Phosphorylation-dependent conformational changes and domain rearrangements in Staphylococcus aureus VraR activation

Abstract

Staphylococcus aureus VraR, a vancomycin-resistance-associated response regulator, activates a cell-wall–stress stimulon in response to antibiotics that inhibit cell wall formation. X-ray crystal structures of VraR in both unphosphorylated and beryllofluoride-activated states have been determined, revealing a mechanism of phosphorylation-induced dimerization that features a deep hydrophobic pocket at the center of the receiver domain interface. Unphosphorylated VraR exists in a closed conformation that inhibits dimer formation. Phosphorylation at the active site promotes conformational changes that are propagated throughout the receiver domain, promoting the opening of a hydrophobic pocket that is essential for homodimer formation and enhanced DNA-binding activity. This prominent feature in the VraR dimer can potentially be exploited for the development of novel therapeutics to counteract antibiotic resistance in this important pathogen.

The spread of Staphylococcus aureus strains that are resistant to many of the commonly administered antibiotics is a significant threat to public health throughout the world. In the United States alone, there were estimated to be 478,000 S. aureus related hospitalizations in 2005, with 58% of these infections involving methicillin-resistant S. aureus (MRSA) (1). There is a pressing need to understand the signaling pathways that allow S. aureus to respond to and resist antibiotic chemotherapy treatment and to find new drug targets, in novel biochemical pathways, to combat these infections.

In recent years, there has been growing interest in the S.aureus vancomycin-resistance–associated sensor and response regulator (VraSR) two-component signal transduction system. Mutations or deletions within genes encoding either the sensor histidine kinase VraS or the response regulator VraR restrict antibiotic resistance to both β-lactam and glycopeptide antibiotics across a range of S. aureus strains (2–4). Furthermore, deletion of the vraSR operon in a clinical USA300 MRSA strain generated an isogenic mutant, that unlike the parent strain, could be treated effectively with the β-lactam oxacillin in mouse models of both skin and lung infections (5).

The VraSR two-component signaling pathway contains a histidine kinase, VraS, and its cognate response regulator, VraR. The VraSR pathway is activated by exposure to cell-wall–active antibiotics or by genetic disruption of cell-wall synthesis (6). Cell-wall damage appears to be detected by VraT, a transmembrane protein encoded by an upstream gene in the vraSR operon. VraT, previously named YvqF (7), physically interacts with VraS (3), but the exact stimulus that it detects is unknown. VraS can act as both a kinase and a phosphatase to control the phosphorylation state of the cytoplasmic transcription factor VraR (8). Phosphorylation increases the DNA-binding activity of VraR (9), leading to increased expression of the vraSR operon and a regulon of at least 42 additional genes, many of which are involved in cell wall synthesis (10).

Here, we report the structure of full-length VraR in both an inactive, dephosphorylated state and an active conformation stabilized by the phosphoryl analog beryllofluoride. Together, the structures provide a description of the conformational changes that accompany VraR activation, leading to domain rearrangements and dimerization that promotes DNA binding at target promoters. A notable feature of activation is the substantial remodeling of the receiver domain, creating a deep hydrophobic pocket that accommodates a protruding methionine side chain from the complementary protomer at the center of the dimer interface.

Results and Discussion

VraR Dimerization Is Strongly Phosphorylation Dependent.

