Atypical Response Regulator ChxR from Chlamydia trachomatis Is Structurally Poised for DNA Binding

ChxR is an atypical two-component signal transduction response regulator (RR) of the OmpR/PhoB subfamily encoded by the obligate intracellular bacterial pathogen Chlamydia trachomatis. Despite structural homology within both receiver and effector domains to prototypical subfamily members, ChxR does not require phosphorylation for dimer formation, DNA binding or transcriptional activation. Thus, we hypothesized that ChxR is in a conformation optimal for DNA binding with limited interdomain interactions. To address this hypothesis, the NMR solution structure of the ChxR effector domain was determined and used in combination with the previously reported ChxR receiver domain structure to generate a full-length dimer model based upon SAXS analysis. Small-angle scattering of ChxR supported a dimer with minimal interdomain interactions and effector domains in a conformation that appears to require only subtle reorientation for optimal major/minor groove DNA interactions. SAXS modeling also supported that the effector domains were in a head-to-tail conformation, consistent with ChxR recognizing tandem DNA repeats. The effector domain structure was leveraged to identify key residues that were critical for maintaining protein - nucleic acid interactions. In combination with prior analysis of the essential location of specific nucleotides for ChxR recognition of DNA, a model of the full-length ChxR dimer bound to its cognate cis-acting element was generated.


Introduction
Two-component signal transduction systems (TCS) are a fundamental mechanism employed by bacteria for rapid adaptation to environmental changes. TCS typically consist of a membrane-bound sensor histidine kinase (HK) and an associated response regulator (RR). Upon sensing stimuli, the sensor kinase undergoes an autophosphorylation event from which the phosphoryl group is then transferred to the receiver domain of a cognate RR. Phosphorylation of the RR promotes oligomerization, stabilizing the active form of the protein. The majority of response regulators contain a DNA-binding domain that alters gene expression in response to phosphorylation [1]. The functions of RRs involve a diverse array of responses, including drug resistance, motility, osmoregulation, pathogenic host invasion and phosphate uptake, among others [2]. RRs are subdivided into families based upon the structure/function of their DNA binding domains. The largest subfamily of RRs (OmpR/PhoB) is comprised of a winged helix-turn-helix domain [3].
Members of the OmpR/PhoB RR subfamily are composed of a receiver domain that contains the site of phosphorylation and homodimerization, and an effector domain that interacts with DNA through the subfamily-defining winged helix-turn-helix motif and RNA polymerase machinery through a transactivation loop [4]. Effector domains of OmpR/PhoB subfamily members share a common tertiary structure, which results in a conserved DNA binding mechanism. The typical OmpR/PhoB effector domain is comprised of an N-terminal four strand b-sheet, a helixturn-helix motif and a C-terminal b-hairpin wing [4]. DNA interaction is achieved primarily through electrostatic interactions between residues within the helix-turn-helix motif and the DNA major groove. DNA binding is further stabilized through interactions between residues within the wing of the effector domain and the adjacent DNA minor groove. While the overall topology of effector domains is conserved, the distinct functional characteristics (e.g. specific DNA binding residues) associated with individual OmpR/PhoB effector domains are predominately provided by differences in key residues and side chain orientation.
OmpR/PhoB interdomain interactions and overall conformations are diverse and appear to reflect the relative DNA 'on-off' equilibrium for an individual RR [5]. For instance, full-length protein structures and functional studies of MtrA [6] and PrrA [7] from M. tuberculosis and DrrB [8] from T. maritima support that the receiver domain forms an extensive intramolecular interface with the effector domain effectively occluding the a4-b5-a5 dimerization interface and resulting in an equilibrium skewed towards an inactive (off) state [5]. In contrast, RegX3 [9] and PhoP [10] from M. tuberculosis and DrrD [11] from T. maritima, have relatively limited interdomain interfaces for which the DNA binding regions are in solvent accessible positions. These structures reflect the ability of unphosphorylated forms to bind DNA, albeit at lower affinity than phosphorylated forms, in a weak on-state that is enhanced by phosphorylation and stabilized by homodimeriza-tion. Importantly, these observations are from a collection of fulllength structures of unphosphorylated OmpR/PhoB response regulators limiting our understanding of the structural and functional diversity employed by this large subfamily.
A relatively new subset of response regulators are the atypical RRs which do not require phosphorylation for activity and have been described in a broad range of phylogenetically diverse organisms. The receiver domain active site typically lacks conserved residues involved in phosphorylation, yet maintains structural homology with prototypical OmpR/PhoB RRs [12,13]. Recent reports have revealed that atypical RRs can exist as monomers [14] or dimers [12] and exhibit a relatively strong affinity for target DNA in the absence of phosphorylation. It remains unclear which structural aspects allow atypical RRs to function in a phosphorylation-independent state, however it is likely these mechanisms retain a large degree of similarity to canonical OmpR/PhoB subfamily members. Of note is HP1043 from H. pylori for which a full-length NMR solution structure has been determined [12]. This solitary atypical OmpR/PhoB structure revealed that the effector domain is in a distinct, freeopen state with virtually no interactions with the receiver domain. These structural and functional observations support that this atypical response regulator is predominantly in an 'on state' in the absence of phosphorylation.
ChxR is an atypical OmpR/PhoB subfamily response regulator encoded by the medically important bacterial pathogen Chlamydia [13,15,16]. Similar to HP1043, ChxR lacks several conserved active site residues, including the phospho-accepting Asp and is able to activate transcription in a phosphorylation independent manner, leading to its classification as an atypical RR [13,15]. Additionally, ChxR exists as a stable homodimer in the absence of phosphorylation, while recognizing multiple sites within its own promoter [13]. ChxR shares 30% identity across the entire HP1043 polypeptide and only 22% identity within the effector domain. Importantly, a contrasting feature between ChxR and HP1043 is the binding of direct or inverted repeats, respectively, indicating a difference in DNA binding domain orientation. Previous studies have suggested that ChxR has a more global role in Chlamydia gene expression based upon the relatively high number of potential binding sites [17]. Additionally, ChxR expression analysis supports that it likely exerts its biological role during developmental stages that include infectious Chlamydia formation [17]. Despite previous studies that have characterized numerous characteristics of ChxR, including DNA recognition sequences and motif [13,15], the residues and regions critical for DNA binding have not been identified. Solution structure studies on both the effector domain and full-length ChxR were carried out, in order to better understand how ChxR interacts with cognate DNA. These observations guided additional functional analyses and the development of a proposed model of full-length ChxR interacting with its cognate direct repeat DNA.

