Substrate-mediated Stabilization of a Tetrameric Drug Target Reveals Achilles Heel in Anthrax*

Bacillus anthracis is a Gram-positive spore-forming bacterium that causes anthrax. With the increased threat of anthrax in biowarfare, there is an urgent need to characterize new antimicrobial targets from B. anthracis. One such target is dihydrodipicolinate synthase (DHDPS), which catalyzes the committed step in the pathway yielding meso-diaminopimelate and lysine. In this study, we employed CD spectroscopy to demonstrate that the thermostability of DHDPS from B. anthracis (Ba-DHDPS) is significantly enhanced in the presence of the substrate, pyruvate. Analytical ultracentrifugation studies show that the tetramer-dimer dissociation constant of the enzyme is 3-fold tighter in the presence of pyruvate compared with the apo form. To examine the significance of this substrate-mediated stabilization phenomenon, a dimeric mutant of Ba-DHDPS (L170E/G191E) was generated and shown to have markedly reduced activity compared with the wild-type tetramer. This demonstrates that the substrate, pyruvate, stabilizes the active form of the enzyme. We next determined the high resolution (2.15 Å) crystal structure of Ba-DHDPS in complex with pyruvate (3HIJ) and compared this to the apo structure (1XL9). Structural analyses show that there is a significant (91 Å2) increase in buried surface area at the tetramerization interface of the pyruvate-bound structure. This study describes a new mechanism for stabilization of the active oligomeric form of an antibiotic target from B. anthracis and reveals an “Achilles heel” that can be exploited in structure-based drug design.

assembles as a tetrameric protein (16), best described as a dimer of tight dimers (Fig. 1B). However, a native active dimer has recently been described (14), which makes the quaternary structure of this enzyme of particular interest. DHDPS from plants (17,18) and some Gram-negative bacterial species (19) is feedback-inhibited by (S)-lysine, acting as an allosteric modulator through partial inhibition of catalytic activity. However, allosteric regulation by lysine at biologically relevant concentrations does not occur in DHDPS from Gram-positive species (14,20), including B. anthracis (21).
Recent work in our laboratory has shown that the purification of recombinant DHDPS from B. anthracis (Ba-DHDPS) in the presence of its substrate pyruvate increases the yield and specific activity of the final product (22). Therefore, the aim of this study was to thoroughly characterize the effect of pyruvate on the solution properties and structure of Ba-DHDPS. Here, we report solution and structural studies that unravel the mechanism for substrate-mediated stabilization of the active quaternary structure of Ba-DHDPS.

EXPERIMENTAL PROCEDURES
Cloning, Expression, and Purification of Ba-DHDPS-The dapA gene encoding DHDPS from B. anthracis (Sterne strain) was amplified by PCR and cloned into the pET11a expression vector as described elsewhere (22). Briefly, recombinant protein was produced in the host strain E. coli BL21-DE3 as follows. Cells harboring vector were cultured at 37°C in Luria broth containing 100 g ml Ϫ1 ampicillin to an A 600 of 0.6. Expression of recombinant Ba-DHDPS was induced by addition of isopropyl 1-thio-␤-D-galactopyranoside to a final concentration of 1.0 mM. Cells were harvested 3 h post-induction and resuspended in 20 mM Tris-HCl, pH 8.0, before lysis by sonication. Ba-DHDPS was subsequently isolated by anion-exchange and hydrophobic interaction liquid chromatography as described elsewhere (22).
DHDPS-DHDPR Coupled Enzyme Kinetic Assay-Enzyme kinetic analyses of Ba-DHDPS and Ba-DHDPS (L170E/G191E) were performed using the DHDPS-DHDPR coupled assay as described in a previous study (23). Assays were routinely performed in duplicate at a constant temperature of 30°C with reaction mixtures allowed to equilibrate in a temperature-controlled Cary 4000 UV-visible spectrophotometer for 12 min before initiating the reaction with 60 nM DHDPS. Prior to the experiment, pyruvate and ASA concentrations were routinely quantified by the addition of limiting amounts of substrate by measuring the consumption of NADPH at 340 nm in the Cary 4000 UV-visible spectrophotometer. Initial velocity data were best fitted to a Ping Pong Model using ENZFITTER. 4 Rate versus enzyme concentration assays were also conducted with Ba-DHDPS concentrations ranging from 2.5 to 120 nM.
