Insights into Protein Allostery in the CsoR/RcnR Family of Transcriptional Repressors

Abstract

CsoR/RcnR transcriptional repressors adopt a disc-shaped, all α-helical dimer of dimers tetrameric architecture, with a four-helix bundle the key structural feature of the dimer. Individual members of this large family of repressors coordinate Cu(I) or Ni(II)/Co(II) or perform cysteine sulfur chemistry in mitigating the effects of metal or metabolite toxicity, respectively. Here we highlight recent insights into the functional diversity of this fascinating family of repressors.

Introduction

CsoR (copper-sensitive operon repressor)¹ and RcnR (resistance to cobalt and nickel repressor)² are founding members of a new structural family of transcriptional repressors that feature a four-helix bundle and adopt an all α-helical dimer of dimers structure.¹ Other functionally characterized members of this family include RicR (regulated in copper repressor),³ DmeR (divalent metal efflux repressor),⁴ InrS (internal nickel-responsive repressor),⁵ CstR (CsoR-like sulfur transferase respressor),⁶ and FrmR (a formaldehyde responsive repressor).⁷ The metal sensing members of this family drive transcriptional derepression of genes encoding metal effluxers upon direct coordination of a cognate metal effector, thus far limited to Cu(I) or Ni(II)/Co(II). All members of this family have a signature fingerprint, W-X-Y-Z, named for the ligands that coordinate the metal in the metal-sensing members of this family, (Figure 1).^8,9 In the case of CsoR, the fingerprint is x-C-H-C (where x is any amino acid), whereas in RcnR it is H-C-H-H with residues in the exact analogous positions relative to CsoR in a multiple sequence alignment (Figure 1). There are other members or clades of this large class of repressors that clearly define new functional signature motifs, but have not yet been characterized.⁵

Figure 1.

Sequence alignment of CsoRs from M. tuberculosis, B. subtilis, T. thermophiles, S. aureus, L. monocytogenes, S. lividans, and C. glutamicum and G. thermodenitrificans with M. tuberculosis RicR, E. coli RcnR, R. Leguminosarum DmeR, Synechocystis PCC 6803 InrS, S. aureus CstR, and E. coli FrmR. Fingerprint residues found in both CsoR and RcnR^8,9 are shown in orange while those that are unique to CsoR and RcnR are in blue and green, respectively. The residues that appear to play an important role in allosteric switching in M. tuberculosis CsoR^1,10 are shown in red. The corresponding residues are also shaded in other proteins. This sequence alignment was generated using ClustatalW2.¹¹

CsoR

CsoR is a Cu(I) responsive transcriptional regulator.¹ CsoR proteins and have been characterized using biological, structural and biophysical methods. In the apo form, CsoR represses the transcription of the cso operon by binding a GC rich pseudopallindromic sequence (5’-GTAGCCCACCCCCAGTGGGGTGGATAC-3’) that overlaps the putative −10 and −35 regions of the cso promoter.¹ The cso operon encodes CsoR, an uncharacterized middle gene and a Cu(I) effluxing P-type ATPase, CtpV.¹ CsoR-like repressors are present in all five major classes of eubacteria¹ and have been characterized in Bacillus subtilis (B. subtilis),^12–14 Thermus thermophilus HB8 (T. thermophilus),¹⁵ Staphylococcus aureus (S. aureus),^6,16 Listeria monocytogenes (L. monocytogenes),¹⁷ Streptomyces lividans (S. lividans),¹⁸ Corneybacterium glutamicum (C. glutamicum),¹⁹ and Geobacillus thermodenitrificans (G. thermodenitrificans) (Figure 1).²⁰

