Protein Signatures That Promote Operator Selectivity among Paralog MerR Monovalent Metal Ion Regulators*

Background: Two nucleotide bases distinguish promoters controlled by paralog MerR monovalent metalloregulators, avoiding cross-activation. Results: Specific residues within the DNA-binding region of the regulators were identified as responsible for the selectivity in the operator recognition. Conclusion: Co-evolution of both the regulator and its target operator sequences prevents cross-activation of paralog regulatory circuits. Significance: The basis for regulator/operator specificity among MerR monovalent metalloregulators is described. Two paralog transcriptional regulators of the MerR family, CueR and GolS, are responsible for monovalent metal ion sensing and resistance in Salmonella enterica. Although similar in sequence and also in their target binding sites, these proteins differ in signal detection and in the set of target genes they control. Recently, we demonstrated that selective promoter recognition depends on the presence of specific bases located at positions 3′ and 3 within the operators they interact with. Here, we identify the amino acid residues within the N-terminal DNA-binding domain of these sensor proteins that are directly involved in operator discrimination. We demonstrate that a methionine residue at position 16 of GolS, absolutely conserved among GolS-like proteins but absent in all CueR-like xenologs, is the key to selectively recognize operators that harbor the distinctive GolS-operator signature, whereas the residue at position 19 finely tunes the regulator/operator interaction. Furthermore, swapping these residues switches the set of genes recognized by these transcription factors. These results indicate that co-evolution of a regulator and its cognate operators within the bacterial cell provides the conditions to avoid cross-recognition and guarantees the proper response to metal injury.

Two paralog transcriptional regulators of the MerR family, CueR and GolS, are responsible for monovalent metal ion sensing and resistance in Salmonella enterica. Although similar in sequence and also in their target binding sites, these proteins differ in signal detection and in the set of target genes they control. Recently, we demonstrated that selective promoter recognition depends on the presence of specific bases located at positions 3 and 3 within the operators they interact with. Here, we identify the amino acid residues within the N-terminal DNAbinding domain of these sensor proteins that are directly involved in operator discrimination. We demonstrate that a methionine residue at position 16 of GolS, absolutely conserved among GolS-like proteins but absent in all CueR-like xenologs, is the key to selectively recognize operators that harbor the distinctive GolS-operator signature, whereas the residue at position 19 finely tunes the regulator/operator interaction. Furthermore, swapping these residues switches the set of genes recognized by these transcription factors. These results indicate that co-evolution of a regulator and its cognate operators within the bacterial cell provides the conditions to avoid cross-recognition and guarantees the proper response to metal injury.
Transcriptional regulators of the MerR family modulate transcription in response to different environmental signals, including heavy metal ions, organic compounds, or oxidative stress (1). These proteins have a structurally conserved Nterminal DNA-binding domain with two helix-turn-helix (HTH) 4 motifs separated by a two-stranded antiparallel ␤-sheet, in a ␣1-␣2-␤1-␤2-␣3-␣4 topology. This region is connected to a variable C-terminal effector-binding domain by an extended ␣-helix, forming a coiled-coil dimerization region. Usually, these regulators recognize pseudopalindromic operator sequences in 70 -targeted promoters with longer (19-or 20-bp) spacers between the Ϫ35 and Ϫ10 elements that prevent open complex formation by RNAP without an activator (2)(3)(4).
According to the current model of MerR-mediated induction, after the signal is detected by the regulator, the information is transduced to the DNA-binding domain, triggering localized base pair breaking and base sliding in the operator. This results in a realignment of the promoter elements that now allows proper RNAP-DNA interaction and transcription initiation. The solved structure of three MerR homologues bound to their target operators, the drug-binding BmrR and Mta proteins and the oxidative stress sensor SoxR, envisages a conserved mechanism for DNA recognition (3)(4)(5). Protein-DNA interactions involve hydrogen bonds and van der Waals contacts between amino acid residues mostly belonging to the ␣2-helix and the ␤2 strand of the N-terminal DNA-binding domain and the DNA backbone of the target operator sequence (3)(4)(5). It has been proposed that these contacts serve both to stabilize the distorted DNA conformation and to provide regulator/operator selectivity.