Protein Purification
The effector domain of ChxR (ChxR Eff ) and full-length ChxR (ChxR) were purified as previously described [17,18]. Briefly, each protein was expressed in E. coli BL21(DE3) (Invitrogen, Carlsbad, CA) and initially purified through metal (Co 2+ ) affinity chromatography. Following their elution from the affinity column, each protein was further purified by size exclusion chromatography and determined to be .95% pure by Coomassie staining after SDS-PAGE.

NMR Spectroscopy
ChxR Eff was overexpressed in E. coli BL21(DE3) cells and 13 C/ 15 N labeled using a previously established method [19]. Following expression, ChxR Eff was purified as described above. The purified protein was equilibrated in 20 mM Na 2 HPO 4 pH 6.5, 20 mM KH 2 PO 4 , 20 mM NaCl, and 1 mM DTT and then concentrated to 1.5 mM using an Amicon (Millipore, Billerica, MA) 3,000 molecular weight cut-off centrifugal device.
For the ChxR Rec -ChxR Eff chemical shift titration experiment, ChxR Eff was 15 N labeled and purified as described herein and ChxR Rec was expressed and purified as described previously [13]. After an initial 2D-1 H-15 N-HSQC was taken of 1 mM ChxR Eff , unlabeled ChxR Rec was titrated into the protein sample. 2D-1 H-15 N-HSQCs were taken at 2:1 and 1:1 molar ratios of ChxR Eff to ChxR Rec .

NMR Structure Calculation
Structures were calculated by restrained simulated annealing using NOE based distance restraints and TALOS+ [23] based dihedral angle restraints. The torsion angle dynamics protocols of CNS 1.2 [24] were used to calculate 50 structures that were then refined using Cartesian dynamics. The 25 structures with the lowest total energies were selected for subsequent analysis. None of the 25 structures violated any distance restraints more than 0.5 Å or any dihedral angle restraints more than 5.0u. Structures were analyzed using PROCHECK-NMR [25]. Approximate interproton distances were obtained from 15 N and 13 C edited NOESY-HSQC experiments. The mixing time was 100 ms for 15 N-edited NOESY and 120 ms for 13 C-edited NOESY NMR experiments. The distance restraints were subdivided into four groups on the basis of NOE peak intensities: 1.8-2.8 Å for strong NOEs, 1.8-3.4 Å for medium NOEs, 1.8-5.0 Å for weak NOEs and 1.8-6.0 Å for very weak NOEs. In addition, backbone dihedral angles Q, and y determined using TALOS+ were restrained to 260u635u (Q) and 240u630u(y) for a-helical regions. For b-strands the values were taken as Q = 2120u630u and y = 120u630u. Final statistics are listed in Table 1 Values for T1 and T2 were determined by plotting the peak heights versus delay times and fitting the curve to a mono exponential nonlinear least squares fit, available in NMR data analysis software SPARKY [22]. The rotational correlation time (t C ; ns time scale) of a monomeric protein (,25 kDa) in solution is approximately 0.6 times its molecular weight (kDa). The t C value was calculated from the following equation [27]:

Electrophoretic Mobility Shift Assay
Electrophoretic mobility shift assays were performed as described for ChxR with DNA corresponding to the high-affinity (DR2) binding site within the chxR promoter [17]. The binding reactions contained 1 nM DNA and 50 nM, 100 nM, 500 nM, 1 mM, 5 mM or 10 mM ChxR Eff . The assay was performed in triplicate and the amount of DNA shifted was visualized and quantified using the Odyssey Infrared Imaging System (LI-COR Biosciences, Lincoln, NE).
Site-directed Mutagenesis -Mutations were introduced into the fulllength ChxR plasmid [17,18] using the QuikChange II XL sitedirected mutagenesis kit by following the manufacturer's protocol (Agilent Technologies, La Jolla, CA). All clones were verified by DNA sequencing analysis (ACGT, Inc., Wheeling, IL).

