The Structural Basis of Signal Transduction for the Response Regulator PrrA from Mycobacterium tuberculosis*

The structure of the two-domain response regulator PrrA from Mycobacterium tuberculosis shows a compact structure in the crystal with a well defined interdomain interface. The interface, which does not include the interdomain linker, makes the recognition helix and the trans-activation loop of the effector domain inaccessible for interaction with DNA. Part of the interface involves hydrogen-bonding interactions of a tyrosine residue in the receiver domain that is believed to be involved in signal transduction, which, if disrupted, would destabilize the interdomain interface, allowing a more extended conformation of the molecule, which would in turn allow access to the recognition helix. In solution, there is evidence for an equilibrium between compact and extended forms of the protein that is far toward the compact form when the protein is inactivated but moves toward a more extended form when activated by the cognate sensor kinase PrrB.

The use of regulatory systems to sense and respond to changing environmental conditions is an intrinsic feature that enables bacteria to survive and adapt to a variety of external challenges. Two-component signaling (TCS) 2 systems are the principle mechanism used by bacteria to perform this task (1). A typical TCS consists of a sensor histidine kinase and a response regulator (RR). Histidine kinases are usually membrane-anchored proteins with a characteristic core consisting of a histidine-containing dimerization domain and a catalytic domain. In response to extracellular, and in a few cases, intracellular, conditions, the histidine kinase autophosphorylates at a histidine residue, usually in the dimerization domain, by phosphotransfer from the catalytic (ATPase) domain of the adjacent protomer. It then acts as a phosphodonor to a universally conserved aspartic acid residue in the response regulator. RRs typically consist of two domains with the N-terminal (receiver) domain being the phosphoacceptor domain and the C-terminal domain being the effector. The effector domain is, in most cases, DNA binding and is involved in transcriptional regulation of the genes necessary to respond to the sensed environment. Phosphorylation of the aspartic acid residue located in the receiver domain activates the effector domain in a manner that is still incompletely understood, not least because there is little structural information on full-length response regulators and no structural information on activated full-length (i.e. multidomain) response regulators.
TCSs have been identified as potential antibacterial targets because they play a key role in controlling cellular processes (2). The lack of these systems in higher eukaryotes makes them potentially selective and unique antibacterial drug targets, There is, however, little biochemical information available on the TCSs of pathogenic bacteria such as Mycobacterium tuberculosis (MtB). The H37Rv strain of MtB has 12 putative TCSs including the recently discovered Rv3220-Rv1626 pair (3) as well as five putative orphan response regulator and sensor kinase proteins. The PrrA-PrrB pair (Genome locii Rv0903c-Rv0902c) is one of five TCS gene pairs conserved in all mycobacterial species (4) and can consequently be assumed to play (a) fundamental role(s) in mycobacteria.
The function and signals sensed by TCS proteins in M. tuberculosis are still poorly characterized despite their apparent importance, although some initial findings have been made (5)(6)(7)(8). The PrrA-PrrB TCS has been implicated in the early intracellular multiplication of M. tuberculosis (9,10). The cDNAs corresponding to co-transcripts from the PrrA-PrrB locus have been recovered from MtB cultured in human blood monocyte-derived macrophages but not from cells cultured in standard laboratory medium, indicating that the genes are transcribed in response to host-cell interaction (11). Thus the ability to survive within a host cell makes these TCS proteins prime targets for the development of new antibacterial agents.
Response regulators all contain structurally similar N-terminal domains with a ␤1-␣1-␤2-␣2-␤3-␣3-␤4-␣4-␤5-␣5 topology that contains the conserved phosphorylation site. They can be subdivided into families depending upon the expected structure of the C-terminal, or effector, domain. The three families, which can also have subdivisions, are named after representative proteins, namely OmpR/PhoB, NarL/ FixJ, and NtrC/DctD. The first two classes are 70 -dependent activators, whereas the last family consist of 54 -dependent activators and contains an additional central ATPase domain necessary for open complex formation of the RNA polymerase. The RR studied here (PrrA) belongs to the OmpR/PhoB family. OmpR and PhoB are members of the wingedhelix-turn-helix family of DNA-binding proteins (12). There are, however, important structural differences between OmpR and PhoB, for example, in the length of interdomain linker and in the transactivation loop that proceeds the DNA recognition helix (13).