In Salmonella enterica serovar Typhimurium, two metal ion sensors of the MerR family, CueR and GolS, control the transcription of genes coding for factors responsible for monovalent metal ion resistance (6 -8). These paralog transcription factors probably evolved from a common ancestor by gene duplication followed by divergence that rewired both signal recognition and the set of controlled genes (9,10). The copper sensor CueR, present in most Gram-negative bacteria, induces the expression of its target genes in response to either Cu(I), Ag(I), or Au(I) (11,12), whereas the horizontally acquired GolS regulator evolved to preferentially sense Au(I) ions (6,7). As we mentioned before, the set of genes controlled by each transcription factor also differs. CueR activates the expression of the P-type ATPase CopA, the multicopper oxidase CueO/CuiD, and the periplasmic copper-binding protein CueP (7,8,13), whereas GolS induces transcription of the Salmonella-specific genes encoding for another P-type ATPase, named GolT, the metal-binding protein GolB and the CBA-type GesABC efflux system (6,14). We recently demonstrated that although these regulators recognize similar operator sequences at their target promoters, particularly at the Ϫ35 promoter region, the presence of only two distinctive nucleotide bases at the 3Ј-and 3-positions relative to the center of the operator determines regulator/operator selective recognition (15).
GolS-controlled promoters have an A at the 3Ј-position and a T at the 3-position relative to the center of the operator, whereas operators recognized by CueR have either a C or a G at these positions (Fig. 1A). Switching these nucleotide bases in gol-like or cue-like operators is sufficient to swap the regulator dependence (15). In other words, the mutant golB CC promoter, which harbors a C at the operator positions 3Ј and 3 as PcopA, decreases the affinity for its native regulator GolS but has an increased binding to CueR compared with the wild-type golB promoter. A similar switch in regulator dependence is observed when the mutant copA AT and the wild-type copA promoters were compared. In fact, the A and T signature nucleotide bases at 3Ј-and 3-positions from the center of the dyad operator sequence are conserved in promoters that are proposed to be transcriptionally controlled by GolS-like regulators, whereas operators predicted to be controlled by the CueR group harbor C or G at these positions (15).
The presence of selective nucleotide bases at the operators must correlate with specific amino acid residues within the DNA-binding domain of the transcription factors that direct selective recognition. To identify the amino acid residues involved in operator discrimination, here we constructed a series of GolS and CueR hybrid proteins replacing different portions of the DNA-binding domain of each regulator by the same region of the paralogous protein and tested their ability to direct expression from the golB or copA promoter. Together with site-directed mutagenesis and in silico modeling, these studies demonstrate that the residue at position 16 from the ␣2-helix of both CueR and GolS is essential for selective recognition. Our results provide additional evidence of the co-evolution of both the MerR regulators and the regulated genes to avoid cross-recognition and guarantee a proper response to metal injury.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Growth Conditions-Bacterial strains (all derivatives of Escherichia coli W3110 strain or S. enterica serovar Typhimurium 14028s, except when indicated) and plasmids used in this study are listed in Table 1. Oligonucleotides are listed in Table 2. Cells were grown at 37°C in Luria-Bertani broth (LB) or on LB agar plates. Ampicillin, tetracycline, kanamycin, and chloramphenicol were used when necessary at 100, 15, 50, and 20 g ml Ϫ1 , respectively. Reagents, chemicals, and oligonucleotides were from Sigma, except for the Luria-Bertani culture medium, which was from Difco.
