The Flexible Motif V of Epstein-Barr Virus Deoxyuridine 5′-Triphosphate Pyrophosphatase Is Essential for Catalysis

Deoxyuridine 5′-triphosphate pyrophosphatases (dUTPases) are ubiquitous enzymes essential for hydrolysis of dUTP, thus preventing its incorporation into DNA. Although Epstein-Barr virus (EBV) dUTPase is monomeric, it has a high degree of similarity with the more frequent trimeric form of the enzyme. In both cases, the active site is composed of five conserved sequence motifs. Structural and functional studies of mutants based on the structure of EBV dUTPase gave new insight into the mechanism of the enzyme. A first mutant allowed us to exclude a role in enzymatic activity for the disulfide bridge involving the beginning of the disordered C terminus. Sequence alignments revealed two groups of dUTPases, based on the position in sequence of a conserved aspartic acid residue close to the active site. Single mutants of this residue in EBV dUTPase showed a highly impaired catalytic activity, which could be partially restored by a second mutation, making EBV dUTPase more similar to the second group of enzymes. Deletion of the flexible C-terminal tail carrying motif V resulted in a protein completely devoid of enzymatic activity, crystallizing with unhydrolyzed Mg2+-dUTP complex in the active site. Point mutations inside motif V highlighted the essential role of lid residue Phe273. Magnesium appears to play a role mainly in substrate binding, since in absence of Mg2+, the Km of the enzyme is reduced, whereas the kcat is less affected.

Deoxyuridine 5-triphosphate pyrophosphatases (dUTPases) are ubiquitous enzymes essential for hydrolysis of dUTP, thus preventing its incorporation into DNA. Although Epstein-Barr virus (EBV) dUTPase is monomeric, it has a high degree of similarity with the more frequent trimeric form of the enzyme. In both cases, the active site is composed of five conserved sequence motifs. Structural and functional studies of mutants based on the structure of EBV dUTPase gave new insight into the mechanism of the enzyme. A first mutant allowed us to exclude a role in enzymatic activity for the disulfide bridge involving the beginning of the disordered C terminus. Sequence alignments revealed two groups of dUTPases, based on the position in sequence of a conserved aspartic acid residue close to the active site. Single mutants of this residue in EBV dUTPase showed a highly impaired catalytic activity, which could be partially restored by a second mutation, making EBV dUTPase more similar to the second group of enzymes. Deletion of the flexible C-terminal tail carrying motif V resulted in a protein completely devoid of enzymatic activity, crystallizing with unhydrolyzed Mg 2؉ -dUTP complex in the active site. Point mutations inside motif V highlighted the essential role of lid residue Phe 273 . Magnesium appears to play a role mainly in substrate binding, since in absence of Mg 2؉ , the K m of the enzyme is reduced, whereas the k cat is less affected.
Epstein-Barr virus, a human ␥-herpesvirus, is the causative agent of infectious mononucleosis and establishes a lifelong persistent infection in over 90% of the world's population. EBV 3 is implicated in a number of cancers, such as Burkitt's lymphoma, undifferentiated nasopharyngeal carcinoma, or Hodgkin disease. The large DNA genome of this virus codes for about 86 proteins implicated in a large num-ber of functions related to viral latency or the lytic cycle, during which the virus replicates.
One protein of the lytic cycle is deoxyuridine 5Ј-triphosphate pyrophosphatase, a ubiquitous enzyme catalyzing the cleavage of dUTP into dUMP and pyrophosphate (PP i ). This enzyme not only provides the precursor for the formation of dTMP by thymidylate synthase, but it also has a crucial role in maintaining a low dUTP/dTTP ratio in the cell in order to limit the incorporation of deoxyuridylate into DNA by DNA polymerases. Based on their oligomerization state, dUTPases can be divided into three families.
The first family of dUTPases contains homodimeric enzymes with the prototype structure of Trypanosoma cruzi dUTPase (1).
dUTPases of the second and largest family form homotrimers. Their structure is unrelated to dimeric dUTPases. Trimeric dUTPases are found in most eukaryotes, in prokaryotes, in some DNA viruses, such as poxvirus, and in a number of retroviruses. Several x-ray structures of dUTPases of this family are available: Escherichia coli (2), Homo sapiens (3), equine infectious anemia virus (4), feline immunodeficiency virus (5), Mason-Pfizer monkey virus (6), Mycobacterium tuberculosis (7), Plasmodium falciparum (8), Arabidopsis thaliana (9), and vaccinia virus (10). The active site is formed by five conserved motifs that are distributed over the entire sequence (11) (Fig.  1A). Each subunit of the trimer contributes to the formation of each of the three active sites (2). Whereas the first four motifs are well ordered, motif V is most often disordered and is only observed in some structures after inhibitor binding such as in the feline immunodeficiency virus dUTPase structure (Protein Data Bank entry 1f7r) (12) and the human dUTPase structure (Protein Data Bank entry 1q5h) (3), both in complex with dUDP. Recently, several structures showing motif V and ␣,␤imido-dUTP in a conformation close to the situation during catalysis became available: one of the human enzyme (Protein Data Bank entry 3ehw) (13) and one from M. tuberculosis (Protein Data Bank entry 2py4) (14). Together with kinetic information about the human enzyme (15), a model of enzymatic action became available: (i) rapid, probably diffusion-limited binding of the substrate; (ii) a substrate-induced structural change required for the formation of the catalytically competent conformation; (iii) the rate-limiting hydrolysis step; and (iv) a rapid release of the reaction products PP i and dUMP. Hydrolysis occurs through a nucleophilic in-line attack on the ␣-phosphate by a water molecule activated by an aspartic acid residue (16).
