These data were fit using a competitive binding model with DynaFit (42) to determine step-wise zinc binding affinities, K_Zni. For mag-fura-2 experiments, 25 µM BxmR dimer (12.5 µM monomer) was mixed with 16.0 µM mag-fura-2 and titrated with a known stock solution of Zn^II; for quin-2 experiments, 20 µM BxmR dimer was mixed with 15 µM quin-2 and titrated with Zn^II (see Table 1). Co^II and Cd^II binding experiments were carried out essentially as described previously (30, 43) with the solution conditions given in the figure legends. The metal concentrations of all metal titrants used in individual experiments was quantified by atomic absorption spectroscopy as described (8).

Table 1.

Step-wise zinc binding affinities, K_Zni, derived from mag-fura-2 and quin-2 chelator competition experiments^a

BxmR protein	chelator	KZn1 (M−1)	KZn2 (M−1)	KZn3 (M−1)	KZn4 (M−1)
wild-type	mag-fura-2	≥109 a	≥109	2.2 × 107	n.d.b
	quin-2	9.4 × 1012	1.1 × 1012	n.d.c	n.d.c
α3NΔ	mag-fura-2	≥109 a	6.0 × 107	–	–
	quin-2	6.4 × 1012	n.d.c	–	–
α5Δ	mag-fura-2	≥109 a	3.0 × 106	–	–
	quin-2	4.6 × 1010	n.d.c	–	–

^aFrom nonlinear least-squares fits to data like those shown in Fig. 2 to a model assuming four (wild-type) or two (α3NΔ and α5Δ) stepwise binding events to a nondissociable dimer, each characterized by an affinity of K_Zni.

^bn.d., binding not detected, or K_Zni≤10⁵ M⁻¹ (a lower limit for mag-fura-2).

^cn.d., binding not detected, or K_Zni≤10⁹ M⁻¹ (a lower limit for quin-2).

Fluorescence Anisotropy-based DNA Binding Experiments

These experiments were carried out essentially as described previously (23, 41) using an ISS PC1 spectrofluorometer fitted with polarizers in the L format. 52-base pair oligonucleotide duplexes were prepared by mixing one unlabeled “bottom” strand at a slight excess relative to a 5′-fluorescein labeled “top” strand, annealed by heating to 95 ºC and slow-cooling at room temperature. Duplexes corresponding to the bxa1 and bxmR operator-promoter (O/P) regions contained a 26-base pair TGAA-containing inverted repeat (see ref. (36) for sequences) centered in the 52-base pair duplex. Titrations were carried out with 10 nM duplex in Buffer B (10 mM Hepes, 0.4 M NaCl, pH 7.0, 1.0 mM dithiothreitol) at 25 (±0.1) ºC with either apo-BxmRs or BxmRs preincubated with the indicated amount of a metal ion, e.g., Cu^I, Ag^I, Zn^II or Cd^II. Experimental binding curves were fit using DynaFit (42) to a binding model that assumes that two dimers bind to the DNA with identical binding affinities (K₁=K₂) (36), linked to a monomer-dimer equilibrium constant, K_dim, of 5.0 × 10⁶ M⁻¹ (44) The characteristic anisotropy of the 2:1 dimer:DNA species, r_(P2)2D, was optimized during the fit, with the r_{(P2) D} value fixed to exactly halfway between r_(P2)2D and r_D. In this work, the allosteric coupling free energies, ΔG_c, was operationally defined as the logarithmic ratio K₁^metal/K₁^apo according to ΔG_c = −RTln(K₁^metal/K₁^apo) with T=298 K.

RESULTS

Sequence alignment of selected ArsR regulators

O. brevis BxmR contains complete predicted ligand donor sets for both the interprotomer α3N and α5 metal sites previously characterized in other ArsR family sensors (6, 21). The α3N site is thiophilic site with a diagnostic ⁷⁵CVC⁷⁷ metal-binding sequence (30), and a number of potential donor residues in the N-terminal region, including by Cys23, His27 and Cys31 (36); the multiple sequence alignment of α3N sensors reveals that Cys31 in BxmR is unique (Figure 1). Cys23, Cys31, Cys75 and Cys77 are the only Cys in BxmR. The α5 site is a well characterized tetrahedral Zn^II-sensing site composed of carboxylate and imidazole ligands (3, 23); in BxmR, these are Asp119, His121, His132′ and Glu135′ with the His corresponding to His132 the key allosteric residue in other Zn^II sensors CzrA and SmtB (23). α3NΔ BxmR corresponds to C77S BxmR; an analogous non-metal liganding substitution abolishes Cd^II binding and sensing by S. aureus CadC (30, 45) and Cd^II/Zn^II sensing by Anabaena AztR (8). α5Δ BxmR is a quadruple mutant (H132N/D134T/E135Y/E136A) that changes two of the four α5 site ligands, His132 and Glu135 to non-metal liganding residues found in Anabaena AztR (8); these substitutions abolish Zn^II/Co^II binding to this site in AztR.

Figure 1.

Multiple sequence alignment of BxmR with other ArsR/SmtB family divalent metal ion-sensing regulatory transcriptional repressors (6, 8, 21, 30, 52). The proposed primary Cu^I binding ligands in α3N site and the α5 site for Zn^II are indicated for BxmR (this work).