Bacterial Genetic and Molecular Biology Techniques-E. coli strains carrying deletions on the lacZ, cueR, or copA genes were generated in the W3110 strain by Lambda Red-mediated recombination (16) using the appropriate primer pairs listed in Table 2. The deletions were individually transferred to different W3110 derivatives by P1-mediated transduction (17) to generate the PB10305 strain. When necessary, the antibiotic resistance cassette inserted at the deletion point was removed using the temperature-sensitive plasmid pCP20 carrying the FLP recombinase (18).
Construction of hybrid cueR-N S , cueR-HTH1 S , cueR-(HTH1ϩL) S , and golS-N R alleles was carried out by using mutagenesis by PCR overlap extension or SOE-PCR (19) (for the chimeric protein nomenclature used, see Figs. 1B and 2A). Briefly, we performed two independent PCRs using the complementary primers carrying the desired hybrid junction and oligonucleotides homologous to the 5Ј and 3Ј wild-type gene of interest harboring BamHI or HindIII restriction sites ( Table 2). Then the products of both PCRs were purified and combined in a third PCR with the appropriate forward or reverse primers to generate the final product. To construct the cueR-(LϩHTH2) S and cueR-HTH2 S alleles, we amplified a first PCR product using GolS-Loop-Fw/GolS-HTH2-Rv or GolS-HTH2-Fw/ GolS-HTH2-Rv primer pairs (Table 2). Each product was employed as a primer in two independent PCRs along with CueR-ORF-Fw or CueR-ORF-Rv to generate the overlapping fragments used in the final PCR, as described above. The cueR-␣2 S , golS-␣2 R , or the alleles carrying point mutations were amplified in a single PCR using the forward oligonucleotides harboring the modification and the BamHI restriction site and the appropriate reverse primer. The final PCR fragments were purified and individually cloned into BamHI/HindIII-digested pUH21-2laqI q or the pSU36 vector to generate the expression plasmids listed in Table 1.
The plasmids and the linear DNA fragments were introduced to E. coli strains by electroporation using a Bio-Rad device following the manufacturer's recommendations. All constructs were verified by DNA sequencing.
Metal Induction Assays-The levels of expression of the lacZ reporter gene under the control of the native (PcopA or PgolB) or mutated (PcopA AT or PgolB CC ) promoters in the presence of 10 M AuHCl 3 or 100 M CuSO 4 or without metal added were assessed essentially as previously described (15) using overnight cultures on LB. ␤-Galactosidase assays were carried out essentially as described (17).
Protein Purification-GolS, CueR, or the hybrid proteins GolS-␣2 R and CueR-␣2 S were overexpressed and purified from E. coli XL1-Blue strain essentially as described previously (14). 0.1 or 0.5 mM IPTG (for CueR or GolS variants, respectively) was added to the cultures at an A 600 of 0.6 to promote protein expression. All procedures were carried out at 4°C. The protein profile of purified samples was determined by SDS-PAGE, and the concentration was calculated, recording absorbance at 280 nm with an ⑀ of 4320 M Ϫ1 cm Ϫ1 (CueR), 6585 M Ϫ1 cm Ϫ1 (CueR-␣2 S ), 11,835 M Ϫ1 cm Ϫ1 (GolS), or 10,345 M Ϫ1 cm Ϫ1 (GolS-␣2 R ), or by Bradford assay, using bovine serum albumin as a standard.
Protein-DNA Interaction Analysis-Electrophoretic gel mobility shift assays were performed using purified wild-type or mutant regulators as described previously (15). The primers used to amplify the PcopA or PgolB promoter region are listed in Table 2.
