Structural Basis of a Physical Blockage Mechanism for the Interaction of Response Regulator PmrA with Connector Protein PmrD from Klebsiella pneumoniae *

Background: PmrD binds to phospho-PmrA and sustains its phosphorylation state. Results: Phospho-PmrA interacts with PmrD via several specific intermolecular interactions. Conclusion: A steric inhibition mechanism was proposed for protecting phospho-PmrA against dephosphorylation. Significance: This work provides novel data revealing how a connector protein protects an activated response regulator. In bacteria, the two-component system is the most prevalent for sensing and transducing environmental signals into the cell. The PmrA-PmrB two-component system, responsible for sensing external stimuli of high Fe3+ and mild acidic conditions, can control the genes involved in lipopolysaccharide modification and polymyxin resistance in pathogens. In Klebsiella pneumoniae, the small basic connector protein PmrD protects phospho-PmrA and prolongs the expression of PmrA-activated genes. We previously determined the phospho-PmrA recognition mode of PmrD. However, how PmrA interacts with PmrD and prevents its dephosphorylation remains unknown. To address this question, we solved the x-ray crystal structure of the N-terminal receiver domain of BeF3−-activated PmrA (PmrAN) at 1.70 Å. With this structure, we applied the data-driven docking method based on NMR chemical shift perturbation to generate the complex model of PmrD-PmrAN, which was further validated by site-directed spin labeling experiments. In the complex model, PmrD may act as a blockade to prevent phosphatase from contacting with the phosphorylation site on PmrA.

For all organisms, sensing and responding to environmental cues is essential to their survival. Compared with eukaryotic signaling cascades, which are mainly involved in mediating signals through specific Ser-, Thr-, and Tyr-phosphorylated residues among the related sub-tracts, in prokaryotic signaling systems, a distinct phosphorylation scheme predominates, the two-component system (TCS) 2 (1)(2)(3). Hundreds of TCSs have been found in eubacteria, archaea, and a few eukaryotic organisms (4), and the TCS is the most prevalent system in bacteria for transducing external information into the cell and coping with environmental stresses (2,5). A PmrA-PmrB TCS, dealing with cationic antimicrobial peptides and polymyxin resistance in pathogens (6), is responsible for sensing the external stimuli caused directly or indirectly by excess Fe 3ϩ , low Mg 2ϩ , and mild acidic environments (7)(8)(9). Under this TCS, pathogens can alter the composition of their cell walls to resist neutralization of polymyxin for drug resistance and allow bacterial survival within macrophages by reducing the affinity with cationic antimicrobial peptides (10,11).
The increasing antibiotic resistance to Klebsiella pneumonia bacteria, a common cause of nosocomial bacterial infections causing pneumonia and urinary tract infections, especially in immunocompromised patients (12), led us to investigate how the virulence and drug resistance persist via the associated TCS. Similar to most response regulators, the PmrA of K. pneumoniae belonging to the OmpR-PhoB family is composed of an N-terminal receiver domain (Fig. 1) and a C-terminal effector/ DNA-binding domain. Interestingly, at low Mg 2ϩ concentration, the PhoP-PhoQ system promotes the expression of a small basic protein, PmrD (7), which can prevent the intrinsic dephosphorylation of phospho-PmrA and enhance the expression of PmrA-activated downstream genes (13). Therefore, PmrD is termed a "connector" because it connects the PmrA-PmrB and PhoP-PhoQ systems (14). Besides PmrD, other small-sized proteins that connect TCSs for Escherichia coli include SafA, which connects the PhoP-PhoQ and EvgS-EvgA systems (15); IraM, which connects PhoP-PhoQ and binds to response regulator RssB, thereby preventing the proteins from recruiting targets for degradation (16); and MzrA, which connects CpxR-Cpxa and OmpR-EnvZ systems (17). To date, the complicated mechanisms of cross-regulation among these TCSs are still not clear (14), and no detailed structural information is available about how the phosphorylation site of the * This work was supported by grants from Academia Sinica and the National Science Council (NSC 97-2311-B-001-013-MY3) in Taiwan and in part by the Ministry of Education (Taiwan) under the ATU plan. The atomic coordinates and structure factors ( response regulator is protected by the connector protein in the interconnecting signal transduction pathways such as PmrD and phospho-PmrA. Previously, we determined the solution structure for K. pneumoniae PmrD and mapped its binding sites when it bound to PmrA N (18). In this study, we present the structure of the BeF 3 Ϫactivated N-terminal receiver domain of PmrA (PmrA N ) solved by x-ray crystallography. The backbone assignment of PmrA N involved transverse relaxation-optimized spectroscopy-type three-dimensional NMR experiments with 2 H, 13 C, 15 N-labeled PmrA N . With the chemical shift perturbations of PmrA N when binding with PmrD and structures of PmrD and PmrA N , we used HADDOCK (19) to generate a complex structure of the PmrD-PmrA N heterotetramer. Furthermore, using NMR hydrogen/deuterium exchange (HDX) exchange studies combined with a paramagnetic spin-labeling approach, we further confirmed the reliability of this complex model. We found that PmrA N interacts with PmrD via the residues near the active sites that are widely used by response regulators for intermolecular protein-protein interactions with histidine kinase. The results of this study can help reveal the molecular mechanisms of the interaction mode between connector proteins and phosphorylated response regulators in TCSs.