The first crystal structure of CsoR from M. tuberculosis was published in 2007 by Liu et al.¹ This structure revealed that CsoR is an all α-helical protein with a four-helix bundle architecture (α1-α2-α1’-α2’) flanked by C-terminal α3/α3’ helices that define much of the tetramer interface. CsoR binds one Cu(I) per protomer in an interprotomer trigonal S₂N complex via two cysteine residues Cys36 (α2-helix) and Cys65’ from opposite subunit within the dimer and a histidine residue from the same subunit as Cys65’, His61’ (α2-helix) (Figures 2A and 3A).¹ Additionally, the crystal structure suggested a hydrogen-bonding network between Nε2 of His61’, and Glu81’ from one subunit and Tyr35 from another subunit (Figure 2B).¹ Subsequent NMR studies conducted on G. thermodenitrificans CsoR in both the apo and Cu(I)-bound states revealed that Cu(I) binding results in an discontinuity or kink in the long α2-helix located between the Cu(I) binding residues His75 and Cys79.²⁰ Crystal structures of apo-CsoRs have more recently been solved from S. lividans¹⁸ (Figure 2C) and T. thermophilus.¹⁵ These structures reveal a similar architecture, with the S. lividans structure characterized by a continuous α2 helix, just as in apo G. thermodenitrificans CsoR in solution.²⁰

Figure 2.

Representations of effector binding to CsoR/RcnR proteins. Fingerprint residues are denoted, W, X, Y and Z. (A) Cognate metal site structures of CsoR, RcnR, and InrS. (B) Intersubunit disulfide and selenotrisulfide formed when CstR is reacted with selenite; the analogous disulfude and trisulfide products are formed with sulfite.⁶ (C) Proposed mechanism for the reaction of FrmR with formaldehyde based on known reactions of cysteine derivatives with formaldehyde.²¹ The amino group shown could correspond to a Lys residue or the α-amine.

Figure 1.

(A) Cu(I) binding site from M. tuberculosis CsoR; Cu(I) (orange) (PDB code 2HH7)¹ is coordinated in a trigonal geometry by Cys36, His61’ and His65’ (X-Y-Z). The N-terminus of CsoR is extended to include W, which corresponds to His3 of RcnR. (B) The hydrogen bonding network involving His61’, Tyr35 and Glu81’. (C) Apo CsoR from S. lividans (PDB code 4ADZ),¹⁶ showing the all α-helical dimer of dimers architecture. (D) A homology model of a B. subtilis CsoR onto the crystal structure of apo S. lividans. The lysine residues are shaded in blue, the projected Cu(I) binding residues are shown in red and second shell hydrogen-bonding residues are shaded in yellow.

Mutagenesis studies and histidine analog substitution experiments coupled with DNA binding studies reveal clearly that this hydrogen bonding network is important in linking Cu(I) binding to CsoR with DNA release in M. tuberculosis and B. subtilis CsoRs. Initial work done by Liu et al. on M. tuberculosis CsoR determined that Glu81 was important for Cu(I)-dependent regulation of DNA binding.¹ They showed that an E81A mutation resulted in a protein that binds Cu(I) with an affinity similar to that of the wild-type protein but was compromised in the regulation of DNA binding.¹ This effect was also confirmed by Ma et al. with the analogous E90A mutation in B. subtilis CsoR.¹² Further studies on M. tuberculosis CsoR using unnatural substitutions of His61 to Nε2-methyl-histidine (MeH) or (4-thiazolyl)-L-alanine (Thz) showed that Cu(I) binding affinity was wild-type like but CsoR-DNA interactions were no longer significantly regulated by Cu(I) binding.¹¹ Additionally, mutating the other two residues involved in the hydrogen bonding network, Tyr35 to Phe and Glu81 to Ala, Gln, Asp, and Asn had no effect on Cu(I) binding but these mutations resulted in a decrease in the allosteric coupling free energy, ΔG_c. Collectively, these studies highlight the critical role this hydrogen-bonding network plays in driving negative allosteric regulation of DNA binding by Cu(I) in M. tuberculosis CsoR. Although the degree to which this hydrogen-bonding network (Figure 2B) is functionally important in other distantly related Cu(I)-sensing CsoRs is not yet established, the Tyr is invariant, while the Glu is highly conserved in all Cu(I)-sensing CsoRs.