Fluorescence anisotropy assays were carried out essentially as described (20,21). Fluorescein-labeled double-stranded DNA fragments harboring the copA or golB operator sequences were generated by incubating pairs of single-stranded oligonucleotides (Table 2) at 85°C for 10 min and then allowing the mix to cool at room temperature. The binding of native and mutant transcriptional regulators to PcopA and PgolB promoters was measured using a Varian Cary Eclipse fluorescence spectrophotometer (Agilent Technologies). All fluorescence anisotropy titrations were done in 10 mM Tris buffer with 10 mM NaCl, 2 mM MgCl 2 , and 5 mM DTT, pH 7.3. The fluorescence was excited at 495 nm. Anisotropy (r) was calculated as r ϭ ( , where I ͉ and I Ќ are the fluorescence intensity parallel and perpendicular to the excitation polarization, respectively, and G is the correction factor for the instrument's different responses to light of parallel and vertical polarizations. Binding isotherms were fitted to the equation, , where r f and r b are the anisotropy values for free and protein-bound DNA, respectively, x is the total protein concentration, and k A is the association constant of the protein-DNA complex, by nonlinear regression using the Sigma Plot software. Dissociation constants (k D ) were estimated from the equation, In Silico Modeling-A structural model for CueR, CueR-␣2 S , and GolS was generated by homology modeling using Rosetta version 3.1. The structure of BmrR transcription factor bound to its promoter (Protein Data Bank code 3Q5R) was used as a template. 100 initial structures were generated using a cyclic coordinate descent algorithm combined with fragment assembly for modeling the CueR sequence with gaps. The 10 lowest energy structures were subsequently refined, and the lowest score structure was used as a model. The model structures were then superimposed on the structure of the complex used as a template in order to locate the position of the residues of interest (Ala-16 and Phe-19) with respect to the promoter DNA and compare their relative conformation in the different proteins.

Selective Regulator/Operator Recognition Is Directed by
Amino Acids within the ␣2-Helix of the Sensor Protein-Salmonella CueR and GolS can distinguish their target operators from those of the paralog regulator by selectively recognizing two nucleotide bases located at the 3Ј-and 3-positions from the center of the operator (Fig. 1A) (15). To identify the amino acid residues within the DNA-binding domain of the transcription factor that direct operator discrimination, we designed chimeric proteins, in which the N-terminal DNA-binding domain of CueR (from Met-1 to Asn-68) was replaced by the equivalent region of GolS and vice versa (Fig. 1B). The encoding mutant alleles were cloned in plasmids and introduced either into an S. enterica serovar Typhimurium ⌬gol ⌬cueP ⌬cueR-copA or into an E. coli W3110 ⌬cueR-copA ⌬lacZ strain harboring lacZ reporter fusions to copA or golB promoters. (The use of strains deleted in copA helped to minimize differences in intracellular copper levels derived from partial activation of the copper transporter by the regulator variants.) The cells were grown in LB in the absence or presence of either 10 M AuHCl 4 or 100 M CuSO 4 , concentrations required to attain the maximal induction of the reporter genes (7,14). The chimeric CueR-N S and GolS-N R regulators were functional to activate transcription of the reporter genes in response to the metals, but their induction pattern resembled that attained by the paralog regulator ( Fig. 1C) (data not shown). In other words, CueR-N S activated the expression of the reporter gene more efficiently under the PgolB promoter than under the native PcopA promoter, resembling wild-type GolS, and GolS-N R switched its operator recognition preference, acting as a better inducer of CueR-regulated promoters. Because essentially identical results were obtained using either Salmonella or E. coli, we choose to continue the analysis in the latter species.   To delimit the region of CueR responsible for operator discrimination, we replaced different portions from the N-terminal DNA-binding domain of CueR by the equivalent regions of GolS and analyzed the ability of each chimeric regulator to activate the expression of the reporter genes under either a GolS-or a CueR-controlled promoter (Fig. 2). Like CueR-N S , CueR-HTH1 S and CueR-␣2 S , carrying the first HTH motif (from Met-1 to Lys-23) or the ␣2-helix (from Thr-13 to Lys-23) of GolS, respectively, drove a reporter gene expression like GolS (Fig. 2B). These hybrid proteins showed an improved activation of PgolB-driven expression and a decreased induction of transcription from the native PcopA promoter compared with wildtype CueR, suggesting that the operator selectivity resides within the ␣2-region. Furthermore, the equivalent ␣2-helix replacement in GolS resulted in a transcriptional regulator that exhibited a CueR-like expression pattern with an increased metal-activated expression of the reporter gene from PcopA and almost no induction of transcription from the PgolB promoter (Fig. 2B). The other CueR chimeric constructions with the replacement of either the ␤1-␤2 loop or of the HTH2 domain from GolS (i.e. CueR-HTH2 S , CueR-(HTH1ϩL) S , or CueR-(LϩHTH2) S ) resulted in less active or inactive regulators  that could not induce the expression from either native PcopA or PgolB (Fig. 2B).
