Unique GMP-binding site inMycobacterium tuberculosis guanosine monophosphate kinase

Unique GMP-Binding Site in Mycobacterium tuberculosisGuanosine Monophosphate KinaseGuillaume Hible,1 Petya Christova,1,2 Louis Renault,1 Edward Seclaman,3 Andrew Thompson,4 Eric Girard,4

Helene Munier-Lehmann,3* and Jacqueline Cherfils,1*1Laboratoire d’Enzymologie et Biochimie Structurales, CNRS, Gif sur Yvette, France2Bulgarian Academy of Sciences, Sofia, Bulgaria3Unite de Chimie Organique, Institut Pasteur-Institute of Organic Chemistry, Paris, France4SOLEIL Synchrotron, St Aubin, France

ABSTRACT Bacterial nucleoside monophos-phate (NMP) kinases, which convert NMPs to nucleo-side diphosphates (NDP), are investigated as poten-tial antibacterial targets against pathogenicbacteria. Herein, we report the biochemical andstructural characterization of GMP kinase fromMycobacterium tuberculosis (GMPKMt). GMPKMt is amonomer with an unusual specificity for ATP as aphosphate donor, a lower catalytic efficiency com-pared with eukaryotic GMPKs, and it carries tworedox-sensitive cysteines in the central CORE do-main. These properties were analyzed in the light ofthe high-resolution crystal structures of unbound,GMP-bound, and GDP-bound GMPKMt. The latterstructure was obtained in both an oxidized form, inwhich the cysteines form a disulfide bridge, and areduced form which is expected to correspond tothe physiological enzyme. GMPKMt has a modulardomain structure as most NMP kinases. However, itdeparts from eukaryotic GMPKs by the unusualconformation of its CORE domain, and by its par-tially open LID and GMP-binding domains whichare the same in the apo-, GMP-bound, and GDP-bound forms. GMPKMt also features a unique GMPbinding site which is less close-packed than that ofmammalian GMPKs, and in which the replacementof a critical tyrosine by a serine removes a catalyticinteraction. In contrast, the specificity of GMPKMt

for ATP may be a general feature of GMPKs becauseof an invariant structural motif that recognizes theadenine base. Altogether, differences in domaindynamics and GMP binding between GMPKMt andmammalian GMPKs should reveal clues for the de-sign of GMPKMt-specific inhibitors. Proteins 2006;62:489–500. © 2005 Wiley-Liss, Inc.

Key words: Guanosine monophosphate kinase;nucleoside monophosphate kinase;structure; X-ray crystallography; Myco-bacterium tuberculosis

INTRODUCTION

With more than three million deaths per year, Mycobac-terium tuberculosis, the causative agent of tuberculosis,remains a major public health problem. One of the re-search strategies in fighting against this disease is the

identification of new targets for more effective drugs.1

Nucleotide metabolism enzymes lead to the production ofnucleotide triphosphates (NTP) which are DNA and RNAprecursors and critical regulators in multiple cellularpathways. They have been identified as essential inM. tuberculosis2 and other pathogens3,4 and have begun tobe explored as potential antibacterial targets.5,6 An obliga-tory step in NTP synthesis is the conversion of nucleosidemonophosphates (NMP) to nucleoside diphosphates (NDP),which involves the reversible transfer of the �-phosphorylgroup of ATP to an NMP acceptor by NMP kinases(NMPKs). In bacteria, five NMPKs are responsible forthese activities: adenylate kinase (AMPK) for the phosphor-ylation of AMP/dAMP, CMPK for CMP/dCMP, GMPK forGMP/dGMP, UMPK for UMP, and TMPK for dTMP (re-viewed in Yan and Tsai7). Representative structures ofbacterial NMPKs are available for each family8–10,47 in-cluding, for M. tuberculosis, crystal structures of TMP-KMt

11–13 and NMR structures of AMPKMt.14 Bacterial

NMPKs share with their eukaryotic counterparts a modu-lar architecture consisting of three subdomains, but inmany cases they also have features that are unique tobacteria or even bacterial subgroups. For instance, CMPKfrom Escherichia coli has a unique insert in the CMP-binding subdomain, which is critical for catalysis.9 An-other example is TMPKMt, in which the position of Mg2�

and an invariant catalytic arginine are permuted com-pared with the human enzyme.12 Such differences ofbacterial NMPKs to their eukaryotic counterparts may beessentially silent as long as natural substrates are used,but may prove valuable for the discovery of specific antibac-

Grant sponsors: French Ministry of Research; Direction des Rela-tions Internationales, CNRS; INSERM; Institut Pasteur.