CD Spectroscopy-CD spectra were recorded using an AVIV 410-SF CD spectrometer. Wavelength scans were performed between 190 and 250 nm in 20 mM Tris, 150 mM NaCl with 0.15 mg ml Ϫ1 Ba-DHDPS and Ba-DHDPS (L170E/G191E) in 1-mm quartz cuvettes. Data were analyzed using the CONTINLL algorithm from the CDPro software package (25) and the SP29 protein data base. For thermal denaturation scans, ellipticity at 222 nm was monitored between 20 and 90°C in 1°C steps.
Analytical Ultracentrifugation-Absorbance-based sedimentation velocity and equilibrium experiments were performed in a Beckman model XL-I analytical ultracentrifuge FIGURE 1. Ba-DHDPS. A, DHDPS catalyzes the condensation of pyruvate and ASA into hydroxytetrahydrodipicolinic acid, which is the first committed step in the lysine biosynthesis pathway in bacteria. B, crystal structure of apo-Ba-DHDPS showing the position of one active site and the self-association interfaces. Two monomers come together at the tight dimer interface to form the dimeric unit, and then the two dimeric units dock at the weak dimer interface to form a tetramer.
using a 4-hole An-60 Ti or an 8-hole An-50 Ti rotor. Doublesector quartz cells were loaded with 380 l of sample and 400 l of reference (20 mM Tris, 150 mM NaCl, pH 8.0) for sedimentation velocity, or a 100-l sample and a 120-l reference for sedimentation equilibrium. Experiments were conducted at 4°C using a rotor speed of 40,000 rpm (sedimentation velocity) or 12,000 and 18,000 rpm (sedimentation equilibrium) with absorbance measured at 227 nm (1.6 M enzyme) or 235 nm (4.8 M enzyme). Solvent density, solvent viscosity, and estimates of the partial specific volume of Ba-DHDPS and Ba-DH-DPS (L170E/G191E) were calculated using SEDNTERP (26). Initial scans were carried out at 3,000 rpm to determine optimum wavelength and radial positions for the experiments. Samples monitored in the presence of 0.6 mM pyruvate contained pyruvate in both the reference and sample channels. Sedimentation velocity data were fitted to a single discrete species or a continuous sedimentation coefficient [c(s)] model (27)(28)(29) using the program SEDFIT. Additionally, van Holde-Weischet analysis (30) was performed using the ULTRASCAN software package (31), whereas sedimentation equilibrium data were fitted to various self-associating equilibrium models using the program SEDPHAT (32).
Crystallization of Ba-DHDPS and X-ray Diffraction Data Collection-Ba-DHDPS was crystallized according to methods described previously (22) using sitting-and hanging-drop, vapor diffusion. The crystals used for diffraction analysis were soaked with 20 mM pyruvate overnight to facilitate ligand binding. For x-ray data collection, crystals were transferred to reservoir solution containing 20% (v/v) glycerol with 20 mM pyruvate and directly flash frozen in liquid nitrogen. Intensity data were collected at the Australian synchrotron using the MX1 beamline as described before (22).
Phasing and Model Refinement-Diffraction data sets were processed and scaled using the package MOSFLM (33) and SCALA (34). Initial phase estimates were solved by molecular replacement using PHASER with the ligand-unbound structure (PDB ID: 1XKY) as the search model. Structural refinement was performed using REFMAC5 (35) with iterative model building using WINCOOT (36). Water, glycerol, sodium ion, and the pyruvate-bound lysine atoms were added at later stages using WINCOOT. A round of simulated annealing was performed with a starting temperature of 5000 K to assign the R free set of reflections from the apo structure (PDB ID: 1XL9) that shared the same crystal properties with the newly solved pyruvate-bound structure (PDB ID: 3HIJ). The final model was checked with PROCHECK (37).