Structural studies investigating individual receiver domains of response regulators in their active and inactive forms provide some information about conformational changes occurring upon phosphorylation, but how this change is transmitted to activate the C-terminal effector domain is, in most cases, still a mystery. This is because structural information is much more limited in the case of intact proteins and totally absent on full-length proteins in the activated form. Consequently, there is a paucity of knowledge about interactions between the regulatory and effector domains. At the time of writing, there are only five structures available on full-length RRs in the Protein Data Bank (PDB). Two of them, DrrB (14) and DrrD (15), belong to the OmpR/ PhoB subfamily and were crystallized from Thermatoga maritima. Although it is often the case that proteins from thermophilic bacteria are more amenable to crystallization than their mesophilic counterparts (15), we present here the structure of PrrA from M. tuberculosis. We show that although in the crystal it exhibits a compact conformation, inhibitory for DNA binding, in solution, there is probably a percentage of a more open conformation that would be more favorable for DNA binding.

EXPERIMENTAL PROCEDURES
The PrrA gene (genomic location tag Rv0903c) was cloned by PCR from genomic M. tuberculosis (H37Rv strain) DNA. DNA flanking primers 5Ј-TATACCATGGGCGGCATGGACACTGGTGTG-3Ј and 5Ј-GATTATCACTGCA TACGCAGCACGAATCCG-3Јwere engineered to provide melting temperatures of ϳ65°C with extensions encoding NcoI and XhoI restriction sites and ligated with T4 DNA ligase into a pETM11 (EMBL) expression vector with an N-terminal His 6 tag and a recognition sequence for tobacco etch virus protease. Following sequence confirmation, the plasmid was transformed into Rosetta(DE3)pLysS cells. Cells were induced with 1 mM isopropyl-1thio-␤-D-galactopyranoside at A 600 ϭ 0.7 and grown for 5 h at 25°C. The cells were harvested by centrifugation and lysed by sonication in a buffer containing 50 mM Tris⅐HCl, 200 mM NaCl, 5 mM ␤-mercaptoethanol, pH 7.5 (buffer A), and two protease inhibitor tablets (Roche Applied Science). The lysate was spun down and loaded onto a 5-ml nickel (HiTrap) column equilibrated in buffer A. The protein was eluted with a 100-ml linear gradient of 10 -300 mM imidazole in buffer A. The fractions containing PrrA were pooled and diluted to a final imidazole concentration of 50 mM. The protein was then mixed with tobacco etch virus protease in a 1:20 ratio and left at 4°C overnight to cleave the His 6 tag. The digested protein (which has an additional Gly and Ala at the N terminus resulting from the cloning procedure) was applied to the HiTrap column equilibrated in buffer A containing 50 mM imidazole. The flow-through fractions were collected, concentrated to a final volume of 5 ml, and further purified on HiLoad 26/60 Superdex 75 equilibrated in 20 mM Tris⅐HCl, 100 mM NaCl, 2 mM dithiothreitol, pH 7.5 (buffer B). The peak fractions were concentrated using a 10-kDa cut-off centricon (Vivaspin) to a concentration of 12 mg/ml as measured by Bradford's assay (16).
Crystallization and X-ray Data Collection-Initial crystallization conditions were identified using the sparse matrix Cryo screen from Hampton Research. Following optimization, the best crystals were obtained after 1 week by hanging drop vapor diffusion, in which 2 l of protein were mixed with 2 l of reservoir buffer, containing 130 mM calcium acetate, 15% polyethylene glycol 8000, 20% glycerol, and 100 mM sodium cacodylate at pH 6.5. To obtain crystals containing Mg 2ϩ in the active site, we substituted calcium acetate by magnesium acetate in the crystallization buffer.
The crystals grew as intertwined clusters, making it necessary to break them to obtain single crystal fragments. For data collection, crystals were directly transferred from the drop and flashed-cooled to 100 K in the cold gas stream. Measurements were carried out on the X11 and X13 EMBL beam lines at Deutsches Elektronen Synchrotron (DESY) Hamburg using a MAR165 mm CCD detector.