E. Seclaman’s present address is University of Medicine andPharmacy, Biochemistry Department, 1900 Timisoara, Romania.

G. Hible and P. Christova contributed equally to this work.*Correspondence to: Jacqueline Cherfils, Laboratoire d’Enzymologie

et Biochimie Structurales, CNRS, Avenue de la Terrasse, 91198 Gifsur Yvette, France. E-mail: [email protected] or Helene Munier-Lehmann, Unite de Chimie Organique, Institut Pasteur-URA CNRS2128, 28 Rue du Dr Roux, 75724 Paris Cedex 15, France. E-mail:[email protected]

Received 11 March 2005; Revised 12 May 2005; Accepted 27 May2005

Published online 15 November 2005 in Wiley InterScience(www.interscience.wiley.com). DOI: 10.1002/prot.20662

PROTEINS: Structure, Function, and Bioinformatics 62:489–500 (2006)

© 2005 WILEY-LISS, INC.

terial compounds. In TMPK, these differences turn 3�-azido-deoxythymidine from a substrate in human TMPK to acompetitive inhibitor in TMPKMt,

15 and were subse-quently used to design specific and potent inhibitors of themycobacterial enzyme.5,16–19

In an effort to identify novel properties of mycobacterialNMPKs, which could be the target for specific antitubercu-losis drugs, we report herein the structural and biochemi-cal characterization of GMPKMt, an essential enzyme formycobacterial growth.2 Structures of GMPKs are cur-rently known for the enzyme from Saccharomyces cerevi-siae (GMPKSc) in the apo form20 and in complex withGMP,21 for the mouse (Mus musculus) enzyme (GMPKMm)bound to ADP and GMP, and for the enzyme from E. Coli(GMPKEc).

48 GMPKs consist, as all NMPK structuressolved to date except UMPK, of a CORE domain with theLID domain and the NMP-binding domain (called GMPdomain hereafter) attached to it (reviewed in Yan andTsai7). The CORE domain is a five-stranded �-sheetflanked by three �-helices, and it carries the P-loop (orWalker A motif), a highly conserved motif that binds the�-phosphate of the ATP donor in NMPKs.23 It is flanked onone side by the GMP domain, which carries the phosphateacceptor binding site with invariant residues that deter-mine the enzyme specificity for GMP. This domain alsoprovides residues to the phosphate transfer catalytic site,including two arginines that are invariant in NMPKs anda tyrosine that is specific to GMPKs. Opposite to the GMPdomain, the LID domain carries conserved residues thatcomplete the ATP-binding site and provides arginines tothe catalytic site. Although no GMPK has been crystal-lized with all combinations of acceptor and donor nucleo-tides, comparison of the GMPK structures to each otherand to the more extensively characterized AMPK andTMPK structures, have led to a model of activation ofGMPKs by step-wise closure of the LID and GMP domainsonto the CORE domain upon ATP and GMP binding, thusassembling their catalytic machinery.

Herein, we show that GMPKMt is a monomer, with acatalytic efficiency that is lower than that of eukaryoticGMPKs and an unusual specificity for ATP as a phosphatedonor. GMPKMt carries redox-sensitive cysteines in theCORE domain, whose physiological thiol form was ana-lyzed in the crystal structure of GMPKMt-GDP. Thisstructure was shown to be similar to that of oxidizedGMPKMt-GDP, thus validating the use of the oxidizedapo-, GMP-bound, and GDP-bound crystal structures forthe analysis of specific features of the mycobacterialenzyme. These structures and their comparison to knownGMPK and related NMPK structures identify the GMP-,ATP-, and Mg2�-binding sites, and the residues that areinvolved in phosphoryl transfer catalysis. They show thatalthough the specificity for ATP may be a general featureof GMPKs because of a conserved motif that recognizes theadenine base, the mycobacterial enzyme departs fromother known GMPKs in several unexpected aspects. Theseinclude an unusual domain conformation and the replace-ment of a critical tyrosine by a serine in the GMP domainthat creates a unique GMP-binding site. Altogether, our