Zn^II binding by wild-type, α3NΔ and α5Δ BxmRs

Like other previously characterized ArsR sensors (20, 44), BxmR is a stable dimer in solution as revealed by dynamic light scattering and gel filtration chromatography (data not shown). As a result, wild-type BxmR is predicted to contain four subunit-bridging metal binding sites per dimer (or two per protomer), two α3N and two α5 sites. We used standard dye competition experiments with two different chelators of widely different affinities, mag-fura-2 (K_Zn=5.0 × 10⁷ M⁻¹) and quin-2 (K_Zn=2.7 × 10¹¹ M⁻¹), in order to quantify the stoiochiometry and affinity of the Zn^II binding sites in wild-type, α3NΔ and α5Δ BxmRs (3, 21). Representative mag-fura-2 and quin-2 competition titrations carried out with wild-type BxmR are shown in Fig. 2, with parameter values obtained from a simple step-wise binding model shown in Table 1 for all BxmR species investigated here. Wild-type BxmR binds three Zn^II ions with an affinity that is either comparable to, or far greater than, that of mag-fura-2, with the fourth site transparent in this assay (K_Zn≤10⁵ M⁻¹) (Table 1). Using the higher affinity competitor dye, quin-2, reveals just two high affinity sites, with K_Zn values of ≈10¹³ and 10¹² M⁻¹; this result is consistent with the mag-fura-2 titrations. Carrying out the same experiments with α3NΔ and α5Δ BxmRs, where one pair of symmetry-related metal sites is inactivated in each case, clearly shows that each mutant binds two metal ions, one with very high affinity (K_Zn>10⁹ M⁻¹), measurable with quin-2 (Table 1), and one with an affinity comparable to or less than that of mag-fura-2 (K_Zn≤10⁷ M⁻¹) (see Supplementary Figure S1; Table 1). These binding parameters are diagnostic for strong negative homotropic cooperativity of Zn^II binding to each pair of metal sites, previously documented in the α3N sensor CadC (44) and structurally validated in the α5 sensors CzrA (3, 41) and SmtB (21).

Figure 2.

Zn^II binding by wild-type BxmR homodimer. (A) Titration of a mixture of 25 µM (monomer) BxmR and 16 µM mag-fura-2 with Zn^II with the absorption monitored simultaneously at 366 nm (μ) and 325 nm (open squares). (B) Titration of a mixture of 20 µM (monomer) BxmR and 15 µM quin-2 with Zn^II, with the absorption monitored at 261 nm. In both cases, the solid lines represent a fit to a multiple site binding model, with the parameters compiled in Table 1. Analogous titrations were carried out with α3NΔ and α5Δ BxmRs under the same conditions (see Supplementary Figure S1) with these parameters compiled in Table 1.

Taking all of the data together, the simplest interpretation is that Zn^II fills the metal binding sites on the wild-type BxmR homodimer in the following order: α5 H α3N ⇒ α5′ H α3N′. Thus, one of the α5 sites and one of the α3N sites possesses very high intrinsic zinc affinities K_Zn≥10¹² M⁻¹ or nearly equivalent to that found previously in the Zn^II sensor CzrA (3, 41); as a result, the dominant Zn^II complex that would persist over a wide range of [Zn^II]_free is the BxmR α5₁/α3N₁ homodimer. Further evidence for this formation of an α5₁/α3N₁ metallated homodimer comes from Zn^II binding isotherms monitored by a change in the steady-state Tyr fluorescence of 5.0 µM BxmRs (Figure 3). BxmR contains the same three Tyr residues found in the known structure of SmtB (2, 3), two in each strand of the β−wing, and one in the α5 helix (Fig. 1). As previously found for SmtB, Zn^II binding to the α5 site in α3NΔ BxmR gives rise to small but measurable enhancement or increase in the Tyr fluorescence intensity (21). In contrast, binding to the α3N site in α5Δ BxmR gives rise to the large quenching of the Tyr fluorescence not observed in SmtB; this spectral change is likely reporting on changes in the β-wing conformation due to α3N site binding. Both α3NΔ and α5Δ BxmR Zn^II binding isotherms reach maximum (or minimum) F/F_o values at ≈0.5:1 Zn^II to monomer consistent with the idea that half the sites drive the conformational change in the dimer as found previously for the α5 site in CzrA (41). Further, the conformational changes may well be distinct since the fluorescence yields of Tyr are different in each case. Strikingly, the addition of Zn^II to wild-type BxmR gives rise to a binding curve that is best described as a linear combination of Zn^II bound to either the α5 or α3N site, i.e., added Zn^II partitions between the two sites, which then reaches a minimum F/F_o value at ≈1:1 Zn^II to monomer (2 per dimer); residual binding of the third Zn^II to the second α3N site results in a small upward curvature of the F/F_o values (Figure 3).

Figure 3.