The VraR protein has been shown to be phosphorylated in vitro by both the small molecule phosphodonor acetyl phosphate, and also by the cytoplasmic domain of the histidine kinase VraS (8). Electrophoretic analysis of native VraR has provided qualitative evidence that phosphorylation promotes homodimerization of the protein in vitro (8), but the labile nature of phosphoaspartate has prevented quantitative analysis of the affinity of this interaction. Beryllofluoride (BeF₃⁻), a noncovalent phosphoryl mimic, provides an alternative method to stabilize receiver domains in active conformations (11). To characterize the oligomeric state of inactive and active VraR, analytical ultracentrifugation (AUC) analysis was performed in the absence or presence of molar excess of BeF₃⁻ (Fig. 1A). Sedimentation equilibrium profiles generated in the absence of BeF₃⁻ were globally fitted to a single species model, with the molar mass fixed as that of a VraR monomer (23428.2 Da). There is no improvement in the quality of the fit if a more complex monomer-to-dimer equilibrium model is applied. In contrast, the sedimentation equilibrium profiles generated in the presence of BeF₃⁻ could only be fit with a monomer-to-dimer equilibrium model, with the best fit obtained using a dissociation constant of 230 nM. At a 95% confidence interval, we estimate at least a four order of magnitude higher dimerization propensity when BeF₃⁻ is present compared with its absence (Table 1). A shift in the protein distribution from a sedimentation coefficient of 2.2 Svedberg (S) units when BeF₃⁻ is absent, to 3.2 S when BeF₃⁻ is present, was observed using sedimentation velocity AUC analysis (Fig. 1B). This corresponds to a shift in the apparent molar mass from 24 ± 6 kDa to 39 ± 8 kDa, consistent with a transition from monomer to homodimer, and provides further confirmation of the strong dependence of dimerization on the presence of the phosphoryl mimic, BeF₃⁻. There is no evidence for the formation of larger oligomeric species, as no peaks are observed in the sedimentation coefficient [c(S)] distribution plot at values greater than 3.2 S.

Fig. 1.

BeF₃⁻ promotes VraR dimerization. (A) Sedimentation equilibrium AUC profiles for 100 µM VraR in the absence (dark gray) or presence (light gray) of BeF₃⁻, recorded at 20,000 rpm using a Beckman Coulter (Brea, CA) AnTi50 rotor, fitted using a monomer only or a monomer-to-dimer equilibrium model, respectively. The line of best fit is shown in black. (B) The c(S) distribution for 100 µM VraR in the absence (dark gray) and presence (light gray) of BeF₃⁻, established using sedimentation velocity AUC analysis.

Table 1.

Dimerization affinities for homodimer and heterodimer associations

Protein(s)	BeF3−	Kd, µM	95% confidence limits, µM
VraR	−	ND	Kd > 2,000*
VraR	+	0.230	0.02 < Kd < 0.53
VraRREC	−	ND	Kd > 2,000*
VraRREC	+	10	4 < Kd < 19
VraRDBD	−	720	510 < Kd < 1,100
VraRREC + VraRDBD	−	134	78 < Kd < 236
VraRM13D	−	ND	Kd > 5,000*
VraRM13D	+	ND	Kd > 5,000*

SEQ data and nonlinear regression fits are shown in Figs. S1–S5. ND, not determined.

^*The 95% confidence intervals, determined by F statistics (29) using a monomer–dimer equilibrium model, were greater than 10 mM. A conservative estimate for the limit of the dissociation constant for homodimerization is provided, restricting the value to within one order of magnitude of the concentration range observed in the experimental data.

Unphosphorylated VraR Has a Closed Conformation.