RESULTS
Effect of Pyruvate on the Secondary Structure Stability of Ba-DHDPS-We recently reported that the raw enzyme activity and yield of recombinant Ba-DHDPS are significantly increased when the enzyme is purified in the presence of its substrate, pyruvate (22). To examine the effect of pyruvate on the stability of Ba-DHDPS in aqueous solution, thermal denaturation experiments monitored by CD spectroscopy were conducted in the presence and absence of the substrates pyruvate or ASA over the temperature range of 20 -90°C. Initially, wavelength scans were performed at 20°C to monitor global secondary structure in the presence of substrates, which revealed no change in response to either pyruvate or ASA ( Fig. 2A). Ba-DHDPS appears to follow a three-state mechanism for thermal unfolding in the absence of pyruvate and in the presence of ASA, with a potential intermediate persisting at temperatures of ϳ50 -60°C (Fig. 2B). However, in the presence of 2.0 mM pyruvate, the thermal denaturation of Ba-DHDPS is delayed with respect to temperature and unfolding appears to occur via a two-state mechanism without the propagation of an intermediate, indicating that pyruvate stabilizes the folded state of the enzyme. Accordingly, the effect of pyruvate on the quaternary and tertiary structure of Ba-DH-DPS was sought.
Effect of Pyruvate on the Quaternary Structure of Ba-DHDPS-To characterize the quaternary structure of Ba-DHDPS in solution in the absence and presence of pyruvate, absorbance-detected sedimentation velocity and equilibrium analyses were conducted in the analytical ultracentrifuge. The absorbance versus radial position profiles of Ba-DHDPS (1.6 M) during sedimentation velocity in the absence of pyruvate are shown in Fig. 3A, whereas Fig. 3B shows the equivalent data sets in the presence of pyruvate. Two predominant boundaries were observed for Ba-DHDPS in the absence of substrate (Fig. 3A). By contrast, the radial absorbance profiles showed a single predominant boundary in the presence of pyruvate (Fig. 3B). These data were analyzed initially using the enhanced van Holde-Weischet method (30), which is a model-independent approach to analyzing sedimentation velocity data. The resulting integral distribution for Ba-DHDPS in the absence of pyruvate (Fig. 3C, white circles) suggests it exists in a reversible selfassociation with sedimentation coefficients ranging from 2 S through to 6 S. This is consistent with previous reports of DHDPS from E. coli, which was determined to exist in equilibrium between a dimer and tetramer (38). In the presence of pyruvate (Fig. 3C, black circles), the equilibrium appears to be shifted greatly in favor of the larger species. To define the oligomeric species of Ba-DHDPS in solution, the data were subsequently fitted to a continuous sedimentation coefficient (c(s)) distribution model (27)(28)(29) (Fig. 3D). In the absence of pyruvate, the c(s) distribution shows two non-baseline resolved species, in approximately equal proportions, with standardized sedimentation coefficients (s 20,w ) of 4.0 and 6.5 S (Fig. 3D and Table 1). These values correspond to the DHDPS dimer and tetramer (38), respectively (Table 1). By contrast, in the presence of pyruvate, a significantly greater proportion of the tet-ramer, compared with the dimer, was observed (Fig. 3D, solid line). The resulting c(s) distributions shown in Fig. 3D therefore agree well with the van Holde-Weischet analyses (Fig. 3C). This effect was subsequently quantified using sedimentation equilibrium analysis of Ba-DHDPS in the absence and presence of 0.6 mM pyruvate. Samples containing 0.80, 1.6, and 3.2 M Ba-DHDPS were centrifuged at 12,000 and 18,000 rpm until sedimentation equilibrium was attained at each speed (supplemental Fig. 1). Consistent with previous studies of DHDPS (38), the global nonlinear least squares best fit of the data (global 2 ϭ 0.6) was obtained from a dimer-tetramer association model and revealed that the dissociation constant (K D 4 -2 ) of Ba-DHDPS in the absence of pyruvate was 1.90 M, compared with K D 4 -2 of 0.66 M in the presence of the substrate (Table 1 and supplemental Fig. 1). That is, the tetramerization constant was 3-fold tighter in the presence of pyruvate compared with the unliganded enzyme. Together, the results of CD thermostability, sedimentation velocity, and sedimentation equilibrium analyses demonstrate that the secondary and quaternary structure of Ba-DHDPS was significantly stabilized in the presence of pyruvate. To gain further insight into the importance of the tetrameric structure of Ba-DHDPS, we subsequently set out to generate a dimeric mutant of the enzyme to assess the activity of the dimer in comparison to the wild-type tetramer.