SAXS Data Collection and Evaluation
SAXS data were collected at the ALS beamline 12.3.1 (SIBYLS) LBNL Berkeley, California [28]. Data were collected using a wavelength l = 1.0 Å and with the sample-to-detector distance set to 1.5 m resulting in scattering vectors, q, ranging from 0.01 Å 21 to 0.33 Å 21 . The scattering vector is defined as q = 4p sinh/l, where 2h is the scattering angle. All experiments were performed at 20uC and data was processed as previously described [28].
SAXS data at short and long time exposures (0.5, 1 and 4 s) were merged to define the entire scattering profile. Different protein concentrations were tested for aggregation and examined by Guinier plots. The radius of gyration (R g ) was derived by the Guinier approximation I(q) = I(0)*exp(-q 2 R g 2 /3) with the limits qR g ,1.3. The curves measured for different protein concentrations (1.25, 2.5, 5.0 mg/ml) displayed a concentration dependence arising from inter-particle interaction (attractions) at q ,0.05 Å 21 and interference free scattering profiles were estimated by extrapolating the measured scattering curves to infinity dilution (see Fig. S4A and [29]). The program SCATTER was used to compute the pair-distance distribution functions, P(r) and to perform Porod-Debye analysis to obtain the P coefficient and Porod Volume [30], which indicated a dimeric state of ChxR with estimated MW = ,62 kDa (calculated MW = ,56 kDa).
The overall shape was restored from the experimental data using the program DAMMIF with P 1 symmetry operator [31]. In our rigid body modeling strategy BILBOMD, molecular dynamics (MD) simulations were used to explore conformational space adopted by the ChxR C-terminal effector domain, which we connected to the N-terminal receiver domain via an 8 residue-long flexible linker. For each registered conformation, the theoretical SAXS profile and the corresponding fit to the experimental data were calculated using the program FoXS [32]. Two sets of ChxR models were generated, one with zero constraints on the orientation of the two effector domains (unconstrained) and one requiring that a head-to-tail orientation of the effector domains be maintained (constrained). The unconstrained model set allowed for all possible orientations of the linker and effector domain, relative to the receiver domain dimer. All possible orientations were also generated for the constrained model set, with the requirement that a head-to-tail orientation be maintained for the effector domains. A Minimal Ensemble Search (MES) was ultimately used to select two conformers from a pool of all generated constrained conformers that achieved the best fit (x) to the experimental curve [33]. Chi is defined by the following equation: Comparison of the structural properties of the selected conformers allowed us to distinguish the degree of flexibility and heterogeneity of the experimental system [34]. Additional SAXS parameters are available in Table 2.

Accession Numbers
The atomic coordinates and structure factors (code 2M1B) have been deposited in the Protein Data Bank, Research Collaboratory for Structural Bioinformatics, Rutgers University, New Brunswick, NJ (http://www.rcsb.org/) as well as Biological Magnetic Resonance Data Bank (http://www.bmrb.wisc.edu/) (code 17014).

ChxR Eff alone can Bind DNA
We have previously shown that ChxR interacts with DNA corresponding to the DR2 (59-TCGATCA-N 5 -TAGATAA-39) binding site in the chxR promoter with a dissociation constant (K d ) of approximately 44 nM [17]. To determine if ChxR Eff (residues 115-227) alone can to bind to DNA, an electrophoretic mobility shift assay (EMSA) was performed with the DR2 DNA sequence. Indeed, ChxR Eff can bind to DNA (Fig. S1), albeit at a much weaker affinity than full-length ChxR. Increasing concentrations of ChxR Eff (50 nM-10 mM) were quantified with respect to DNA interaction and a K d was calculated (450675 nM). The calculated K d assumes that two ChxR Eff molecules are bound to the DNA as the DR2 sequence contains two binding sites. The approximate 10-fold decrease in DNA affinity for ChxR Eff relative to ChxR likely results from the lack of receiver domain-mediated dimerization to stabilize the protein-nucleic acid complex. This result is in agreement with previous studies on OmpR/PhoB RRs that demonstrated dimerization promotes DNA interaction [13].

