A Hidden Active Site in the Potential Drug Target Mycobacterium tuberculosis dUTPase Is Accessible through Small Amplitude Protein Conformational Changes*

dUTPases catalyze the hydrolysis of dUTP into dUMP and pyrophosphate to maintain the proper nucleotide pool for DNA metabolism. Recent evidence suggests that dUTPases may also represent a selective drug target in mycobacteria because of the crucial role of these enzymes in maintaining DNA integrity. Nucleotide-hydrolyzing enzymes typically harbor a buried ligand-binding pocket at interdomain or intersubunit clefts, facilitating proper solvent shielding for the catalyzed reaction. The mechanism by which substrate binds this hidden pocket and product is released in dUTPases is unresolved because of conflicting crystallographic and spectroscopic data. We sought to resolve this conflict by using a combination of random acceleration molecular dynamics (RAMD) methodology and structural and biochemical methods to study the dUTPase from Mycobacterium tuberculosis. In particular, the RAMD approach used in this study provided invaluable insights into the nucleotide dissociation process that reconciles all previous experimental observations. Specifically, our data suggest that nucleotide binding takes place as a small stretch of amino acids transiently slides away and partially uncovers the active site. The in silico data further revealed a new dUTPase conformation on the pathway to a relatively open active site. To probe this model, we developed the Trp21 reporter and collected crystallographic, spectroscopic, and kinetic data that confirmed the interaction of Trp21 with the active site shielding C-terminal arm, suggesting that the RAMD method is effective. In summary, our computational simulations and spectroscopic results support the idea that small loop movements in dUTPase allow the shuttlingof the nucleotides between the binding pocket and the solvent.

developed the Trp 21 reporter and collected crystallographic, spectroscopic, and kinetic data that confirmed the interaction of Trp 21 with the active site shielding C-terminal arm, suggesting that the RAMD method is effective. In summary, our computational simulations and spectroscopic results support the idea that small loop movements in dUTPase allow the shuttling of the nucleotides between the binding pocket and the solvent.
Commonly, nucleotide-hydrolyzing enzymes harbor a binding pocket for their cognate ligand at interdomain or intersubunit clefts. This arrangement facilitates proper solvent shielding for the catalyzed reaction. Many well described enzymes bind nucleotides in a relatively open (e.g. DNA and RNA helicases, molecular motors, and small G-proteins) or completely open (ATP-dependent DNA ligases) binding pocket. Conformational changes associated with nucleotide binding result in the closure of the active site, enabling the subsequent reaction steps (1)(2)(3)(4)(5). The conformational change of the enzyme upon nucleotide binding can be understood by comparing the available apo and ligand-containing crystal structures. However, using single molecule optical spectroscopy in addition to the static structural data, the nucleotide binding pocket of DNA polymerase I was shown to exhibit dynamic open-closed equilibrium in the binary complex with DNA that is shifted toward the closed conformation upon dNTP binding (6). In such a case, the ligand may bind to a pocket that appears relatively closed on a millisecond-second time scale.
Homotrimeric dUTPases have three equivalent active sites in the clefts constituted by two neighboring subunits ( Fig. 1) (7,8). The C-terminal arm of the third subunit folds over the active site and shields the substrate from the solvent (Fig. 1A). This segment of the polypeptide is only visible in protein crystals that contain substrate analogues. In the apo state, the C-terminal arm is undefined in the crystal structures of dUTPases from widely different species (9 -11). This observation has been interpreted as an open-closed transition of the C-terminal arm and the active site. The steady-state conformational states of the C-terminal arm were probed by a fluorescent sensor (Trp 158 in human dUTPase). The measured solvent accessibility of this sensor in various nucleotide-bound states suggested that the arm resides in a relatively closed conformation during the largest fraction of the reaction cycle time (12). A presumed openclosed conformational transition upon ligand binding could not be observed. Because of the seeming contradictory evidence between the crystallographic and fluorescence spectroscopy data, we could not establish a comprehensive model of substrate binding or product release. To fill this gap, we implemented the random acceleration molecular dynamics (RAMD) 4 method for its ability to gain high resolution information on the enzyme-nucleotide interaction. Under physiological conditions, ligand binding and dissociation usually takes place on a microsecond-millisecond time scale and requires several microsecond long molecular dynamics simulation for accurate modeling (13)(14)(15)(16). Therefore, in case of buried active sites, guided or biased molecular dynamics simulations such as steered molecular dynamics (17,18), RAMD (19), and metadynamics (20,21) are preferably used to accelerate the process of ligand binding or dissociation. The RAMD method (19) applies an artificial random oriented force that acts on the ligand and accelerates the dissociation process, offering a reasonable time scale for computational simulations. Because of the reduced simulation time scale and the randomly oriented force, a larger sample size of ligand dissociation pathways are explored at atomic resolution. Considering that ligand binding is a reversible process, this method can be used to explore substrate dissociation and binding pathways as well (19). The RAMD method has been successfully employed to study ligand dissociation paths in cytochrome P450 (19), rhodopsin (22), haloalkane dehalogenase (23), histone deacetylase (24), heme oxygenase (25), thyroid hormone receptor (26), P-glycoprotein (27), and B-RAF kinase (28), among many others.
