Rgg protein structure–function and inhibition by cyclic peptide compounds

Significance

Peptide pheromones regulate developmental processes, including virulence, in Gram-positive bacteria. Immature propeptide pheromones are synthesized, secreted, and undergo proteolytic maturation to serve as intercellular signals. The regulator gene of glucosyltransferase (Rgg) transcription factors are a large family of receptors that directly bind pheromones transported to the cytosol. Here we report X-ray crystal structures of a Streptococcus Rgg protein alone and complexed with cyclosporin A, which is a potent inhibitor of pheromone signaling. Based on these structures and extensive genetic and biochemical studies, we mapped the pheromone-binding site, discovered mechanistic aspects of Rgg regulation, and determined how cyclosporin A and its nonimmunosuppressive analog valspodar function to inhibit pheromone-mediated receptor activation. We conclude that similar compounds targeting bacterial pheromone receptors have potential for therapeutic applications.

Abstract

Peptide pheromone cell–cell signaling (quorum sensing) regulates the expression of diverse developmental phenotypes (including virulence) in Firmicutes, which includes common human pathogens, e.g., Streptococcus pyogenes and Streptococcus pneumoniae. Cytoplasmic transcription factors known as “Rgg proteins” are peptide pheromone receptors ubiquitous in Firmicutes. Here we present X-ray crystal structures of a Streptococcus Rgg protein alone and in complex with a tight-binding signaling antagonist, the cyclic undecapeptide cyclosporin A. To our knowledge, these represent the first Rgg protein X-ray crystal structures. Based on the results of extensive structure–function analysis, we reveal the peptide pheromone-binding site and the mechanism by which cyclosporin A inhibits activation of the peptide pheromone receptor. Guided by the Rgg–cyclosporin A complex structure, we predicted that the nonimmunosuppressive cyclosporin A analog valspodar would inhibit Rgg activation. Indeed, we found that, like cyclosporin A, valspodar inhibits peptide pheromone activation of conserved Rgg proteins in medically relevant Streptococcus species. Finally, the crystal structures presented here revealed that the Rgg protein DNA-binding domains are covalently linked across their dimerization interface by a disulfide bond formed by a highly conserved cysteine. The DNA-binding domain dimerization interface observed in our structures is essentially identical to the interfaces previously described for other members of the XRE DNA-binding domain family, but the presence of an intermolecular disulfide bond buried in this interface appears to be unique. We hypothesize that this disulfide bond may, under the right conditions, affect Rgg monomer–dimer equilibrium, stabilize Rgg conformation, or serve as a redox-sensitive switch.

Gene expression in bacterial populations is coordinated by pheromone-regulated cell-to-cell signaling networks. This intercellular communication, commonly referred to as “quorum sensing,” regulates diverse behaviors across the microbial world (1). Quorum sensing among Gram-positive bacteria is commonly mediated by peptide pheromones (reviewed in refs. 2 and 3). The pheromones either are detected at the cell surface by membrane-bound receptors or are transported across the membrane by oligopeptide permeases, whereupon the pheromones engage cytoplasmic receptors (Fig. 1A). Gram-positive cytoplasmic pheromone receptors include Bacillus response regulator aspartate phosphatases (Rap), neutral protease regulator (NprR), and phosphatidylinositol-specific phospholipase C gene regulator (PlcR), Enterococcus pheromone-responsive transcription factor (PrgX), and Streptococcus regulator gene of glucosyltransferase (Rgg) (as well as the homologous MutR and GadR) (4–11). The Rap proteins are phosphatases and transcriptional antiactivators, whereas NprR, PlcR, and PrgX are DNA-binding transcription factors. Structure–function studies revealed that Rap, NprR, PlcR, and PrgX (the RNPP family proteins) use a structurally similar C-terminal tetratricopeptide (TPR)-like repeat domain to bind their cognate peptide pheromones (12–19). The 3D structure of Rgg proteins was unknown; however, based on their functional similarity to NprR, PrgX, and PlcR and their remote sequence similarity to RNPP family proteins, Rgg proteins preliminarily were included in this group (4, 6, 7, 20).

Fig. 1.