Cloning, Expression, and Purification of the Recombinant
Proteins-The coding regions of pmr and PmrA N (N-terminal 123 residues) were PCR-amplified from the genomic DNA of K. pneumoniae. The amplified gene products were cloned as an NdeI/XhoI fragment into the pET29b vector (Novagene). The resulting plasmids (pET-PmrD and pET-PmrA N ) allowed for the in-frame fusion of each coding region containing an additional LEHHHHHH sequence at the C terminus to facilitate protein purification. The recombinant proteins were overexpressed in the host E. coli strain BL21(DE3) (Novagen) induced with 1 mM isopropyl 1-thio-␤-D-galactopyranoside at 37°C (for PmrD) or 30°C (PmrA N ). After lysis with use of an M-110S microfluidizer (Microfluidics) and subsequent centrifugation, the overproduced proteins were purified from the soluble fraction by affinity chromatography on His-Bind resin (Novagene) and size-exclusion chromatography with a Superdex75 10/300 GL column. The eluted samples were dialyzed against 50 mM Tris (pH 8.0), 100 mM NaCl, and concentrated by use of Amicon (molecular weight of 5000; Millipore). For isotopically enriched samples, cells were grown in M9 minimal medium (20) containing 1 g/liter 15 NH 4 Cl and 2 g/liter [ 13 C]glucose. PmrD and PmrA N mutants were generated according to QuikChange TM protocol (Stratagene). The authenticity of the recombinant proteins was verified by SDS-PAGE and mass spectrometry.

Crystallization of BeF 3
Ϫ -activated PmrA N -For crystallization trials, the BeF 3 Ϫ -activated PmrA N was first concentrated to 5 mg/ml in buffer containing 20 mM sodium phosphate, pH 6.0, and 30 mM NaCl with use of a Centricon concentrator (Millipore). The BeF 3 Ϫ -activated PmrA N was obtained by adding 5.3 mM BeCl 2 (Fluka), 35 mM NaF, and 7 mM MgCl 2 . The hanging drop vapor diffusion method was then performed at 298 K by mixing 1 l BeF 3 Ϫ -activated PmrA N with an equal volume of crystal screening solutions. Hexagonal crystals appeared after 3 days under the condition of 100 mM imidazole and 1 M sodium acetate at pH 6.5. The crystals belong to space group P41 and contain one PmrA N dimer per asymmetric unit. Collection of the cryogenic multiwavelength anomalous dispersion data involved use of an Area Detector Systems Quantum-315 charge-coupled device detector with a synchrotron radiation x-ray source at beamline BL13B1 at the National Synchrotron Radiation Research Center in Taiwan. X-ray diffraction data integration and scaling involved use of the HKL2000 package (21). The extension of initial phases to 1.99 Å and the preliminary auto-model building involved use of RESOLVE (22). XtalView (23) was used to examine electron density maps and for manual model building. Further refinement involved use of CNS (24) and PHENIX (25). After completion, the R factor of the final model for all reflections above 1 between 25.38 and 1.70 Å resolution was refined to 16.5%, and an R free value of 17.5% was obtained with 5.2% randomly distributed reflections (26). The structures were visualized and plotted with use of PyMOL (27).  