Structural Analysis of ChxR Eff
We have previously elucidated the structure of the ChxR receiver domain (ChxR Rec ), which has many unique features compared to typical OmpR/PhoB subfamily members [13]. To determine if the atypical features of ChxR are limited to the receiver domain or if the effector domain also has unique features, we determined the solution structure of the effector domain (ChxR Eff ) through NMR Spectroscopy. Analysis of the structure of ChxR Eff will help facilitate the identification of residues important to DNA binding.
To determine whether an NMR approach was suitable for investigating the structure of ChxR Eff , an initial Heteronuclear Single Quantum Coherence (HSQC) spectrum was analyzed with a 1 H-15 N-labeled sample of ChxR Eff . Resonance signals for 105/ 112 residues were detected (Fig. 1). The signals were well resolved and dispersed, which was a positive indication that the structure of the protein could be determined using this method. Following data acquisition and analysis as described in the Materials and Methods section, the 25 lowest total energy structures (Fig. S2) displaying good Ramachandran plot statistics and low restraint violations were selected for further analysis. An average of 15 NOEs per residue (1637/112) constrains the ChxR Eff structure, while 84% (94/112) of the dihedral angles were defined. The RMSD of the backbone atoms of the mean structure was 0.7660.14 Å , indicating a high degree of structural similarity across the 25 lowest energy structures. All relevant NMR statistics are listed in Table 1 and the final structure of ChxR Eff was deposited in the Protein Data Bank under the identification number 2M1B and the Biological Magnetic Resonance Data Bank under code 17014.
The backbone dynamics of ChxR Eff. were investigated by 15 N T1, T2 and Heteronuclear-NOE, with an average T1 value of 0.4360.06 s and an average T2 value of 0.06460.018 s. Het-NOEs for well-defined regions in the structure were found to be in the range of 0.75-0.85 indicating that internal motions on the ps (fast motion) time scale are restricted. Het-NOEs for the N-and Ctermini as well as loop residues 174-182 were significantly smaller than average. The overall rotational correlation time (t C ) was estimated to be 8.6660.40 ns from T1/T2 ratios of residues selected from well-defined regions of the structure. This value indicates that ChxR Eff exists in a monomeric state in solution (MW in kDa60.6; expected t C for monomeric ChxR Eff (12.7 kDa) = 7.6 ns) and is in good agreement with those reported in the literature for proteins of similar size [26,27]. In additional support of a monomeric state, additional and/or broad cross peaks were not observed in the 2D-HSQC spectrum.
The electrostatic surface potential of ChxR Eff (Fig. 2B) reveals several regions of positive charge on both faces of the protein. Of potential importance is a cohort of Arg (121, 181, 191, 200 and 205) and Lys (165 and 192) residues within the putative DNA binding region (helix-turn-helix and wing motifs). Positive surfaces appear to be conserved within OmpR/PhoB subfamily effector domains [3,41,42], as would be expected from a region that interacts with the negatively charged phosphate backbone of DNA. Unsurprisingly, several of these residues have previously been implicated in DNA binding studies [12,41,[43][44][45] (Arg191, Lys192 and Arg205) and are highly conserved across this subfamily (Fig. 3).

ChxR Eff Resembles Prototypical OmpR/PhoB Subfamily Members
In contrast to the atypical receiver domain dimerization interface, the structure of ChxR Eff closely resembles that of typical OmpR/PhoB subfamily effector domain structures. While the primary sequence similarity between effector domains of the OmpR/PhoB subfamily varies from 20-65% [3], the secondary and tertiary structure of this domain is highly conserved throughout the subfamily. Structural superposition with previously determined OmpR/PhoB subfamily effector domains (OmpR [3,43,46], PhoB [42], PhoP [47], KdpE [45] and HP1043 [12]) reveals ChxR Eff aligns with an RMSD no greater than 2.50 Å to each structure ( Table 3, Fig. S3). The high structural similarity between OmpR/PhoB subfamily effector domains has been proposed to be a result of the conservation of 13 hydrophobic residues that comprise the hydrophobic core of the effector domain [48]. Indeed, twelve of the thirteen residues are conserved in ChxR Eff .
Regions of high variability between ChxR Eff and structurally characterized OmpR/PhoB effector domains exist within the transactivation loop and the disordered C-terminal extended coil. The transactivation loop is a site of variable function within this subfamily of DNA binding proteins. Not surprisingly, sequence conservation is completely absent within this region for the aligned OmpR/PhoB subfamily members (Fig. 3). Interactions with sigma factors (PhoB [49,50]) or the a-subunit of RNA Polymerase (OmpR [51]) have been documented, albeit based largely upon genetic studies. Loop orientation is most similar between ChxR and OmpR, however differences do exist within this region between the two proteins. The length of the loop in ChxR Eff is one residue shorter than OmpR and the residues that comprise their respective loops are quite different. Polar and charged residues primarily comprise the transactivation loop in OmpR, whereas the loop in ChxR Eff is almost entirely composed of hydrophobic residues. Furthermore, the four loop residues within OmpR that are important for interaction with the a-subunit of RNA polymerase and for transcriptional activation [3] are not conserved in ChxR, suggesting an alternative site of transcriptional machinery interaction might occur.