Peptide pheromone signaling. (A) Intercellular vs. extracellular detection of peptide pheromones. Members of the RRNPP protein family modulate gene expression in response to direct binding of specific peptide pheromones that are translocated to the cytoplasm. Rgg proteins such as NprR, PrgX, and PlcR are DNA-binding transcription factors. Rap proteins govern gene expression indirectly through protein–protein interactions with other regulators, e.g., Spo0F (11) and ComA (5). The RRNPP proteins are depicted here as dimers, but Rap proteins were shown to be monomeric in solution (16, 44), and PrgX likely forms tetramers (18). In contrast to the RRNPP systems, extracellular pheromone detection occurs by two-component signal transduction (TCST) pathways that control transcription through phosphorylation of a response regulator. The different peptide colors highlight the fact that multiple pheromone types can be produced by single or multiple species. (B) SHP2_Sd, SHP2_Sp, and SHP3_Sp amino acid sequences.

Rgg proteins are widespread in Firmicute species, including but not limited to the Streptococcaceae, Lactobacillales, Listeriaceae, and Enterococcaceae (6). It also is common for organisms to express multiple paralogous Rgg proteins putatively serving nonredundant regulatory functions. For example, Streptococcus pyogenes, which contains a thoroughly studied Rgg regulatory system, expresses four Rgg paralogs: Rgg2_Sp, Rgg3_Sp, ComR_Sp, and RopB_Sp. Rgg2_Sp, Rgg3_Sp, and ComR_Sp are transcription factors whose activity is regulated via interactions with pheromones (4, 21, 22). RopB is a transcription factor as well, and its role in pathogenesis has been thoroughly documented; however, the exact identity of RopB’s cognate regulatory pheromone has not been determined (23–25). Thus far, there are two known families of Streptococcus peptide pheromones, the SigX-inducing peptides (XIPs) and the short hydrophobic peptides (SHPs). In S. pyogenes, Rgg-XIP and Rgg-SHP pairs regulate diverse developmental processes, including biofilm formation and induction of a cryptic competence regulon (4, 22).

Rgg2_Sp and Rgg3_Sp are the most similar of the four S. pyogenes Rgg paralogs. In fact, Rgg2_Sp and Rgg3_Sp bind to the peptide pheromones SHP2 and SHP3 (Fig. 1B) with similar affinities, and Rgg2_Sp and Rgg3_Sp bind to identical DNA-regulatory elements upstream of their target promoters (21, 26). The net response of S. pyogenes to SHP pheromone is the robust expression of Rgg2/3-controlled promoters; however, Rgg2_Sp and Rgg3_Sp affect transcription by different mechanisms. More specifically, Rgg3_Sp represses transcription in the absence of pheromone, and pheromone binding triggers derepression by dissociating Rgg3_Sp from DNA. Conversely, Rgg2_Sp induces transcription only when bound to an SHP pheromone (4, 27). Therefore, these regulators work systematically to down-regulate target gene transcription in the absence of pheromone, and they up-regulate transcription in the presence of pheromone. Although exceptions almost certainly exist, Rgg proteins whose amino acid sequences are more similar to Rgg3_Sp than to Rgg2_Sp generally function as repressors, and Rgg proteins whose sequences are more similar to Rgg2_Sp than to Rgg3_Sp function as SHP-dependent activators (28, 29).

Here we report X-ray crystal structures of Streptococcus dysgalactiae Rgg2 (Rgg2_Sd) alone and in complex with an inhibitor, the cyclic undecapeptide cyclosporin A (CsA). To our knowledge, these are the first Rgg protein X-ray crystal structures reported. In addition to identifying the SHP-binding site, the structural, genetic, and biochemical studies presented here enabled us to show how CsA and its nonimmunosuppressive analog valspodar function to inhibit SHP-mediated regulation of Rgg activity in both S. pyogenes and S. dysgalactiae. Based on these results, and because Rgg-SHP signaling systems regulate diverse developmental responses in Firmicutes, which includes widespread human and animal commensal and pathogenic bacteria, we conclude that Rgg-modulating compounds similar to those described here have potential for therapeutic application.

Results

Rgg2 X-Ray Crystal Structure.

The X-ray crystal structure of full-length Rgg2_Sd alone was determined to a resolution of 2.05 Å (Fig. 2 and Table S1). There are four Rgg2_Sd protomers (protomers A, B, C, and D) in the crystallographic asymmetric unit (Fig. S1). Gel filtration analysis established that Rgg2_Sd forms homodimers in solution (Fig. S2), and the noncrystallographic dimer interface (crystallographic interfaces A–B and C–D) is large, burying more than 5,600 Ų of surface area (Fig. S1).