Diffraction images were processed using DENZO, and the intensities were scaled and reduced using SCALEPACK (17) and converted to structure amplitudes using the program TRUNCATE within the CCP4 program suite (18). Crystals (calcium-containing) belong to space group P1 with unit cell parameters a ϭ 37.2, b ϭ 46.5, c ϭ 65.7 Å and ␣ ϭ 100.3, ␤ ϭ 101.7, ␥ ϭ 92.5°with two molecules in the asymmetric unit and an estimated solvent content of 48%. Data collection details are summarized in Table 1.
Structure Determination-The sequence with accession code Q10531 was used for homology modeling with the SWISS-PROT server, which is based on an automated comparative protein modeling (19). The resulting model was derived from the PDB entries 1KGS (15), 1MVO (20), 1NXW, 1NXV, and 1NXP (21). The N-and C-terminal domains of the homology model were placed in two coordinate files containing residues 7-119 and 138 -223, respectively. Residues 120 -137 were omitted on the assumption that the interdomain linker would be flexible. The model of the N-terminal domain was used initially as the search model for molecular replacement, with data in the range 10 -4 Å used in the rotation and translation function searches. Two molecules per asymmetric unit were expected on the basis of the Matthews parameter (22), and consequently, a multiple copy search for the domain was performed using the program MOLREP (23). The R-factor and correlation coefficient for the first domain were 0.56 and 0.20, respectively, which became 0.55 and 0.22, respectively, upon location of the domain in the second molecule. The resultant model containing both N-terminal domains was then fixed, and the C-terminal domain was used as a search model. Again, a multiple copy search was performed for two domains; however, only one was easily found dropping the R-factor to 0.49 and increasing the correlation coefficient 0.43. The C-terminal domain of the second molecule was created based on the relative orientations of the N-and C-terminal domains of the first molecule and the complete solution was checked in molecular modeling program O (24) for good packing contacts before proceeding to refinement. The model was then subjected for rigid body refinement in CNS (25), considering each domain as a separate entity. Further positional and B-factor refine- where F o is the observed structure factor and F c is the calculated factor. d R free was calculated as for R work but on 5% of the reflections excluded from the refinement. ment were performed using CNS when the R-factor and R free (26) dropped to 0.43 and 0.47, respectively. At this stage, sigmaa-weighted maps were calculated using the program SIGMAA (27), and careful examination of the maps allowed corrections to be incorporated into the model. Finally, the model was used as an initial model for automated model building using the program ARP/wARP (28) against the 1.77 Å data. 96% of the model was correctly built in 10 building cycles. Further refinement was continued using REFMAC5 (29). The model was iteratively improved by combination of refinement and manual building using Coot (30). In the final stages of refinement, a bulk solvent correction, non-crystallographic symmetry restraints, and anisotropic scaling were used in REFMAC. The refinement was monitored throughout using the R-free, which was calculated using 5% of the unique reflections.
In the final model of PrrA, complexed with either Ca 2ϩ or Mg 2ϩ , two N-terminal residues (Gly and Ala) introduced during the cloning procedure together with the first six residues of the protein are not visible in the electron density and, because they extend into the solvent region, are assumed to be disordered. Other regions with uninterpretable electron density are residues 87-90 of molecule A (PrrA-Ca and PrrA-Mg), Arg-88 of the molecule B (PrrA-Ca and PrrA-Mg), residues Gly-214 and Gly-215 of molecule A (PrrA-Ca), and Gly-215 and Ser-128 of molecule B (PrrA-Mg). The overall geometric quality of the model was assessed using PROCHECK (31). 89.6% and 90.4% of the amino acid residues, for PrrA-Ca and Prra-Mg, respectively, were found in the most favorable region of the Ramachandran plot with one residue (Val-63) in the disallowed region.
The refined coordinates and the structure factors for the PrrA structure complexed with Ca 2ϩ and Mg 2ϩ have been deposited with the RCSB (PDB ID: 1YS6 and 1YS7, respectively). The figures have been produced with Molscript (32), CCP4MG, and WebLab Viewer 2.0.