To understand how the phosphorylation state of VraR controls dimerization of the protein, crystal structures of VraR were determined in the absence and presence of BeF₃⁻. Data collection and refinement statistics are shown in Table 2. These provide a pair of structures of a full-length response regulator transcription factor in both inactive and active states. The core receiver domain (Thr2 to Arg121) with its canonical (βα)₅ topology and a four α-helix DNA-binding domain that defines the NarL/FixJ subfamily of response regulator transcription factors are separated by an extension of the receiver domain that includes a short loop followed by helix α6. In the absence of BeF₃⁻, VraR crystallized with four monomers in the asymmetric unit of the crystal, with the receiver and DNA-binding domains (DBDs) packed together in a closed conformation (Fig. 2A). Receiver and DNA-binding domain packing arrangements are not conserved in available structures of inactive NarL/FixJ subfamily response regulators (12, 13), suggesting that a variety of closed conformations have evolved as has been observed in the more widely characterized OmpR/PhoB winged-helix transcription factor response regulator subfamily (14). The association of the isolated VraR receiver domain (VraR_REC) and DNA-binding domain (VraR_DBD) is relatively weak, with a dissociation constant of 134 µM (Table 1). However, this interdomain interaction will be greatly stabilized in the wild-type protein by the linker that tethers the domains. Helix α10 of the DNA-binding domain is packed against α3 and α4 of the receiver domain through a predominantly hydrophobic interface, burying a total of 942 Ų of protein surface area. This closed conformation is further stabilized by a salt bridge between Arg137 and Glu191 and a network of hydrogen bonds connecting the DNA-binding domain with both the core receiver domain (Gln204 hydrogen bonds with Lys71, Gly98, and Asp100) and the α6 linker helix (Tyr201 hydrogen bonds with Arg141 and Tyr145) (Fig. 2B). The closed conformation that exists in the crystal structure, with helix α10 buried in the interdomain interface, provides an explanation for the slow rates of hydrogen–deuterium exchange observed for residues of α10 in previous analyses of unphosphorylated VraR (15).

Table 2.

Data collection and refinement statistics

Dataset	VraR, apo	VraR Mg-BeF3−
Data collection
Space group	P212121	R3
Cell dimensions
a, b, c; Å	78.15, 103.80, 114.42	110.60, 110.60, 282.29
α, β, γ; °	90, 90, 90	90, 90, 120
Resolution, Å	50.00–2.03 (2.07–2.03)	40.00–2.35 (2.39–2.35)
Rsym	0.096 (0.607)	0.060 (0.777)
I/σI	25.1 (3.3)	33.5 (2.0)
Completeness, %	99.8 (99.5)	98.6 (97.7)
Redundancy	7.4 (7.2)	5.8 (5.2)
Refinement
Resolution, Å	46.16–2.03	36.98–2.35
No. reflections	57522	49310
Rwork/Rfree	0.172/0.229	0.181/0.226
No. atoms
Protein	6,601	6,556
Ligand/ion	—	40
Water	746	159
B-factors, A2
Protein	29.8	59.8
Ligand/ion	—	79.0
Water	37.1	51.5
R.m.s. deviations
Bond lengths, Å	0.007	0.009
Bond angles, °	0.968	1.030

One crystal was used per protein structure. Values in parentheses are for highest-resolution shell.

Fig. 2.

Structural overview of the unphosphorylated VraR monomer. (A) Cartoon depiction of the VraR inactive monomer showing the receiver domain (green) and DNA-binding domain (cyan). The linker helix (α6) and the putative DNA-recognition helix (α9) are colored orange and magenta, respectively. (B) Expanded view of the interdomain interface. Hydrogen bonds and salt bridge interactions that contribute to stabilizing the packing of the DNA-binding domain against the receiver domain are indicated by dashed black lines.

Rearrangements in Activated VraR.

The crystals obtained in the presence of BeF₃⁻ contain four protomers of VraR in the asymmetric unit, but the arrangement of the domains is very different from that of inactive VraR. In its activated state, the receiver domain is dissociated from the DNA-binding domain (Fig. 3A). A single turn of helix α6 is unwound between Lys140 and Glu143, creating a flexible linker between the receiver and DNA-binding domains that allows greater freedom of movement between the domains. A VraR tetramer is observed in the asymmetric unit of the crystal, made up of two receiver domain dimers and two DNA-binding domain dimers. The flexible linker between the domains allows domain swapping such that the protomer pairs that form receiver domain dimers (assigned as chains A with C or B with D) are different from the protomer pairs that form the DNA-binding domain dimers (A with D and B with C). VraR has been found to bind cooperatively to nucleotide sequences designated as the R1 and R2 binding sites within the vraSR operon promoter (16), but it is not clear whether this cooperativity results from changes in the DNA structure or from direct protein–protein interactions between VraR dimers. Whereas the tetrameric complex that we observe in our activated VraR structure, stabilized by helix–helix packing of two α6 helices, could potentially explain the observed cooperativity, it seems likely that the homotetramer complex is an artifact of crystallization, as we found no evidence of tetramer formation under solution conditions (Fig. 1B). The data suggest that any protein–protein interactions that might exist between VraR dimers are weak (K_d > 100 µM), and if such interactions are responsible for cooperative binding, they result from high local concentrations of VraR dimers generated by recruitment to DNA.