To verify the role of the ␣2-helix in the selectivity of operator recognition, we performed electrophoresis mobility shift assays (EMSAs) using amplified fragments derived from copA or golB promoter regions and the purified wild-type regulators or chimeric CueR-␣2 S or GolS-␣2 R proteins (Fig. 3A). A 7-fold increase in CueR-␣2 S amounts was required to similarly affect the mobility of the PcopA promoter compared with CueR. On the other hand, CueR-␣2 S substantially gained an apparent affinity for the PgolB promoter. GolS-␣2 R shows much lower affinity for the PgolB promoter compared with the parental GolS regulator, whereas it gained affinity for the PcopA region, resembling wild-type CueR (Fig. 3A).
The binding affinity for each regulator/operator pair was estimated by fluorescence anisotropy titration. In these assays, we used fluorescein-labeled 41-base pair double-stranded DNA containing the operator sequences from copA or golB. Each labeled dsDNA probe was titrated with increasing amounts of CueR, GolS, or the mutant variants CueR-␣2 S or GolS-␣2 R . Representative titration curves for each pair are shown in Fig.  3B. The curves for GolS/PgolB and GolS-␣2 R /PcopA did not reach saturation because these proteins aggregate at high concentrations. This behavior is evidenced by a sharp increase in the anisotropy values induced by the contribution of light scattering from higher molecular weight particles to the measured anisotropy at protein concentrations higher than those reported. The saturation levels of anisotropy observed in both CueR variants (that reached saturation) are 0.102 and 0.133. The values obtained from fitting the available data on the complexes with GolS variants are 0.133 and 0.147, which are close to the values obtained for CueR. Assuming that similar kinds of complexes are formed by both proteins, we can conclude that the extrapolated saturation values are in good agreement with the physical system. With all of this information, the dissociation equilibrium constants (k D ) were calculated for the interaction of the CueR/PcopA and GolS-␣2 R /PcopA as 141 Ϯ 15 and 198 Ϯ 49 nM, respectively. Similarly, the estimated k D values for GolS/PgolB and CueR-␣2 S /PgolB interactions were 12 Ϯ 3 and 52 Ϯ 12 nM, respectively. The equilibrium constants for the GolS/PcopA, CueR-␣2 S /PcopA, CueR/PgolB, and GolS-␣2 R /PgolB pairs could not be estimated because the low affinity of these interactions precluded the acquisition of binding data at protein saturating concentrations and thereby hindered fitting the experimental points (Fig. 3B). Nevertheless, previously estimated binding affinities for the GolS/PcopA and CueR/ PgolB heterologous interactions (by resonant mirror biosensor technology) were at least 1 and 2 orders of magnitude lower than those of the native CueR/PcopA and GolS/PgolB interactions, respectively (15). These data indicate that, compared with the parental proteins, both the CueR-␣2 S and the GolS-␣2 R mutants exhibit stronger binding affinities for the otherwise heterologous promoters in detriment of the native target sequences, resembling the paralog regulators. Overall, these results indicate that the regulator/operator specificity resides in the ␣2-helix of these transcription factors.