SAXS Experiments and Data Analysis-The synchrotron radiation x-ray scattering data were collected on the X33 camera (33,34) of the EMBL (on the DORIS III storage ring at DESY). Solutions of PrrA were measured at several protein concentrations (from 2 to 13 mg/ml using a MAR345 two-dimensional image plate detector (Marr Research), in the range 0.12 Ͻ s Ͻ 4.5 nm Ϫ1 . To check for radiation damage, two successive 2-min exposures were compared, but no radiation effects were observed. The data were averaged after normalization to the intensity of the incident beam, corrected for the detector response, and the scattering of the buffer was subtracted. The difference data were extrapolated to zero solute concentration following standard procedures. The PrrBdependent activation was studied in solutions containing 4 mg/ml PrrA and 0.1 mg/ml PrrB HCD (see Ref. 35 for the nomenclature) with 5 mM Mg 2ϩ and 10 mM ATP. The scattering from the PrrB HCD /Mg 2ϩ /ATP solution was then subtracted as background. All data manipulations were performed using the program package PRIMUS (36).
The forward scattering I(0) and the radius of gyration R g were evaluated by the Guinier approximation (37) and by using the indirect transform package GNOM (38). The molecular masses of the solutes were evaluated by comparison of the forward scattering with that from reference solutions of bovine serum albumin (molecular mass ϭ 66 kDa) and were compatible with monomeric PrrA in solution.  The scattering from the high resolution models was calculated using the program CRYSOL (39) Given the atomic coordinates, the program fits the experimental intensity I(s) by adjusting the excluded volume of the particle and the contrast of the hydration layer surrounding the particle in solution to minimize the discrepancy where N is the number of experimental points, I exp (s), I calc (s), and (s j ) are the experimental and calculated intensity and the experimental error at the momentum transfer s j , respectively, and c is a scaling factor. Rigid body modeling of PrrA was done by freely rotating its C-terminal domain around the residue Ser-128 and computing the fits to the experimental data with CRYSOL. The volume fractions of the component in the mixtures best fitting the small x-ray angle scattering (SAXS) data from inactivated and activated PrrA were evaluated using the computed scattering curves from crystallographic models of PrrA, DrrB, and DrrD by the program OLIGOMER (36).

RESULTS AND DISCUSSION
PrrA crystallizes in the triclinic space group P1 with two molecules in the asymmetric unit. The structure of the protein was solved by molecular replacement and refined at 1.77 Å resolution when complexed with Ca 2ϩ (referred to as PrrA-Ca) and at 1.58 Å resolution when complexed with Mg 2ϩ (referred to as PrrA-Mg).
The structure of molecule B from PrrA-Mg is shown in Fig. 1, which shows the N-terminal receiver domain and C-terminal effector domain connected by a linker region. The receiver domain has the expected (␣/␤) 5 topology, with five parallel strands ␤2-␤1-␤3-␤4-␤5 forming the hydrophobic core surrounded by two helices (␣1 and ␣5) on one side and three (␣2-␣4) on the other. The C-terminal domain has a wingedhelix fold and is composed of a four-stranded antiparallel ␤-sheet followed by a three-helix bundle and a C-terminal ␤-hairpin. The two domains are connected by a long, well ordered linker between ␣5 and ␤6 and pack together, forming an extensive buried interface. The two molecules in the asymmetric unit superimpose with an r.m.s.d. of 0.56 Å (for 204 C␣ atoms). The largest difference is observed in the linker region (Fig. 2) and the region containing the ␣4 helix, which have r.m.s.d. values of 8.0 and 8.8 Å, respectively.
The Receiver Domain-The active site of PrrA is defined by a crevice formed by loops ␤1-␣1 and ␤3-␣3 and contains all residues essential for phosphorylation. These are Asp-14, Asp-15, Asp-58, and Lys-108 (Fig.  3). In some response regulators, Asp-14 is replaced by glutamate. Bacterial response regulators are phosphorylated by the cognate sensor kinase in a divalent cation-dependent reaction, and divalent metal ions are usually observed in the catalytic site. Ca 2ϩ and Mg 2ϩ (depending on crystallization conditions) ions were found in the Asp-15/Asp-58 pocket (Fig. 3B). Both ions exhibit octahedral coordination, interacting in the same way with the carboxylate groups of Asp-15 and Asp-58 as well as the carbonyl oxygen of Asn-60. The three remaining coordination sites are occupied by water molecules. The distances between Ca 2ϩ and Mg 2ϩ and their coordinating atoms lie in the ranges 2.20 -2.54 and 1.96 -2.14 Å, respectively, in good agreement with equivalent distances in other protein structures (40). The amino group of conserved Lys-108 forms a salt bridge to the carboxylate groups of Asp-58 (2.9 -3.1 Å) and Asp-14 (2.6 -2.8 Å). Asp-14 is not directly involved in coordinating the metal ion but forms a strong hydrogen bond to a water molecule of the octahedral metal coordination sphere. Similar coordination of the metal ion has been observed for CheY (41). The electron density for the region of the protein surrounding the metal coordination site is mostly well defined, in contrast to the case in DivK (42), where a destabilizing effect on the protein structure was observed due to metal binding, which resulted in the disordering of the ␣4-␤4 region. The absence of electron density for residues 87-90 in molecule A of the PrrA-Ca structure is unlikely to result from metal binding because the same region in molecule B with metal bound in the active site is well ordered.