a molar ratio of 1:2) in a Shigemi NMR tube (Allison Park, PA) were performed at 310 K with the use of Bruker AVANCE 600 and 800 NMR spectrometers (Bruker, Karlsruhe, Germany) equipped with a triple ( 1 H, 13 C, and 15 N) resonance cryoprobe including a shielded z-gradient. All heteronuclear NMR experiments were performed as described (28). Sequence-specific assignment of the backbone atoms was achieved by independent connective analysis of transverse relaxation-optimized spectroscopy-type three-dimensional HNCA, HN(CO)CA, CBCA(CO)NH, HNCACB, HNCO, and HN(CA)CO experiments. The chemical shifts of individual spin systems (HN, N, C␣, C␤, CЈ, and H␣) were collected manually, and the backbone resonances were assigned by visual inspection. Transverse relaxation-optimized spectroscopy-based NOESY and HNCO spectra were acquired on the triple-labeled PmrA N in complex with PmrD. From the resonance assignments for free PmrA N , most of the resonances in the bound PmrA N could be easily identified because their cross-peaks were well superimposed. For the residues with significant shift perturbations, their backbones were assigned by comparing NOE connectivities and carbonyl carbon chemical shifts between the free and bound forms. All NMR spectra were processed and analyzed by use of Topspin (Bruker Biospin), NMRPipe (29), and NMRView (30) software packages.
HDX Experiments-HDX experiments were initiated by adding D 2 O (99.9%) to lyophilized protein, which had been prepared at the required pH and buffer conditions. The concentration of 15 N, 13 C, 2 H-labeled PmrA N was ϳ100 M, and the unlabeled PmrD was ϳ200 M. The 15 N-1 H selective optimized flip-angle short-transient heteronuclear multiple quantum coherence procedure (31) was used to obtain the 15 N-1 H correlation spectra with 1024 (t 2 ) and 64 (t 1 ) complex points and 4 scans at 310 K.
Chemical Shift Perturbation Experiments-To map the binding sites on PmrA N with PmrD, we collected 1 H-15 N HSQC spectra for 2 H, 15 N-labeled PmrA N with unlabeled PmrD at a molar ratio of 1:2. We investigated and ruled out the possibility that the shift changes of PmrD were due to constituents of BeF 3 Ϫ . All spectra processing were analyzed by use of XWIN-NMR (Bruker Biospin) and SPARKY (32). Normalized chemical shift changes were calculated as follows.
The cut-off (0.135 ppm) was set as the S.D. for all chemical shift changes. All residues with values above the cut-off were considered affected by interaction with PmrD.
Complex Structure Determination using HADDOCK-The information-driven docking program HADDOCK (version 2.0) (19) was used to generate a PmrD-PmrA N complex model. The starting structures for docking were an x-ray crystal structure of PmrA N and the 10 lowest energy structures of PmrD, which were refined with the constraints measured from the chemical shift perturbations (18). From NMR titration data, different sets of ambiguous interaction restraints were used to generate the complex model. PmrD contained sets that included the active residues 4, 26, 48, 49, 50, 52, 67, 70, 71, 73, and 74 and passive residues 1, 24, 47, 64, and 66. The PmrA N model contained sets that included the active residues 8, 9, 53, 59, 99, 104, 105, 138, 139, 183, 189, 229, 234, and 236. The active residues were chosen on the basis of the chemical shift perturbation data and high solvent accessibility (Ͼ50%). The passive residues are the solvent-accessible surfaces neighboring active residues. During the rigid body energy minimization, 10,000 structures were calculated, and the 199 best solutions based on the intermolecular energy were used for the semi-flexible, simulated annealing, followed by explicit water refinement. The best 199 docked models were clustered by use of a cut-off r.m.s.d. of 3.5 Å. The clusters were ranked by the averaged HADDOCK score of their top 10 complex models (Table 1).