The third family contains monomeric dUTPases encoded by avian and mammalian herpesviruses, which include important human pathogens, such as Epstein-Barr virus and herpes simplex virus. They have limited sequence homology to the trimeric enzymes; the five motifs forming the active site are present but reshuffled and spread out over a single polypeptide that is twice as long as the sequence of the subunit in trimeric enzymes. Working on the enzyme from EBV, we reported the first crystal structures of a monomeric dUTPase determined in complex with the product dUMP or the non-hydrolyzable substrate analogue ␣,␤-imido-dUTP (17) (Fig. 1B). Like most dUTPase structures, those of the EBV enzyme showed four of the five conserved sequence motifs (Fig. 1A), with motif V invisible due to its flexibility, and revealed an active site that is extremely similar to those of trimeric dUTPases. EBV dUTPase furthermore exhibited similar kinetic parameters (17,18). This therefore implies that the conclusions of studies of the enzymatic mechanism for trimeric dUTPases are also valid for the EBV enzyme and vice versa.
In the previously published EBV dUTPase structures (17), the N and C termini of the ordered structure were linked by a disulfide bridge between Cys 4 and Cys 246 . We therefore wished to check the influence of this bridge on activity and its possible regulatory role by studying mutant C4S.
EBV dUTPase structures (17) revealed that residues 114 -133 linking the two domains of monomeric dUTPase and containing part of motif I show different structures in the ␣,␤imido-dUTP-bound form (Fig. 1B) and in the dUMP-bound form (Fig. 1C). This leads indirectly to a non-productive conformation of catalytic residue Asp 76 of motif III (Fig. 1B). Motif I is always well ordered in trimeric dUTPases. We hypothesized first that the clustering of negative charges of two aspartic acid residues (catalytic residues Asp 76 and Asp 131 of the linker region) and the ␥-phosphate of the inhibitor leads to destabilization of this part of the structure. When dUTPase structures showing motif V in a productive conformation became available (13,14), they showed an interaction of the residue corresponding to Asp 131 with the conserved arginine residue of motif V. On the other hand, Asp 131 is only partially conserved in sequence alignments. These revealed two classes of dUTPases, the first containing EBV dUTPase with a conserved aspartic acid residue in a position corresponding to Asp 131 and the second group containing human and E. coli enzymes with a conserved aspartic acid residue at the position corresponding to residue 78 in EBV (Fig. 1A). We decided to study three single mutants of Asp 131 (D131S, D131N, and D131E) along with a double mutant (D131S/G78D), making the EBV enzyme more similar to the second group.
Finally, we report the structure of a dUTPase mutant with motif V deleted (⌬V), constructed in order to facilitate binding studies and crystallographic studies in view of potential antiviral drug design. This mutant showed the presence of intact Mg 2ϩ -dUTP in its active site, posing the question of why this mutant is completely inactive although most of the catalytic machinery is in place. In order to understand the precise mechanism of action of motif V, we decided to study two conserved residues of motif V, Arg 268 and Phe 273 , since it has been shown that they interact with the bound substrate analogue ␣,␤-imido-dUTP (13,14).
Since binding of an Mg 2ϩ -substrate complex appeared possible without catalysis, we wanted to further characterize the role of Mg 2ϩ . Previous studies carried out on E. coli dUTPase (16,19,20) indicated an important role of Mg 2ϩ for catalytic activity, by promoting enzyme-substrate complex stabilization (21). Various dUTPase structures in the presence of ␣,␤-imido-dUTP show the metal ion coordinating all three phosphate groups.

MATERIALS AND METHODS
Site-directed Mutagenesis-Mutagenesis was carried out according to the manufacturer's instructions, using the QuikChange II kit (Stratagene) on EBV wild type (WT) dUTPase coding sequence cloned into pPROEx Htb plasmid (17). Proteins were expressed in E. coli BL21 (DE3) strain and purified on Ni 2ϩ -nitrilotriacetic acid resin, as described previously (17). ⌬V mutant is a 257-residue construct with the last 22 C-terminal residues removed.
Single crystals were harvested from the drop, dipped into paraffin oil from the panjelly kit (Jena Biosciences), and frozen directly at 100 K in a nitrogen gas stream (Oxford Cryosystems). Diffraction data (Table 1) were collected at the European Synchrotron Radiation Facility (Grenoble, France). Data were integrated using MOSFLM (22) and further processed using the CCP4 package through ccp4i (23). Structures were solved by molecular replacement using MOLREP (24). Structures were refined with REFMAC (25), and models were built using COOT (26). Illustrations were made with PyMol (27).