Like M. tuberculosis CsoR, B subtilis CsoR also binds Cu(I) with a trigonal planar geometry and a S₂N ligand set.¹² DNA binding studies determined that the protein binds its operator with a stoichiometry of two tetramers per DNA.¹² The protein binds Cu(I) as well as Ni(II), Zn(II) and Co(II) with binding affinities of ≈10¹⁹, 10⁹, 10⁸ and ≤10⁵ M⁻¹, respectively.¹² The noncognate metals adopt coordination geometries that are distinct from Cu(I), with Ni(II) adopting a square-planar-like coordination geometry and Co(II) forming a tetrahedral or distorted tetrahedral complex.¹² The metal binding affinities for B. subtilis CsoR reveal that there is a strong thermodynamic preference for Cu(I) binding; however, ΔG_c determined for both cognate and non-cognate metal ions reveals that the low coordination number and trigonal geometry of the metal complex is strongly linked to metal responsiveness in B. subtilis CsoR.

The stoichiometry of CsoR:DNA operator binding is 2:1¹² although details remain elusive as to how CsoR binds to DNA. However, pulse-chase amidination experiments used to measure the change in residue-specific lysine accessibility in Cu(I)-bound and DNA-bound complexes relative to the reference apostate reveal significant protection of the C-terminal Lys residues that localize to the center of the tetramer (K96, K97 and K100) on DNA binding, as well as those near the N-terminus of the α1 helix in B. subtilis CsoR (Figure 2D).²² These studies suggest that there is a remodeling of the CsoR tetramer upon Cu(I) binding that perturbs these regions of protection and ultimately leads to dissociation of CsoR from the DNA operator.

RicR

RicR is a second Cu(I)-responsive transcriptional repressor to be identified in M. tuberculosis.³ RicR regulates a five-operon regulon vs. a single operon for CsoR and thus may have a larger regulatory impact on the bacterium under Cu(I) stress. Many of the genes under the transcriptional control of RicR are unique to pathogenic mycobacteria and include MymT, a copper-protective metallothionein; LpqS, a putative lipoprotein; Rv2963, a putative permease; the socAB (small ORF induced by copper A and B) operon; a putative multicopper oxidase, and the ricR gene itself.³ A palindromic sequence, 5’-TACCC-N₅-G/AGG-TA-3’ promoter sequences, and is likely the site of RicR binding.³ Like CsoR, RicR possesses all three Cu(I) binding residues of CsoR (Cys38, His63 and Cys67, Figure 1) and is predicted to adopt the same overall fold as CsoR. RicR is a tetramer in the presence and absence of Cu(I) (F. Chang, B. Kester, D. Giedroc, unpublished) and retains all the second coordination shell residues that potentially interact with His63. It is currently unknown why M. tuberculosis possesses two CsoR-like Cu(I)-specific sensors, but this could be linked to differential set-points or cellular sensitivities for Cu(I) sensed by CsoR vs. RicR. Pathogenic mycobacterial strains encode one additional CsoR/RcnR repressor, e.g., M. tuberculosis H37Rv Rv1766, that has not yet been characterized but bears strong sequence similarity to CstR (see below) but may well be representative of a functionally distinct CsoR subfamily.

RcnR

E. coli RcnR is a 40 kDa tetramer that regulates the expression of the Ni(II) and Co(II) exporter, RcnA and an associated periplasmic protein, RcnB.^2,23 RcnR coordinates a variety of metal ions but only the binding of Ni(II) and Co(II) results in conformational changes that allosterically negatively regulate DNA binding interactions.⁸ The protein binds one Ni(II) or Co(II) per protomer with a binding affinity of 4×10⁷ and 2×10⁸M⁻¹, respectively.⁸ Iwig et al. determined that RcnR recognizes a TACT-G₆-N-AGTA sequence of which there are two in the rcnA-rcnR intergenic regions.²⁴ RcnR also interacts nonspecifically with the flanking DNA regions (ca. 50 base pairs) leading to some DNA wrapping.²⁴