Ser-86 is one of the residues known to be crucial in signal transduction (43), although in the other response regulators of the OmpR/PhoB family, it is normally threonine (Fig. 4). This residue has a slightly different environment in each of the crystallographically independent molecules of PrrA, and this environment also differs slightly when comparing the Mg 2ϩ -and Ca 2ϩ -containing structures. In molecule A, the carbonyl group of Ser-86 forms a water-mediated interaction with the main chain nitrogen of Lys-108, whereas in molecule B, the carbonyl oxygen of Ser-86 and main chain nitrogen of Lys-108 are directly hydrogen-bonded (2.7 Å). In addition, for molecule B, the hydroxyl oxygen of Ser-86 makes a salt-bridge interaction with the side chain of Arg-94 in One noticeable difference between the two molecules in the asymmetric unit is the length of helix ␣4, which is unwound in molecule A relative to molecule B (Fig. 2). The position of unwinding is at Gly-97, which is conserved in all response regulators of the OmpR/PhoB family. Comparing the secondary structures shown in the alignment in Fig. 4, it is also apparent that ␣4 is extended in some and stops at the glycine in others. This helix has been shown to undergo structural changes upon activation, at least in structures of isolated regulatory domains (45,46). In particular, for FixJ (45), such a structural change produces a dimerization interface. This observation is probably of limited relevance here because the FixJ family binds palindromic DNA, whereas the OmpR/ PhoB family binds to (minimally) a tandem repeat of DNA (47), and dimer formation is dependent upon DNA binding. Dimers of isolated receiver domains (both activated and inactivated) with diad symmetry are observed, for example, in PhoB (48) and TorR (49), although the nature of the dimer interface differs in the two cases. In the crystals of PrrA, there are also two molecules per asymmetric unit, but the molecules pack head-to-tail with the principle interactions being between activation and effector domains of adjacent molecules. It has been postulated that dimerization could be part of the activation process, in part because the residues at the dimer interface observed in the structure of the receiver domains of TorR and KdpE are highly conserved in the OmpR/PhoB family (49). They are also largely conserved in PrrA; however, the small angle x-ray scattering data, either for inactivated or for activated PrrA, show no evidence of dimer formation. Even the very open structure observed in the crystals of DrrD would not allow, for steric reasons, the dimerization interface observed in the structure of the isolated receiver domain of TorR. Although these observations would not seem to support dimer formation involving the receiver domains prior to DNA binding, they would not preclude a subsequent conformational change stabilizing the binding to a tandem repeat by interaction between receiver domain ␣4-␤5-␣5 faces.
One residue (Val-63) in each molecule is found in the disallowed region on the Ramachandran plot (Table 1). This residue occurs in the loop extending from the phosphorylation site at the tip of ␤3, and it is fully visible in the 2F o Ϫ F c map contoured at the 1 level. Interestingly, in the response regulator DrrD (15), the same residue was also observed to be in the disallowed region of the Ramachandran plot, and in PhoB and OmpR, it is glycine.
The Effector Domain-The crystal structures of isolated effector domains of OmpR (50,51) and PhoB complexed with DNA (47) as well as the uncomplexed full-length response regulators DrrB (15) and DrrD (14) have been described in the literature. A structure-based sequence alignment is shown in Fig. 4, illustrating that the members of this family characteristically possess a four-stranded ␤-sheet preceding the interdomain linker. Comparing these structures shows that although the secondary structure elements superimpose quite well, there are large deviations in the loops extending from both ends of the recognition helix (␣8). The r.m.s. deviations for the compared structures are given in Table 2, and the superpositions of both effector and receiver domains are shown in Fig. 5.