Reduction of the Disulfide Bridge-Reducing conditions were obtained by adding 3 mM DTT to the protein in buffer for 1 h at room temperature. 1 h prior to gel electrophoresis, free thiol groups were blocked with 10 mM N-ethylmaleimide. Loading buffer did not contain ␤-mercaptoethanol for non-reducing conditions; otherwise it contained 10 mM ␤-mercaptoethanol.
dUTPase Activity Assay-The pH drop due to proton release during dUTP hydrolysis was followed using cresol red as a pH indicator by monitoring absorbance at 573 nm. The reaction was carried out at 25°C at pH 7.8 in quartz cells (Hellma), 1-cm path, using a Varian Cary 50 Bio spectrophotometer. The mixture contained 25 M dUTP, 25 M cresol red, 150 mM KCl, 1 mM MgCl 2 , 0.5 mM Bicine-NaOH, pH 7.8. The reaction was initiated by adding dUTPase (in 5 mM Hepes, pH 7.5, 250 mM NaCl, 10 mM MgCl 2 ) to a final concentration of 31 nM. Kinetic values for K m and k cat were determined by linearization of the Essential Role of Motif V of EBV dUTPase progress curve using the integrated Michaelis-Menten equation (19).
Less active mutants were studied using the EnzChek pyrophosphate assay kit (Invitrogen), allowing photometric detection of PP i released during dUTP hydrolysis. It consists of a cleavage of PP i to phosphate by pyrophosphatase and a detection of inorganic phosphate by the conversion of 2amino-6-mercapto-7-methylpurineribonucleoside to ribose 1-phosphate and 2-amino-6-mercapto-7-methylpurine with an absorption maximum at 360 nm by purine ribonucleoside phosphorylase. Reactions were carried out at 25°C in 96-well plates in volumes of 250 l with variable dUTP and protein concentrations. The buffer contained 50 mM Tris, pH 7.5, 1 mM MgCl 2 , 0.05 mM NaN 3 . Absorbance was continuously monitored at 355 nm using a Thermo Labsystems iEMS Reader MF plate reader. Enzyme concentrations were chosen so that the initial velocity of the reaction with the dUTP substrate was at least 10 times slower than the initial velocity of the chromogenic reaction using the same concentration of a PP i standard. Kinetic constants were determined using Lineweaver-Burk plots (supplemental Fig. 1). WT dUTPase concentrations were determined using an ⑀ 280 ϭ 30,160 M Ϫ1 cm Ϫ1 .
Mg 2ϩ -free Activity Assay-To assess the importance of Mg 2ϩ for dUTPase activity, the "cresol red" assay was carried out after removing Mg 2ϩ . WT protein solution and reaction mixture without cresol red (25 M dUTP, 150 mM KCl, 0.5 mM Bicine-NaOH, pH 7.8) were passed onto a column containing 1 ml of Chelex 100 resin (Sigma) to eliminate metal cations. 25 M cresol red was then added, and the assay was performed as described above. The remaining Mg 2ϩ in the assay was quantified after dilution by flame atomic absorption spectrometry (PerkinElmer Life Sciences) at the Grenoble University Hospital. Samples and standards were prepared in a solution of 1 g/ liter lanthanum oxide, and concentrations were obtained using external calibration. The Mg 2ϩ concentration in the Chelextreated assay was below the detection limit of 0.08 M. The absence of a lasting effect of Mg 2ϩ depletion was checked by the addition of 1 mM Mg 2ϩ to the mixture after Chelex treatment. Kinetic constants were the same as without the treatment (data not shown).