Site-directed Spin Labeling and Paramagnetic Relaxation Enhancement (PRE) Experiments-PmrDC54 constructs containing a single cysteine residue were generated by use of sitedirected mutagenesis to convert existing extra cysteine residues to serine. Mutations were confirmed by DNA sequencing over the entire open reading frame and were found to be properly folded and capable of binding BeF 3 Ϫ -activated PmrA N as for wild-type PmrD. Purified mutants were modified with the thiospecific spin-label reagent MTSL ((1-oxyl-2,2,5,5-tetramethyl-3-pyrroline-3-methyl) methanethiosulfonate)) purchased from Affymetrix-Anatrace. The protein solution was added to a 20-fold molar excess of MTSL solution dissolved in acetone. The spin-labeling reactions were performed in the dark at room temperature overnight. Excess MTSL was removed by extensive dialysis, and the sample was concentrated to 150 M. For NMR HSQC experiments, the complex was prepared with a molar ratio of 1:2 of 15 N, 13 C, 2 H triple-labeled PmrA N : PmrDC54 with MTSL or 15 N, 13 C, 2 H triple-labeled PmrD: PmrA N D56C with MTSL. All experiments were repeated after the spin label was reduced with 5-fold excess concentrated ascorbic acid and DTT (2-5 l from a concentrated stock; dilution Ͻ 1.25%) to NMR samples. Samples were placed at 25°C for at least 1 h after the addition of reducing agent. The intensity ratios of paramagnetic versus diamagnetic spectra were obtained from the peak heights and further normalized before distance constraints were converted.

Structure of PmrA N and Dimer
Interface-Beryllium fluoride, a phosphoryl analog (33), was used to produce an activated form of PmrA N for structural studies. The structure of PmrA N was solved with experimental phases from molecular replacement data at 1.70 Å. The initial phases of PmrA N were solved with use of the molecular replacement software MOLREP from the CCP4 suite (34). The overall structure of PmrA N adopts a typically (␤␣) 5 topology and follows the signature CheY-like ␣/␤-fold architecture of receiver domains shared by all response regulators (35). Alternating ␤-strands and ␣-helices in the primary structure fold into a central five-stranded parallel ␤-sheet surrounded by two ␣-helices on one side and three on the other side ( Fig. 2A). The asymmetric unit has only one molecule that forms a dimer with another crystallographic symmetry-related molecule using the face formed by helix ␣4, strand ␤5, and helix ␣5 (Fig. 2A). The presence of a large number of intra-and intermolecular interactions involving hydrophobic and polar residues was revealed by analysis of the interface. The ␣4 helix of one monomer packs against the ␣5 helix of the other monomer through a hydrophobic patch formed between Leu-84 (␣4), Ile-88 (␣4), and Leu-91 (␣5) and the aliphatic portions of Ala-104 (␣4), Ala-110 (␣5), and Ala-114 (␣5) side chains (Fig. 2B). The interface is further stabilized by a number of side chains involved in attributive salt bridges, especially toward the center region, which creates an extensive network that surrounds the hydrophobic packing between the ␣4 and ␣5 helices. Intermolecular salt bridges are formed between Arg-87(␣4)-Glu-107(␣5), Asp-92(␣4)-Arg-113(␣5), and Asp-97(␤5)-Arg-111(␣5) at the outer sides (Fig. 2C). The x-ray structures of other members of the activated form of receiver domains in the OmpR-PhoB subfamily were reported previously, including PhoB and PhoP from E. coli and DrrB and DrrD from Thermotoga maritime (36 -38). In all cases, the receiver domains form a dimer with an interface at ␣4-␤5-␣5, which reveals a common dimerization mechanism on activation with members of the OmpR-PhoB subfamily.