Dimeric Mutant of Ba-DHDPS Shows Significantly Attenuated Activity-To highlight the importance of the homo-tetrameric structure of Ba-DHDPS, a double mutant, Ba-DHDPS (L170E/G191E), was designed to break apart the "weak dimer" interface. Leu 170 forms an important hydrophobic interaction with Gly 191 from its neighboring chain at the weak dimer interface. Therefore, by mutating both these residues to glutamic acid, it was thought that the resulting charge repulsion at this site would stabilize the dimer. The purified recombinant Ba-DHDPS (L170E/G191E) product (supplemental Fig. 2, A-C) retained native secondary structure (supplemental Fig. 2D) and not surprisingly was shown to be dimeric in solution by sedimentation velocity analysis (Fig. 4A, dashed line). Furthermore, enzyme kinetic analysis of the dimeric mutant revealed that it retained only 1.8% of the total catalytic activity of the wild-type tetrameric enzyme (Fig. 4B (Fig. 3D). c Frictional ratio calculated using the v method from SEDNTERP (26). d Dissociation constant for tetramer to dimer. Note that the rate of dissociation (k off ) is 10 Ϫ5.3 s Ϫ1 in the absence of pyruvate (supplemental Fig. 4A) and 10 Ϫ5.1 s Ϫ1 in the presence of pyruvate (supplemental Fig. 4B).
0.05 mM, and 3.7 mM for Ba-DHDPS (L170E/G191E) ( Table 2). These results show that the Ba-DHDPS tetramer is significantly more active than the dimeric form of the enzyme, which highlights the significance of the substrate-mediated stabilization phenomenon described above. To support this we show that the rate versus enzyme concentration profile of wild-type Ba-DHDPS was nonlinear at low enzyme concentrations, which indicates that the native dimer was significantly less active than the tetrameric species (supplemental Fig. 2E).
Crystal Structure of Ba-DHDPS Bound to Pyruvate-To elucidate the structural mechanism behind the substrate-mediated stabilization phenomenon observed in solution, we solved the crystal structure of Ba-DHDPS in complex with pyruvate. Interestingly, the newly solved pyruvate-bound crystal structure (PDB ID: 3HIJ) shared the same space group, unit cell parameters, and a similar resolution to the structure of substrate unbound Ba-DHDPS (PDB ID: 1XL9). Crystallization and preliminary diffraction analysis were reported recently (22). The initial structure revealed four monomers in the asymmetric unit, arranged in the biologically relevant tetramer (Fig.  5A). The first round of refinement gave an R cryst of 21.4% (R free of 23.2%), which also revealed election density associated with Lys 163 in the active site of all four monomers that wasn't accounted for by the search model. This is the site where pyruvate binds DHDPS (39). The pyruvate-bound molecule was modeled manually at Lys 163 (Fig. 5B), and later rounds of structural refinement were performed using REFMAC5. The iterative model-building tool, WINCOOT, was used to model in waters, Na ϩ ions, and glycerol molecules. To make a direct comparison to the Ba-DHDPS substrate-unbound structure (PDB ID: 1XL9), a round of simulated annealing was performed on the substrate-bound structure (PDB ID: 3HIJ) with a starting temperature of 5000 K to ensure both structures possessed the same R free set of reflections. The final molecule had an overall R cryst of 15.3% (R free of 21.0%) to 2.15-Å resolution. The resulting model was examined using PROCHECK, which revealed that 99.2% of residues in Ba-DHDPS bound to pyruvate were in the favored regions of a Ramachandran plot. The 8 residues in the "disallowed" regions were Tyr 109 and Ile 142 from chains A, Top, residuals resulting from the c(s) distribution best fits shown in panel A plotted as a function of radius from the axis of rotation. B, the initial velocity is plotted as a function of (S)-ASA concentration for Ba-DHDPS (circles) and LG-DHDPS (triangles). The nonlinear best fit to the ping-pong model is presented as solid lines and results in the kinetic parameters summarized in Table 2.  