XAS and UV-vis studies determined that RcnR forms six-coordinate complexes with its cognate metals, which adopt a (N/O)₅S ligand environment (Figure 3A) and three-/four- and four-coordinate complexes with non-cognate metals, Cu(I) and Zn(II), respectively.^8,25 Mutagenesis studies coupled with lacZ assays and XAS determined that the cognate metal ion binding could be distinguished from non-cognate metal ion binding as the former involved the direct coordination by the α-amine.²⁵ RcnR is characterized by a H-C-H-H metal binding motif, different from that observed in CsoR (Figure 1). The metal binding sites for the cognate metal ions differ in their M-S bond distances as well as the number of imidazole ligands. For Co, the Co-S bond is ca. 2.3 Å, whereas in the case of the Ni(II), the Ni-S bond is ca. 2.6 Å (Figure 3A).^8,25 Additionally, Co(II) binds one more imidazole ligand than Ni(II) (ca. 3 for Co(II) and ca. 2 for Ni(II)).²⁵ LacZ expression experiments determined that His3 was essential for coupling cognate metal binding with the dissociation from DNA; however, XAS data revealed that His3 is a ligand for Co(II) but not for Ni(II).^8,25

A H3E mutation in RcnR altered the metal response resulting in an RcnR that was still responsive to Ni(II) and Co(II) but now also responsive to non-cognate Zn(II).²⁵ This substitution transforms the four-coordinate Zn(II) site observed in wild-type RcnR to an unusual six- or seven- (bidentate carboxylate) site with (N/O)_5–6S ligand donor set, resembling the Ni(II) wild-type RcnR with a Zn(II)-S bond distance of 2.61 Å.²⁵ These findings suggest that His3 might be considered a “negative evolutionary design element” the integrity of which prevents illegitimate activation by non-cognate Zn(II). Remarkably, a similar conclusion was reached for His3 in the Ni(II) sensor M. tuberculosis NmtR, a member of the ArsR-family of repressors that adopt an entirely different fold but, like RcnR, seems to incorporate an α-amine and His3 into the Ni(II) coordination complex.^26,27

RcnR possesses three other His residues, His60, His64 and His67, the first two of which are equivalent to His61 and Cys65 of M. tuberculosis CsoR (Figure 1). These residues were mutated to Cys residues with the expectation that if these residues were ligands to the metal sites then the additional S donor ligand would be apparent in the EXAFS spectrum.²⁸ The His60C and His67C substitution mutants were still responsive to Ni(II) and Co(II); however, the Co(II) site resembled that of the Ni(II) wild-type as it features a single Co-S bond with coordination distance of ca. 2.6Å.²⁸ Interestingly, the H64C mutant protein was not responsive to Cu(I) and the Cu(I) site remained a four-coordinate (N/O)₂SBr complex as observed in wild-type RcnR, but distinct from the S₂N ligand set observed in a bona fide CsoR.^1,28 This finding reveals that other features beyond the first coordination site formed by the correct Cu(I) ligands is required to elicit non-cognate Cu(I)-mediated allostery in RcnR that leads to DNA dissociation. Finally, H64C RcnR was found to be selective for Co(II) over Ni(II). Co(II) coordinates the additional S-donor provided by this substitution in an (N/O)₃S₂Br ligand set while Ni(II) site was 4-/5-coordinate with (N/O)_2–3SBr.²⁸

DmeR

Like RcnR, Rhizobium leguminosarum UPM791 (R. leguminosarum) DmeR is a Ni(II)- and Co(II)-responsive transcriptional regulator that regulates the transcription of the dmeRF operon.⁴ The dmeRF operon is composed of two genes, dmeR and dmeF, the latter of which encodes a putative cation-diffusion facilitator (CDF) transporter.⁴ DmeR has an RcnR-like fingerprint, H-C-H-H, and is therefore expected to bind Ni(II) and Co(II) similar to that of RcnR. Like CsoR and RcnR, DmeR is predicted to harbor three α-helices and adopt a dimer of dimers tetrameric architecture.⁴