Fig. 3.

The activated VraR protein structure. (A) Two VraR protomers form a dimer through dimerization of their receiver (REC) domains. The DNA-binding domains (DBDs) form homodimers with alternative protomer chains within the crystal lattice as highlighted by the dashed oval. This domain swapped arrangement is presumed to be a consequence of crystallization and is not detected in solution studies (see text). (B) VraR receiver domain dimer with Met13 docked into a hydrophobic pocket within the dimerization interface. The residues that form the dimerization interface are highlighted. (C) VraR DNA-binding domain homodimer shown as both a cartoon and electrostatic surface representation. The two α9 helices in the dimer are positioned to create a large electropositive DNA-binding surface.

The receiver domains dimerize through a hydrophobic interface of 786 Ų that juxtaposes the α1 and α5 helices of two protomers (Fig. 3B). At the center of this interface, Met13 from one protomer docks into a pocket within the opposing protomer. This dimerization pocket, lined by the hydrophobic amino acids Val14, Ile18, Tyr21, Leu22, Leu82, Lys105, Ala109, and Ile112, is not present in inactive VraR (Fig. 4 A and B). A phosphorylation-induced conformational change in the receiver domain is required to promote homodimer formation.

Fig. 4.

Phosphorylation-dependent opening of the receiver domain dimerization pocket. (A and B) Electrostatic surface representations of the VraR receiver domain in inactive (A) and active (B) states showing the hydrophobic pocket that is present at the center of the dimerization interface in the active VraR protein. (C and D) 2F_o–2F_c map at 1.5σ cutoff for the VraR active site in the absence (C) and presence (D) of Mg²⁺ and BeF₃⁻. For clarity the electron density for BeF₃⁻ (Be, chartreuse; F, cyan) and Mg²⁺ (green) are not shown. (E) Stereo images showing a superposition of an inactive (orange) and an active (green) receiver domain showing the key residues that change conformation to transition from a closed dimerization pocket conformation to the open dimerization pocket conformation required for homodimer formation. The α1-helix from the other protomer chain in the active VraR dimer is shown (blue) with Met13′ docked into the hydrophobic pocket.

Additional conformational changes are observed on a face of the receiver domain opposite to the dimerization interface. The angle of the linker helix α6 on the α4–β5–α5 face of the core receiver domain changes by 35°, accompanied by a rearrangement of salt bridge interactions between α4 and α6. The position of helix α6 is conserved in all four protomers in the asymmetric unit of the active VraR crystal, suggesting that the movement of α6 is a feature of the activation mechanism and not an artifact of crystal lattice constraints. In the inactive state, Asp88 and Asp96 in α4 form salt bridges with Lys133 and Arg141 in α6, respectively. In the activated state, the Asp88–Lys133 salt bridge is maintained, but Asp96 forms a strong salt bridge with Arg137 instead of Arg141. Repositioning of helix α4 in active receiver domains has been described for several response regulators (17, 18). In VraR, as in other response regulators, perturbations to α4 contribute to release of the DNA-binding domain from the receiver domain. However, in contrast with members of the OmpR/PhoB subfamily in which the altered α4–β5–α5 surface forms the dimerization interface of the active receiver domains, VraR receiver domains use a different surface for dimerization.