Anaerobic titration of 5.0 µM wild-type BxmR monomer (2.5 µM dimer) (μ), α3NΔ (λ) and α5Δ (υ) BxmRs with Zn^II as monitored by change in the Tyr fluorescence intensity, F_i/F_o. The solid lines through the data correspond to a nonlinear least squares fit to model that invokes sequential binding of Zn^II to a single high affinity site (for α3NΔ and α5Δ BxmRs) or two high affinity sites (for wild-type BxmR) each with a characteristic F_i/F_o signature. After filling these high affinity zinc binding sites (one α5 site with an F_i/F_o>1; one α3N site with an F_i/F_o <1), Zn(II) binds to each of the remaining low affinity α5′ and α′ sites on the dimer. Fitted parameters are as follows, with K_Zn1>10¹⁰ M⁻¹ (see Fig. 2A; Table 1): α3NΔ BxmR: F_i/F_o values, 1.045 (±0.001), 1.03 (fixed) for the high affinity and low affinity α5 Zn(II) sites, respectively; α5Δ BxmR: F_i/F_o values, 0.753 (±0.002), 0.915 (±0.011) for the high affinity and low affinity α3N Zn(II) sites, respectively; wild-type BxmR, F_i/F_o constrained to the above values for first three binding steps, with F_i/F_o for filling the final binding site, 0.80 (±0.03).

Co^II binding by wild-type BxmR

While Co^II is not an inducer of the bxa1 or bmtA genes (36), Co^II is often used as a structural surrogate for Zn^II to obtain insight into coordination geometry and nature of the ligand donor atoms. Interestingly, the sequence of metal binding steps for Zn^II is in contrast to that observed for Co^II, which largely fills both α5 sites, followed by both α3N sites in wild-type BxmR, an interpretation facilitated by the distinct spectral features of the Co^II complexes of the α5 and α3N sites (8, 21, 30) (Supplementary Figure S2). Both sites adopt tetrahedral or distorted tetrahedral coordination geometries consistent with S₃(N/O) α3N and N₂O₂ α5 ligand donor sets (8, 21). As expected, addition of stoichiometric Zn^II effectively bleaches the Co^II-α5 visible absorption spectrum in α3NΔ BxmR, consistent with K_Zn^α5>>K_Co^α5 as expected for tetrahedral symmetry (43).

Cd^II binding by wild-type, C31S, α3NΔ and α5Δ BxmRs

Representative Cd^II binding curves are shown for wild-type and C31S BxmRs as monitored by Cys S⁻⇒ Cd^II ligand-to-metal (or metal-to-ligand) charge transfer absorption at 235 nm (Supplementary Figure S3). The binding curves in both cases are essentially stoichiometric and reveal a single Cd^II binding site per monomer (2 per dimer) characterized by a molar absorptivity at 235 nm of ≈20,000 M⁻¹ cm⁻¹ for wild-type BxmR, most consistent with a tetrahedral S₄ complex (30). This binding curve is reporting on filling of the α3N sites with Cd^II, with any binding to the carboxy-terminal α5 sites not detected in this assay since this site contains no thiolate ligands. Cd^II does bind here, but with a lower affinity than Co^II since a 4-fold molar excess of Cd^II retains significant Co^II absorption deriving from the α5 site in α3NΔ BxmR (data not shown). Since we observe a lower molar absorptivity with the C31S BxmR (ε₂₃₅=16,000 M_Cd⁻¹ cm⁻¹), the simplest interpretation is that Cd^II coordinates to all four Cys in BxmR (Cys23, Cys31, Cys75, Cys77) to create a distorted tetrahedral S₄ site. As expected from prior studies of CadC (30, 46), Cd^II binds to the mutant α3N sites in α3NΔ BxmR, but with a lower affinity and lower molar absorptivity (ε₂₃₅=7000 M⁻¹ cm⁻¹); this is indicative of a significantly perturbed first coordination shell (data not shown).

Cu^I binding by wild-type, C31S and α5Δ BxmRs

Previous studies have established expression of both the bxa1 and bmtA genes is responsive to Cu/Ag and that the addition of reduced Cu^I or Ag^I to preformed BxmR-bxa1 and BxmR-bxmR/bmtA operator-promoter fragments in vitro resulted in dissociation of these complexes (36). Although these results are consistent with a functionally important and direct interaction of the Cu^I/Ag^I with BxmR, they provided no insight into the nature of the coordination complexes formed by Cu^I/Ag^I.