analysis identifies structural differences between GMP-KMt and mammalian GMPKs that may prove valuable forthe design of GMPKMt-specific inhibitors.

MATERIALS AND METHODSChemicals

Nucleotides, restriction enzymes, T4 DNA polymerase,and coupling enzymes for determination of GMK activitywere from Roche Applied Science. T4 DNA ligase and VentDNA polymerase were from New England Biolabs, Inc.Affi-Gel Blue Gel were obtained from Bio-Rad, UltrogelAcA54 from Sigma, and ultra pure urea from ICN Bio-chemicals, Inc. GMPKEc is a gift from Anne-Marie-Gilles(Institut Pasteur, Paris); NDP kinase is a gift from IoanLascu (IBGC, Bordeaux, France); the antibody directedagainst histone-like nucleoid-structuring protein from E.coli (H-NS) is a gift from Evelyn Krin (Institut Pasteur,Paris, France).

Cloning, Purification, and BiochemicalCharacterization

The 627-bp fragment corresponding to the gmk genefrom M. tuberculosis (RV1389 gene in the M. tuberculosisgenomic information database Tuberculist, http://genolist.pasteur.fr/TubercuList) was amplified by polymerase chainreaction using M. tuberculosis genomic DNA as the matrix.The amplified gmk gene was cloned into the pET22bplasmid (Novagen, Inc.) digested with the same enzymes.Three clones containing the M. tuberculosis gmk gene andoverexpressing GMPKMt were characterized and one ofthem (harboring plasmid named pHL40-2) was kept forfurther studies. Cultures were grown in 2YT mediumsupplemented with 100 mg/mL ampicillin and 30 mg/mLchloramphenicol. Production of recombinant GMPKMt wasinduced with isopropyl-1-thio-�-D-thiogalactoside (1 mMfinal concentration) when cultures reached an absorbanceof 1.5 at 600 nm. Bacteria were harvested by centrifuga-tion 3 h after induction.

Bacteria suspended in 50 mM Tris-HCl pH 7.4, 1 mMEDTA, and 0.5 mM DTT (buffer A) were disrupted bysonication. After centrifugation at 14,000 rpm for 30 min,the bacterial extract was loaded onto a Blue-Sepharosecolumn equilibrated with the same buffer (10 mg ofprotein/mL of swollen gel). The column was washed with10 volumes of 0.5 M NaCl in buffer A. GMPKMt was elutedwith 4 volumes of 1 M NaCl in buffer A. Fractionscontaining GMPKMt were pooled, concentrated, and loadedonto a 1 � 110 cm Ultrogel AcA54 column equilibratedwith buffer A and fractions of 1.8 mL were collected at aflow rate of 9 mL/h. Protein concentration was measuredwith a Bio-Rad Bradford kit or by amino acid analysis on aBeckman system 6300 high-performance analyzer after6 N HCl hydrolysis for 22 h at 110°C. Titration of thiolgroups in native GMPKMt were determined with aliquotsof enzyme free of exogenous reducing agent by passageover a Sephadex G-25 M column (PD-10, Amersham).Thiol estimations were calculated on a basis of a molarabsorbance of 13,500 for 5-carboxy-4-nitrophenol at 412nm.24 The absorbance at 412 nm was read against a blanktreated in parallel without GMPKMt.