Surface Plasmon Resonance Experiment-Surface plasmon resonance (SPR) measurements were performed with a BIAcore T100 (Biacore AB) operated with BIAcore T100 evaluation Software 1.1. All measurements were performed at 25°C, at FIGURE 1. EBV dUTPase sequence and structure. A, sequence alignment of different dUTPases with known structures and the one from C. glutamaticum. Residue numbers are given above the sequence for the human enzyme and below the sequences for the mycobacterial and EBV enzymes. The linker region is highlighted with cyan letters. The five conserved motifs are highlighted. Key residues of motif V are printed in red. A pink background highlights residues of motif V where side chain hydroxyl and main chain amide contact ␥-phosphate in the crystal structure (13,14). Green background, main chain amide interaction with ␤and ␥-phosphate; residues that have been deleted in the ⌬V mutant are underlined. The conserved residues interacting with the motif V arginine of human-like dUTPases are printed in blue; those interacting with the motif V arginine of mycobacterium-like dUTPases are shown in magenta. B, structure of EBV dUTPase in complex with ␣,␤-imido-dUTP and Mg 2ϩ (Protein Data Bank entry 2bt1). The catalytic residue Asp 76 is shown together with Asp 131 next to it. The four visible sequence motifs of the active site are colored according to Fig. 1A, and the loop connecting the two domains (residues 114 -125) is colored in cyan. The bound ␣,␤-imido-dUTP molecule is shown as sticks and colored according to atom types. C, the ⌬V structure (pink) in complex with dUTP (sticks, atom colors) and Mg 2ϩ . The disordered part of the connecting loop is located between the two arrows. Superposed is the dUMP-bound WT dUTPase structure (Protein Data Bank entry 2bsy; gray, connecting loop in cyan, motif I in green, disordered part dotted).  SEPTEMBER 11, 2009 • VOLUME 284 • NUMBER 37 pH 7.5, 150 mM KCl, 1 mM MgCl 2 . The carboxymethylated dextran layer of a CM5 sensor chip (Biacore AB) was activated by a 7-min pulse of a 1:1 mixture of freshly prepared 50 mM N-hydroxysuccinimide and 200 mM N-ethyl-NЈ-(dimethylaminopropyl)-carbodiimide. Recombinant human CSF1 (colony-stimulating factor-1), produced according to a modified protocol (28), and streptavidin (Sigma) in 10 mM acetate buffer, pH 4, WT dUTPase, and truncated C-terminal mutant V in 20 mM Hepes, pH 7.5, 250 mM NaCl, 10 mM MgCl 2 were bound to the activated surfaces during several pulses. Remaining N-hydroxysuccinimide ester groups were blocked by a 7-min injection of 1.0 M ethanolamine hydrochloride, pH 8.5. CSF1-and streptavidin-modified surfaces were used as reference for the WT dUTPase (7000 resonance units)-and for the ⌬V mutant (8000 resonance units)-modified surfaces, respectively. Curves obtained on the reference surface are deduced from the curves recorded on the recognition surfaces, allowing elimination of refractive index changes due to buffer effects. Substrate dUTP and inhibitor ␣,␤-imido-dUTP diluted in running buffer were injected at concentrations ranging from 1 to 1000 M. No regeneration procedure was required. The absence of mass transport effects was checked on each surface by separately running one injection of ␣,␤-imido-dUTP (5 M) for 120 s at different flow rates ranging from 10 to 70 l/min. It is possible to overlay the obtained curves, confirming the kinetic control of the experiments (not shown). K D values are obtained by fitting the steady-state response versus the concentration according to a 1:1 interaction model using BIAcore T100 evaluation software version 1.1.

Role of the Disulfide Bridge-
The presence in solution of the disulfide bridge between residues 4 and 246 could be verified by SDS-PAGE under non-reducing conditions in the presence of the blocking agent N-ethylmaleimide compared with a C4S mutant (Fig. 2) or with the WT protein under reducing conditions. The disulfide-linked form migrates faster than the reduced protein and represents the major fraction of dUTPase. Incubation of dUTPase under reducing conditions followed by the same analysis on non-reducing PAGE showed an increase of the reduced form in solution, although most of the protein remains oxidized. The enzymatic activity of the native enzyme, the one partially reduced with DTT, and the C4S mutant were virtually identical (Table 2) excluding a regulatory role of the disulfide bridge. The structure of the C4S mutant was almost identical to that of WT dUTPase with dUMP. The remaining disulfide bridge partner Cys 246 stayed in place; only Ser 4 became disordered, and Pro 5 adopted a different conformation.
Stabilization of the Linker Region-We speculated earlier (17) that the disorder of the loop containing motif I in the ␣,␤-imido-dUTP-bound structure might be due to a clustering of negative charges contributed by Asp 76 , Asp 131 , and the ␥-phosphate group of ␣,␤-imido-dUTP (Fig. 1B) observed so far only in the EBV dUT-Pase structure. Site-directed mutagenesis of the apparently EBVspecific residue Asp 131 to serine, the residue present in the E. coli enzyme, or to asparagine or glutamic acid led to a large loss in catalytic activity (Table 2). This activity was undetectable in the cresol red assay and could only be measured in the more sensitive EnzChek-based assay. With the EnzChek assay, WT k cat could be estimated as 1.2 s Ϫ1 , allowing the values determined with this assay to be validated, although the K m was too low to be determined, since substrate concentrations had to be above 5 M. Mutants D131S and D131N showed a k cat of 0.04 s Ϫ1 and K m value of 5 or 7 M, respectively, whereas activity of the D131E mutant remained undetectable. A double mutation D131S/G78D, which will be discussed below, showed some recovery of enzymatic activity, with a 5-fold increase of its k cat to 0.18 s Ϫ1 compared with the single mutants and a doubled K m .
Crystallization of these mutants in the presence of dUTP was attempted using a robotic screen of 576 crystallization conditions. Only for mutant D131N and the double mutant D131S/G78D was it possible to obtain crystals that turned out to be isomorphous, with the WT crystals containing dUMP. Structural changes concern only the mutated residues (Fig. 3) and the degree of disorder of the flexible region 114 -133, which varies between structures (not shown).