Despite 24% sequence identity between Sinorhizobium meliloti FixJ and VraR receiver domains and a conserved domain fold, phosphorylation promotes dimerization through very different surfaces in these two proteins; the FixJ receiver domain dimerizes at its α4–β5 surface (19), whereas VraR dimerizes through a hydrophobic surface located between the α1 and α5 helices. Additional complexity is introduced by evidence of oligomerization of some NarL/FixJ transcription factors on DNA, implying that multiple surfaces of a receiver domain might be involved in protomer–protomer contacts. Indeed, mutagenesis studies suggest that the α1 helix of Escherichia coli UhpA, a response regulator that mediates sugar phosphate uptake, is involved in oligomerization (20). However, there are some indications from the crystal packing arrangements of more closely related VraR homologs, that the mode of receiver domain dimerization seen in the active VraR structure is not simply an outlier, but could represent a conserved dimerization mode shared by a subset of the NarL/FixJ family. In the orthorhombic crystal form of E. coli NarL, a nitrate-responsive response regulator, there is antiparallel packing of two α1 helices from adjacent receiver domains with Met17 from one protomer buried in a hydrophobic pocket between the α1 and α5 helices of the other protomer chain (21). Although it has only been speculated that this interface might represent the receiver domain dimerization interface of NarL, the common interface features observed in the active VraR structure suggest that E. coli NarL uses the same dimerization mode as S. aureus VraR. Further analysis of crystal packing arrangements for VraR homologs, Mycobacterium tuberculosis NarL (22), the cell-density-responsive transcription factor S. aureus LuxR (PDB ID 3B2N) and a regulator of transporter gene expression S. pneumoniae spr1814 (23) shows that all three receiver domain structures crystallized with previously unreported α1–α5 dimerization interfaces, with Leu19, Met14, and Met11, respectively, being the hydrophobic amino acid that is buried in a pocket between the α1 and α5 helices of the other protomer chain. The consistent dimerization features observed for several members of the NarL/FixJ family strongly suggest that the VraR active structure might define the active state for a subset of this response regulator family. However, the α1–α5 interface is controversial in the case of spr1814. Park et al. (23) reported an alternative interpretation of their crystal lattice, suggesting that spr1814 dimerizes via a 374-Ų interface consisting of the α4 helix, the α5–α6 loop, and part of the α6 helix. Although we note that the interface proposed by Park et al. (23) is much smaller than the 760-Ų α1–α5 interface, and only the α1–α5 interface is scored by PISA (24) as being essential for complex formation, caution should be exercised when interpreting crystal packing arrangements. Future experiments will need to be performed to establish whether the α1–α5 interface is used by other VraR homologs, under solution conditions.

Phosphorylation Promotes Dimerization of the DNA-Binding Domains.

Domain rearrangements in inactive and active VraR provide an explanation for the greater affinity of phosphorylated VraR for its target DNA sequences relative to the unphosphorylated protein (9). The DNA-binding domain consists of four α helices (Leu148 to Gln209) with a classical helix-turn-helix (HTH) DNA-binding motif that defines members of the NarL/FixJ response regulator subfamily, as previously reported for the NMR structure of the isolated domain (25). Based on similarity to homologous HTH domains, helix α9 (residues Ile176 to Leu190) is predicted to be the DNA-recognition helix that slots into the DNA major groove (25). The DNA-binding domains form dimers in the active VraR protein crystals, burying 397 Ų of protein surface area per protomer within the dimerization interface. The hydrophobic face of α10 (Thr196, Gln197, Val199, Ile200, Phe203, and Gln204) contributes 82% of the buried surface area with the C-terminal end of α7 (Ile159 to Gly162) providing minor contributions. Residues that comprise the hydrophobic face of α10 are not conserved among close homologs of VraR, suggesting that this surface contributes to the specificity for homodimer formation of the active response regulator.