InrS

InrS is a Ni(II)-responsive transcriptional repressor from Synechocystis PCC6803.⁵ InrS conserves all the Cu(I) ligands identified in M. tuberculosis CsoR: Cys53 and Cys82 (equivalent to Cys36 and Cys65 in M. tuberculosis CsoR) and His78 (equivalent to His61 from M. tuberculosis CsoR) although it is unknown if all three are required for Ni(II) coordination Ni(II)-mediated allostery. InrS also specifically lacks the hydrogen-bonding tyrosine residue that is characteristic of Cu(I)-sensing CsoRs (Figure 1). InrS, like CsoR and RcnR, bind Co(II), Ni(II), and Zn(II) with distinct metal coordination geometries with the coordination geometries of these metals in InrS similar to that observed for B. subtilis CsoR.^5,12 Co(II) and Ni(II) bind to InrS with tetrahedral and square-planar or distorted square-planar geometries, respectively.⁵ The fact InrS has a similar ligand set as CsoR but specifically lacks the second-shell Tyr next to the X-Cys (Figure 1) and is Ni(II)-responsive suggests that some aspect of this residue, either directly or indirectly, makes Cu(I)-sensing CsoRs more specific for Cu(I) over other bioavailable transition metals.

CstR

S. aureus strain Newman encodes two CsoR/RcnR family repressors.⁶ CstR is a paralog of a bona fide Cu(I)-sensing CsoR in S. aureus and regulates the expression of the cst operon, which encodes genes that may be associated with sulfur assimilation/trafficking and/or sulfide detoxification. CstR binds two operator sites with a stoichiometry of two tetramers per operator and an average tetramer DNA association equilibrium constant, K_tet of ≈2.7 × 10⁸ M⁻¹.⁶ S. aureus CstR is 35% identical and 65% similar to S. aureus CsoR and is characterized by an x-C-x-C fingerprint that does not bind Cu(I).⁶ Instead, disulfide and trisulfide linkages are formed between the two Cys residues, Cys31 (X) and Cys60’ (Z) on opposite protomers within the dimer as revealed by tandem mass spectrometry when reacted with sodium sulfite anaerobically (Figure 1).⁶ The formation of these crosslinked species markedly reduces DNA binding affinity. Similarly, CstR will form disulfides, selenotrisulfides, tellurotrisulfides when reacted with the calcogen oxyanions selenite and tellurite, respectively (Figure 3B).²⁹ These crosslinked products were not observed in CsoR.^6,29 Work done by Luebke et al. showed that Cys31 initiates the reaction with sulfite through the formation of S-sulfocysteine (Figure 3B).²⁹ They also determined that modify-cation of Cys31 (in a Cys60A background) with the cysteine-modifying reagent methylmethanethiosulfonate (MMTS) resulted in a decrease in DNA binding affinity while similar modifications to Cys60 had no effect on DNA binding affinity.²⁹ Collectively, these studies show that CstR and CsoR function via different derepression mechanisms and that Cys31 is the functionally important cysteine residue in CstR.

The actual in vivo inducer(s) of the cst operon has not yet been unambiguously identified, although microarray experiments reveal that hypochlorous acid³⁰ and sodium nitrite³¹ in biofilm-favoring conditions are potent, albeit non-selective inducers. It is interesting to note that an operon encoding a similar set of genes that is regulated by an ArsR/SmtB-family protein (BigR) in the plant pathogen Xylella fastidiosa is inducible by hydrogen sulfide under hypoxic, biofilm-promoting conditions.³²

Structural studies of CstR are complicated by self-association beyond the tetramer, to octameric and higher order assembly states; as a result, the reaction mechanism governing the formation of intersubunit disulfide, trisulfide and selenotrisulfide linkages (Figure 3B) as well as how these chemistries drive changes in CsoR structure that leads to transcriptional derepression is unknown. Nor is it clear how a nucleophilic Cys31 thiol(ate) in CstR is activated to react with negatively charged sulfite or selenite oxyanions; the involvement of a transiently formed sulfenated cysteine intermediate which would make Cys31 more electrophilic seems possible but not yet established.^33,34 In any case, a high resolution structure of CstR may provide insights into what may be an unusual microenvironment around Cys31.