Structure of the Deletion Mutant of Motif V-Using the same conditions as for the WT enzyme in the presence of dUTP, after 2  days, crystallization trials of ⌬V mutant yielded crystals belonging to tetragonal space group I4 (Table 1) containing unhydrolyzed dUTP in the active site. After 2-3 weeks, orthorhombic dUTPase crystals appeared and were isomorphous with the WT-dUMP complex crystals described previously (17). These crystals diffracted X-rays to 1.5 Å resolution and showed a bound dUMP molecule in the active site, obviously arising from slow spontaneous or enzymatic hydrolysis of dUTP during the 2-3-week period needed for crystallization and a protein structure virtually identical to the one of the WT protein-dUMP complex (not shown). The tetragonal crystals diffracted X-rays to 2 Å resolution and showed a bound Mg 2ϩ -dUTP complex (Fig. 4, A, C, and D) and a largely disordered linker region (residues 114 -133), where residues 115-128 are invisible (Fig. 1C). Still, the catalytic residue Asp 76 is correctly in place to interact with a water molecule, which is supposed to carry out the nucleophilic attack on the ␣-phosphate (16) (Fig.  4A). The Mg 2ϩ ion is coordinated in a tridentate way by the three phosphate groups and three additional water molecules, leading to an octahedral coordination. The coordinating water molecules are hydrogen-bonded to Asp 135 of motif I, to the catalytic Asp 76 of motif III, and to the three phosphate groups, whereas Asp 131 contributes to the negative charge of the environment (Fig. 4D). An arginine residue from a symmetry-related molecule forms a salt bridge with the ␥-phosphate (Fig. 4A). The hydroxyl group of Ser 172 is turned away from the substrate and hydrogen-bonds to a water molecule (Fig. 4A) in the same way as observed in complexes with dUMP or in the D90N mutant of E. coli dUTPase (Protein Data Bank entry 1syl) (16) with bound Mg 2ϩ -dUTP. This differs from the conformation observed in complexes with ␣,␤-imido-dUTP, as illustrated in Fig. 4B.
Kinetics and Surface Plasmon Resonance Binding Assays in the Presence or Absence of Mg 2ϩ -Interaction of dUTPase with substrate  dUTP or analogue ␣,␤-imido-dUTP in the presence or absence of Mg 2ϩ was further studied by SPR analysis using ⌬V mutant or WT dUTPase. The signal recorded for binding of dUTP or ␣,␤-imido-dUTP onto ⌬V mutant allowed fitting of the data and determination of K D values of 37 and 62 M, respectively, in the presence of Mg 2ϩ (Fig. 5, A and B, and Table 3). Using ␣,␤-imido-dUTP and the WT enzyme (Fig. 5C), we obtained a value of 7 M. The dissociation constant of dUTP (37 M) determined by SPR for the ⌬V mutant is 50 times increased compared with the K m of the wild type enzyme (0.7 M); an increase from 7 to 62 M is observed for ␣,␤-imido-dUTP. As shown in Fig. 5, binding and release of dUTP and ␣,␤-imido-dUTP were very rapid, making determination of k on and k off values impossible in these conditions. In the absence of Mg 2ϩ , the signal is of insufficient quality to obtain interpretable data (below 10 resonance units) even at high concentrations of ligand (up to 5 mM) (data not shown).
We determined enzyme kinetics of the WT enzyme in the absence of Mg 2ϩ . Whereas K m ϭ 0.8 M and k cat ϭ 1.4 s Ϫ1 for the WT enzyme in the presence of magnesium, these values are 6.0 M and 1.2 s Ϫ1 , respectively, in its absence in the cresol red assay based on proton release during the reaction. The absence of magnesium affects thus almost exclusively the K m by a 7.5-fold increase and has only a small effect on the k cat .
Site-directed mutagenesis of the two key residues in motif V, with Arg 268 or Phe 273 changed to alanine, led to a nondetectable enzymatic activity when the phenylalanine was replaced, even using the more sensitive EnzChek-based assay. Catalytic efficiency of mutant R268A was largely affected, being reduced by a factor of 300. The mutation had a much lower effect on binding, with a 10-fold increase of K m to 8 M.

DISCUSSION
Role of the Disulfide Bridge-All of the different EBV dUT-Pase crystal structures determined so far show the presence of a disulfide bridge involving cysteines 4 and 246. Since the second residue is close to the disordered part of the structure carrying essential motif V, we wanted to verify that there is no regulatory role of this unexpected disulfide bridge. The difference in mobility in non-reducing PAGE (Fig. 2) of the disulfide-linked and the reduced form allowed us to confirm that the link is already largely present in solution and only to a small degree influenced by a reduction with DTT. The kinetic parameters in the presence of DTT (Table 2) changed insignificantly, but only a small fraction of the reduced form could be obtained by reduction with DTT. We constructed mutant C4S, which would correspond to the situation of a reduced disulfide bridge. Upon mutation of Cys 4 to serine, this residue actually becomes disordered in the structure. The kinetic parameters of the C4S mutant were similar to those of the WT enzyme (Table 2), indicating that there is no regulatory role of the disulfide bridge but that it is more likely an artifact obtained under the rather oxidizing conditions of protein preparation.