ChxR Eff Residues that are Critical for DNA Interaction
Residues important for direct interaction with DNA in OmpR/ PhoB subfamily members are generally located within the helixturn-helix or wing of the effector domain (Fig. 2) [41]. To begin identification of ChxR residues that might interact with DNA, Ala substitutions were generated within the full-length protein in surface-exposed residues of the DNA binding helix (a6) or wing (b11-b12 loop) (Fig. 2C). Proper folding of ChxR mutants was assessed by size exclusion chromatography and the ability of each mutant to bind DNA was evaluated using EMSAs. The amount of DNA bound by ChxR and shifted with each substitution was quantified and compared to wild-type ChxR (Fig. 4). Substitutions in three residues (Asn182, His186, and Lys192) within the recognition helix (a6) and a residue (Arg205) within the wing (b11-b12 loop) significantly reduced DNA interaction, while Asp184, Ile187 and Val207 substitutions appeared to bind with near wild type affinity (Fig. 4B). Substitution of Arg191 resulted in protein expression localized to inclusion bodies, potentially reflecting a role in structural stability of the effector domain. Importantly, residues implicated in DNA binding are not conserved amongst OmpR/PhoB subfamily members (Fig. 3), supporting that these may provide DNA sequence specificity for ChxR [4].

Small-Angle X-Ray Scattering (SAXS) of ChxR Supports Conformation of Activated OmpR/PhoB Subfamily
Due to the paucity of full-length structures of OmpR/PhoB subfamily RRs in the active state, little is known structurally about receiver-effector domain interactions and effector domain orientation upon phosphorylation. While ChxR is a member of the OmpR/PhoB subfamily, it is an atypical response regulator, meaning that dimerization and function are retained in the absence of phosphorylation [15,52,53]. Previous biochemical characterization of ChxR demonstrated that stable homodimers, primarily through receiver domain interactions, were maintained in solution both in vitro and in vivo [17]. The results described here, in concert with previous receiver domain structural studies [13], lead us to hypothesize that ChxR exists as a constitutively active dimer in solution. Full-length ChxR has been recalcitrant to crystallization, aggregating rapidly at concentrations higher than ,5 mg/ml (unpublished data), however, the addition of 5% (v/v) glycerol to the protein solution described within the Materials and Methods section resulted in conditions that give monodisperse samples. Thus, in order to test this hypothesis Small-angle X-ray Scattering (SAXS) was used to analyze the solution state of ChxR. SAXS, with recent technical and computational advances, has become a robust and effective technique for analyzing macromolecular structures, including the generation and experimental validation of relatively high-resolution models of proteins in solution [29,[54][55][56]. This process is particularly effective when the atomic structures of individual protein components (e.g. domains) have been determined [57], which allows the generation of extensive collections of computational models reflecting many possible conformations of the full-length protein. 'Best fit' model(s) of the complete macromolecule can then be selected based upon the experimental SAXS data. Together, these processes (e.g. smallangle scattering and computational modeling) provide complementary information about flexibly linked domains [34,58,59], shape, conformation, and assembly state in solution [29,30,60]. trachomatis ChxR Eff (residues 114-229) colored blue (N-terminus) to red (C-terminus). The topology of ChxR Eff is b6-b7-b8-b9-a4-b10-a5-a6-a7-b11-b12. B, Surface representation of electrostatic potential, generated by DelPhi [40], of ChxR Eff . Color scheme represents regions of negative (red) and positive (blue) charge density contoured at 63 e/kT. Structure is oriented the same as panel A. C, Surface representation of ChxR Eff with surface exposed side chains targeted for site directed mutagenesis colored magenta. Structure is oriented the same as panel A. All three panels are rotated 180u on the right. doi:10.1371/journal.pone.0091760.g002 Clearly, a fundamental strength of SAXS analysis is that it provides an efficient and powerful way to experimentally test models for different macromolecular assemblies and conformations in solution, as evidenced by the wealth of recent solution studies on multidomain proteins [60][61][62][63][64][65].
SAXS data were collected and analyzed on full-length ChxR. Concentration dependent scattering (from interparticle interference or aggregation) can be revealed by superposition of scaled scattering curves at multiple concentrations (1.0 mg/ml to 5.0 mg/ml, Fig. S4A). Linear dependence between Intensities, I(q), and concentrations indicated a systematic influence of individual scattering factors, S(q), up to q ,0.05 Å 21 (Fig.   