We measured Cu^I binding to apo-forms of wild-type and C31S BxmR anerobically using two different spectroscopic techniques. These include UV-visible absorption which reports on Cys S⁻⇒ Cu^I ligand-to-metal charge transfer absorption (LMCTs), i.e., formation of cysteine-copper coordination bonds, and luminescence in the visible region (λ_ex=300 nm; λ_em=580 nm). The results of these experiments are shown for wild-type and C31S BxmRs in Fig. 4A, B and Fig. 4C, D, respectively. Cu^I binding isotherms monitored in the UV region are consistent with a limiting stoichiometry of two Cu^I/monomer or four per dimer in both cases (Fig. 4A, C). While the C31S Cu^I binding curve is essentially stoichiometric under these conditions (pH 6.3, 25 ºC), the binding curve for wild-type BxmR can be fit to two step-wise equilibrium binding affinities, K_Cu1 and K_Cu2 of 3.6 × 10⁷ M⁻¹ and 2.7 × 10⁶ M⁻¹. The luminescence spectra are also consistent with a limiting stoichiometry of 2.0 Cu^I/monomer (4 per dimer) as judged by point at which the maximum luminescence is obtained. At excess Cu^I, the luminescence intensity decreases significantly which is likely reporting on rearrangement of the chelate to a less solvent shielded environment concomitant with the formation of nonnative multinuclear copper clusters (38). This feature also characterizes Cu^I binding to some but not all unrelated copper regulatory proteins known to form multinuclear copper clusters (38, 47), but not in a mononuclear site of relatively low luminescence intensity (37). We see virtually the same luminescence titration curve with α5Δ BxmR (data not shown), revealing that not surprisingly, the carboxylate/imidazole-containing α5 site does not bind Cu^I, at least in a way that contributes to the luminescence intensity. Thus, unlike the case for the Zn^II/Cd^II, these studies reveal that the α3N site of wild-type and C31S BxmRs binds Cu^I with a maximal stoichiometry of 2:1 (4 per dimer). We also note the Cu^I-saturated forms of wild-type and C31S BxmR contain a fully accessible α5 site since each is capable of binding Co^II to this site (Supplementary Figure S4); this result suggests that these two classes of metal sites are structurally independent.

Figure 4.

Anaerobic Cu^I binding properties of 10.5 µM (monomer) WT BxmR (panels A-B) and 11.2 µM (monomer) (panels C) or 10.0 µM C31S BxmR (monomer) (panel D) as monitored by UV absorption at 265 nm (panels A, C) or Cu^I luminescence (panels B, D) (38). Conditions: 10 mM Mes, 0.1 M NaCl, pH 6.3.

BxmR forms a binuclear Cu^I₂-S₄ complex

Copper K-edge X-ray absorption spectroscopy (XAS) was used to investigate the structure of the wild-type Cu^I complex formed at a 2:1 Cu^I:BxmR monomer molar ratio, and compared with that formed by C31S and α5Δ BxmRs, both at 1:1 metal-monomer ratios. The near edge spectra of all three samples are essentially superimposable and each is fully compatible with an average trigonally coordinated cuprous (Cu^I) ion (Figure 5A) (37). Experimental copper K-edge extended X-ray absorption fine structure (EXAFS) spectra are shown in the inset to Figure 5B with Fourier transforms of these data shown in the main body of the figure; structural parameters derived from the EXAFS curve fitting compiled in Supplementary Table S1. Virtually identical coordination environments are obtained for all three samples and each is well-described by three Cu-S interactions at 2.25-2.26 Å and an intense Cu-Cu scattering peak at ≈2.70 Å. These spectra are therefore consistent with a limiting Cu^I₂-S₄ complex, involving all four Cys of BxmR (Cys23, Cys31, Cys75 and Cys77) in wild-type and α5Δ BxmRs. Since the XAS near-edge and EXAFS spectra are virtually identical at saturating Cu^I for C31S BxmR this suggests that if Cys31 does indeed donate a thiolate ligand, it can be readily replaced with a molecule from solvent (H₂O or Cl⁻; note that Cl⁻ can not be distinguished from S⁻) (30), the Ser side chain itself, or another side chain, e.g., His27 (see Fig. 1). It is important to note that XAS near-edge and EXAFS spectra derived from coordination by a single His within a tetracopper Cu^I₄•S₅N complex could not be experimentally distinguished from Cu^I₄•S₆ complex (48); thus, recruitment or coordination by His27, if present, may be difficult to ascertain from these spectra of C31S BxmR.

Figure 5.

X-ray absorption spectroscopy of Cu-BxmRs. (A) Cu K-edge X-ray absorption near-edge spectra of WT BxmR (solid line), α5Δ BxmR (dashed line) and C31S BxmR (dotted line). (B) Cu—S phase-corrected EXAFS Fourier transforms of WT BxmR (solid line), α5Δ BxmR (dashed line) and C31S BxmR (dotted line). Inset, k³-weighted EXAFS spectra for each Cu^I-complex. Parameters that derive fitting these data to various coordination models are compiled in Supplementary Table S1.

Ag^I forms a mononuclear trigonal planar structure that is distinct from the Cd^II complex

In contrast to Cu^I, titration of apo-BxmR with Ag^I gives rise to significant absorption in the UV region that appears to be saturable with just one mol•equiv of the ion (spectra not shown). Perturbed angular correlation (PAC) spectroscopy was used to investigate the structure of the Ag^I complex in more detail and determine how this structure differs from that of Cd^II complex. ¹¹¹Ag decays to ^111mCd and in the PAC measurement the nuclear quadrupole interaction (NQI)¹ between the cadmium nucleus and the surrounding charge distribution is recorded (49). It is assumed that the electronic relaxation after the nuclear decay is very fast on the PAC (nsec) timescale, but the structural relaxation from Ag^I coordination geometry to Cd^II coordination geometry may appear fast, intermediate or slow on this timescale. This can be tested by comparing ¹¹¹Ag and ^111mCd PAC spectra on the same protein. If they differ from one another, it is an indication that the structural relaxation is slow. Inspection of Fig. 6A-C reveals that there appears to be no trace of the ^111mCd-BxmR PAC spectrum (Fig. 6C) in the ¹¹¹Ag-BxmR PAC spectrum (Fig. 6A,B) ; this suggests that the lifetime of the Ag^I-coordination geometry is sufficiently long as to allow for observation of the Ag^I-coordination complex without complications of decay to the Cd^II complex.