High resolution crystal structures (10, 29 -35), a detailed kinetic mechanism (12,36,37) and additional regulatory functions (38 -40) have already been revealed for dUTPases in a wide variety of species. Previously, we have investigated the detailed mechanism of substrate binding, hydrolysis, and product release in dUTPase using a combination of structural and transient kinetics methods (12,41). These results obtained from crystal structures and from spectroscopic measurements on the millisecond time scale movements did not allow insights into the possible substrate binding pathways to the hidden active site of dUTPase. This process may have pharmacological significance because we have recently shown that the dUTPase enzyme may be a selective drug target in mycobacteria (42). This enzyme plays a crucial role in DNA integrity by catalyzing the hydrolysis of dUTP into dUMP and pyrophosphate, producing the precursor for dTTP biosynthesis. At the same time, the dUTPase reaction is also responsible for balancing a low intracellular dUTP:dTTP ratio required in most organisms to avoid DNA uracilation that may have serious consequences (8,43).
In this study, we determine the putative ligand shuttling pathway of dUTP between the solvent and the active site of the Mycobacterium tuberculosis dUTPase using RAMD simulations in combination with experimental methods including X-ray crystallography, fast kinetics, ligand binding, and solvent accessibility experiments. The results suggested that active site residues are predominately in the closed conformation over the entire reaction cycle and during the nucleotide binding process. We observe a small stretch of amino acids that slides transiently away with small amplitude movements to partially uncover the active site. In addition,  (33)). The subunits are shown as white, cyan, and blue ribbons, whereas the fifth conserved motif located on the C-terminal arm is highlighted as thick yellow ribbon in all subunits. B, three-dimensional structure of the active site of M. tuberculosis dUTPase enzyme complexed with dUPNPP (PDB code 2PY4 (33)). The protein backbone is shown as white ribbon, and the conserved motifs are highlighted with stick representation of different colors (violet, green, orange, blue, and yellow for first, second, third, fourth, and fifth conserved motif, respectively). dUTP is visualized by atomic colored stick representation. Mg 2ϩ ion, the water molecules of the Mg 2ϩ ion coordination sphere, and the catalytic water molecule are shown as green and red spheres, respectively. C, amino acid sequence of the M. tuberculosis dUTPase. The conserved motifs are color-coded according to B.
this is the first study that demonstrates direct experimental validation of the powerful RAMD method.

RAMD Simulations Reveal the High Spatiotemporal Resolution Movements of the Active Site upon Substrate Unbinding-
Of 72 individual RAMD simulations performed, 55 led to productive dissociation of the ligand from the active site. In the remaining 17 runs, the ligand did not exit the active site during the entire 300-ps length of the simulations. A representative of unproductive simulations can be seen in Fig. 2.
Based on the order of dissociation of the substrate from the various interaction points within the protein listed below, the 55 productive RAMD runs could be classified into five major unbinding pathways A-E (Fig. 3). The difference between these paths can be best appreciated visually by watching supplemental Movies S1-S5. To understand the classification criteria in the text format analysis in Table 1, it is crucial to describe the interactions of the substrate within the active site. The first conserved motif (violet in Fig. 1, B and C) coordinates the Mg 2ϩ of the physiological substrate-Mg 2ϩ complex through water molecules. The second and fourth motifs (green and blue, respectively in Fig. 1, B and C) coordinate the phosphate chain of the substrate, whereas the third motif (orange in Fig. 1, B and C) maintains interactions with the uracil and deoxyribose moieties of the nucleotide (35). The fifth motif (yellow in Fig. 1, B and C) located on the C terminus coordinates the ␥-phosphate and the uracil ring of the substrate and is responsible for diphosphate/triphosphate ligand discrimination mainly by promoting hydrolysis of the triphosphate nucleotide (41). In agreement with this, it has been shown in a wide variety of dUTPases that the enzymatic activity of C-terminally truncated mutants is drastically lower (practically 0) than that of the WT (11,34,37,41,44,45).