B, C, and D, which is consistent with the crystal structure of substrate-unbound Ba-DHDPS (8). Tyr 109 is an imperative catalytic site residue, and Ile 142 interacts with its equivalent neighbor at the tight dimer interface providing structural stability. A full table of statistics is provided (Table 3). Both the pyruvatebound and the previously reported ligand-free structures are very similar (Fig. 5). The two tetramers align at alpha carbon atoms with a root mean square deviation of 0.50 Å, and little difference could be observed in active site residues (Fig. 5B). However, the "weak" dimer interface is more extensive in terms of buried surface area in the presence of pyruvate (Table 4). This is the interface where the two tight dimers dock to form the tetramer (Fig. 1A). At this interface the buried surface area increases from 1735 Å 2 in the absence of pyruvate to 1826 Å 2 when bound to pyruvate. By comparison, the buried surface area at the "tight" dimer interface has only changed from 2780 Å 2 in the apo structure to 2782 Å 2 in the presence of pyruvate. Additionally, the number of interactions is significantly greater at the weak dimer interfaces as calculated by STING MILLENIUM (40) ( Table 4). There is also an increase in hydrogen bonds at the tight dimer interface in the pyruvatebound structure (Table 4). In total, there were twelve more hydrogen bond interactions at both interfaces in the crystal structure of pyruvate-bound Ba-DHDPS (PDB ID: 3HIJ) relative to the apo structure (PDB ID: 1XL9).

DISCUSSION
The recently described DHDPS structure from methicillinresistant S. aureus, the first reported native dimeric DHDPS enzyme, was observed to have greater solution stability in the presence of pyruvate (14). It is thus of interest to probe the role of this substrate in stabilizing the active quaternary structure of DHDPS. Disruption of quaternary structure offers a new approach for inhibitor design of active oligomeric enzymes, revealing an "Achilles heel" to target in pathogenic bacteria such as B. anthracis. Therefore, we examined the effect of pyruvate in stabilizing the Ba-DHDPS tetramer.
The solution stability of Ba-DHDPS was initially investigated at the secondary structure level using CD spectroscopy. Neither substrate of DHDPS, pyruvate nor ASA, had an effect on the overall secondary structure of Ba-DHDPS at 20°C (Fig. 2A). This agrees with previous x-ray crystallographic studies of apoand substrate-bound intermediate states of E. coli DHDPS (41). However, the thermostability of Ba-DHDPS was significantly stabilized in the presence of pyruvate compared with the unliganded enzyme (Fig. 2B). By contrast, the presence of ASA, which is the second substrate to bind, showed no effect on the stabilization of the enzyme (Fig. 2B). Interestingly, the unfolding of ligand-free Ba-DHDPS revealed the presence of an intermediate state that persisted over a temperature range of 50 -60°C (Fig. 2B). The intermediate could potentially represent a collapsed molten globule state of Ba-DHDPS that has native-like secondary structure, but a more dynamic tertiary structure, a phenomenon that has been previously observed in other bacterial enzymes (42). This intermediate state was not propagated during thermal denaturation in the presence of pyruvate (Fig. 2B). In fact, the presence of pyruvate significantly stabilized the folded state of the recombinant enzyme for a further 20°C relative to the thermal denaturation profile in the absence of pyruvate (Fig. 2B). Conversely, ASA had no effect on the thermostability of the enzyme with respect to secondary structure (Fig. 2B), which demonstrates that the stabilization phenomenon is specific to pyruvate and thus associated with Schiff-base formation at the active site lysine (Lys 163 ).