Figure 6.

¹¹¹Ag and ^111mCd Perturbed Angular Correlation (PAC) spectroscopy of Ag^I-substituted and Cd^II-substituted wild-type (WT) BxmR, respectively. Top (panels A and B, ¹¹¹Ag PAC; panel C, ^111mCd PAC), the experimentally determined perturbation function (with error bars, in blue), and fit (bold faced black line). Bottom (panels D-F), Fourier transformations of the experimental perturbation function in panels A-C, respectively (thin line), and fit (bold faced line). The data from ¹¹¹Ag PAC experiments in panels A and B are identical and merely show the results of two independent fits (1 and 2; see Table 2); these are multiplied by −1 to allow for easy comparison with ^111mCd PAC data (panel C). Parameters that derive from the fitted curves are compiled in Table 2.

The Fourier transform of the Ag^I-BxmR perturbation function (Fig. 6D-E) reveals two main features, one at 0.23 rad/ns and one at ≈0.4 rad/ns. Since the ≈0.4 rad/ns peak has a higher amplitude than the peak at about 0.23 rad/ns, this strongly indicates that more than one NQI and thus more than one coordination environment is present in the Ag^I complex. The NQI with ω₀ about 0.220 rad/ns (Table 2), can be extracted with reasonable certainty, as all three peaks (at about 0.220 rad/ns, 0.410 rad/ns and 0.630 rad/ns, respectively) are present beyond the noise in the Fourier transform, and they obey the rule ω₁ + ω₂ = ω₃; this is valid as the intermediate nuclear level has a spin of 5/2 (50). The second NQI shown in Table 2 is less reliable and difficult to extract, and two possibilities (Fit 1 and Fit 2; Fig. 6D-E) give almost the same reduced χ² when fit. These two NQIs both reproduce the low intensity peak at about 0.820 rad/ns observed at approximately twice the frequency of the most intense peak in the Fourier transform. This indicates that the corresponding NQI either has η close to 1 (Table 2 and Fig. 6D, Fit 1) or η close to 0 (Table 2 and Fig. 6E, Fit 2). It is a classic problem in PAC spectroscopy that these two cases are difficult to distinguish (51).³ The fact that two significantly different NQIs are observed reveals two different coordination environments are present in solution for Ag^I-BxmR, at least at Ag^I:protein ratio is about 1:10. Although the NQIs with ω₀ at 0.220 rad/ns may represent a variety of coordination geometries, the NQI with ω₀ about 0.420 rad/ns and low η is diagnostic for a trigonal planar coordination geometry composed of three coordinating cysteines.⁴

Table 2.

PAC fitted parameters for ¹¹¹Ag and ^111mCd bound to O. brevis BxmR.

Protein	ω0 (rad/ns)	η	Δω0/ω0 ×100	A ×100	χr2
Ag-BxmR: Fit 1	0.217(3)	0.33(4)	2f	2.1(3)	1.11
	0.245f	1f	2f	1.6(4)
Ag-BxmR: Fit 2	0.221(3)	0.32(3)	2f	3.0(2)	1.19
	0.420f	0.05f	2f	1.8(2)
Cd-BxmR	0.97(4)	0.78(19)	19(3)	5.0(6)	1.14
	0.101(1)	0.89(3)	6(1)	2.9(6)

The numbers in parentheses represent the standard deviations of the fitted parameters. “f” indicates that the parameter was fixed in the fit.

Although visual inspection of the Cd-BxmR spectrum (Fig. 6C, F) seems to be indicative of only one NQI with a relatively high value of η, it is not possible to make a satisfactory fit the data with just one NQI. Indeed, addition of a second NQI reduced χ² is reduced significantly (51) (by about 0.4). Thus the fit was carried out with two NQIs which, not surprisingly, turn out to be very similar (Table 2). This indicates that the two symmetry-related α3N sites in Cd^II-BxmR adopt similar coordination geometries at a Cd^II:protein ratio or ≈1:5. The relatively low values of ω₀ might indicate the coordination geometries deviate markedly from perfect tetrahedral symmetry, assuming that only cysteine ligands are present (see Figure S2).

Regulation of operator-promoter (O/P) DNA binding of BxmR by different metal ions

The above metal binding experiments with wild-type and mutant BxmRs establish the stoichiometry and affinity by which individual metal ions bind to the α3N and α5 sites on BxmR, as well as insights into their coordination environments. Previous studies establish that BxmR binds to a typical ArsR-family operator containing a 12-2-12 inverted repeat sequence situated just upstream of the bxa1 gene encoding a CPx-ATPase (36). Fig. 7 shows representative anisotropy-based bxa1 O/P isotherms for wild-type (Fig. 7A), α3NΔ BxmR (Fig. 7B) and α5Δ BxmR (Fig. 7C) carried out with apo-BxmR or with BxmRs preincubated with 1:1 (Zn^II, Cd^II or Ag^I) or 2:1 (Cu^I) (per BxmR monomer) of the indicated metal ion (25 ºC, pH 7.0, 0.2 M NaCl). The data were fit to a model assuming that two BxmR dimers bind to the operator with equal affinities (36), with the fitted parameters compiled in Table 3.