Coordination of the Active Site in PmrA N -PmrA is phosphorylated at Asp 51 , the residue that corresponds to the conserved phosphorylation site shared by the receiver domains of response regulators (39). In the crystal structure, Mg 2ϩ binds to F 1 of BeF 3 Ϫ , the side chain carboxyl oxygens of Asp-8 and Asp-51; the main chain carbonyl oxygen of Gly-53, a water molecule hydrogen-bonded to Asp-8 and another water molecule hydrogen-bonded to Asp-9, thus satisfying the octahedral coordination of Mg 2ϩ (Fig. 2D). The BeF 3 Ϫ is non-covalently bound to Asp-51 centered within the active site. The conserved Lys-101 is involved in salt bridges with Asp-9 and F 3 of BeF 3 Ϫ . The conserved Thr-79 forming a hydrogen bond with the F 2 of BeF 3 Ϫ , involved in the activation mechanism in other response regulators, is in an inward orientation. The rotation of Thr-79 is accompanied by a rotation of the backbone N of the next residue, Ala-80, which forms a hydrogen bond with F 3 of BeF 3 Ϫ . Ser/Thr and Phe/Tyr located in the ␤4-␣4 loop and ␤5 (Thr-79 and Tyr-98 in PmrA, Fig. 2D) are the signature aromatic "switch" residues involved in rearrangements associated with phosphorylation (3,40). In the PmrA N structure, Thr-79 is oriented toward the active site to coordinate with BeF 3 Ϫ , whereas Tyr-98 adopts an inward position, forming a hydrogen bond with the main chain carbonyl oxygen of Arg-81 situated at the middle of the ␤4-␣4 loop. Crystallographic data are summarized in Table 2, and electron density for the active site of the refined model is shown in Fig. 2E. 1 H, 13

C, and 15 N Assignments and Secondary Structure Comparisons between Solution and X-ray PmrA N Structures-We
could not grow good crystals of PmrA N in complex with PmrD and therefore used NMR techniques to obtain detailed structural information and gain insight into the protein-protein interactions (41). Two-dimensional 15 N-1 H HSQC spectrum of PmrA N is well dispersed, which indicates a well folded structure. Following a standard sequential assignment procedure, all of the observed cross-peaks in the HSQC spectrum were completely assigned, which correspond to 97% of 1 H N -15 N pairs (115 of 119). The four unassigned residues were His-44, Leu-105, Gln-122, and Gly-123. Their cross-peaks disappeared, presumably because of solvent exposure. For the other backbone atoms, 98% of 1 H␣ resonances (116 of 119) and 98% of 13 CЈ, 13 C␣, and 13 C␤ resonances (351 of 360) were assigned. H␣ of Arg-87 was shifted up-field at 3.54 ppm, which is likely due to the shielding effect by Tyr-98 as seen in the crystal structure. The chemical shift index (29) based on H␣, C␣, C␤, and CЈ chemical shifts showed that PmrA N contains four ␤-strands (residues 2-7, 47-51, 74 -79, and 95-100) and five ␣-helices (residues 10 -22, 34 -41, 59 -70, 84 -92, and 104 -120), which are similar to those in the crystal structure.
Interaction between PmrD and PmrA N -Previously, we studied the binding stoichiometry of the PmrD-PmrA N complex in the presence of BeF 3 Ϫ (18). The binding ratio of 1:1 was determined by size-exclusion chromatography and 1 H transverse relaxation time measurements. The equilibrium dissociation constant of ϳ1 M was measured by surface plasma resonance. To further identify the sites of interaction of PmrA N with PmrD, we performed chemical shift perturbation experiments, in which the chemical shift changes on the 15 N, 1 H N correlation plot of the BeF 3 Ϫ -activated PmrA N were detected with the addi-tion of 2-fold unlabeled PmrD (Fig. 3A). The addition of PmrD caused significant chemical shift changes, the intermediate exchange line broadening on several residues, and emerging cross-peaks for backbone 15 N and 1 H N resonances. At a molar ratio of 1 to 0.9, two separate NMR signals, one for the free and the other for the bound form, were clearly seen for some residues such as Asp-9, Val-49, Asp-56, and Gln-73. Therefore, we suggested that the interaction belongs to the slow exchange regime on the NMR timescale. The backbone resonances of Lys-2 and Leu-12 were missing because of peak broadening beyond detection. Leu-105 was not observed in the free form but appeared in the presence of PmrD, which suggests lower solvent accessibility of the amide proton of Leu-105 in the complex. We found no significant intermolecular NOEs in the three-dimensional NOESY-HSQC spectra, which may be due to weak binding of the protein complex. A total of 14 residues of PmrA N showed significant perturbations with the addition of a 2-fold molar ratio of PmrD (Ͼ 0.15 ppm) and were labeled on the chemical shift perturbation bar plot (Fig. 3B). Mapping these residues onto the structure showed that most of them are located near the phosphorylation site of PmrA N , which comprises the ␤1-␣1 loop, N-terminal ␣1, ␤3-␣3 loop, N-terminal ␣3, C-terminal ␤4, ␤4-␣4 loop, C-terminal ␤5, ␤5-␣5 loop and N-terminal ␣5 (Fig. 3C).