The Amino Acid Residues at Positions 16 and 19 of the Regulator Determine Target Recognition-We analyzed the sequence differences within CueR or GolS ␣2-helix in order to identify the amino acid residues that direct operator recognition specificity (Fig. 4A). Considering that different members of the MerR family employ a similar DNA distortion mechanism for transcriptional activation (reviewed in Refs. 1 and 22-24), we also included in the analysis the sequence for the predicted ␣2-region from MtaN, BmrR, and SoxR, non-metal binding MerR proteins for which detailed structural information of the  k A ϫ x)) by nonlinear regression. Association equilibrium constants (k A ) for each sensor/operator duplex are indicated. The binding isotherm curves are shown at the bottom. Dissociation equilibrium constants (k D ) were determined, expressing normalized anisotropy values ((k A ϫ x)/(1 ϩ k A ϫ x)) against protein concentration at logarithmic scale. JULY 12, 2013 • VOLUME 288 • NUMBER 28

JOURNAL OF BIOLOGICAL CHEMISTRY 20515
regulator bound to their target DNA sequences is available (3)(4)(5), as well as the predicted region from the metal-responsive regulators MerR and ZntR. We focused on those residues that, according to the crystallographic studies performed on MtaN, BmrR, and SoxR, could establish hydrogen bonds or van der Waals contacts with DNA (residues at positions 15,16,18,19, and 20 relative to CueR (Fig. 4A)). We observed that the Tyr-20 of CueR is conserved in these proteins, and the Ile-17 and Arg-18 residues are present in all metal-binding sensors (and not in the homologs responding to other signals), whereas the Lys residue at position 15 of both CueR and GolS is not conserved in other metal sensors. The residues at positions 16 and 19 differ between GolS and CueR. Thus, we carefully examined the identity of these residues in the different CueR-like and GolS-like homologs that we previously characterized for their operator selectivity (15). As shown in Fig. 5, all GolS-like proteins that were shown or proposed to recognize gol-like operator sequences harbor a conserved Met residue at position 16, whereas Ala, Ser, or Thr, but not Met, is present in all CueR-like proteins. The identity of the residue at position 19 was less conserved, but interestingly, the closer CueR homologs harbor a Phe at this position, whereas close GolS xenologs have a Tyr.  In view of these observations, we constructed single and double mutant versions of CueR and GolS at these positions, replacing GolS residues at position 16 and/or 19 by those present in CueR and vice versa and assayed their ability to activate transcription from PcopA and PgolB in the presence of copper. Although residues at positions 14 and 22 are not predicted to interact with the DNA, we also analyzed the role of these residues in the selective operator recognition because they differ between CueR and GolS (Fig. 4A).
The activation profile driven by either CueR A16M/F19Y or GolS M16A/Y19F mutants resembled the ␣2-variants of CueR and GolS, respectively. These mutants have an improved induction of transcription from the heterologous promoter and a diminished induction from their innate target promoter compared with the wild-type regulators (Fig. 4B). Furthermore, CueR A16M and GolS M16A displayed patterns of transcriptional induction similar to the double mutant proteins. Replacement of the residue at position 19 (CueR F19Y and GolS Y19F ) had only minor effects on their operator recognition pattern (Fig. 4B). These results clearly indicate that the identity of the residue at position 16 is a main determinant of regulator/operator selectivity among gol and cue regulons. Moreover, our observations also pointed out that the residue at position 19 finely tunes the selectivity. As expected, mutant regulators with replacements at positions 14 and 22 displayed wild-type patterns of transcriptional induction (Fig. 4B), indicating that these residues are not involved in operator discrimination.