Fig. 2.

Rgg2_Sd crystal structure. (A) The Rgg2_Sd dimer formed by protomers A (Rgg2_A) and B (Rgg2_B). The α-helices are depicted as cylinders. The repeat domain α-helices are labeled according to the following convention: R1_A is the A helix of HTH repeat 1, and R1_B is the B helix of the HTH repeat 1. For simplicity, the protomer A repeat domain α-helices are not labeled. (B) Surface representation of the Rgg2_Sd dimer. To obtain this view of the SHP- (and CsA)-binding surface, the Rgg_Sd dimer depicted in A was rotated as indicated by the arrow. The protomer B residues highlighted in green displayed reduced SHP2-dependent activity in vivo. The corresponding residues in protomer A are not visible in this orientation.

Rgg2_Sd monomers within a homodimer are related by approximate twofold noncrystallographic symmetry, and the monomers are domain swapped (Fig. 2). More specifically, each monomer consists of an N-terminal DNA-binding domain (DBD) (residues 1–65) connected by a short linker region (residues 66–70) to a large C-terminal repeat domain (residues 71–284). Both the N- and C-terminal domains mediate contacts across the dimer interface, and the N- and C-terminal domains are swapped around the approximate twofold axis.

Comparison of the Rgg2_Sd DBD structure with all previously determined structures in the Protein Data Bank (PDB) database showed that the Rgg2_Sd DBD structurally is most similar to members of the XRE family of helix-turn-helix (HTH) DBDs (30, 31). Members of this family of DBDs contain five α-helices (Fig. 2A and Fig. S3) in which helices α2 and α3 and the linker connecting them form the DBD HTH fold. Helix α3 is the principal DNA-binding helix, and, as detailed below, residues in helices α4, α5, and the loop connecting α3 and α4 commonly mediate DBD homodimerization (32, 33). The Rgg2_Sd DBD dimerization interface observed in the Rgg2_Sd crystal structure is essentially identical to those observed in many prototypical members of the XRE family, such as Bacillus subtilis SinR (Fig. S3 A–C) (33) and the RNPP proteins PlcR and PrgX (Fig. S3 D–F) (15, 34); however, in the Rgg2_Sd structure (and the Rgg_Sd–CsA structure described below) the Rgg2_Sd DBDs are covalently linked across the dimerization interface by a disulfide bond between the α4 helices (Fig. 2A and Fig. S3 A, C, and F). Cys45 forms the disulfide bond, and this cysteine is absolutely conserved among more than 120 Rgg2 and Rgg3 orthologs from 27 different species. In contrast, this cysteine is absent from the more than 400 proteins that comprise the other Rgg subfamilies, including the group II, group III, ComR, and RopB proteins (6).

The Rgg2_Sd C-terminal repeat domain contains five HTH folds and a capping helix that together form a right-handed superhelical structure with a concave inner surface and convex outer surface (Fig. 2). This structure resembles a TPR domain superhelix, but the primary amino acid sequence does not contain TPR sequence motifs as determined using TPRpred (35). Ten additional C-terminal amino acids (residues 275–284) follow the capping helix, but there was insufficient electron density to model these residues.

Structural alignment of the full-length Rgg2_Sd protomers (rmsd for modeled C_α carbons = 0.48–1.195 Å for all pairwise comparisons) revealed only very subtle rigid-body differences in the positions of the essentially identical DBDs relative to the C-terminal repeat domains (Fig. S4A). The linker region connecting the DNA-binding and repeat domain is flexible, and clear electron density corresponding to this entire region existed for only one of the four modeled protomers (protomer B) in the crystallographic asymmetric unit (Fig. 2A and Fig. S1). The conformational flexibility in the linker region could facilitate rigid-body movements of the DBDs relative to the repeat domains (Fig. S4A).