Figure 7.

Representative normalized fluorescence anisotropy-based bxa1 operator/promoter DNA binding isotherms for various WT (A), α3NΔ (B) and α5Δ (C) BxmRs in the absence (apo-) and presence of the indicated metal ions. The continuous line through each binding isotherm represents a nonlinear least squares fit to a two-dimer DNA binding model (see Experimental Procedures), with the fitted parameters compiled in Table 3.

Table 3.

bxa1 O/P binding parameters for apo- and various metal loaded states of wild-type, α3NΔ and α5Δ BxmRs^a

BxmR protein	metal	rD	r(P2)2·D	K1 (M−1)	ΔGc (kcal/mol)b
wild-type	apo	0.108	0.135	1.0 (±0.2) × 108	–
	CuI1.8	0.106	0.135c	7.1 (±0.1) × 106	1.6
	AgI	0.108	0.135c	3.7 (±0.4) × 106	1.9
α3NΔ	apo	0.108	0.138	1.2 (±0.1) × 108	–
	CuI	0.109	0.138	1.0 (±0.2) × 108	0.1
	CdII	0.113	0.135	7.2 (±0.1) × 107	0.3
	ZnII	0.109	0.138c	≤5.0 (±0.3) × 105	≥3.2d
α5Δ	apo	0.109	0.137	1.0 (±0.2) × 108	–
	CuI	0.106	0.137c	5.0 (±0.1) × 106	1.8
	AgI	0.105	0.137c	5.6 (±0.1) × 106	1.7
	CdII	0.111	0.137c	5.0 (±0.1) × 106	1.8
	ZnII	0.110	0.137c	1.8 (±0.4) × 106	2.4

^aParameters determined from nonlinear least squares fits the data shown in Fig. 7 using the model described under Experimental Procedures.

^bEstimated uncertainty in ΔG_c is ±0.1 kcal/mol.

^cThis value of r_(P2)2·D was fixed to the value obtained for the appropriate apoprotein in each case during the parameter optimization. The value of r_D was fixed to the experimentally measured value during parameter optimization.

^dReflects a lower limit given the upper limit on K₁ (see Fig. 7B for binding curve).

The data taken collectively reveal that bxa1 O/P DNA binding activity of BxmR is allosterically negatively regulated by all four metals tested. Monovalent Cu^I and Ag^I as well as divalent Cd^II are equally effective in inhibiting DNA binding, with an allosteric coupling free energy, ΔG_c, of +1.7 (±0.1) kcal mol⁻¹ under these conditions (Table 3). All three of these metals function through the thiophilic α3N metal binding site, since α3NΔ BxmR shows no regulation by Cd^II or Cu^I (Fig. 7B). In striking contrast to Cu^I and Cd^II, Zn^II is unique in that this metal is a potent effector of bxa1 O/P binding by α3NΔ BxmR, with a lower limit for ΔG_c, of +3.4 (±0.1) kcal mol⁻¹; this result shows that Zn^II is capable of functioning through the α5 metal binding sites. Binding to the α5 sites is not obligatory however, since Zn^II also negatively regulates DNA binding in α5Δ BxmR, i.e., through the α3N binding sites. Thus, BxmR retains a functional redundancy in its response to Zn^II, unlike all other metals that induce bxa1 expression in O. brevis (32).

DISCUSSION

O. brevis BxmR is the fifth primary zinc-sensing member of the ArsR/SmtB family of metalloregulatory repressors (33) to be purified and characterized, four of which derive from different cyanobacterial species. The others are Synechococcus SmtB (17), Synechocystis ZiaR (52), Anabaena PCC7120 AztR (8) and S. aureus CzrA (20, 53). BxmR is most closely related on the basis of global amino acid sequence similarity to ZiaR, followed by AztR and SmtB (see Figs. 1, 8A), and these four proteins are part of a small cluster of sequences within the Group 2 clade of ArsR-family repressors (7) (our data not shown; see Fig. 8A). These sensors represent four of the ≈15 ArsR/SmtB sensors for which significant biochemical and cell biological data are now available, which collectively respond to an impressive range of metal ions including Zn^II, Zn^II/Co^II, Ni^II/Co^II, Cd^II/Pb^II, As^III/Sb^III, Bi^III and Cu^I/Ag^I (6, 14) (see Fig. 8B). In this report, we establish that BxmR is unique in this family of proteins in that despite the fact that Zn^II is the most potent inducer of bxa1 expression (33), BxmR mediates a clear response to Cu^I/Ag^I both intracellularly and in vitro (36) in striking contrast to ZiaR (52).

Figure 8.