S4B). As this q value was outside the first Shannon channel (q min = 0.031 Å 21 for D max = 100 Å ; [66]), infinity dilution [29] was applied to the SAXS profile at q ,0.05 Å 21 and merged with the higher concentration (5 mg/ml) SAXS profile at q .0.05 Å 21 . This scattering profile was used for subsequent data analysis. The resulting Guinier plot (Fig. S4C) was linear, which indicated the sample was relatively free of aggregation and gave a radius of gyration of 30.860.3 Å . Estimated molecular mass using the Porod Volume was ,62 kDa, which is consistent with the dimeric state of ChxR (calculated MW = ,56 kDa). All further scattering analyses were determined from this interference free (e.g. aggregation) SAXS curve. Analysis of the Kratky and Porod- Figure 3. Limited structure-based sequence alignment of OmpR/PhoB subfamily Response Regulator effector domains. Numbers above the sequences correspond to C. trachomatis ChxR. The secondary structure of ChxR is shown above the alignment. Residues are colored according to conservation (cyan = identical and purple = similar) as judged by the BLOSUM62 matrix. Red triangles below the sequences correspond to amino acid side chains identified by ChxR mutagenesis that are involved in DNA binding, blue triangles correspond to DNA interacting side chains identified within a single OmpR/PhoB subfamily member while black stars represent DNA interaction sites within multiple OmpR/PhoB subfamily members [41,[43][44][45]. Sequences used within alignment are comprised of OmpR/PhoB subfamily members with extensive structural and/or functional studies. Accession numbers are detailed in the Materials and Methods section. doi:10.1371/journal.pone.0091760.g003 Denotes the number of Effector Domain residues that superimpose within 5.0 Å distance of an equivalent position in ChxR Eff. b The LGA_S parameter represents a scoring function to evaluate the overall levels of structural similarity between two sets of coordinates. For each set of corresponding residues, it combines information pertaining to both the fraction of residues that overlap within a given RMSD window as well as those that overlap within a given distance cutoff (5.0 Å ) [37]. doi:10.1371/journal.pone.0091760.t003 Debye plots (Fig. S5) reveals that ChxR may have partially unfolded or flexible regions (Porod-Debye Exponent, P = 3.8).
Additional SAXS data collection parameters are available in Table 2.
Reconstitution of the solvated molecular envelope was carried out with a full-length model of ChxR generated by MODELLER [67] and Chimera [68], which was comprised of the previously determined [13] ChxR receiver domain homodimer (PDB ID: 3Q7R), ChxR Eff mean minimized NMR solution structure described herein and models for the 24 residue N-terminal Histag and 8 residue interdomain linker. The molecular dynamics simulation program, BILBOMD [33], facilitated the determination of ,20,000 different conformations of full-length ChxR dimer models (,10,000 each of constrained and unconstrained model sets; described in Materials and Methods section). The entire ensemble of both ChxR model sets was used to calculate theoretical SAXS profiles with FoXS [32]. In support of previous biological ChxR data indicating a ''head-to-tail'' effector domain orientation would be required to bind direct DNA repeats [13,15,17], the single best fit constrained conformer (Fig. 5) was in agreement with the experimental scattering curve (x = 3.86). However, the single best fit unconstrained conformer (Fig. 6) produced a poor fit to the experimental scattering curve (x = 8.05). This unconstrained conformer model is characterized by a ''head-to-head'' orientation for both effector domains, maintaining the two-fold symmetry present within the receiver domain. Additionally, the majority of this poor fit (Fig. 6B) occurs within the medium resolution range of the experimental SAXS profile (q = 0.1-0.2 Å 21 ), indicative of an incorrectly modeled domain conformation [29]. Analysis of the experimental scattering data with constrained versus unconstrained conformers indicated a better relative fit across all constrained models (Fig. 7). Overall, the observations from the comparative (e.g. constrained vs. unconstrained) analyses provided stronger support for the ChxR DNA binding domain being in a 'head-to-tail' orientation, reflective of the direct symmetry required to interact with the DNA repeats found within ChxR promoter sites [13,15,17].
Using the constrained dataset, Minimal Ensemble Search (MES) was applied to determine the level of conformational heterogeneity in ChxR and develop refined conformers that better match the experimental scattering (reviewed in [34]). Briefly, MES is a weighted genetic algorithm that generates a subset of conformers based upon multiple iterative modifications of highly representative models and best-fit selection with experimental data. Two conformers representative of the ChxR constrained dataset (including the single best fit conformer) were together compared to the experimental curve (Fig. 5B, C), achieving a slightly better fit (x = 3.64) than the single best fit conformer alone to the experimental curve (x = 3.86). The slightly improved x score reflects a better fit to the experimental scattering by accounting for the coexistence of multiple solution conformations following MES. Importantly, the addition of more than two conformations failed to increase the quality of fit to the experimental SAXS curve, indicating that ChxR adopts a compact, dimeric conformation with a minimal degree of flexibility between each effector domain (Fig. 5C). As such, the entirety of the scattering profile can essentially be attributed to ChxR in a compact state with effector domains in a ''head-to-tail'' orientation. Moreover, the DNA binding helices within each effector domain are ,36 Å apart, as measured from the Ala188 Ca of each chain (Fig. 5A). This orientation of ChxR potentially allows for each effector domain to interact within the DNA major groove of two recognition sites of the ChxR promoter, as the distance between major grooves is ,34 Å [69].