Stabilization of the Linker Region-Mutants of Asp 131 were designed in order to destroy a cluster of negative charges formed by Asp 76 , Asp 131 , and the ␥-phosphate in the EBV dUT-Pase active site, since we suspected that it was responsible for the disordering of the active site residues in the ␣,␤-imido-dUTP-EBV dUTPase structure. The mutation of Asp 131 to serine, the corresponding residue in the E. coli structure, resulted in a much reduced enzymatic activity. This loss of activity has been confirmed with the more conservative mutants D131N and D131E (Table 2).
An initial comparison of the structure with that of feline immunodeficiency virus dUTPase (12) suggested a possible interaction of Asp 131 with the conserved arginine residue in the RGXXGFG sequence of motif V (Fig. 1A), which was confirmed by the structure of M. tuberculosis dUTPase in the presence of ␣,␤-imido-dUTP (14) (Fig. 6A). The loss of this interaction could explain the drastic effect of a substitution of Asp 131 , which could play a central role in the ordering of motif V. This idea is in agreement with the mutations to serine and asparagine having a smaller effect than the change to glutamic acid, which conserves the charge but would lead to a steric hindrance excluding Arg 268 from its position between the ␥-phosphate and the acidic residue.  The interaction of equivalents of Asp 131 with the arginine residue of motif V is the only ionic interaction between motif V and the rest of the protein. The interactions of motif V mainly involve the bound substrate and some hydrogen bonds with residues of the body of the enzyme. Despite this apparently essential role, it appears that Asp 131 is not conserved (Fig. 1A), for example, in the human or E. coli enzymes. Inspection of human dUTPase structures that show motif V points to a possible salt bridge between the arginine residue and Asp 81 with a distance between 3.5 and 3.8 Å in the ␣,␤-imido-dUTP-bound structure (Protein Data Bank entry 3ehw) (13) or 3.0 Å in the dUDP-bound enzyme (Protein Data Bank entry 1q5h) (3). In addition, there is a hydrogen bond with the hydroxyl group of Ser 19 .
From the sequence alignment (Fig. 1A), it became obvious that in the sequences that do not conserve the equivalent of EBV Asp 131 , the aspartic residue equivalent to human Asp 81 is conserved, along with the hydroxyl group of the serine or threonine in position 19. The only exceptions are the sequences of a number of Corynebacterium species, such as Corynebacterium glutamicum, which show the presence of both aspartic acid residues, the one corresponding to the conserved residue of EBV, and the one corresponding to human dUTPase. Therefore, two solutions become obvious for the interaction with the arginine of motif V, either a bidentate salt bridge with an equivalent of EBV Asp 131 as in the prototype structure of M. tuberculosis dUTPase (Fig. 6A) or a weak salt bridge with Asp 81 as in the prototype structure of human dUTPase combined with a hydrogen bond with the hydroxyl group of residue 19 (Fig. 6B). This defines two groups of dUTPases. A first group containing EBV dUTPase is formed by the monomeric enzymes from herpesviruses and the trimeric dUTPases from lentiviruses, P. falciparum, and some bacteria, such as mycobacteria, corynebacteria, and clostridia. A second group contains the majority of trimeric dUTPases, in particular most eukaryotic ones and, for example, the enzymes from poxviruses and E. coli. Corynebacteria often contain both aspartic acid residues (Fig. 1A).
The large reduction of activity in the D131S mutant prompted us to construct a double mutant of the EBV enzyme in order to make it more E. coli-like. An aspartic acid residue is placed at position 78 in EBV numbering, replacing the glycine residue normally present at this position (Fig. 1C), whereas Asp 131 is mutated to serine, the residue present in the E. coli enzyme. A comparison of E. coli dUTPase belonging to the second family with the EBV enzyme showed that there is no structural equivalent at the position of Ser 19 and that thus we would only be able to obtain a partial effect. Indeed, when comparing the double mutant with the D131S simple mutant, we observed an approximately 5-fold increase of k cat , whereas the K m changed by a factor of 2 ( Table 2). This showed the importance of the partner aspartic acid residues of the arginine residue of motif V, and as expected, the recovery of activity was only partial, since one of the interactions could not be restored.
Role of Motif V-The ⌬V mutant allowed us to obtain the first structure of an Mg 2ϩ -dUTP substrate productively bound to a dUTPase containing no mutations other than the deletion of motif V. dUTP binds in the same gauche conformation as in the ␣,␤-imido-dUTP complex (17) (Fig. 4A), thus confirming the value of this compound as a model for substrate binding. The same binding mode is observed in a number of dUTPase structures (Protein Data Bank entries 1six, 1sjn (7), 1rn8, 1syl (16), 2py4 (14), and 3ehw (13)) and is generally believed to correspond to the catalytically competent form. Previous structures containing dUTP lacked magnesium and showed different dUTP conformations (Protein Data Bank entries 1f7q (12) and 1sm8 and 1smc (7)), which were essentially trans (29) or carried a mutation of the catalytic aspartic acid residue (16).