Varied strategies for regulating DNA-binding activity are not surprising, given the differences in domain arrangements found in NarL/FixJ family members. In E. coli NarL, the DNA-recognition helix α9 is buried against the receiver domain in the inactive state, occluding interaction with DNA (12). In contrast, in both inactive and active VraR, α9 helix is solvent exposed. In inactive VraR, the interface with the receiver domain buries the hydrophobic face of α10, occluding the surface of the DNA-binding domain that forms the dimerization interface in active VraR. Phosphorylation of Asp55 in the receiver domain stabilizes a conformation of VraR that promotes receiver domain dimerization, disrupting the interdomain interface and allowing formation of a DNA-binding domain dimer (Fig. 3A). In solution, the isolated DNA-binding domains have a low propensity for dimerization (K_d = 720 µM) (Table 1). However, this relatively weak affinity is significant in the context of full-length VraR, where dimerization of the phosphorylated receiver domains would increase the local concentration of the DNA-binding domains. The observation that the K_d for dimerization of full-length VraR (230 nM) is an order of magnitude tighter than that of the isolated receiver domain (10 µM) in the presence of saturating levels of BeF₃⁻ is consistent with this model for a VraR dimer stabilized by both receiver domain and DNA-binding domain dimerization interfaces. The arrangement of the two DNA-binding domains in the dimer creates a positively charged surface and positions the two α9 DNA-recognition helices appropriately for interaction with adjacent major grooves of DNA (Fig. 3C). This DNA-binding domain dimer configuration is consistent with all of the reported crystal structures of helix-turn-helix DNA-binding domains when bound to their target DNA sequences (26–29). The DNA-binding domain dimer, with its twofold rotational symmetry, suggests that VraR would bind with the highest affinity to a palindromic DNA sequence. However, the flexible linker between the receiver and DNA-binding domains, coupled with the observation that receiver domain dimerization is 72-fold stronger than DNA-binding domain dimerization, could allow for binding to suboptimal target sequences where the spacing and orientation of the VraR target sequence half sites dictate alternative arrangements of the VraR DNA-binding domains on the DNA. Flexibility in the relative orientations and spacing of the DNA-binding sites has been observed for other members of the NarL/FixJ subfamily of response regulators. E. coli NarL binds with highest affinity to a 7-2-7 tail-to-tail palindromic sequence, where each NarL DNA-binding domain recognizes one heptameric consensus site. Weaker DNA-binding activity is observed for NarL when the heptamer consensus sequence is arranged as a head-to-head palindrome, as tandem repeats or where only a single consensus heptamer is present (27).

Phosphorylation of VraR Promotes Opening of the Dimerization Pocket.