490 G. HIBLE ET AL.

GMPK activity was determined at 30°C using the coupledspectrophotometric assay at 334 nm as previously de-scribed25 (0.5 mL final) on an Eppendorf ECOM 6122photometer. The reaction medium (0.5 mL final volume)contained either 50 mM Tris-HCl pH 7.4, 0.2 mM NADH,0.5 mM phosphoenolpyruvate, 50 mM KCl, 2 mM MgCl2,and 2 units of lactate dehydrogenase, pyruvate kinase andNDP kinase (forward reaction), or 50 mM Tris-HCl pH 7.4,50 mM KCl, 1 mM glucose, 0.4 mM NADP�, 2 mM MgCl2,and 2 units of hexokinase and glucose-6-phosphate dehy-drogenase (reverse reaction). One unit of enzyme activitycorresponds to 1 �mole of the product formed in 1 min at30°C and pH 7.4. AMPK activity was assayed with andwithout preincubation of GMPKMt for 20 min at roomtemperature with antibodies directed against AMPKEc.Antibodies against H-NS from E. coli were used as acontrol. IMPK activity was assayed by overnight incuba-tion of GMPKMt with IMP and ATP at 1 mM, followed byhigh-performance liquid chromatography (HPLC) separa-tion of the resulting nucleotides. The thermal stability wastested by incubating purified GMPKMt (1 mg/mL) in bufferA at temperatures between 30° and 60°C for 10 min, anddetermining the residual activity of the sample. GMPKMt

exhibits a low thermal stability (Tm � 45°C) in contrast toother mycobacterial NMPKs.

Equilibrium sedimentation experiments were per-formed at 20°C on a Beckmann Optima XL-A analyticalcentrifuge using an An-60 Ti rotor and a cell with a 12-mmoptical path length. Samples (150 �L) in buffer A with orwithout 0.4 M KCl at approximately 0.5 �g/�L werecentrifuged at 20,000 and 25,000 rpm. Radial scans ofabsorbance at 280 nm were taken at 2-h intervals. Equilib-rium was reached after 14 h. Data were analyzed by theXL-A program supplied by Beckman. The partial specificvolume of GMPKMt was calculated from the sequence. Themolecular mass of GMPKMt was determined by electros-pray ionization mass spectrometry (21,963.71 1.47 Da)and corresponds to that calculated from the deducedprotein sequence without the initiation methionine(21,964.10 Da). All fluorescence measurements were per-formed using a Jasco FP-750 spectrofluorometer thermo-stated at 25°C. Emission spectra were measured between310 and 410 nm (bandwidth 5 nm). Upon excitation at 295nm, GMPKMt exhibits a fluorescence emission spectrumwith a maximum at 347 nm, indicating that the uniqueTrp89 of GMPKMt is exposed to the solvent.

Crystallization, Data Collection, and StructureRefinement

GMPKMt was concentrated to 7.5 mg/mL and incubatedwith GMP, GDP, or ganciclovir-monophosphate added at aconcentration of 10 mM before crystallization. In oneexperiment, 5 mM Tris (2-carboxyethyl) phosphine hydro-chloride (TCEP) was added in the protein sample in thepresence of GDP and in the reservoir to prevent theformation of the disulfide bridge between Cys40 andCys193. We refer to the TCEP-treated crystal structure asGMPKMt-GDP or reduced GMPKMt-GDP, and to the otherstructures as oxidized apo-GMPKMt, GMPKMt-GMP, and

GMPKMt-GDP. All crystallizations were done by the vapordiffusion method at room temperature using hangingdrops of 1 �L of protein and 1 �L of a reservoir solutioncontaining 15% (w/v) xylitol, 3.5 M sodium chloride ineither 0.1 M bicine pH 9 (GMP and ganciclovir-monophos-phate) or 0.1 M sodium HEPES pH 7.5 (GDP) buffers.Crystals grew within 2–3 days to dimensions of 150–200�m. Before freezing in liquid ethane, crystals were trans-ferred to a cryoprotectant solution containing 35% (w/v)xylitol, 2.5 M sodium chloride, and 0.1 M of the crystalliza-tion buffer without added nucleotide.