The inactivity of the ⌬V mutant during the 2-day co-crystallization period allowed the observation of the substrate in the active site and was even more surprising, since apparently most elements necessary for catalysis were in place (Fig. 4A): (i) a productive conformation of the dUTP molecule, (ii) the Mg 2ϩ ion, and (iii) an arginine residue contributed by a crystal contact mimicking the conserved arginine of motif V. This arginine residue in the tetragonal crystal form and the position of the conserved arginine of motif V in the M. tuberculosis dUTPase structure (14) are shown in Fig. 4, A and B. The hydrogen bonding pattern is similar, although the orientation of the guanidinium group is different. These observations question the role of magnesium in catalysis, which was investigated further as discussed below along with the role of the arginine residue in the polarization of the substrate's phosphate groups. This prompted us to assess the relative importance of the two principal residues of motif V, Arg 268 and Phe 273 , by site-directed mutagenesis. Surprisingly, the R268A mutant still had a significant activity similar to the mutants of residue Asp 131 interacting with Arg 268 (Table 2), whereas the mutation affecting lid residue Phe 273 led to an activity below the detection limit. In most structures of trimeric dUTPases, as in the monomeric one of EBV, motif V is disordered. The recent structures of the human or M. tuberculosis enzyme in complex with ␣,␤imido-dUTP (13,14) (Protein Data Bank entries 3ehw and 2py4) give a view onto this part of the enzyme and show best the interactions of motif V with the substrate analogue (Fig. 6, A  and B) and its role in catalysis. In other structures, only parts of motif V are visible, and/or an inhibitor is bound with its phosphate groups in a non-productive conformation that does not allow us to analyze all of the interactions of motif V. The structures by Varga et al. show that the phosphate groups of ␣,␤imido-dUTP interact with hydroxyl groups of conserved serine residues and amide nitrogen atoms of the backbone of motif V adjacent to the lid residue (Fig. 1A). The conserved arginine residue also interacts with the ␥-phosphate (Fig. 6). Varga et al. (13) could show that the folding of motif V onto the bound analogue ␣,␤-imido-dUTP leads to a decrease of the distance (2.8 Å) of the catalytic water molecule needed for nucleophilic attack from the ␣-phosphorus atom compared with an active site with a disordered motif V, where this distance is 3.2 Å. In the EBV dUTPase, the distance between the phosphorus and the catalytic water molecule is 3.5 Å in the dUTP-⌬V dUTPase complex, the same as in the EBV wild type dUTPase-dUMP complex. The EBV dUTPase structure with bound ␣,␤-imido-dUTP showed a partially disorganized active site with the absence of the catalytic water molecule due to a large movement of the activating residue Asp 76 (17) (Fig. 1B). Based on the results of Varga et al. (13), folding of motif V with the numerous interactions of its backbone atoms with the phosphate groups leads to an intimate contact of the substrate's ␣-phosphorus atom with the catalytic water molecule. The relatively large distance of the catalytic water molecule from the ␣-phosphorus atom explains the stability of the substrate in our ⌬V mutant. When the lid residue Phe 273 covers the active site and the rest of motif V becomes ordered, expulsion of the solvent from the active site results in a change in the dielectric constant and electrostatics, probably promoting catalysis. Another consequence of binding of the hydrophobic lid residue is the displacement of the water molecule normally hydrogen-bonded to Ser 172 (Fig. 4A). This probably reorients Ser 172 toward an interaction with the bridging oxygen atom between the ␣and ␤-phosphate groups in the same way as it interacts with ␣,␤imido-dUTP (Fig. 4B). A mutation of the corresponding residue in the E. coli enzyme to alanine led to a 3-order of magnitude reduction in the catalytic rate, whereas binding of the substrate was enhanced by a factor of 30 (30). The enhanced substrate binding is most easily explained by a binding site for the lid residue, which has been rendered more hydrophobic due to the loss of the hydroxyl group binding to the water molecule in the absence of the lid residue. This would enhance the contribution of motif V to binding, which is discussed below. The reduction of the catalytic rate points to a second role of the hydroxyl group with a direct involvement in catalysis, probably through the interaction with the bridging oxygen atom. The hydroxyl group is likely to either stabilize the charge of this atom or to donate a proton. The complete inactivity of the F273A mutant of EBV dUTPase supports these ideas.
Despite the flexible character of motif V, it appears to contribute to substrate affinity. When motif V is deleted, the dissociation constant of dUTP increases 50 times, and the one of ␣,␤-imido-dUTP increases 10 times, an effect which has to be taken into account in approaches of structure-based drug design. It is likely that the interaction of motif V with the substrate largely takes place via the phenylalanine lid residue interacting with the uracil ring and the ribose, thus explaining the reduced affinity of the ⌬V mutant for the substrate. One difference in the binding of the two compounds is that in the ␣,␤-imido-dUTP complex, the bridging atom between ␣and ␤-phosphate groups (-NH 2 -) interacts with Ser 172 , a conserved residue within motif II, whereas no hydrogen bond was formed by the corresponding oxygen atom in the substrate. This would suggest rather better binding of ␣,␤imido-dUTP, but preferred conformations of the phosphate groups in solution may be different for the two compounds, resulting in a lower population of the active gauche conformation for ␣,␤-imido-dUTP in solution.