Features of phosphorylated VraR that contribute to formation of the dimerization pocket are revealed by comparison of the structures of the inactive and active receiver domains. The switch residues, Thr83 and Tyr102, adopt a characteristic active conformation, with the hydroxyl group of Thr83 and the phenol ring of Tyr102 forming direct hydrogen bonds with the beryllofluoride and backbone carbonyl of Phe85, respectively, in the active VraR structure. However, in the inactive VraR structure these residues are not oriented away from the active site and a subtle 13° rotation of Tyr102 is sufficient to preclude the hydrogen bond with Phe85 that occurs in the active conformation. The opening of the dimerization pocket requires a number of small conformational changes throughout the receiver domain, involving rearrangements of both the hydrophobic core and the protein surface. Within the active site itself, two substantial rearrangements accompany the binding of the magnesium ion and the phosphoryl group mimic, BeF₃⁻ (Fig. 4 C and D). First, in contrast with the inactive structure in which Lys105 forms salt bridges with Asp55 and Asp9, in the active structure, Lys105 retains a salt bridge with Asp9, but the salt bridge with Asp55 is replaced by a salt bridge with a fluoride of BeF₃⁻. Lys105 and the β4–α5 loop in which it resides (Lys105 to Ser108) are positioned further from the active site. Coordinately, two residues from α5, Ala109 and Ile112, which partially block the entrance to the dimerization pocket in the inactive state, are repositioned away from the opening of the dimerization pocket. Second, at the other side of the active site, Asp9 and Asp10, located in the β1–α1 loop, are positioned closer to the active site where they coordinate, indirectly or directly, with the Mg²⁺ ion and α1 is shifted 1.6 Å toward the active site. The sidechain of Arg15 is shifted toward the active site to maintain a hydrogen bond to the backbone amide of Asp9, and the sidechain of Glu33 is rotated to maintain its salt bridge contact with Arg15. The repositioning of these two polar groups accommodates rearrangement of two key “pocket gating” amino acids, Phe7 and Ile18 (Fig. 4E). In the inactive state, the phenyl ring is confined by Leu57 and Arg15 on either side so that the only conformation of phenylalanine that is allowed is a rare high energy state where the plane of the phenyl ring is aligned with the plane of the Cα–Cβ and Cβ–Cγ bonds. In the active state, the sidechain of Phe7 is flipped from its inactive rotamer conformation, pointing toward the active site, to an alternative rotamer, pointing away from the active site, where Phe7 is able to adopt the energetically favorable rotamer conformation with the phenyl ring aligned perpendicular to the plane of the Cα–Cβ and Cβ–Cγ bonds, a conformation that would clash with the carboxylate group of Glu33 if it was not itself rotated toward the active site by interaction with Arg15. To accommodate the active conformation of Phe7, Ile18 is rotated to a conformation that no longer blocks the entrance to the dimerization pocket.

As described above, a coordinated rearrangement propagated across the phosphorylated receiver domain promotes opening of the dimerization pocket. The constraint of Phe7 in a high-energy rotamer conformation in the inactive VraR protein reduces the energetic cost of rearranging this bulky amino acid within the hydrophobic core of the protein. However, rearrangement of the hydrophobic core of the protein likely poses a substantial energy barrier to formation of the open dimerization pocket conformation in the absence of Mg²⁺ and BeF₃⁻, providing a molecular explanation for the unusually strong dependence on BeF₃⁻, or phosphorylation, for VraR dimer formation. BeF₃⁻ binding or phosphorylation at the active site is required to shift the equilibrium toward the open dimerization pocket conformation that enables homodimerization of receiver domains.

Dimerization of Receiver Domains Is Required for DNA Binding.

As previously described, the DNA-recognition helix α9 is solvent exposed in the inactive VraR monomer and inhibition of DNA binding appears to result from occlusion of the dimerization interface. To test this assumption, it is important to establish whether blocking dimerization is sufficient to reduce the affinity of VraR for its target DNA. To disrupt VraR receiver domain dimerization, Met13, the residue that docks into the hydrophobic dimerization pocket, was substituted with aspartate, generating VraR_M13D. The sedimentation equilibrium AUC profiles for VraR_M13D, recorded in the absence and presence of BeF₃⁻, both fit to a single monomeric species with no indication of dimer formation (Fig. 5A). To determine whether the M13D substitution altered phosphorylation of the receiver domain, autophosphorylation using ammonium phosphoramidate as a chemical phosphodonor was assessed using reverse phase chromatography. Both wild-type VraR and VraR_M13D were found to phosphorylate stoichiometrically, as demonstrated by a complete shift in the elution peak to a longer retention time during reverse phase HPLC. Although, as expected, the VraR_M13D protein eluted earlier than the wild-type protein from the reverse phase column because of additional negative charge on its surface, both proteins exhibited comparable shifts upon phosphorylation. The phosphorylation of VraR is unaffected by the M13D substitution, providing evidence that the active site of VraR_M13D is correctly folded and thus should bind Mg²⁺ and BeF₃⁻ similarly to wild-type VraR (Fig. 5B). Taken together, the AUC and HPLC analyses demonstrate that substitution of Met13 with Asp is sufficient to prevent VraR dimer formation without adversely affecting the overall fold and catalytic activity of the receiver domain. Therefore, differences observed between the DNA-binding activities of the substituted and wild-type VraR proteins can be directly attributed to the effect of the M13D substitution on the dimerization of VraR.