The structural element in the focus of the present study is the fifth motif as it shields the nucleotide binding cavity and restricts nucleotide shuttling. Fig. 4 highlights the position of this motif relative to the observed exit routes and also shows its overall conformational change between the initial and final time points of the RAMD runs. In pathway A, the N-terminal part of the C-terminal arm transiently moves away from the enzyme core as the substrate leaves the active site, whereas the FIGURE 2. A representative of the 17 unproductive RAMD runs. The enzyme can be seen with gray surface representation. For clarity, the C-terminal arm is omitted from this picture. dUTP is visualized as atomic coloring stick model, and Mg 2ϩ is an orange sphere. The path of the center of mass of the dUTP-Mg 2ϩ complex is visualized as a blue, white, and red line.  rest of the C-terminal arm keeps its position ( Fig. 4A and supplemental Movie S1). In pathway C, partial unfolding of the C-terminal arm can be observed, whereas all intermolecular contacts are broken between the C-terminal arm and the protein core ( Fig. 4C and supplemental Movie S3). In contrast, the C-terminal arm moves on the surface of the protein without losing its interactions with the rest of the protein in pathways B, D, and E ( Fig. 4, B, D, and E, and supplemental Movies S2, S4, and S5). In three of the runs (shown in green, orange, and gray in Fig. 4E), the route of ligand exit changes its direction after its initial progress, which is due to a steric clash with the first conserved motif forcing the ligand to exit at different directions. Another subtle difference that was observed during the analysis of the RAMD runs was that in pathways A and C, but not in pathways B, D, or E, the coordination geometry between Mg 2ϩ and the phosphate chain of dUTP ( Fig. 5) becomes distorted as the ligand exits the active site (Table 1). This distortion coincides with a larger conformational change of the C-terminal arm as also reflected by the root mean square deviation (RMSD) for the C-terminal arm between the initial and final conformations (Table 1).
Importantly, the C-terminal arm remains close to the enzyme core in each RAMD simulation as opposed to opening up completely and exposing the active site to the solvent (supplemental Movies S1-S5). A quantitative analysis of the displacement of amino acids within the C-terminal arm can be viewed in Fig. 6. The small RMSD values characteristic to amino acids 130 -135 and 150 -155 indicate that the proximal and distal regions of the arm stay in contact with the enzyme core. The fifth conserved motif is situated in between these sequences and takes active part in the ligand exit process as reported by the elevated RMSD values. His 145 acts as the aromatic lid over the active site present in all dUTPases by engaging in astacking interaction with uracil ( Fig. 1B) (35,46). This interaction was earlier exploited to study the enzymatic cycle as it reports on ligand binding and product unbinding using various spectroscopic methods (12). The high RMSD values of His 145 explain the observed spectroscopic changes and reinforce the importance ofstacking interactions in directing ligand exit (47). Arg 140 , responsible for ␥-phosphate coordination, also displays relatively high RMSD values in pathways A and E. This phenomenon will be discussed in the subsequent sections.
These RAMD results provide the first body of evidence that reconciles all the previous ensemble spectroscopic measurements and crystallographic data concerning the movements and role of the C-terminal arm during substrate binding. Upon substrate binding or product release, the nucleotide moves through small and transient spaces available as a stretch of the C-terminal polypeptide chain moves away from the shielding position. Such a mechanism involves relatively small conformational changes (12) of a yet dynamic C-terminal arm (37,48) and a relatively closed active site even without bound substrate (12).
New Substrate-dependent Interaction Point Detected in the RAMD Runs-During detailed analysis of the pathways, some less conserved residues in the neighborhood of the conserved motifs appeared to take active part in the substrate unbinding    DECEMBER 16, 2016 • VOLUME 291 • NUMBER 51 process. Particularly, a yet undescribed stacking interaction between the imidazole ring of His 21 (N-terminal of the first conserved motif) and the guanidino group of Arg 140 (fifth conserved motif) residues displayed characteristic changes. This interaction in pathways A and E seems to break up as the ligand exits the active site ( Fig. 7 and supplemental Movies S1 and S5), also explaining the large displacement of Arg 140 indicated in Fig. 6. In the other three pathways, this interaction persists during the entire simulation. We found this phenomenon interesting because it directly involves the opening of the active site and the anchoring of the C-terminal arm (Arg 140 ) to the enzyme core (His 21 ). We carried out the further detailed biochemical and structural biology experiments to validate this in silico observation and to learn more about the substrate binding process.

Nucleotide Dissociation from a Hidden Active Site
Implementation of a Reporter to Study the His 21 -Arg 140 Interaction-To continuously monitor the interaction between His 21 and Arg 140 during the reaction cycle, we introduced a single Trp fluorescent sensor into position 21. We already had experience with His to Trp changes in this protein that left the protein fold and activity unaffected and provided useful signals (33,42,49). Position 21 is conserved in the bacterial phylum Actinobacteria. Tyr may replace His in some sequences (Atopobium, Olsonella, and Actinoplanes), which also suggests that the aromatic nature of the amino acid in this position is important. The H21W point mutant M. tuberculosis dUTPase was thus generated to study the dynamic behavior of the interaction between the protein core and the C-terminal arm of the protein.