HDX Study and Solvent Accessibility between the Free and Bound
Forms of PmrA N -HDX rates can provide information about the solvent accessibility for different backbone amide protons. Here we used HDX to map the binding interface in PmrD-PmrA N . First, we used the free form of 15 N-labeled PmrA N for the HDX study and found that ϳ41 cross-peaks were water-inaccessible (Fig. 4A). Mapping these residues onto the PmrA N structure showed that they were mainly located on the central five-stranded parallel ␤-sheet or the interface mediated by the ␣4-␤5-␣5 face but not on the surface near the phosphorylation site. Next, we added unlabeled PmrD into 15 N-labeled PmrA N and repeated the same experiment under the same conditions and compared the results with the HDX data for free PmrA N . Most of the water-inaccessible residues were similar to those in the free PmrA N (Fig. 4B), so these residues are not involved in binding with PmrD. However, three residues, Leu-11, Phe-103, and Arg-81, showed a lower water-accessible rate in the presence of PmrD (Fig. 4B), presumably because of the steric occlusion effect of PmrD. Interestingly, Leu-11, Arg-81, and Phe-103 are all close to the phosphorylation site and are consistent with the binding interface determined by chemical shift perturbation (Fig. 4C).
Modeling of the PmrD-PmrA N Complex-With formation of the PmrD-PmrA N complex, we observed significant broadening of NMR signals for both PmrD and PmrA N , which may be due to the slower tumbling of the complex and/or a contribution from intermediate timescale chemical exchange broadening. These significantly broadened NMR resonances limited the acquisition of intermolecular NOEs critical for determining the structure of multiprotein complexes, although this effect aided in identifying interfacial residues. Because of a lack of intermolecular NOEs, we generated the complex model of a PmrD-PmrA N heterotetramer using the experimental data-driven docking method of HADDOCK (19). The generated complex  AUGUST 30, 2013 • VOLUME 288 • NUMBER 35 model is dominated by intermolecular interactions between residues located on the active site pockets of PmrA N and the solvent-exposed face of the ␤-barrel of PmrD (Fig. 5A).  5B). In addition, two intermolecular salt bridges were found between Lys-64 -Glu-31 with 2-fold symmetry (Fig. 5B). There-fore, the hydrogen bonds and salt bridges for each partner interact with appropriately charged residues on the corresponding interface, providing a specificity of structural arrangement for the interaction between PmrD and PmrA N . PRE Effect of Site-directed Spin Labeling on PmrDC54 and PmrA N D56C-To double-confirm the orientations of the PmrD-PmrA N complex model, we used site-directed spin labeling with a nitroxide spin-label compound, MTSL, covalently attached to a Cys residue (42) for PRE experiments, which  can be used as an independent biophysical method to obtain intermolecular long distance (Ͻ30 Å) information. PmrD contains three Cys residues (Cys-17, Cys-35, and Cys-54), and PmrA N contains only one Cys residue (Cys-27). To determine which Cys residues could be labeled under the conditions described under "Experimental Procedures," we initially covalently linked the stable mutant 15 N-labeled PmrD C35S and wild-type 15 N-labeled PmrA N using MTSL. The backbone HN of Cys-54 in PmrD C35S was significantly shifted as compared with PmrD, so Cys-54 was able to bind with MTSL. However, Cys-27 of PmrA N did not show any shift perturbation on adding MTSL, so Cys-27 is unlikely to bind to MTSL. We therefore produced two mutants, C17S/C35S on PmrD (PmrDC54) and D56C on PmrA N (PmrA N D56C), for PRE experiments. Spin labeling did not change the protein-protein interaction because the 1 H-15 N HSQC spectrum of PmrA N in complex with MTSLlabeled PmrDC54 with a reduced proxyl group was similar to that of PmrA N in complex with unlabeled PmrD (Fig. 6A) and vice versa. The unpaired electron of MTSL caused line-broadening effects on nearby residues in a distance-dependent manner. Line-broadening effects on the 15 N-labeled PmrA N were observed in the complexes with MTSL-labeled PmrDC54 (Fig.  6A). The superimposed spectra of the reduced (black) and oxidized (red) states showed significant R 2 relaxation enhancement effects. With the spin label on Cys-54 of PmrD, the signals on several residues of PmrA N , such as Asp-82, Leu-84, Val-100, Phe-103, and Leu-105, were significantly reduced in the oxi-dized state (Fig. 6, A and B). Residues Asp-82, Leu-84, Val-100, and Phe-103 were all close to the active site of PmrA N , as seen in the crystal structure. With MTSL attached to D56C of PmrA N , three peaks corresponding to PmrD residues Gly-24 -Ala-25 and the side chain amide peak of Trp-3 were severely attenuated (data not shown). These data suggest that the interaction brings the binding surface of PmrD (the open ␤-barrel area) (18) and the active site of PmrA N close to each other. In conclusion, the observed intermolecular PRE data are generally consistent with the HADDOCK-derived complex structure (Fig. 6C) and hence can provide an independent biophysical method to confirm the long range distance information between corresponding regions of the complex model.