The Residues at Position 16 and 19 of the Regulator Recognize the Selective Operator Bases 3Ј and 3-There is currently no structural information about the target operator recognition by MerR metalloregulators. Thus, assuming that all MerR homologs interact with DNA in a similar manner, we used the available crystallographic structure of the drug-binding homolog BmrR bound to its target promoter (3) to simulate the interaction of CueR or GolS with DNA. We selected BmrR because, like CueR and GolS, it recognizes target promoters with a 19-bp spacer between the Ϫ35 and Ϫ10 elements and interacts with dyad-symmetric sequences separated by 1 bp (4,15). In the models, we replaced the amino acid residues present in the N-terminal region of BmrR (from residue 1 to 88) by those present either in CueR, CueR-␣2 S , or GolS. As shown in Fig. 4C, the side chain of the residue at position 16 in both CueR and GolS approached the DNA backbone toward the nucleotide bases located at position 3 from the center of the operator (15). Substitution of the small methyl side chain of the alanine by a bulkier methionine will probably shift the position of the HTH motif with respect to the operator DNA. The side chain of the residue at position 19 was also oriented toward the interface between nucleotide bases at positions 3 and 4, supporting the experimental data about the role of the residue at position 16 in directing selectivity in the operator recognition.
To verify the role of amino acid residues at positions 16 and 19 in the distinction of the operator nucleotide bases at positions 3Ј and 3, we compared the promoter preferences of the mutant regulators CueR A16M/F19Y and GolS M16A/Y19F with the wild-type regulators, using innate PcopA and PgolB promoters or the mutated versions, PcopA AT and PgolB CC , in which the operator nucleotide bases at positions 3Ј and 3 were switched by those present in the heterologous operators (15) (also see Fig. 6A). Here again, we observed that the induction of expression from the wild-type and mutated gol and cue promoters by CueR A16M/F19Y is similar to that obtained using GolS, whereas the induction of the analyzed promoter by GolS M16A/Y19F mimicked that of wild-type CueR (Fig. 6B). In other words, CueR A16M/F19Y was more efficient in activating expression from PcopA AT than from wild-type PcopA. Conversely, it better recognized wild-type PgolB, which harbors A and T at position 3Ј and 3, respectively, than PgolB CC , which harbors C at these two positions, exhibiting a pattern of induction similar to GolS. Similarly, GolS M16A/Y19F better recognized promoters with Cs at positions 3Ј and 3, such as the native PcopA or the mutant PgolB CC promoters, than promoters with A and T at these positions (i.e. PgolB and PcopA AT ). Overall, these results established the importance of the amino acid residue at position 16 and, to a lesser extent, at position 19 in the selection of the Salmonella cue and gol regulons target operators.

DISCUSSION
Transcription factors must be capable of locating their specific target sequences along the chromosome, avoiding their unproductive or even harmful interaction at ectopic places. This appears particularly relevant when similar regulatory proteins with almost identical target recognition operators coexist in a single cell. Salmonella has two structurally related but functionally distinctive metalloregulators of the MerR family that orchestrate the cellular response to the presence of toxic amounts of monovalent metal ions in the environment (25). By rewiring both the input signal detection and the recognized operator sequences, the horizontally acquired gold sensor GolS is able to induce the expression of its target genes without interfering with the function of the copper homeostasis cue regulon, controlled by the enterobacterial ancestral regulator CueR. In a previous report, we demonstrated that selectivity in the recognition of GolS or CueR target operators is achieved by subtle modifications of the operator sequences (regulon signature nucleotide bases are displayed at the 3 and 3Ј operator positions (15)), which are accompanied, as is shown in this work, by subtle modifications at the DNA-binding motif of these transcriptional regulators.
We constructed a set of hybrid proteins between GolS and CueR to identify the region that directs operator recognition ( Fig. 1 and 2). These studies allowed us to focus on the ␣2-helix (from residue 14 to 22), which can be defined as the minimal region necessary for operator discrimination among the gol and cue regulons. In vitro experiments confirmed these observations, showing that the solely replacement of this motif in each regulator lowered its affinity for its innate promoters and increased its affinity for the paralog operators (Fig. 3).