Schematic comparison of ArsR/SmtB family metal sensors with well-characterized metal binding properties. (A) Dendogram that illustrates the degree to which the four cyanobacterial ArsR/SmtB zinc sensors are related to one another and to previously characterized α5 (Sa CzrA) and α3N (Sa pI258 CadC) sensors on the basis of a global sequence similarity (% identity to Obr BxmR also indicated). (B) Schematic rendering of ArsR/SmtB sensors that highlights known regulatory metal sites (filled symbols) with their associated trivial designations (third column) and metal ions that are sensed in vivo (fourth column) by these sites. Other known metal sites that play no regulatory role are represented by open symbols (in Syn SmtB and Sa pI258 CadC). Sensors are: Synechococcus (Syn) SmtB (31, 57), Anabaena (Ana) AztR (8), Synechocystis (Sync) ZiaR (6, 52), O. brevis (Obr) BxmR (this work), S. aureus (Sa) pI258 CadC (30, 46), Listeria monocytogenes (Lm) CadC (30), E. coli R773 (Ec) ArsR (59), S. aureus (Sa) CzrA (3), M. tuberculosis (Mtb) NmtR (20, 60), M. tuberculosis (Mtb) CmtR (22, 24, 25), M. tuberculosis (Mtb) KmtR (7), Xylella fastidiosa (Xf) BigR (27, 29). For Obr BxmR, the metals sensed by the α3N site are shaded green; Zn^II is sensed by both α5 and α3N sites (this work); for ZiaR, both sites are apparently required to sense Zn^II in vivo (52). Although the inducer identity of Xf BigR is not yet known (?), BigR contains conserved Met, Cys and Cys residues in the projected α1, α2 and α5 helices (27). See text for additional details.

Extensive biochemical, crystallographic and NMR structural studies of CzrA, a Zn^II/Co^II sensor from S. aureus, the Cd^II/Pb^II sensor CadC from S. aureus pI258 and L. monocytogenes, Synechococcus SmtB and NmtR, the Ni^II/Co^II sensor from M. tuberculosis, establishes the evolution of two distinct metal sensing sites (α3N and α5) on this homodimeric winged helical scaffold (Fig. 8B) (1, 3, 20, 21, 26, 30). The α3N site is anchored by a pair of Cys likely positioned at the N-terminus of the α3 helix (Cys75 and Cys77 in BxmR), but shows considerable variability in the number and type of donor atoms that are known or predicted to make coordination bonds to the metal. Previous studies of S. aureus CadC established that a Gly substitution of the Cys corresponding to Cys77 in BxmR abrogates allosteric coupling of Cd^II/Pb^II/Zn^II binding to DNA binding by CadC and rendered CadC unable to sense metals in vivo (45); the analogous Ser substitution in AztR abolishes Zn^II sensing in vivo and Pb^II binding in vitro (8). Interestingly, the same Gly substitution occurs naturally in SmtB (see Fig. 1); this substitution makes the prediction that the α3N site in SmtB would be capable of binding metals, but would not be regulatory and this is exactly in accord with experiment (31). Interestingly, S. aureus CadC contains both intact α3N and α5 metal sites, and the crystal structure of “apo-CadC” revealed that both α5 sites were filled with Zn^II (1), a finding consistent with its very high equilibrium affinity (L. Busenlehner and D. Giedroc, unpublished results). However, this site is not metalloregulatory for cad O/P binding, since an α3N inactivating mutation analogous to the one characterized here in α5Δ BxmR (C60G in CadC; C77S in BxmR) abolishes regulation of DNA binding by all metal ions (30). This property is in strong contrast to BxmR which retains potent Zn^II regulation through its α5 metal sites to a degree that is comparable to that of bona fide α5 zinc sensor S. aureus CzrA (23, 41).

It is striking that the four cyanobacterial zinc sensors have evolved distinct metal binding properties in a way that seems to track largely with differences in the N-terminal region. CadC contains an additional N-terminal α-helix (α0)¹ not found α5 sensors SmtB (3), CzrA (54) or NmtR (A. Arunkumar, H. Reyes, and D. Giedroc, unpublished observations), with the most N-terminal residue visible in electron density maps Tyr12 (1). Tyr12 is just C-terminal to Cys7 and Cys11, the former of which donates a thiolate ligand to the Cd^II/Pb^II ion (30) and may well be required to drive a conformational change in the dimer that lowers the DNA-binding affinity of the repressor (55). ZiaR and AztR contain longer N-terminal domains (see Fig. 1). Each contains a single Cys that aligns with the key allosteric residue Cys7 of CadC, as well as one conserved His (His27 in AztR). The α3N chelates of ZiaR and AztR are structurally indistinguishable and are consistent with an S₃(N/O) tetrahedral coordination environment as determined by the Co^II coordination complex, easily distinguished from the highly distorted S₄ site of CadC (8). BxmR contains the longest N-terminal extension of any cyanobacterial ArsR-family Zn^II sensor to date (see Fig. 1). BxmR conserves the key Cys corresponding to Cys7 in CadC (Cys23), but also contains three additional candidate metal ligands: a His analogous to His18 in SmtB (this His is proposed to donate a ligand to the α3N-bound Zn^II in SmtB (21, 43)), a unique Cys (Cys31) not found in the other sensors, and His15, analogous to Cys12 in ZiaR (see Fig. 1). The Co^II spectrum of Co^II-substituted α5Δ BxmR is very similar to Co^II-AztR, and is thus consistent with S₃(N/O) α3N site (8); however, the Cd^II complex contains 3-4 thiolate ligands, with the unique Cys31 apparently making a coordination bond (Supplementary Figures S2-S3).