FrmR

Formaldehyde-sensing E. coli FrmR is characterized by a x-C-H-x fingerprint. FrmR regulates an operon that encodes frmR itself, frmA, which encodes a glutathione-dependent dehydrogenase and frmB, encoding a putative S-formylglutathione hydroxylase.⁷ Expression of the frmAB operon is induced by formaldehyde, but the degree to which other carbon electrophiles, e.g., methyl glyoxal, induce this operon is not known. It is interesting to note that B. subtilis and Streptococcus pneumoniae encode single-cysteine containing MerR family activators that can be induced by electrophile stress via an unknown mechanism.^35,36 Although it is not yet known how FmrR mediates formaldehyde sensing, formation of an S-formyl adduct via Micheal addition by the “X” Cys in FrmR is a reasonable possibility, which upon further reaction with a nearby primary amine could result in the formation of a thiazolidine-like adduct (Figure 3D).⁷ It is unknown if FrmR binds metals. In any case, understanding the structural basis for formaldehyde responsiveness in FrmR will be of high interest.

Concluding Remarks

As the number of characterized CsoR/RcnR repressors increases, trends are beginning to emerge in metal or inducer selectivity. The work on mutant RcnRs²⁸ and InrS⁵ is consistent with the idea, first raised for Cu(I)-sensing CsoRs,¹⁰ that factors beyond the first metal coordination sphere and/or differing bioavailable concentrations of competing metals in the cell,⁵ may well dictate inducer selectivity in this family of proteins. Remarkably, the two Ni(II) sensors thus far characterized appear to have evolved independently of one another, as they employ different metal coordination geometries and Ni(II) affinities^5,8,25 to perform the same biological function. A more extensive phylogenetic analysis than performed previously^5,8 with a statistical coupling analysis of covarying residues may well provide new insights into the evolution of this protein family as well as identify key residues that are important for protein allostery.³⁷

Like the ubiquitous ArsR/SmtB and MerR families of transcriptional repressors,⁹ nature has once again utilized a single protein scaffold to elicit a diversity of biological and regulatory outcomes, in this case employing an overlapping W-X-Y-Z fingerprint region to evolve functional specificity (Figure 1A). Metals bind to these proteins with a particular geometry and ligand set while non-metals exploit cysteine sulfur chemistry to allosterically negatively regulate DNA operator binding affinity. In contrast to the ArsR/SmtB and MerR families however, transition metals known to be sensed by CsoR/RcnR metal-losensors are thus far limited to Ni(II), Co(II) and Cu(I). As new members of this family are identified and characterized, one might anticipate that other metal and non-metal inducer specificities will arise within this family of repressors.

Since the CsoR/RcnR family repressors lack a winged helix and three-helix “homeodomain” bundle present in other bacterial transcriptional regulators, it remains unknown how any member of this family of proteins binds to DNA in the apo-state. Such a structure would aid in understanding the architecture of DNA-protein complex, detailed interactions at the interface, and how the binding of various effectors leads to conformational changes in the tetramer that drives DNA operator dissociation. Some clues have emerged from mutagenesis and pulsed chase amidination studies with CsoRs²² but how this or other protein-DNA interfaces are disrupted upon metal coordination or sulfur chemistry promises to provide important clues into the evolution of this class of transcriptional repressors.

Biographies

David Giedroc was born in central Pennsylvania (1959) and graduated from the Pennsylvania State University in 1980. He earned his Ph.D. in Biochemistry at Vanderbilt University School of Medicine in 1984 and following an NIH postdoctoral appointment at Yale University, he joined the Department of Biochemistry and Biophysics at Texas A&M University (1988). In 2007, he moved to the Department of Chemistry at Indiana University in Bloomington where he is currently Professor and Chair. He studies transition-metal homeostasis in bacteria, sulfur homeostasis, and assimilation in the major hospital pathogen Staphylococcus aureus and RNA structure and protein-RNA interactions important for the replication of human coronaviruses. He was elected a Fellow of the American Association for the Advancement of Science in 2013.

Khadine A. Higgins was born in St. Andrew, Jamaica (1982). She graduated in 2004 from Lawrence University in Appleton, WI with her B.A. degree in Chemistry. She completed her Ph.D. work (December 2011) at the University of Massachusetts, Amherst under the guidance of Professor Michael J. Maroney where she studied wild-type and mutant RcnR proteins using X-ray absorption spectroscopy (XAS) to examine the metal site structures. She started her postdoctoral research with Professor David Giedroc at Indiana University Bloomington in February 2012 where she studies a putative sulfur transferase from Staphylococcus aureus.