Structural alignment of the C-terminal repeat domains revealed conformational differences near the C-terminal portion of the protomers in a subdomain consisting of repeat 5 α-helix B (residues 241–257), the capping helix (residues 260–274), and the loop (residues 258–259) connecting these helices (Fig. 2A and Fig. S4 B and C). We refer to this region (residues 241–274) as the “cap subdomain.” The cap subdomains of protomers A and C are positioned proximal to the concave surface of the repeat domain, whereas the cap subdomains of protomers B and D are positioned distal to the concave surface of the repeat domain (Fig. S4 B and C). As detailed below, the conformational differences in the cap subdomains result from flexibility in the loop connecting repeat 5 α-helices A and B (Fig. 2A and Fig. S4 B and C), and the position of the cap subdomain has important implications for ligand binding to the concave surface of the repeat domain.

Identification of the SHP2-Binding Site.

A search (30) of the PDB database for proteins structurally similar to Rgg2_Sd identified PrgX (PDB ID code 2AXZ, Z score = 19.2) (18), PlcR (PDB ID code 3U3W, Z score = 13.1) (15), NprR (PDB ID code 4GPK, Z score = 11.7) (19), and RapI (PDB ID code 4I1A, Z score = 10.7) (17). In previous studies, the concave surface of the repeat domain of peptide pheromone receptors was identified as the pheromone-binding site (Fig. S5 A–D) (13, 14, 16–18). We hypothesized that the Streptococcus Rgg receptors similarly used the concave surface of their repeat domain as the SHP-binding site. To test this hypothesis, we developed a test-bed assay using a Δrgg2 Δshp strain of S. pyogenes (BNL200) that is unable to produce or respond to SHP pheromones (Fig. 3A). When S. pyogenes rgg2 or S. dysgalactiae rgg2 were transferred to the test-bed strain, response to synthetic SHP peptide was restored, as indicated by luminescence activity produced by the integrated Pshp-luxAB reporter (Fig. 3A). The response was specific for SHP2, because the negative control peptide Rev-SHP2 did not trigger light expression above that of the Δrgg2 Δshp strain.

Fig. 3.

In vivo and in vitro functional analysis. (A) Relative luminescence activity of the P_shp2-luxAB reporter in response to exogenous SHP2. Rgg2 variants were expressed from a plasmid under their native promoters in group A Streptococcus (GAS) strain BNL200 (∆rgg2 Δshp2Δshp3, attP::P_shp2-luxAB); empty vector (blue), Rgg_Sp (red) and Rgg_Sd (green). (B) Luciferase response of Rgg2_Sd mutants in GAS test bed in response to 10 nM SHP (red), 10 nM SHP + 10 μM CsA (green), or vehicle (blue). (C) Direct FP of 10 nM FITC-SHP2 synthetic peptide titrated with purified Rgg_Sd. (D) CsA competes directly with FITC-SHP2 for binding to 500 nM Rgg_Sd in the FP assay. Plots indicate the means of at least three independent experiments. K_d values were determined by applying linear-regression on dose–response curves using GraphPad Prism (version 6.01).

To begin to map the SHP2-binding site, Rgg2_Sd mutants containing targeted alanine substitutions in surface-exposed residues of the concave surface of the repeat domain were expressed in the test-bed strain and assessed for their response to synthetic SHP2 pheromone (Fig. 3B). In comparison with the wild-type Rgg2_Sd control, Rgg2_Sd-N150A, Rgg2_Sd-R153A, Rgg2_Sd-N190A, and Rgg2_Sd-Y222A were insensitive to pheromone, and, to different degrees, Rgg2_Sd-R81A, Rgg2_Sd-I146A, Rgg2_Sd-K178A, Rgg2_Sd-L183A, Rgg2_Sd-L187A, Rgg2_Sd-D217A, Rgg2_Sd-L219A, and Rgg2_Sd-L262A displayed reduced SHP2-dependent activity (Fig. 3B). We used EMSA to measure the DNA-binding activity of the mutants that were insensitive to pheromone (Fig. S6). Rgg2_Sd-Y222A displayed wild-type–level activity, whereas Rgg2_Sd-N150A, Rgg2_Sd-R153A, and Rgg2_Sd-N190A displayed a partial loss of function. Only one of the substitution mutations introduced into the concave surface of the repeat domain, Y84A, had no effect on SHP2-triggered Rgg2_Sd activity (Fig. 3B). Based on these data and additional evidence outlined in Discussion, we conclude that SHP2 activates Rgg2_Sd by binding to the concave surface of the Rgg2_Sd C-terminal repeat domain.