Crystal Structure of the H21W Mutant-The crystal structure of the H21W mutant has been solved to 1.5 Å resolution and deposited to the PDB database under code 4GCY (for crystallographic data see Table 2). The global structure of the mutant enzyme is highly similar to the WT (PDB code 2PY4 where the sum is calculated over all observations of a measured reflection (I i ), and ϽIϾ is the mean intensity of all the measured observations (I i ). c R cryst ϭ ⌺ (͉F obs Ϫ F calc ͉)/⌺ (F obs ), where F obs is the observed structure factor amplitude, and F calc is that calculated from the model. d R free is equivalent to R cryst but where 5% of the measured reflections have been excluded from refinement in thin shells and set aside for cross-validation purposes. e Average B values for all atoms/ ligand atoms/water/Mg 2ϩ ions. f Ramachandran plot analysis was from MolProbity. (33)), which is clearly demonstrated by the 0.2 Å RMSD of the protein backbone atoms (Fig. 8A). It should be mentioned that parts of the flexible termini of the structure are affected by slight model bias and are flagged as real space R value Z score outliers. This, however, does not affect the accurate interpretation of the protein core and the discussed residues of interest. Thus, the active site can be considered as equivalent to that of the WT. The RMSD of the non-hydrolyzable substrate analogue dUPNPP-Mg 2ϩ complexes in the reference and new structure is 0.1 Å. The mutant Trp side chain is in the same position as the His side chain in the WT (Fig. 8B), despite the fact that the indole ring of the Trp residue is larger and more hydrophobic than the imidazole ring of the His. The indole ring establishes similar interactions with the backbone atoms of Ser 139 and Arg 140 and with the side chain of Arg 140 (Fig. 8C) to those of the imidazole ring (Fig. 8D). The hydrogen bond with the side chain of Ser 139 vanished upon the substitution. Arg 140 is the first residue of the fifth conserved motif. It has a crucial role in the coordination of the ␥-phosphate of the substrate by a bifurcated hydrogen bond (10) and hence in the enzymatic catalysis (41,50). These structural similarities suggest a WT behavior for the H21W mutant similar to a previously described H145W mutant, where the mutated aromatic residue stacks against the uracil ring of the substrate establishing astacking interaction (35).

Enzymatic Mechanism Probed by Transient Kinetics-
The k cat of the H21W mutant was measured to be 2.9 s Ϫ1 (Table 3) which is comparable with that of the WT (3.1 s Ϫ1 according to Pecsi et al. (35)). Next, we assayed whether the fluorescence of the Trp 21 residue reports any events of the reaction cycle. Surprisingly, the H21W mutant displays similar fluorescent signal change during the enzymatic cycle as previously reported for the H145W active site Trp mutant (Fig. 9A) (35,49). In the H21W mutant, however, the Trp residue is outside of the active site, presumably far from the substrate during the chemical reaction. It is only connected to the active site through the interaction with Arg 140 pointed out in our RAMD runs. This suggests that the movements of the C-terminal arm during the reaction are efficiently transmitted to the Trp 21 residue. As a result, the fluorescence intensity of the Trp 21 residue decreases upon substrate binding and remains quenched in the steadystate hydrolysis phase to recover upon products dissociation (Fig. 9A).
Using the large fluorescent signal, we recorded time courses in substoichiometric dUTP concentrations in the stopped flow apparatus to obtain the true single turnover rate constant (k sto ) of the dUTPase reaction (Fig. 9A). The exponential fit to the fluorescence recovery phase yielded k sto ϭ 3.1 s Ϫ1 . The observation that k cat equals k sto with a fluorescent reaction profile similar to that of the H145W dUTPase implies the same enzymatic mechanism in which dUTP hydrolysis is rate-limiting (12).
The kinetics of dUTP binding was also measured using stopped flow (Fig. 9B). For comparison, the yet undescribed substrate binding kinetics of H145W was measured as well ( Fig.  9C) (our previous kinetic investigation (12) was carried out using the human enzyme). The observed ligand binding rate constants are plotted against the ligand concentration and fitted with linear functions to obtain the microscopic association and dissociation rate constants (k on and k off , respectively; Fig. 9, D and E, and Table 3). Compared with the human enzyme, the M. tuberculosis H145W enzyme displays similarly high k off value and somewhat lower k on value. Rate constants of the H21W mutant are in the same range as those for the H145W; however, there is a 3.5-fold decrease in the k on value and a 50% increase in the k off value. The apparent dissociation constants of the enzyme-dUTP complexes could not be determined from the amplitudes of the fluorescence change caused by saturating conditions. In view of the structural similarity between the two   DECEMBER 16, 2016 • VOLUME 291 • NUMBER 51 enzymes (Fig. 8), it was not expected that the ligand binding process would be much slower when a Trp residue is present instead of a His in position 21. This change, however, further proves that this residue is involved in the ligand binding pathway.