Given the observed secondary structure stabilization phenomenon induced by pyruvate, the propensity for the substrate to stabilize the quaternary structure in solution was therefore investigated. Absorbance-based analytical ultracentrifugation experiments revealed that Ba-DHDPS existed in an apparent equilibrium between two oligomeric states, the dimer and tetramer, at low micromolar concentrations at 4°C (Fig. 3). Sedimentation velocity analysis showed that pyruvate significantly shifted the apparent equilibrium in favor of the tetramer in solution (Fig. 3, C and D), thereby enhancing tetramerization in solution. The stabilization of quaternary structure was quantified using sedimentation equilibrium analysis (supplemental Fig. 1), showing that the tetramer was 3-fold tighter in the presence of pyruvate (Table 1). The importance of the tetrameric quaternary structure of Ba-DHDPS is highlighted by kinetic  and solution studies of the dimeric mutant, Ba-DHDPS (L170E/G191E). Our results show that the dimeric mutant was significantly less active than the wild-type tetramer (Table 2 and Fig. 4B). This is consistent with recent studies demonstrating that the enzymatic activity of dimeric mutants of E. coli DHDPS possess Ͻ2.5% of the catalytic activity of the wild-type tetramer (43).
To probe the structural mechanism governing the pyruvatemediated stabilization of the active tetrameric form, x-ray diffraction studies of Ba-DHDPS in the presence of pyruvate were undertaken. Recombinant native Ba-DHDPS, purified as described previously (22), crystallized in identical conditions to those published for the unliganded enzyme (PDB ID: 1XL9) (8). The final structure of Ba-DHDPS in complex with pyruvate was solved to a resolution of 2.15 Å, and initially no significant changes could be observed in comparison to the pyruvate-unbound structure. However, upon close inspection of the weak dimer interface using PISA analysis (44), significant differences were observed. Moreover, an increase in buried surface area of 91 Å 2 resulted when the enzyme was bound to pyruvate relative to the apo structure (Table 4). This agreed with the STING MILLENIUM (40) analysis, which predicted greater inter-residue contacts at the weak dimer (and also the tight dimer) interface with a total of twelve more hydrogen bonds for Ba-DHDPS when bound to pyruvate ( Fig. 5C and Table 4). This is significant given that the pyruvate-bound structure solved in this study (3HIJ) shares the same resolution, unit cell parameters, and space group with the pyruvate-unbound structure (1XL9). Furthermore, this explains the increased thermostability (Fig.  2B) and the tetramerization propensity of Ba-DHDPS (Fig. 3) in the presence of pyruvate given the greater number of noncovalent interactions formed at the tetramerization interface upon substrate binding.
In addition, temperature-dependent dynamic disorder of atoms in the crystal, represented by crystallographic temperature factors, can be used to predict potential conformational dynamics within the structure. Interestingly, the crystal structures of pyruvate-bound and -unbound Ba-DHDPS share a very similar trend across all residues with respect to average mainchain temperature factor, except for a region of 15 residues from position 220 to 235 (supplemental Fig. 3). In this region the temperature factors for the pyruvate unbound structure are much higher than in other positions in the enzyme. Not surprisingly, most of these residues are located at the weak dimer interface, where the greatest structural changes were observed upon substrate binding (Fig. 5C). This suggests that pyruvate reduces conformational dynamics at the weak dimer interface and thus enhances tetramerization and thermostability of the enzyme.
Finally, substrate-mediated stabilization of quaternary structure and enhanced thermostability have been previously reported for another enzyme, namely L-tryptophan 2,3-dioxygenase, from Bacillus brevis (45,46). Thus, this phenomenon could be of much broader utility and biological relevance to oligomeric enzymes.
In conclusion, this study describes the structural basis for substrate-mediated stabilization of the active oligomeric form of Ba-DHDPS. Bacterial DHDPS has long been considered a potential antibiotic target (16,24), which has recently been validated by knock-out studies demonstrating that DHDPS is the product of one of only 271 essential genes in Bacillus species (7). Accordingly, our study provides significant structure-function knowledge that can be applied to rationale drug design strategies in the pipeline to generating novel anti-anthrax agents. This study has thus identified an "Achilles heel" in an essential enzyme and antibiotic target from a significant human pathogen.