Fig. 5.

Disruption of the receiver domain dimerization is sufficient to inhibit the DNA-binding activity of VraR. (A) Sedimentation equilibrium AUC profiles for VraR_M13D in the absence (green) or presence (blue) of BeF₃⁻. Raw data for both data sets have been fitted with a monomer species model, using mass conservation with rotor stretch restraints. Lines of best fit are shown in red. (B) Phosphorylation of VraR (pink and cyan) and VraR_M13D (green and blue) in the absence (pink and green) or presence (cyan and blue) of ammonium phosphoramidate. Phosphorylation of VraR or VraR_M13D is indicated by a shift of the elution peak from the reverse phase column to a later retention time. (C) DNA-binding activity of VraR (pink and cyan) and VraR_M13D (green and blue) in the absence (pink and green) or presence (cyan and blue) of BeF₃⁻ assessed using fluorescence anisotropy.

Fluorescence anisotropy was used to measure the binding of VraR and VraR_M13D to a fluorescently labeled DNA duplex containing the R1 binding site sequence within the vraSR promoter, previously identified as a strong binding site for wild-type VraR (9) (Fig. 5C). Wild-type VraR in the presence of BeF₃⁻ binds to the DNA duplex with a dissociation constant of 29 ± 5 µM; in the absence of BeF₃⁻, no DNA-binding activity is observed with up to 100 µM VraR. In similar experiments with VraR_M13D, no DNA-binding activity was observed either in the absence or presence of BeF₃⁻. The data indicate that disruption of receiver domain dimerization reduces the affinity of VraR for its target DNA sequence by at least an order of magnitude.

Response regulator proteins have been proposed as attractive drug targets because there are no human homologs for these proteins and they are often involved in the control of virulence factors or pathways that help bacteria evade the host immune system. The VraR protein represents a particularly enticing therapeutic target as the VraSR two component system is critical for S. aureus resistance to a wide range of FDA-approved antibiotics that are losing effectiveness due to emergence of resistant S. aureus strains. The VraR structures reported here provide a detailed understanding of the conformational changes that are induced by phosphorylation to promote dimerization that enhances the DNA-binding activity of this important transcription factor. The existence of a deep hydrophobic pocket, in close proximity to the active site pocket, provides a potential target site for novel therapeutics that would inhibit VraR dimerization and reduce the affinity of VraR for its target promoters, potentially restoring sensitivity of S. aureus to many of the antibiotics where resistance has become an urgent clinical concern.

Materials and Methods

S. aureus VraR protein expression constructs were generated in a pET30a vector (Novagen), expressed and purified from E. coli as described in SI Materials and Methods. Purified VraR protein was crystallized in the absence and presence of beryllofluoride. Diffraction data for crystals of apo VraR and VraR in complex with Mg²⁺ and BeF₃⁻ were collected at Cornell High Energy Synchrotron Source beamline F1 and Brookhaven National Synchrotron Light Source beamline X29, respectively, and the structure determined by molecular replacement in both cases. Detailed methods for protein crystallization, diffraction data collection, and structure determination are provided in SI Materials and Methods. Sedimentation equilibrium AUC was used to determine the monomer–dimer equilibrium for VraR, VraR_REC, VraR_DBD, and VraR_M13D as well as the dissociation constant for interaction between VraR_REC and VraR_DBD. Detailed methods for the AUC experiments and nonlinear regression fitting of data are provided in SI Materials and Methods. A fluorescence anisotropy assay was developed to measure the affinity of VraR and VraR_M13D with the putative R1 VraR binding site DNA (9). See SI Materials and Methods for experimental conditions and procedure.