DISCUSSION
Several crystal structures of receiver domains of response regulators have been solved (43)(44)(45)(46), but little structural information is available regarding how they interact with each other in multiprotein complexes. As well, knowledge of the structure of the receiver domain in complex with the connector protein is limited. Several groups have reported that the active site from the receiver domain of the response regulator is widely used for intermolecular protein-protein interactions with histidine kinase. Here, we present the crystal structure of PmrA N , in which a hydrogen bond couples with the carbonyl oxygen atom of Arg-81 situated on the ␤4-␣4 loop and side chain of the aromatic residue Tyr-98, which is similar to the "T-loop-Y" model observed for the activation mechanism of a response regulator (47). In the model, the conformation of the ␤4-␣4 loop was gated by phosphorylation and the position of the active site residue Thr-79, thus resulting in an inward orientation of the aromatic switch residue (47). Next, we demonstrated that PmrA N uses a similar surface to interact with the TCS connector protein PmrD (Fig. 7A). Therefore, PmrD is likely a competitive inhibitor of histidine kinase because it binds with the response regulator at a similar interface surrounding the phosphorylation site (48). In the PmrD-PmrA N complex model, the phosphoacceptor Asp-51 faces the open mouth area of the ␤-barrel of the PmrD structure, which reflects the protection mode of the connector protein on the phosphorylated response regulator. The orientation of PmrA N relative to PmrD can be attributed to three specific interactions. The first interaction is between the ␣1 and the ␤5-␣5 loop region of PmrA N and loop 4 and N-terminal ␤1 in PmrD. The second interaction is between the ␤1-␣1 loop of PmrA N and loop 2 of PmrD. The third interaction is between the ␤4-␣4 loop and ␤3-␣3 loop containing the Asp-51 phosphorylation site of PmrA N and C-terminal ␤6, loop 6, N-terminal ␤1, and ␤3 of PmrD. Interactions 2 and 3 create an environment that hides the phosphoryl analog of BeF 3 Ϫ between PmrD and PmrA N . The binding interface of PmrA N is also similar to that observed in the crystal structure of the DHp domain in the ThkA-TrrA complex and Spo0F in the Spo0B-Spo0F complex from Bacillus subtilis, in which TrrA and Spo0F are the response regulators. In the PmrD-PmrA N complex, the phosphoacceptor Asp-51 in PmrA faces the His-70 of PmrD, similar to the phosphodonor His in the HK DHp domain (His-547 in ThkA) (48,49). Hydrogen bonds and electrostatic interactions are important for stability and specificity of protein-protein interactions (50,51). Consistently, we identified seven intermolecular hydrogen bonds and five salt bridges in our HADDOCK complex model (Fig. 5C), which emphasizes the role these interactions play in the recognition and specificity of PmrD connector proteins.