The Cyclic Undecapeptide CsA Is a Potent Inhibitor of Rgg Function in Vivo and in Vitro.

To identify inhibitors of Rgg function, we developed a fluorescence polarization (FP) assay whereby drug and druglike compounds (Prestwick Chemical) were screened for their ability to disrupt the binding of fluorescent-labeled synthetic SHP peptides (FITC-SHP) to Rgg_Sp (26) or Rgg2_Sd (Fig. 3 C–D). Using this assay, we determined that the cyclic undecapeptide CsA is a potent inhibitor of SHP2 binding to Rgg2_Sp and Rgg_Sd in vitro (IC₅₀ = 0.4 μM) (Fig. 3D). Finally, using the in vivo SHP bioassay described above, we found that CsA inhibits both Rgg2_Sd and Rgg2_Sp activity in vivo (Fig. 3A).

Rgg2–CsA Complex X-Ray Crystal Structure.

To determine how CsA functions to inhibit SHP-triggered Rgg2 transcriptional activity, we determined the X-ray crystal structure of Rgg2_Sd in complex with CsA (Fig. 4 and Table S1). The single-wavelength anomalous diffraction (SAD) method was used to obtain experimental phases, and the Rgg2_Sd–CsA structure ultimately was refined to 1.95-Å resolution (Table S1). There are four Rgg2_Sd protomers in the Rgg2_Sd–CsA asymmetric unit, and each protomer binds to CsA (Fig. 4 and Figs. S4D and S7). As detailed below, CsA makes extensive interactions throughout the concave surface of the Rgg2_Sd repeat domain and a few contacts across the dimer interface (Fig. 4 and Fig. S7).

Fig. 4.

Rgg2_Sd–CsA crystal structure. Surface representation of the Rgg2_Sd–CsA dimer formed by protomers A (Rgg2_A) and B (Rgg2_B). (A, Left) In this orientation, CsA (ball and stick model) bound to Rgg2_B is visible. The conformation of CsA bound to Rgg2_B also represents the conformation of CsA bound to protomers C and D. (Right) An expanded view of the area enclosed by the black dashed lines in the left panel. (B, Left) In this orientation, CsA (ball and stick model) bound to Rgg2_A is visible. (Right) An expanded view of the area enclosed by the black dashed lines in the left panel. In all panels, the Rgg2_Sd surface that interacts with CsA is colored orange or green. The green surface highlights the positions where alanine substitution mutations desensitized Rgg2_Sd to CsA-mediated inhibition of SHP-induced Rgg2 activity. Abu, γ-amino-butanoic acid; Ala, alanine; Dal, d-alanine; MeBmt, 4-methyl-4-[(E)-2-butenyl]-4,N-methyl-threonine; Mle, N-methyl-l-leucine; Mva, N-methylvaline; Sar, sarcosine; Val, valine.

CsA adopts a nearly identical conformation (and makes similar receptor contacts) in complex with Rgg2_Sd protomers B, C, and D (Fig. 4A and Fig. S7); however, CsA adopts a distinct conformation in complex with protomer A (Fig. 4B and Fig. S7B). Although CsA adopts two different conformations, the CsA-binding site in all the protomers is similar (Fig. S7). That is, the great majority of the Rgg protomer A residues that contact CsA in one conformation are also used by protomers B, C, and D to contact CsA in the other conformation; however, important conformational differences in the Rgg2_Sd protomers discussed below explain how they accommodate two CsA conformations.

The Rgg2_Sd–CsA interface consists largely of hydrophobic interactions and a few H-bonds (Fig. S7). One conformational difference between the protomer A and the protomer B, C, and D CsA-binding interfaces is in the position of the cap subdomain (Fig. S4 E and F). The Rgg2_Sd protomer A cap subdomain is positioned proximal to the CsA-binding site, and the protomer B, C, and D cap subdomains are positioned distal to the CsA-binding site. It is important to note that electron density in protomer D was insufficient to model the loop (residues 258 and 259) connecting the C terminus of repeat 5 α-helix B to the N terminus of the cap subdomain (Fig. S4E). This finding is consistent with the idea that the loop is a flexible hinge allowing the cap subdomain to undergo rigid-body conformational changes. Other notable conformational differences include Rgg2_Sd-Arg153, which mediates numerous H-bonds to CsA exclusively in protomer A, and Rgg2_Sd-Tyr222, which forms H-bonds with CsA only in protomers B, C, and D (Fig. S7).