Nucleotide Dissociation from a Hidden Active Site
Ligand Binding-To reveal more mechanistic details of the involvement of residue 21 in the ligand binding/dissociation pathways, we carried out comparative fluorescence and circular dichroism studies. The fluorescent spectra of the apo H21W mutant enzyme and its complexes with dUMP, dUDP, and the dUTP-mimicking, practically non-hydrolyzable dUPNPP were recorded (Fig. 10A). The relative fluorescence quench induced by these ligands differs considerably from that observed previously in the H145W mutant (Fig. 10B). Only dUPNPP induced a significant fluorescence intensity decrease, whereas the dUDP and dUMP complexes of the H21W mutant remained fluorescently silent. In contrast, all ligand-enzyme complexes of H145W show a decrease in fluorescence intensity (49) according to the extent of C-terminal arm rigidity, which stabilizes the stacking interaction between the indole ring of Trp and uracil (12). The Trp incorporated in the H21W mutant thus only reports the binding of the triphosphate nucleotides. This phenomenon is consistent with the RAMD results, suggesting that the Trp 21 residue is in interaction with Arg 140 responsible for coordinating the ␥-phosphate of the ligand exclusively (Fig.  8C). The dissociation constant of the enzyme-dUPNPP complex was determined using fluorescence intensity titrations and was found to be 1 order of magnitude larger than that of the H145W-dUPNPP complex (Fig. 10C).
Binding of dUDP could not be measured using fluorescent titration in a lack of a detectable signal (Fig. 10A). However, to ensure that dUDP indeed binds to the H21W mutant, circular dichroism titration measurements were performed. Determination of the K d of the enzyme-dUDP complex using this method is only possible with relatively high error because the concentrations applied in the measurements are necessarily far above the K d . Nevertheless, it is clear from the data that the H21W mutant binds dUDP significantly tighter than the H145W (p ϭ 0.046) (Fig. 10D).
The results of the nucleoside mono-, di-, and triphosphate binding experiments together with the RAMD data suggest that Trp 21 senses the nucleotide-bound state of the active site through its interaction with Arg 140 . Arg 140 probably remains in the apo state conformation when only monophosphate or diphosphate nucleotides bind to the active site (Fig. 10A). Because the C-terminal arm (including Arg 140 ) is usually only visible in crystal structures that contain the triphosphate ligand, the apo state conformation of Arg 140 has never been reported before. However, the dUDP-complexed structure of the M. tuberculosis dUTPase is available for comparison for the position and mobility of Arg 140 in two different nucleotidebound states (Fig. 11). By comparing the relative B-factors of Arg 140 within the same structure in Fig. 11 (A and B), it is clear that with a lack of the ␥-phosphate, the mobility of the Arg residue is much higher. This phenomenon suggests that only a relatively static Arg 140 is able to change the fluorescence properties of Trp 21 , probably through cation-interaction, and thus yields a specific sensor for nucleoside triphosphate binding. It is interesting that modification of the electrostatic properties of the aromatic interacting partner (i.e. His to Trp change) perturbed the nucleotide affinities to the enzyme (Table 3). A possible explanation to this phenomenon may be that the Arg 140 - FIGURE 9. Fast kinetics of the H21W and H145W mutants. The latter was described earlier to display WT behavior (35). A, the fluorescence time courses of the H21W mutant hydrolyzing sub-and superstoichiometric amounts of substrate show similar progression to those reported for the H145W mutant. 3 M enzyme was mixed with various concentrations of dUTP (post-mix concentrations). Triple exponential function was fitted to the single turnover curve obtained for the lowest, substoichiometric dUTP concentration which yielded 3.1 s Ϫ1 for the third phase (k 1,obs ϭ 69 s Ϫ1 , k 2,obs ϭ 3.6 s Ϫ1 ). In earlier works, this rate constant was identified as the single turnover rate constant (k STO ) (12). B and C, dUTP binding to the H21W (B) or to the H145W (C) mutant. 1 M enzyme was mixed with buffer or with various concentrations of dUTP in pseudo first order conditions. Double exponential function was fitted to the data. D, concentration dependence of the observed catalytic rate constants of the fast phase of dUTP binding (k obs,1 ). E, concentration dependence of the observed catalytic rate constants of the slow phase of dUTP binding (k obs,2 ). This phase seems to be independent of the substrate concentration in agreement with previous data.
Trp 21 interaction is more stable than the Arg 140 -His 21 interaction in the WT. This in turn results in less frequent dislocation of the C-terminal arm (Fig. 7) and finally in the slower kinetics observed for dUTP binding (Table 3).