With response regulators, the phosphatase reaction is generally considered to proceed essentially as a reversal of the phospho-His phosphotransfer reaction, with water or hydroxide taking the place of the histidine imidazole side chain (52). Some of the histidine kinases have phosphatase activities toward their respective phosphorylated response regulators. The dephosphorylated kinases could promote a transition state with water in place of the phosphorylated histidine side chain, leading to hydrolysis instead of phosphorylation (53). For example, the structural basis of the response regulator Spo0F dephosphorylation by RapH phosphatases was revealed by the RapH-Spo0F crystal structure (Fig. 7B). In the structure, the side chain of RapH Gln-47 inserts into the Spo0F active site. RapH Gln-47 may orient water for direct in-line hydrophilic attack on the Spo0F phosphoaspartate 54 phosphorous atom and thus cause hydrolysis of the phosphoryl group (54). In these cases, the interactions are primarily responsible for the specific histidine kinase-response regulator or connector-response regulator binding.
The receiver domain of PmrA could form a cooperative, intermolecular dimer of a response regulatory component that interacts to convert an inherently weak set of protein-protein interactions into a stronger binding force that permits the formation of a functional viable complex. Activation of a response regulator by a "pre-existing population shift" may be a general mechanism for two-component signaling (55). Selective binding to a low populated conformation hidden with traditional structural methods may be a general model for ligand binding (56). Therefore, PmrD may selectively bind to the phosphorylated higher energy active conformation of PmrA N and prevents its dephosphorylation by phosphatase such as PmrB.
In our previous study, we proposed a model of the PmrD-PmrA N protein complex and examined how PmrD regulates phospho-PmrA. Here, we propose that PmrD is a phosphatase inhibitor that specifically binds the phosphorylated active site and acts as a molecular barricade that inhibits interaction between the phosphatase and phosphorylatable Asp residue of PmrA and thus prevents intermolecular dephosphorylation. PmrD might inhibit dephosphorylation by sterically hindering the transfer of phosphate from Asp-51 of PmrA to phosphatase PmrB. The large size of PmrD and its predicted binding interface with PmrA are consistent with the model that inhibition occurs by a steric hindrance. The model shows three H-bonds between PmrD and residues near phosphorylation sites of PmrA N . For example, the HD1 of His-70 in PmrD formed H-bonds with the carbonyl group of Gly-53 in PmrA N . In addition, the H⑀2 of His-70 in PmrD formed an H-bond with OD1 of Asp-9 in PmrA N (Fig. 7C). A divalent cation close to the active site is required to add or remove phosphoryl groups in the receiver domain of a response regulator (57). Therefore, the steric hindrance effect provided by PmrD could prevent the dephosphorylation of PmrA mediated by the cation. As well, nucleophilic attack by a water molecule is the likely mechanism of autodephosphorylation in response regulators (58,59). The side chain of His-70 in PmrD likely maintains the phosphoryl group stability by sterically inhibiting access of the water molecule, which executes a nucleophilic attack on the phosphorus and causes hydrolysis of the phosphoryl group. These interactions may explain how PmrD stabilizes the activated PmrA N protein and thus prevents dephosphorylation of phosphate on Asp-51 from a phosphatase such as PmrB. PmrD may inhibit dephosphorylation of the response regulator by acting as a simple physical blockade to sterically block access to the phosphoacceptor Asp residue on the receiver domain of the response regulator, thus precluding close contact of phosphatase with the phosphorylation site on PmrA N . FIGURE 7. A proposed physical blockage mechanism of PmrA dephosphorylation inhibited by PmrD. A, the monomeric form of the PmrD-PmrA N complex is shown. His-70 of PmrD and Asp-51 from PmrA N are in red and blue, respectively. B, the crystal structure of the response regulator Spo0F in complex with the RapH phosphatase. The phosphorylation site Asp-52 is shown in blue, and the residue Gln-47 in RapH, which is responsible for dephosphorylating Spo0F, is in red. C, a stick diagram shows the intermolecular interactions near the active site residue Asp-51 of PmrA N . The nitrogen and oxygen atoms are in blue and red, respectively. The carbon atoms of His-70 from PmrD are in green and are colored gray for others. The H␦1 and H⑀2 of His-70 in PmrD form H-bonds (yellow dashed lines) with the carbonyl groups of two active site residues (Asp-9 and Gly-53) of PmrA N . The interactions demonstrate how PmrD can protect the phosphate group analog BeF 3 Ϫ (magenta) and the divalent cation Mg 2ϩ (green sphere) on PmrA N and prevent it from dephosphorylation.