Comparison of PrrA with Other OmpR/PhoB Family Members-The high resolution structural data for five full-length response regulators are available: NarL (49), DrrB (15), DrrD (15), CheB (52), and RsbQ (53). Two of these, DrrB and DrrD, belong to the OmpR/PhoB family, and therefore, we make a more detailed comparison with these structures. Fig. 6 shows the superimposition of PrrA with DrrB and DrrD. The superimposition of regulatory and effector domains separately (Fig. 5) shows that although the secondary structures occupy very similar positions, the overall root mean square deviation is quite large (1.6 -2.0 Å, Table 2). In the case of the regulatory domains, the region containing helix ␣4 and the loop ␤4-␣4 shows the most variability. In the case of the effector domains, the largest backbone deviations occur in the loop regions ␣7-␣8 (transactivation loop) and ␣8-␤11. When all three fulllength proteins are superimposed using their regulatory domains (Fig.   6), the effector domains occupy completely different orientations. Of the three proteins, DrrD shows the most extended conformation, with only one end of helix ␣5 in contact with the antiparallel ␤-sheet of effector domain. DrrB is slightly more compact, with the ␣4-␤5-␣5 face of the regulatory domain packing against the ␤-sheet of the effector domain. Both molecules of PrrA, with either Mg 2ϩ or Ca 2ϩ bound, exhibit the same compact conformation despite differences in the position of the interdomain linker (Fig. 2). The platform formed by ␣4-␤5-␣5 packs against helices ␣7, ␣8, and the transactivation loop. The antiparallel ␤-sheet of the effector domain is positioned in such a way that it does not interact with the regulatory domain. This feature of the structure is of interest because the postulation that the ␤-sheet mediates interactions with the regulatory domain (14) would not appear to be supported, at least for PrrA.
The buried surface between domains of PrrA is 820 Å 2 , a bit larger than in DrrB (751 Å 2 ) but much larger than for DrrD (245 Å 2 ). Both hydrophobic and hydrophilic interactions form the interface between N-and C-terminal domains. Hydrophobic interactions involve residues in helices ␣4, ␣5 of the regulatory domain with residues of ␣8 (recognition helix), and the transactivation loop of the effector domain. However, there is no clear hydrophobic core, as is observed in DrrB (14). The side chain of Arg-125 located in helix ␣5 forms hydrogen bonds with the carbonyl oxygens of Leu-187 and Glu-186 present in the ␣6 helix of the effector domain. The hydroxyl group of Tyr-191 hydrogen-bonds to the carbonyl group of Asp-103. The main chain oxygen of Asp-103 also forms a hydrogen bond to Trp-189 located at the base of ␣7. The carboxylate side chain of Asp-104 interacts directly with Tyr-191 and forms a water-mediated hydrogen bond with Thr-197. The strong hydrogen bond between Tyr-105 and Asn-198 in the effector domain has already been described. This is an important observation because this residue upon phosphorylation is thought to undergo a large torsional rearrangement from an outward pointing orientation to one that fills the space in the receiver domain vacated by Ser-86 (Thr in other RRs) on ␣4 (45). Therefore in PrrA, we are able to suggest that the signal passage upon phosphorylation, via Ser-86 and Tyr-105, destabilizes the interface between receiver and effector domains and that, as we explain in the next section, this would make the recognition helix of the effector domain more accessible for DNA binding.