Solvent Accessibility-The solvent accessibility of the incorporated Trp 21 residue was measured by acrylamide quenching to obtain further dynamic information about the vicinity of the studied interaction. The acrylamide quenching of N-acetyl-Ltryptophanamide (NATA) was measured as reference and control for a rotationally free and maximally solvent-accessible Trp residue. The quenching constants (K SV ) for the apo and the dUPNPP-bound enzymes were similar and significantly reduced compared with NATA (Fig. 12). The solvent accessibility of the active site Trp in the apo state and in the dUPNPPbound complex are also shown for comparison (12). The comparative data indicate that the Trp 21 residue is buried similarly to the active site Trp. This suggests that part of the C-terminal arm shielding Trp 21 from the solvent is associated to the protein core in equilibrium even in the apo enzyme. Consistently in all of the RAMD runs, we also observed that the C-terminal arm remains associated with the enzyme core following dUTP dissociation.

Discussion
This study combined computational, biochemical, and biophysical methods to gain insights into the nucleotide binding of a nucleotide-hydrolyzing enzyme at exceptionally high time and structural resolution. The RAMD method provided the high time and structural resolution, whereas the numerous experimental data served to validate the results of the computational simulation. In terms of mechanistic detail, the RAMD runs can be regarded as fast transient kinetics experiments on a  Table 3. D, circular dichroism titration of H145W and H21W with dUDP; quadratic fit to the data yielded dissociation constants (K d,dUDP ) shown in Table 3. FIGURE 11. Thermal mobility of the Arg 140 residue in various ligand bound complexes. The enzyme is shown as thick transparent ribbon, whereas the ligand and Arg 140 are visualized as sticks. Coloring represents the relative crystallographic B factor for each atom (blue, less mobility; red, more mobility). A, the structure of M. tuberculosis dUTPase complexed with dUPNPP (PDB code 1SIX (10)). B, the structure of M. tuberculosis dUTPase complexed with dUDP (PDB code 1SLH (10)). Note that Arg 140 is in different conformation and that Arg 140 has much more mobility compared with the dUPNPP-complexed structure.  DECEMBER 16, 2016 • VOLUME 291 • NUMBER 51 single molecule level. It appears that the assembly of the phenomena observed in single RAMD runs agrees well with the ensemble solution kinetics results presented earlier. The RAMD models reveal how the substrate can leave/enter the active site without completely opening the C-terminal arm. A previously hypothesized mechanism like a "hinge door" in which the C-terminal arm completely opens up at the hinge of a conserved Pro (51) was not observed in the RAMD simulations. Instead, we observed the nucleotide shuttling between the active site and the solvent through small openings of the C-terminal arm. The arm, fixed at both ends to the enzyme core via secondary interactions, moves only by sliding away close to the enzyme surface or by opening partially and locally. These movements involving 10 -15 residues are small compared to the conformational changes observed for other nucleotide binding enzymes (4,5,52). In DNA helicases and polymerases, substrate binding involves displacement of whole domains (1)(2)(3)6). It should be mentioned that the RAMD method may not be capable of modeling large loop motions because relevant reports are missing. However, our computational simulations and spectroscopic results together unequivocally support the finding that small loop movements allow the shuttling of the nucleotides between the binding pocket and the solvent in dUTPase, whereas larger scale movements may not be excluded.

Nucleotide Dissociation from a Hidden Active Site
There may be an apparent contradiction between the C-terminal arm not being visible in the crystal structures and the RAMD simulations showing only small amplitude movements of the C-terminal arm. These phenomena, however, are complementary. The electron density maps represent the average conformation of all atoms in a protein crystal. The inability to localize an entity may be due to movements on a small amplitude scale just exceeding the atomic distance. Molecular dynamics simulations, on the other hand, investigate single molecules rather than a population average and commonly reveal small amplitude movements. The example of dUTPase demonstrates that the comparison of crystal structures in various nucleotide-bound states alone is not fully informative of the nucleotide binding mechanism.
This study yielded novel insights regarding the dynamics and other properties of the interaction between dUTP and Mg 2ϩ as well. The RAMD method offered a unique opportunity to observe that this interaction is maintained during the whole ligand binding/unbinding process, although the octahedral coordination between Mg 2ϩ and dUTP became distorted in some of the pathways.
As proof of principle, we chose the His 21 residue de novo predicted by the RAMD simulations to be involved in substrate dissociation to validate our data. It turned out that the replacement of this residue with a Trp yielded a sensitive fluorescence reporter specific to triphosphate nucleotide binding. As such, Trp 21 reported the hydrolysis reaction similarly to the active site nucleotide sensor, Trp 145 . Using the combination of structural and optical spectroscopic methods, we proposed a possible molecular mechanism for nucleotide sensing by Trp 21 relatively far from the active site. This phenomenon also shines light onto the possibility to exploit this specific interaction in target-based drug design against actinobacterial pathogens.