Flexibility of Interdomain Contacts in OmpR/PhoB Subfamily Members
Full-length RRs, including ChxR, have proven quite recalcitrant to structural determination, which likely is a reflection of the highly flexible interdomain interfaces formed by the receiver and effector domains. As such we propose that the available structures can be classified into three structural and functional subclasses based upon interdomain interactions and the resulting steric hindrance of the DNA recognition helix within the effector domain. The full-length structures of MtrA [6], DrrB [8] and PrrA [7] form extensive contacts between the receiver and effector domains (involving the activated dimer interface, a4-b5-a5) and thus belong to the ''closed'' subclass (Fig. 8A). Interdomain interfaces within these subclass members have been demonstrated to inhibit both in vitro and in vivo autophosphorylation rates [5]. Full-length crystal structures of RegX3 [9] and DrrD [11] have solvent accessible recognition helices as a result of limited interdomain contacts that do not primarily involve the a4-b5-a5 interface, and as such have been classified in the ''open'' subclass (Fig. 8B). Finally, we propose a new subclass of fulllength RRs, termed the ''free'' subclass ( Fig. 8C) whose members completely lack interdomain interactions and readily form phosphoryl-independent homodimers through the a4-b5-a5 interface. The subclass is currently comprised of HP1043 [12] and ChxR. NMR relaxation data on dimeric HP1043 supports its classification as a ''free'' subclass member, as residues potentially involved in interdomain interactions had an increased S 2 (generalized order parameter) relative to the remainder of the protein [12]. In agreement with these data, NMR chemical shift perturbation experiments with varying molar ratios of ChxR Eff and ChxR Rec failed to detect interdomain interactions (unpublished data).
These structural subclasses are merely snapshots of an equilibrium that exists between inactive and active conformational states for prototypical OmpR/PhoB RRs [5]. The active state is stabilized by phosphorylation, which enhances DNA binding affinity, and makes dimerization through the a4-b5-a5 interface energetically favorable. Atypical RRs exist in a constitutively active, dimeric state as demonstrated by previous studies on ChxR [13,17] and HP1043 [12,14,52]. Recent studies by Barbieri et al. demonstrated that interdomain interfaces stabilize the inactive state and inhibit phosphotransfer-mediated activation [5]. OmpR/PhoB subfamily RRs classified in the ''free'' subclass lack interdomain interactions and its members (ChxR and HP1043) are able to bind DNA in a phosphoryl-independent manner [10,12,17]. The structural studies discussed herein provide further support for the delineation of three separate OmpR/PhoB RR structural subclasses.

DNA Interactions in OmpR/PhoB Subfamily Members
OmpR/PhoB subfamily RRs regulate a diverse collection of biological processes involving signaling, metabolism and development, among others. As such, great diversity in the target DNA sequences of these members is not unexpected. This naturally leads to unique protein-DNA interactions among subfamily members. Residues critical for DNA binding have also been identified in OmpR [43], PhoB [41], PhoP [44], KdpE [45] and HP1043 [12]. All of these residues are localized to the recognition helix (labeled a6 in Fig. 2) or the minor groove binding wing. Of the residue positions implicated in multiple OmpR/PhoB subfamily member mutants (black star in Fig. 3), all three predominantly involve Arginine (including Arg205 described within), the most frequent side chain involved in protein-DNA interactions [70]. This suggests these side chains are involved in non-specific DNA interactions found across various OmpR/PhoB subfamily members. Residues identified in only a single member of the OmpR/PhoBs subfamily (blue triangles in Fig. 3) predominantly involve non-conserved positions without a preference for aliphatic or charged side chains, suggesting potential roles in site-specific  interactions. As expected, amino acids within the recognition helix that stabilize effector domain tertiary structure (positions 183, 187, 190 and 194; ChxR numbering) have not been implicated in DNA binding. Based up the previous observations, residues imparting protein specific-DNA base recognition are likely to be found within the few non-conserved side chains of the OmpR/PhoB recognition helix. The effect of substitutions at these specific-DNA base recognition sites in ChxR was not evaluated for their capability to bind to alternate ChxR recognition sites (e.g. DR1, 3-6; [17]). While the proposed ChxR binding motif has an overall low nucleotide conservation, three nucleotides are (TXGAXXX) are highly conserved among ChxR binding sites. Additionally, when these conserved nucleotides were mutated and naturally variant, ChxR binding was severely reduced [17]. These observations would support that the amino acids important for ChxR binding to DR2 site, and these cognate conserved nucleotides, are also important for binding to alternate sites as well. Clearly, experimental analyses will be needed to support this hypothesis.