How can CsA adopt two conformations in the Rgg2_Sd–CsA cocrystal structure? In brief, these two conformations are possible because the Rgg protomers make nonidentical crystal contacts (the contacts between the crystallographic asymmetric units), and some of the Rgg2_Sd–CsA protomers adopt different conformations (Fig. S4 D and E). More specifically, Rgg2_Sd–CsA protomers B, C, and D are in a conformation most similar to Rgg2_Sd protomers B and D, whereas Rgg2_Sd–CsA protomer A is in a conformation most similar to Rgg2_Sd protomers A and C (compare Fig. S4 B–E). It appears that CsA drives the conformational change in protomer C observed in the Rgg2_Sd–CsA structure, which, as discussed above, is enabled by both the flexibility in the loop connecting the C terminus of repeat 5 α-helix B to the N terminus of the cap subdomain (Fig. S4E) and the nonrestrictive (largely solvent-mediated) crystal contacts near the protomer C cap subdomain. Because CsA binding drove the conformationally flexible protomer C into a conformation similar to protomers B and D of Rgg2_Sd and Rgg2_Sd–CsA, we propose that in solution the receptor-bound CsA conformational equilibrium is toward that of Rgg2_Sd–CsA protomers B, C, and D.

Functional Analysis of the Rgg2–CsA Interface in Vivo.

To begin to determine which of the receptor–ligand interactions observed in the Rgg2_Sd–CsA complex crystal structure are functionally important, we measured the ability of CsA to inhibit SHP-induced activity of Rgg2_Sd mutants containing single alanine substitutions in Rgg2_Sd–CsA interfacial positions (Fig. 3 A and B and Fig. S7). The alanine substitutions were engineered at positions where CsA contacts Rgg2_Sd in the concave surface of the repeat domain or at the dimer interface (Fig. S7). Our biochemical, genetic, and structural data suggested that the SHP2- and CsA-binding sites overlap significantly, and, in accordance with these results, we found that many of the Rgg2_Sd residues that interact with CsA also are required for SHP2 to activate Rgg2_Sd (e.g., Asn150, Arg153, Asn190, and Tyr222) (Fig. 3B). However, we also identified a number of CsA-binding residues (e.g., Tyr84, Lys178, Leu183, Leu187, Asp219, and Leu262) that are not absolutely critical for SHP2-mediated Rgg2_Sd activation, and alanine substitution mutations in these positions desensitized Rgg2_Sd to CsA-mediated inhibition of SHP-induced Rgg2 activity (Fig. 3B). These residues mediate critically important interactions with CsA but appear to contribute less to the SHP2-binding energy than Asn150, Arg153, Asn190, and Tyr222.

Consistent with CsA antagonizing SHP binding to both Rgg2_Sd and Rgg2_Sp in vitro and in vivo, 23 of the 24 Rgg2_Sd residues that contact CsA (Fig. S7) are identically conserved in Rgg2_Sp. The one nonidentical CsA-binding site residue, Rgg2_Sd-A261 (Rgg2_Sp-S261) contacts CsA only in the Rgg2_Sd promoter A conformation (Fig. S7B). The great similarity of residues within the concave surfaces of the Rgg2_Sd and Rgg2_Sp repeat domains also is consistent with our observation that the S. dysgalactiae SHP2 and S. pyogenes SHP2 sequences are identical (Fig. 1B).

The Nonimmunosuppressive CsA Analog Valspodar Antagonizes Rgg2_Sd and Rgg2_Sp Activity in Vivo.

To begin to assess the importance of the CsA structural features (namely its side chains) to Rgg inhibitory function, we tested the activity of the CsA analog SDZ PSC 833 (also known as “PSC 833” or “valspodar”) (36, 37) in vivo. Valspodar is a nonimmunosuppressive CsA analog that substitutes 3′-keto-MeBmt and valine in place of CsA 4-methyl-4-[(E)-2-butenyl]-4,N-methyl-threonine (MeBmt) and γ-aminobutanoic acid, respectively. Modeling these substitutions showed that they were accommodated without significant van der Walls overlap in all protomers of the Rgg2_Sd–CsA structure, and, like CsA, valspodar completely inhibited SHP-dependent Rgg2_Sp and Rgg2_Sd activity in vivo (Fig. 3A).