System Preparation for Molecular Dynamics Simulation-
The initial structure for the simulations was the crystal structure of the M. tuberculosis dUTPase complexed with dUPNPP (PDB code 2PY4 (33)). The original dUPNPP ligand in the crystal structure was manually modified to dUTP. The protonation states of the titratable groups were determined by means of the ProPka algorithm (53,54) at pH 7.4, and the optimization of hydrogen bonding interactions was performed using the PDB2PQR program package (55). The structure obtained was solvated using the TIP3P water model (56), and a minimum distance of 14 Å was used between any atom of the protein and the wall of the periodic box. The number of Na ϩ cations and Cl Ϫ anions was 49 and 34, respectively, to neutralize the system and mimic the physiological ionic strength (0.15 M NaCl). For description of the protein and dUTP, the CHARMM27 parameters (57) including the CMAP correction term for proteins (58) were applied, whereas the ions were described by the parameters of Beglov and Roux (59).
Molecular Dynamics Simulations-Molecular dynamics simulations were performed with the NAMD2.7 software (60) applying the following protocol: (i) 1600 steps minimization with restrained protein atoms using a 10 kcal mol Ϫ1 Å Ϫ2 force constant; (ii) 3200 steps minimization; (iii) 32-ps NVT dynamics, where number of atoms (N), volume (V), and temperature (T) are fixed (T ϭ 310 K, Langevin dynamics (61)) with restrained protein atoms using a 10 kcal mol Ϫ1 Å Ϫ2 force constant; and (iv) NPT dynamics, where number of atoms (N), pressure (P), and temperature (T) are fixed (p ϭ 1 atm, T ϭ 310 K) with the Noosé-Hoover Langevin piston method (61,62). The time step was set to 1 fs, a cutoff value of 10 Å was applied (using a smoothing function from 8 Å), and the long range electrostatic interactions were calculated with the Particle Mesh Ewald (PME) method (63). Two independent, 2-nslong simulations with different initial velocities were conducted to relax the system, and the structures of the last frames were used in the subsequent RAMD simulations. During the 300-pslong RAMD simulations, the acceleration was set to 0.066 kcal mol Ϫ1 Å Ϫ1 (atomic mass unit) Ϫ1 , and the successful exit event was accepted if the distance between the center of mass of protein and that of dUTP was larger than 50 Å. Every 80 fs, new orientation was generated if the distance traveled by the ligand was less than 0.1 Å, and the trajectory was saved every 80 fs. No restraints were applied when conducting the RAMD simulations. From both starting structures and every binding cavity, 12 RAMD simulations were started resulting in 72 calculations. Once the ligand escapes into the solvent, the dissociation pathway becomes smooth because of the constant force acting on the substrate. This linearity of the ligand trajectories within the bulk solvent is irrelevant for the ligand finding the possible paths out of the active site.
Analysis of the Trajectories-Analysis of the trajectories was performed with the Visual Molecular Dynamics (64) package, which was used to generate Figs. 1-4, 7, 8, and 11 and supplemental Movies S1-S5.
Reagents-Molecular biology products were from New England Biolabs and Fermentas, electrophoresis and chromatogra-phy reagents were from Bio-Rad and Qiagen. Phenol red was from Merck, dUDP and dUPNPP was from Jena Bioscience, and dUTP and other chemicals were from Sigma-Aldrich.
Cloning, Protein Expression, and Purification-were performed as described previously in Ref. 33. Recombinant dUTPase with an N-terminal His 6 tag was transformed into Escherichia coli strain Rosetta(DE3)pLysS, which was used for protein expression. The final supernatant after cell extraction was loaded on a nickel-nitrilotriacetic acid column (Novagen) and purified according to the Novagen protocol. Because wild type lacks a useful tryptophan to monitor ligand binding, we have used the H145W mutant for kinetic studies. This tryptophan substitution in the active site has been previously shown to not affect catalytic activity (33). Protein concentration was measured using the Bradford method (Bio-Rad protein assay) and is given in subunits. Protein was dialyzed against buffer pH 7.5, containing 20 mM HEPES, 100 mM NaCl, 2 mM MgCl 2 , and 1 mM DTT, and measurements were carried out in this buffer if not stated otherwise.
Crystallization-The H21W mutant was crystallized as described for the WT enzyme using the hanging drop diffusion method (33). 250 M dUTPase and 1.25 mM dUPNPP were mixed with the reservoir solution containing 50 mM Tris-HCl, pH 7.5, 10 mM MgCl 2 , 1.20 -1.75 M ammonium sulfate, and 10% glycerol in a 1:1 ratio.
Data Collection and Structure Determination-High resolution complete crystallographic data set from a single flash frozen crystal was collected at Eötvös University using RIGAKU RU-H2R rotating anode home source at ϭ 1.5418 Å on 100 K temperature. Reflection data were recorded on RIGAKU R-AXIS IVϩϩ image plate detector.