DNA Homology Model of ChxR Bound to Direct Repeat
The large majority of OmpR/PhoB subfamily RRs have been found to bind direct DNA repeats, which thus requires these proteins to form functional dimers [71,72]. Structural studies of full-length OmpR/PhoB RRs bound to their cognate DNA repeats have proven elusive, with only the PhoB effector domain in complex with the pho box having been reported [41]. While a fulllength structure for PhoB has yet to be determined, the structures of each individual domain are available, in addition to the BeF 3 activated a4-b5-a5 receiver dimer [73]. Furthermore, autophosphorylation of PhoB suggests it has a minimal interdomain interface, much like ChxR [5]. Each of these structures demonstrates strong structural similarity with the respective ChxR domain (Fig. S6). Of the 17 amino acid contacts within the PhoB-DNA complex, 9 are conserved within ChxR. Furthermore, the majority of the contrasting side chain interactions can be found within the recognition helix, which is anticipated given the differences in target DNA sites [17,41]. These similarities suggested that the active state ChxR structure could be modeled onto the PhoB-DNA complex (Fig. 9). The four ChxR residues that were demonstrated by site directed mutagenesis to be critical for DR2 interaction are within appropriate distances to bind each direct DNA repeat. Thus, the solution structure of ChxR in a DNA-binding state provides a model for comparison within the OmpR/PhoB subfamily. However, as atypical RRs appear to lack interdomain interfaces, the primary site for regulation of prototypical RRs, further studies are needed to elucidate how these proteins are turned ''on'' and ''off''.
The proposed model of full-length ChxR has the domains (receiver and effector) in different paired orientations. Specifically, the regulatory domains have been determined to interact with two-fold symmetry, which is strongly supported by crystallographic data for ChxR and almost all other OmpR/PhoB subfamily members. However, the DNA binding domains of ChxR appear to have tandem symmetry (head-to-tail) based upon the SAXS analysis presented herein, as well as their monomeric state and reduced DNA binding affinity in the absence of the receiver domain. This matches the orientation of the DNA sequence motif (direct repeat) recognized by ChxR [13]. The resulting orientation is not unexpected for an OmpR/PhoB subfamily member, and was recently highlighted by Bachhawat et al. [73]. They suggested that orientational constraint is lost between the domains of PhoB, based upon the observations of the receiver domain structure (twofold symmetry) and DNA repeat-bound effector domain (tandem symmetry) [41]. This prediction was further strengthened by the recently reported crystal structure of PhoB in complex with RNAP and DNA [74].
Support for the ChxR-DNA model is limited by the relative structural absence of effector domains bound to DNA with only two examples of protein bound to neighboring sites. However, the overall expectation of independent domain orientation is further supported as most OmpR/PhoB subfamily response regulators bind to direct repeats of DNA sequences, which would seemingly require two similarly oriented winged helix DNA binding motifs. One of the few OmpR/PhoB subfamily members that bind to inverted repeats is the atypical response regulator HP1043 [75]. It is also among the few full-length structures from this family that have been determined and, of note, the DNA binding domains are oriented similar to the receiver domain with two-fold symmetry, which is in the best orientation for binding inverted repeats. Clearly, the absence of a full-length structures from the OmpR/ PhoB subfamily bound to tandem repeat DNA limits any strong conclusions regarding the orientation of these molecules, although proposing a model that incorporates domain independent orientation seems best supported by the overall observations for OmpR/PhoB subfamily response regulators, including the structural data presented herein.
Lastly, the recent advances in Chlamydia genetics [76,77] and the development of molecular tools [78][79][80] will enable studies related to the function and role of ChxR to be performed directly in Chlamydia. Observations described herein are essential in directing those future studies, specifically the possibility of generating dominant negative ChxR variants. Substitutions that rendered ChxR incapable of binding to DR2 still retained the ability to form homodimers. Furthermore, prior studies demonstrated that binding to both direct repeats was essential to stabilizing the protein-nucleic acid complex [17]. Thus, conditional expression [79] of DNA binding deficient ChxR could form heterodimers with wild-type ChxR and effectively disrupt the function of ChxR in Chlamydia. This would enable phenotypic and functional studies (e.g. transcriptome analysis) to provide a better understanding of the role of ChxR in the chlamydial developmental cycle and pathogenesis.  Figure S4 Experimental Scattering Profile, Guinier Plot and P(r) function of ChxR. A, Experimental scattering profile of ChxR for 5 mg/ml (blue), 2.5 mg/ml (magenta), 1.25 mg/ml (green) and extrapolated curve at the infinity dilution (cyan). B, Intensities obtained for scaled SAXS profiles (panel A) at q = 0.02, 0.03 and 0.05 Å 21 indicate effect of the Structure factor at higher protein concentration for q ,0.05 Å 21 . To eliminate this effect we used infinity dilution for further data analysis C, Guinier plot with Guinier region. A linear dependence of ln(I(q)) vs. q 2 indicates the sample is free of aggregation. Radius of gyration (R g ) values as obtained from Guinier plot: Glmn R g = 30.860.3 Å . D, Pair distribution function P(r) calculated from the SAXS curve shown in Figure 5B. (TIF) Figure S5 Kratky and Porod-Debye Plot of ChxR. A, Experimental SAXS curve shown as a Kratky plot indicate minimal flexibility. B, We performed a Porod-Debye analysis to obtain direct insights into their flexibility. In a plot of the normalized q 4 NI(q) vs. q 4 , the positive slope and obtained Porod-Debye coefficient of P = 3.8 is consistent with inter-domain flexibility (27). This observation suggests that the ChxR C- Figure 9. Proposed model of full-length ChxR binding to tandem DNA repeat. The solution state of full-length, dimeric ChxR (each polypeptide colored a different shade of gray, surface representation) was manually overlayed onto the cocrystal structure of the PhoB effector domain from E. coli bound to its cognate pho box (PDB ID: 1GXP). The structure is rotated 90u about the protein:DNA interface on the right. Surface exposed residues within ChxR that were implicated in DNA binding by site-directed mutagenesis are highlighted in magenta.