A Disulfide Bond in the Rgg2 DBD Dimerization Interface.

Although the DBD dimerization interface observed in both the Rgg2_Sd and Rgg_Sd–CsA structures is essentially identical to those previously described for members of the XRE family (Fig. S3), the presence of an intermolecular disulfide bond buried in this interface is, at least to our knowledge, unique. As detailed above, Cys45, which forms the disulfide bond, is highly conserved in Rgg2 and Rgg3 orthologs, and the disulfide bond is present in both dimers in the crystallographic asymmetric unit in the CsA-bound and CsA-free structures (Fig. 2A and Fig. S3 A, C, and F).

To begin to explore a possible role for the disulfide bond in the regulation of Rgg2_Sd, we measured the SHP2-dependent activity of the Rgg_Sd-C45S mutant in vivo (Fig. 5A). In comparison with wild-type Rgg2_Sd, Rgg2_Sd-C45S exhibited only a very slightly reduced response to SHP2 (Fig. 5A). In addition, we grew the group A Streptococcus strain BNL178 in both oxidizing and reducing conditions (Fig. 5B). The oxidizing reagent paraquat and the reducing agents DTT and N-acetyl cysteine had little or no effect on SHP2-triggered Rgg2_Sd activity in vivo. In contrast, after denaturation in the absence of DTT in vitro, Cys45 was required for the formation of a species consistent with the dimer form of Rgg2_Sd (Fig. 5C).

Fig. 5.

Functional analysis of the Rgg2_Sd disulfide bond. (A) Luciferase response of Rgg2_Sd–C45S [BNL200(pCA128)] compared with WT Rgg2_Sd [BNL2000(pCA113)] and empty vector [NL200(pLZ12-Sp)]. (B) Luciferase response of WT Rgg2 in presence of reducing agents (10 mM DTT or 15 mM N-acetyl cysteine, NAC) or oxidizing agent (10 mM paraquat). (C) In vitro formation of the Rgg2_Sd disulfide bond requires Cys45. Samples of 10 μg Rgg2_Sd (WT) or Rgg2_Sd-C45S were boiled in the presence (+) or absence (−) of 5 mM DTT before analysis by SDS/PAGE. M, molecular weight standards; Rgg2₁, Rgg2_Sd monomer; Rgg2₂, Rgg2_Sd dimer. The asterisk marks the His-SUMO-Rgg2_Sd contaminant.

Discussion

The Rgg, Rap, NprR, PlcR, and PrgX proteins are cytoplasmic receptors regulated by peptide pheromones. As a result of the Rgg_Sd– and Rgg_Sd–CsA crystal structures, we now know that the Rgg proteins have domain architectures identical to those of the PrgX, NprR, and PlcR proteins, i.e., an N-terminal DBD and a C-terminal repeat domain. Based on the structural similarity of Rgg2_Sd and the RNPP proteins, and because Rgg and RNPP proteins are peptide pheromone receptors, the Rgg proteins now can be considered bona fide members of the RNPP family. Rgg proteins are, in fact, the largest constituents of this family, and, as previously proposed, the family can be renamed “RRNPP” to include these proteins (4, 20).

A longstanding goal has been to identify inhibitors of RRNPP proteins, which commonly regulate the expression of critical developmental phenotypes. These inhibitors could disrupt cell–cell signaling and potentially function as antibiotics or antiinfectives. The Rgg2_Sd–CsA X-ray crystal structure showed that CsA binds to the concave surface of the Rgg2_Sd repeat domain. Based on our genetic analysis of Rgg2_Sd and on studies from our laboratories and others that identified the concave surface of RNPP proteins as the pheromone-binding site, we conclude that SHP peptides bind the concave surface of Rgg protein repeat domains. Additional support for this idea comes from our computational docking studies showing that the concave surface of the Rgg2_Sd repeat domain can physically accommodate the binding of SHP2 peptides (Fig. S5E). X-ray crystal structures of Rgg2–SHP2 and Rgg3–SHP3 cocomplexes are required to reveal the functionally relevant SHP-binding mode and to determine the atomic-level details of Rgg–SHP binding and the basis of their interaction specificity. Studies are underway in our laboratory to determine the X-ray crystal structures of Rgg proteins bound to SHP peptides and/or DNA. Finally, because the great majority—but not all—of the alanine substitutions targeted to the concave surface of the repeat domain disrupted SHP2 and CsA binding, we conclude that SHP2 and CsA bind to largely overlapping, nonidentical regions of the concave surface.