An initial data set was first collected for the ganciclovir-monophosphate incubated GMPKMt crystal at beam lineID29 ( � 0.9756 Å) at the European Synchrotron RadiationFacility (ESRF). Initial phases for these crystals were ob-tained by molecular replacement using the 2.16 Å structureof oxidized apo-GMPKMt (S. Chan, M.R. Sawaya, L.J. Perry,D. Eisenberg, unpublished results deposited as a PDB entry,code 1S4Q) after previous attempts using GMPKSc, GMP-KMm, and GMPKEc as models had failed. No electron densitycould be identified for the nucleotide, indicating that thecrystals contained in fact the oxidized apo-enzyme. Subse-quently, 20 crystals of oxidized GMPKMt co-crystallized witheither GMP or GDP were tested and 11 diffraction data setswere collected on ESRF beamline ID29 ( � 0.9756 Å) as atest case of the DNA project for automatic screening and dataprocessing.26 Data sets with resolution better than 2.3 Åwere indexed and scaled using XDS and XSCALE.27 Thedata set for the reduced GMPKMt-GDP crystals were col-lected on our in-house rotating anode generator ( � 1.541Å). All crystals were found to be isomorphous to oxidizedapo-GMPKMt crystals. The refined oxidized apo-GMPKMt

structure was used for calculating Fourier difference mapsfor the oxidized GMPKMt-GMP and GMPKMt-GDP, andreduced GMPKMt-GDP structures. All refinements wereperformed with CNS28 and Refmac529 using TLS parametri-zation with the CORE, GMP, and LID domains as rigidbodies, and model building was done with O30 and Turbo-Frodo (http://afmb.cnrs-mrs.fr). The 19 N-terminal extensionand the 6–7 C-terminal residues are not visible in theelectron density for any of the four refined structures. Thefinal refined models of oxidized apo-, GMP-bound and GDP-bound, and reduced GDP-bound GMPKMt have respectively93, 91, 92, and 92% residues in the most favored regions, andno residues in the disallowed regions of the Ramachandranplot. The statistics of data collection and refinement aresummarized in Table II. Coordinates and structure factorshave been deposited with the Protein Data Bank (PDB) withentry codes: 1ZNW and RCSB032936 for oxidized apo-GMPKMt, 1ZNX and RCSB032937 for oxidized GMPKMt-GMP, 1ZNY and RCSB032938 for oxidized GMPKMt-GDP,and 1ZNZ and RCSB032939 for reduced GMPKMt-GDP.Figures were drawn with PyMOL (http://www.pymol.org) orMOLSCRIPT.31

RESULTSMolecular and Catalytic Properties of GMPKMt

GMPKMt overexpressed in E. coli was purified by atwo-step procedure that takes advantage of the affinity of

UNIQUE GMP-BINDING SITE 491

the Cibacron Blue dye for nucleotide binding sites32 andhas proved effective on many NMPKs.15,33,34 After gelpermeation chromatography, GMPKMt was recovered atan approximate molecular weight of 30 kDa, correspond-ing to a monomeric protein. This was confirmed by equilib-rium sedimentation ultracentrifugation, which yielded amolecular mass of 24,261 1,014 Da that was indepen-dent of the ionic strength up to 0.4 M KCl. GMPKMt waspurified in the presence of EDTA and a reducing agent,which effectively inhibits the proteolysis of other mycobac-terial NMPKs expressed in the same E. coli strain. Afterdialysis in a buffer devoid of DTT, its two cysteines (Cys40and Cys193) were found to be reactive with 5,5�-dithiobis(2-nitrobenzoic acid) (DTNB) under native conditions. More-over, the GMPK activity was not affected by the binding ofthe Ellman’s reagent to these cysteines.

The kinetic parameters of the forward and reversereactions (Fig. 1) catalyzed by GMPKMt are indicated inTable I. Compared with GMPKEc and GMPKSc, the myco-bacterial enzyme exhibited significantly lower kcat valuesin both the forward and reverse reactions. The Km valuesof GMPKMt for either substrate were in the same range as

those of GMPKSc (except a higher Km value for ADP) buthigher than those of GMPKEc. Unlike GMPKSc, which atconcentrations of GMP above 0.22 mM exhibited a de-crease of the apparent maximum velocity of up to 50%,35