Interestingly, the disorganization of the active site observed for EBV dUTPase ␣,␤-imido-dUTP complex (Fig. 1B) was not observed for the dUTP-⌬V complex despite the similar position of the two ligand molecules (Fig. 1C). This may point to a globally reduced stability of this region in EBV dUTPase and an influence of crystal contacts on the structure, since the ␣,␤imido-dUTP-WT complex crystals belong to a different space group than those of the dUMP-bound form or of the ⌬V-dUTP complex.
Role of the Magnesium Ion in Catalysis-The observation of an Mg 2ϩ -dUTP complex in the active site of the ⌬V mutant raised the question of the importance of the magnesium ion for catalysis. It turned out that the enzymatic activity depends little on the presence of Mg 2ϩ ions, since the rate constant k cat (Table  2) decreases insignificantly, whereas the K m increases 10-fold at an Mg 2ϩ concentration (Ͻ0.08 M from atomic absorption spectrometry) well below the apparent dissociation constant of an Mg 2ϩ -dUTPase/dUTP complex (21). The effect of an Mg 2ϩ depletion in an EDTA-buffered system was studied first by Larsson et al. (19) for E. coli dUTPase. The authors found a 100-fold reduction in the binding affinity of dUTP in the absence of Mg 2ϩ (K m increasing from 0.2 to 16 M), confirmed by Palmén et al. (30). In a later study, a 3-fold reduction of k cat was observed, combined with an increase of K m by 1 order of magnitude (from 0.6 to 9 M) (21). Quesada-Soriano et al. (31) studied P. falciparum dUTPase and observed that the K m increased from 1.9 M in the presence of Mg 2ϩ to 22 M without the metal ion. k cat was also affected, falling from 13 to 0.2 s Ϫ1 without Mg 2ϩ . Concerning the K m , our results are similar; still, the effect on k cat is much smaller in our study. This may be a distinction of the EBV enzyme, which generally shows a lower k cat than other dUTPases. SPR showed a large difference between responses with and without magnesium as well for the ⌬V mutant and dUTP as for the WT and ␣,␤-imido-dUTP, confirming that Mg 2ϩ plays an important role in substrate binding.
In the EBV enzyme, rigidification of the triphosphate group and polarization of the phosphate groups by magnesium coordination are obviously not essential for catalysis. Binding of a preexisting Mg 2ϩ -dUTP complex (32) from solution is proba-bly entropically more favorable and leads to tighter electrostatic interactions of the phosphate groups (Fig. 4D). As a consequence, substrate binding is increased, and in turn, as the catalytic step is rate-limiting (15), the K m drops.
Conclusions-The current study helped to gain a clearer vision of the common mode of catalytic action of trimeric and monomeric dUTPases. The relevance of the evolution from trimeric to monomeric dUTPase with its concomitant increase of the required coding space still remains unexplained. A regulatory role of the disulfide bridge in the EBV enzyme could be rejected. There may also be non-enzymatic roles of EBV dUTPase; e.g. a recent study (33) describes an up-regulation of proinflammatory cytokine secretion due to activation of NF-B through interaction of extracellular dUTPase with TLR-2 (Tolllike receptor 2).
Although monomeric EBV dUTPase has a different subunit organization from trimeric dUTPases, there is no evidence for differences in the catalytic mechanism. Depending on the position of the aspartic acid residue interacting with the arginine residue of motif V, monomeric and trimeric dUTPases can be classified into two families. In the absence of motif V, all elements needed for catalysis appear to be in place, but the substrate remains stably bound to the enzyme. Consequently, motif V and in particular the phenylalanine lid residue appear to be absolutely essential for catalysis. The magnesium ion appears to play a minor role in catalysis. Polarization effects due to Mg 2ϩ -binding and electrostatic interactions with ␤and ␥-phosphate groups contribute surprisingly little.
Together with previous results, our findings contribute to an overall picture of catalysis by dUTPases. A mold formed by the rigid part of the molecule comprising motifs I-IV seizes dUTP and discriminates against similar molecules, such as UTP or dCTP. A transient event folds the C-terminal motif V over the active site, leading to a tighter binding of the substrate, as indicated by the contribution of motif V and particularly the lid residue to binding. This event is very rapid, since it apparently could not be resolved in the kinetic study of Tóth and co-workers (15). At the same time, the hydroxyl group of the conserved serine of motif II changes its position to interact with the oxygen atom bridging ␣and ␤phosphate groups and assists in catalysis, as proposed by Palmén and co-workers (30). The conformational change pushes the substrate farther down into the active site to an intimate contact with the catalytic water molecule, leading to substrate hydrolysis, as suggested previously (13,15). Finally, the products are released rapidly (15).