How do the cyclic undecapeptides CsA and valspodar antagonize SHP signaling? It was shown previously that a C-terminal portion of PrgX rearranges upon pheromone (cCF10) binding and that the PrgX C terminus and cCF10 interact to form a small β-sheet (18). The Rgg2_Sd C-terminal cap subdomain is conformationally flexible (Fig. S4 B–F), as is the Rgg2_Sd C-terminal tail, which was disordered in the crystal structure and could not be modeled. We propose that, like PrgX-cCF10, the SHP peptide and Rgg C-terminal tail can form a β-sheet. Furthermore, we propose that the large undecapeptide CsA (and valspodar) alone can occupy the same 3D space that would be occupied by both the linear SHP peptide and the Rgg C-terminal tail. CsA and valspodar may function not only by competitively inhibiting SHP binding but also by blocking the Rgg2_Sd C-terminal tail from adopting an active (SHP-bound) conformation and, in turn, a potentially important allosteric conformational change in the C-terminal domain that could be required for receptor activation. Based on the structural similarity of the Rgg and RNPP proteins and their similar modes of pheromone binding, we speculate that cyclic peptides could serve as antagonists not of only Rgg proteins but also of other RRNPP family proteins.

Finally, perhaps one of the most interesting mechanistic questions to address going forward pertains to what in vivo role, if any, is played by the Rgg2_Sd intermolecular disulfide bond observed in the Rgg2_Sd– and Rgg2_Sd–CsA crystal structures and in solution (Fig. 5). The disulfide bond forms at the core of the DBD dimerization interface, which is structurally identical to other XRE protein dimer interfaces (Fig. S3). Furthermore, as discussed above, the disulfide is formed by Cys45, which is remarkably well conserved in more than 120 Rgg2 and Rgg3 orthologs from 27 different species. We also note that the Rgg2_Sd and Rgg2_Sd–CsA crystals were grown in the presence of 5 mM DTT and soaked in cryoprotection solutions containing 5 mM DTT immediately before data collection. Therefore, although the disulfide bond can be reduced by boiling the protein in DTT, the bond is sufficiently buried in the natively folded protein to resist chemical reduction. This degree of shielding may enable the disulfide bond to persist in the reducing environment of the bacterial cytoplasm. Under the right conditions, which we have not yet identified, the conserved cysteine may play a regulatory role, perhaps affecting Rgg monomer–dimer equilibrium, stabilizing Rgg conformation, or functioning as a redox-sensitive switch.

Methods

Primers used in this study are listed in Table S2. Bacterial strains and plasmids used in this study are listed in Table S3.

The Rgg2–CsA crystal structure was determined by the SAD method using crystals of selenomethionyl Rgg2–CsA. PHENIX (AutoSol) was used to locate heavy atom positions, calculate phases, and generate an initial model at 1.95-Å resolution (38). The final model was generated through iterative cycles of building in COOT (39) and refinement in PHENIX. The Rgg2 and CsA models were built de novo into the SAD-phased map. The earliest rounds of refinement in PHENIX used simulated annealing, individual atomic coordinates, and individual B-factor refinement. The later rounds of refinement in PHENIX used individual atomic coordinates, individual B-factor refinement, and a TLS model whose initial parameters were guided by the TLS Motion Determination (TLSMD) server (40). During the final rounds of refinement in PHENIX, the ADP weights were optimized, i.e., the weights yielding the lowest R_free value were used for refinement. CsA was modeled only after the Rgg2 models were nearly complete. CsA residues were numbered according to the convention used in the PDB database, which also is in agreement with the proposed CsA biosynthetic reaction mechanism (41). Water and sulfate molecules were built into clear electron density during the final stages of refinement. The structure of Rgg2 alone was determined by molecular replacement using protomer A from the Rgg2–CsA structure as a search model. Ramachandran statistics were calculated in MolProbity (42). Molecular graphics were produced with PyMOL (43).