The distinct N-terminal region of BxmR stably accommodates an α3N Cu^I complex in a solvent-shielded (38) binuclear Cu₂S₄ cluster that we propose involves Cys23, Cys31, Cys75′, Cys77′ as wild-type metal ligands. This proposed binuclear copper cluster structure appears analogous to that proposed for Cu^I-specific sensor Enterococcus CopY (56). Each of the Cu^I ions in BxmR is trigonally coordinated with three Cu-S⁻ distances at 2.26 Å and a single Cu-Cu distance of 2.70 Å; see Fig. 5); essentially identical spectra are obtained with the WT and α5Δ BxmR consistent with the conclusion that the C-terminal α5 site plays no role in Cu^I binding or Cu^I-dependent negative allosteric regulation of DNA binding (Fig. 7). We note however, that Cu^I binds to BxmR with an average step-wise binding affinity of ≈1 × 10⁷ M⁻¹ (Fig. 4A), which is quite weak relative to other known mononuclear bacterial Cu^I sensors E. coli CueR (19) and M. tuberculosis CsoR (37) as well as the multinuclear Cu^I complex in Drosophila MTF-1 (38). This finding is consistent with the idea that the α3N site in BxmR is not selective for Cu^I and may have resulted from the use of copper salts as algaecides for O. brevis used in these studies (33). These physical data show conclusively that residues derived from the N-terminal domain of BxmR allow direct binding of monovalent ions Cu^I/Ag^I in a way that induces negative regulation of DNA binding to a degree that is identical to divalent ions Zn^II/Cd^II. These findings provide additional evidence that the α3N site is structurally plastic, and particularly well-suited to adopt a wide range of coordination chemistries (geometries and stoichiometries) to accommodate a wider range of metal ions and thereby mediate their homeostasis under conditions of metal ion excess (30).

A side-by-side comparison of the physical properties of the four closely related zinc-sensing ArsR/SmtB repressors (Syn SmtB, Ana AztR, Obr BxmR and Sync ZiaR) reveals the remarkable finding that each likely utilizes distinct structural mechanism to mediate resistance to toxic concentrations of Zn^II in the cell (see Fig. 8). A similar conclusion was reached on the basis of a comparison of three As^III-sensing ArsRs, each of which harbor distinct S₃ coordination sites proposed to result from convergent evolution in response to recent environmental pressures (5). We note that Anabaena contains one other ArsR/SmtB regulator, most closely related to Syn SmtB termed AzuR (alr0831) not listed in Fig. 8B; AzuR has been proposed to be a Zn^II uptake repressor (8) but this has not yet been tested experimentally.⁵ In any case, Syn SmtB (and likely Ana AzuR) utilizes a single pair of subunit bridging α5 metal sites (31, 57) much like Sa CzrA (23). In AztR, the α5 metal sites have been lost and as a result, Zn^II regulation falls to the CadC-like S₃(N/O) α3N sites (8). AztR also senses Cd^II/Pb^II through these metal sites, but the metal effluxer whose expression is regulated by AztR, AztA, is largely functionally selective for Zn^II (58). In contrast, Obr BxmR can exploit either α5 or α3N metal sites to effect zinc regulation, with the Zn^II-bound α5 sites far more selective and far more effective in allosterically inhibiting operator/promoter DNA binding. Finally, in vivo experiments suggest that Sync ZiaR requires both α5 or α3N metal sites to effect metalloregulation by Zn^II since inactivation of one or the other abolished Zn^II sensing in the cell (52); the structural basis for this is not yet known but would appear to be in contrast to BxmR. This feature might enhance the metal selectivity of ZiaR for Zn^II relative to AztR and BxmR since Cu^I and Ag^I do not induce the zia operon in Synechocystis cultures (52). It will be interesting to determine the extent to which the metallothionein BmtA and the metal transporter Bxa1 binds and transports, respectively, Ag^I/Cu^I vs. Zn^II/Cd^II.

One common feature of all ArsR/SmtB sensors that have investigated in detail to date is that filling of one of two symmetry-related metal sites is necessary and sufficient to mediate regulation of DNA binding (41). For example, inactivation of one of the two pairs of sensing sites in BxmR to create α5Δ and α3NΔ BxmRs results in retention of a single high affinity metal binding site with the second symmetry-related site on the homodimer characterized by a lower affinity for the metal (Table 1); a similar characteristic was found for all α5 first coordination shell mutants of Sa CzrA (23). Nonetheless, all drive approximately wild-type levels of regulation, at least in vitro. Current efforts are directed toward understanding in molecular terms how metal binding to the α5 vs. the α3N sites drives negative allosteric regulation of DNA binding in BxmR and other ArsR family sensors, and what factors dictate the quantitatively more potent regulation that occurs upon metal binding by the C-terminal α5 sites (54).