PrrA Exhibits a Structure That Is Inhibitory for DNA Binding-Within the OmpR/PhoB family, the mechanisms of regulation differ. For OmpR, which interacts with the ␣-subunit of the RNA polymerase II (54,55), phosphorylation of the regulatory domain enhances its DNA binding activity, whereas unphosphorylated OmpR, and the effector domain alone, do bind DNA but with substantially lower affinity (50). In contrast with PhoB, which interacts with 70, the isolated effector domain is constitutively active for transcription (56) and has a higher   affinity for specific DNA binding than full-length unphosphorylated PhoB (57). In summary, phosphorylation of OmpR enhances DNA binding activity, whereas phosphorylation of PhoB results in relief of the inhibition of DNA binding. Previously determined structures of full-length response regulators of the OmpR/PhoB family (DrrB and DrrD) do not provide a structural explanation for the inhibitory role of the regulatory domain on DNA binding. DrrD has minimal interdomain interactions and, although DrrB has a larger interdomain surface, this does not involve the DNA recognition helix, the transactivation loop, or the dimerization surface (14). For PrrA, we observe that many of the interdomain interactions involve parts of the transactivation loop or recognition helix ␣8 (Fig. 7). Fig. 6 shows that the observed crystal structures exhibit the most compact to the most extended conformations ongoing from PrrA to DrrB to DrrD. It is obvious from the PrrA structure ( Fig. 1) that activation of the protein for DNA binding requires separation of the N-and C-terminal domains to expose the recognition helix and transactivation loop. The observed closed conformation of the protein can thus predict an inhibitory role of the regulatory domain, which is weakened upon phosphorylation via the signal transducing serine and tyrosine.
The Closed Conformation Predominates in Solution-We performed SAXS measurements to check whether the closed conformation observed in the crystal structure is an artifact of crystal packing, although we believed this to be unlikely because we have two independent molecules in the asymmetric unit with different conformations of the interdomain linker (Fig. 2). To our surprise, the experimental radius of gyration R g in solution (20.1 Ϯ 0.4 Å) computed from the experimental SAXS data (Fig. 8, curve 1) noticeably exceeded the value computed from the crystallographic model (18.9 Å). Also, the scattering curve computed from the crystal displayed systematic deviations from the experiment (discrepancy ϭ 2.7; Fig. 8, curve 2). On the other hand, a more open crystallographic model of DrrB having R g ϭ 20.8 Å provided yet worse agreement with the experiment ( ϭ 4.1; Fig. 8, curve 3), whereas the completely open DrrD (R g ϭ 22.4 Å) completely disagrees with the data ( ϭ 8.8: Fig. 8, curve 4). To assess the solution conformation of PrrA, two options were considered. Firstly, we assumed a single species in solution consisting of two domains joined by a flexible linker and optimized the position of the domains to best fit the data ( ϭ 1.30; Fig. 8, curve 5). The refined model (not shown) is somewhat more open than the crystal structure but still much more compact than DrrD or DrrB. Alternatively, we assumed that multiple conformations of PrrA may coexist in solution and tried to fit the data by a mixture of the closed (PrrA) structure and differently opened (DrrB and DrrD) structures. The best fit ( ϭ 1.16; Fig. 8, curve 6) was for a mixture of 23% DrrD and 77% PrrA structure (interestingly, the volume fraction of DrrB was  found to be 0). This model is not only the best overall fit but is the more reasonable of the two because in other systems, regulation is known to involve a change in equilibrium between active and inactive species (58).
The Open Conformation of PrrA Is More Prevalent upon Activation by PrrB-It is reasonable to assume that transcriptional control is governed by the relative concentration of the open, DNA binding-competent, conformation. To investigate this, we performed SAXS experiments on activated PrrA by the addition of a limited amount (0.1 mg/ml) of the constitutively active construct of the cognate histidine kinase, PrrB (35). In these experiments, corresponding background scattering of the PrrB construct together with the Mg 2ϩ and ATP necessary to activate the kinase has to be subtracted from the PrrA scattering. This limits the accuracy of the processed data at higher angles, but a clear change compared with the scattering from inactivated PrrA can still be observed (Fig. 8, inset). The R g of the activated PrrA increases to 22.3 Ϯ 0.5 Å and agrees well with that of DrrD. Allowing for multiple conformations as in the above analysis of inactivated PrrA, the best fit to the scattering from activated PrrA is provided by a mixture of 65% of open (DrrD) and 35% of closed (crystal structure of PrrA) conformations. The errors in the percentages could be as high as 10% due to the limited accuracy of activated PrrA data, but minimally, upon activation, there is a significant increase in the percentage of the open, PrrA, structure relative to the closed form in the crystal structure reported above. Additional evidence is provided by the fact that the experiment on the same PrrA ϩ PrrrB mixture but without the addition of Mg 2ϩ and ATP yielded R g ϭ (20.1 Ϯ 0.5) Å and no noticeable changes to the scattering from the inactivated PrrA (data not shown).