There is no structural information on any metal-sensing MerR regulator bound to its target operator, and based on the available biochemical and genetic data, it was postulated that the DNA distortion mechanism for transcriptional activation is conserved among all family members (reviewed in Refs. 1 and 22-24). Therefore, we assumed that CueR and GolS interact with their target sequences in a similar manner as the nonmetal binding MerR homologs BmrR and MtaN, from which the crystallographic structure of protein-DNA complexes is available (3,5). In other words, the axis of symmetry of the CueR/GolS dimer is facing the minor groove at the center of the palindrome (position 1), whereas the ␣2-helix approaches the DNA at the second minor groove near the adjacent major grooves, where the signature 3 and 3Ј bases are located. An in silico modeling performed for both CueR and GolS pointed out residues 16 and 19 as candidates for directing selectivity toward target operators (Fig. 4C). Amino acid swapping demonstrated that residue 16 is key in directing the selective recognition of the signature operator base 3 (Fig. 4). In the presence of copper ions, the mutant CueR A16M induced the expression of PgolB::lacZ and PcopA::lacZ like GolS (Fig. 4B). A similar trend of operator switch was detected with GolS M16A . PcopA::lacZ achieved higher levels of induction with GolS M16A than with GolS, whereas there was a lower metal-dependent induction of PgolB::lacZ with GolS M16A than with the native gold sensor. On the other hand, the contribution of the residue at position 19 was less notorious per se. CueR F19Y and GolS Y19F regulators exhibited patterns of metal induction intermediate between the parental regulator and the mutant that harbors the ␣2-region replaced. Nevertheless, their contribution to fine tuning the regulator/operator interaction is corroborated by the observation that double mutant regulators CueR A16M/F19Y and GolS M16A/Y19F were as effective as the hybrid CueR-␣2 S and GolS-␣2 R , respectively, to activate the promoters controlled by the paralog regulator (Fig. 4B). In addition, CueR A16M/F19Y and GolS M16A/Y19F activated transcription of the reporter gene from the promoter with its 3Ј and 3 bases replaced more efficiently than from the wild-type promoters (Fig. 6). We hypothesize that the failure of GolS to recognized operators bearing C/G substitutions at positions 3 and 3Ј could be explained by the formation of a third hydrogen bond between the pairing CG bases, which renders a base pair less deformable than the AT base pair. This would result in either steric or electrostatic interference with the interaction of GolS, which has the more voluminous yet hydrophobic methionine residue at position 16.
The presence of the distinctive Met residue at the ␣2-helix of the DNA-binding domain of GolS is extended to all GolS-like proteins having the 3Ј-3 AT signature in the operators of genes predicted to be controlled by them (Fig. 5) (also see Ref. 15). By contrast, CueR homologs that recognize promoters with the signature (C/G)(C/G) have Ala, Ser, or Thr at position 16, and indeed, replacement of the Ala-16 residue of CueR by Thr did not affect its operator selectivity (Fig. 4B). As expected, the residue at position 19 is not conserved, but interestingly, close GolS homologs harbor a Tyr at this position, whereas all CueR xenologs have a Phe. Indeed, the recently characterized Cupriavidus metallidurans CH34 gold sensor CupR, which controls genes with the characteristic 3Ј-3 AT operators (26), harbors the GolS-like Met-Phe signature at the N-terminal DNA-binding domain. Notoriously, several MerR homologs that recognized one central base pair separation between the pseudopalindromic sequences and have the 3-3Ј signature (C/G)(C/G) like CueR (e.g. the non-metal sensors BmrR, Mta, and TipA and the metal sensors CadR and PbrR) have Ala or Thr at position 16, but not Met, and have either a Phe or a Tyr at position 19 (4). Therefore, it is evident that specific operator recognition in CueR-like and GolS-like regulators relies exclusively on the interaction of the amino acid residue at position 16 and the signature nucleotide base at position 3 of the operator.
Overall, these studies suggest that, along evolution, rewiring of both the transcriptional regulators GolS and CueR and the regulatory elements in their target genes confers novel abilities to detect distinct environmental cues, avoiding at the same time cross-regulation that would jeopardize the adequate response to a specific stress.