Data reduction was performed using iMosflm and Scala programs and the CCP4 program suite (65). The structure of the H21W mutant in complex with dUPNPP and Mg 2ϩ was determined by molecular replacement using MolRep, where the model was the WT dUTPase structure (PDB code 2PY4 (33)). R merge ϭ Α(͉I i Ϫ ϽIϾ ͉)/Α(ϽIϾ), where the sum is calculated over all observations of a measured reflection (I i ), and ϽIϾ is the mean intensity of all the measured observations (I i ).
Manual model building was carried out running the Win-Coot program (66). Refmac5 software from CCP4 program package (65) was used for high resolution refinement. R free values were computed from cross-validation throughout the refinement, using 5% random subset of the data set. Ramachandran statistics were as follows: 99.3% favored, 0.7% additionally allowed, and 0% generously allowed. The coordinates and structure factor data have been deposited in the Protein Data Bank under PDB code 4GCY.
Steady-state Colorimetric dUTPase Assay-Steady-state colorimetric dUTPase assay was performed as described in Ref. 35. Phenol red indicator assay was used to detect protons released in the dUTPase reaction. 0.1 M protein was used in a buffer containing 1 mM HEPES, pH 7.5, 100 mM KCl, 40 M phenol red, and 5 mM MgCl 2 . A Specord 200 (Analytic, Jena, Germany) spectrophotometer and 10-mm-path length cuvettes were used at 20°C, and absorbance was recorded at 559 nm. Initial velocity was determined from the first 10% of the progress curve.
Stopped Flow Experiments-Measurements were done using an SX-20 (Applied Photophysics) stopped flow apparatus as described in Refs. 12 and 35. Trp fluorescence was excited at 297 nm, and emission was selected with a 320-nm long pass filter. Time courses were recorded upon mixing sub-and superstoichiometric concentrations of dUTP with 1 or 3 M enzyme at 20°C (post-mix concentrations). Single turnover curves could be fitted with three exponentials (12). The two fast phases represent ligand binding, whereas the third, fluorescence recovery phase represents the hydrolysis-limited product release event and yields the single turnover rate constant (12). The fluorescence decrease of the time courses resulted from mixing excess dUTP over active sites and was fitted with double exponential function. Each time course shown is the mean of five measurements recorded sequentially in the same experiment. The k obs versus dUTP concentration plots show the mean and standard error of three independent measurements. Time courses were analyzed using the curve fitting software provided with the stopped flow apparatus or by Origin 7.5. (OriginLab Corp., Northampton, MA).
Determination of the Enzyme-Ligand Dissociation Constants-Determination of the enzyme-ligand dissociation constants was done by fluorescence intensity titrations and circular dichroism intensity titrations. Fluorescence intensity titrations were recorded at 20°C on a Jobin Yvon Spex Fluoromax-3 spectrofluorometer with excitation at 295 nm (slit, 1 nm), emission at 347 nm (slit 5 nm) using proteins at 4 M concentration as described in Ref. 49. CD spectra were recorded at 20°C on a JASCO 720 spectropolarimeter using a 10-mm-path length cuvette. Protein concentration was 50 M in a buffer containing 20 mM HEPES, pH 7.5, 50 mM NaCl, and 2 mM MgCl 2 . A spectrum between 250 and 290 nm was recorded at each nucleotide concentration as described in Ref. 35. Differential curves were obtained by subtracting the signal of the ligand alone from that of the corresponding complex. Differential ellipticity at 269 nm was plotted against the ligand concentration to obtain the binding curves.
In both types of measurement, protein solution was titrated with dUPNPP or dUDP using small volumes from a concentrated stock solution. The following equation describing 1:1 stoichiometry for the dissociation equilibrium was fitted to the titration data, y ϭ s ϩ A͓͑c ϩ x ϩ K͒ Ϫ ͱ͑c ϩ x ϩ K͒ 2 Ϫ 4cx͔ 2c (Eq. 1) where x is the nucleotide concentration, y is the fluorescence or circular dichroism intensity, s ϭ y at x ϭ 0, A is the amplitude of the fluorescence or circular dichroism intensity change, c is the enzyme concentration, and K is the dissociation constant (K d ).
Acrylamide Quenching-Acrylamide quenching was carried out by the addition of small volumes of a 5 M acrylamide solution to the enzyme, enzyme-dUPNPP, or NATA solutions as described in Ref. 12. Raw data were corrected for the fluorescence arising from the acrylamide solution itself. F 0 /F versus [Q] curves were analyzed using a modified Stern-Volmer equation, where F 0 is the unquenched, F is the quenched fluorescence, [Q] is the quencher, K SV is the dynamic (bimolecular) quenching constant, and V is the static (sphere of action) component of quenching.