GMPKMt was not inhibited by an excess of GMP (2 mM).NMPKs show, in general, little specificity for the phos-

phate donor, but have a narrow specificity for the phos-phate acceptor (reviewed in Anderson et al.36 and VanRompay et al.37). Accordingly, GMPKMt was found todisplay a narrow specificity for GMP as the phosphateacceptor. dGMP was also accepted as a substrate, althoughwith only 8% of the GMP catalytic rate. Phosphorylation ofAMP (8% of GMP rate) and IMP (2.5%) could also bedetected, both of which could be ruled out as enzymicactivities due to GMPKMt by using alternative assays.Phosphorylation of AMP by GMPKMt was completelyabolished by prior incubation of GMPKMt with antibodiesdirected against AMPKEc, indicating that trace amounts ofchromosome-encoded AMPKEc were responsible for thisactivity. Under similar experimental conditions, theGMPKMt activity with GMP was unaffected. Likewise, noIDP could be identified by HPLC analysis after incubationof a large excess of GMPKMt with IMP and ATP. On thecontrary, a GDP peak was identified after incubation ofGMPKMt with GMP and ATP. Thus, the IMPK activitymay be attributable to contamination of the IMP sampleused in the enzymic coupled assay. However, our analysisof the specificity of GMPKMt for the NTP phosphate donorrevealed an unusual specificity for ATP. Only dATP wasalso found to be effective as a phosphate donor (80% of theATP catalytic rate), but the reaction rate decreased to lessthan 2% with GTP and UTP and no activity was detectablefor TTP and CTP.

Crystal Structures of Apo-, GMP-Bound, and GDP-Bound GMPKMt

The crystal structures of apo-, GMP-bound, and GDP-bound GMPKMt, in which the redox-sensitive cysteinesform a disulfide bridge, and of GMPKMt-GDP where the

Fig. 1. Double-reciprocal plots of steady-state kinetic data for theforward (upper panel) and reverse reactions (lower panel) of GMPKMt.Kinetics were measured as described under Materials and Methods andthe parameters are given in Table I.

TABLE I. Kinetic Parameters of GMPKMt, GMPKSc, andGMPKEc

GMPKMt GMPKSc GMPKEc

Forward reaction (formation of ADP and GDP)Km

ATP (�M) 237 204 63Km

GMP (�M) 80 91 37kcat (s�1) 23 394 188Reverse reaction (formation of ATP and GMP)Km

ADP (�M) 137 17 23Km

GDP (�M) 98 97 33kcat (s�1) 17 90 55

Reaction velocities were measured at different ATP concentrations(100, 200, 300, and 400 �M) over a range of GMP concentrations(50–300 �M) for the forward reaction, and at different ADP concentra-tions (100, 200, 300, and 400 �M) over a range of GDP concentrations(25–300 �M) for the reverse reaction. The different parameters werecalculated considering the reaction to follow a random ordered bi bimechanism. The kinetic parameters of GMPKSc are from Li et al.35,those for GMPKEc are a personal communication from Anne-MarieGilles (Institut Pasteur, Paris).

492 G. HIBLE ET AL.

cysteines are reduced, were solved at 2.5, 2.1, 2.35, and2.3 Å, respectively (crystallographic statistics summarizedin Table II). All four crystals are isomorphous, and containone GMPKMt monomer per asymmetric unit [Fig. 2(a)].There is a clear electron density for the GMP and GDPnucleotides, whereas crystallization trials using ganciclo-vir monophosphate, the phosphorylated form of the antivi-ral prodrug ganciclovir, which is an inhibitor of mamma-lian GMPKs,38 yielded the apo-GMPKMt structure.

The reactive cysteines, Cys40 in �1 and Cys193 in �7,face each other in the CORE domain, where they exist asfree thiols in the presence of the reducing agent TCEP [Fig2(b), top panel] and form a disulfide bridge in its absence[Fig. 2(b), bottom panel]. The disulfide bridge is alsopresent in the apo-GMPKMt structure deposited in thePDB (S. Chan, M.R. Sawaya, L.J. Perry, D. Eisenberg,unpublished results, PDB code 1S4Q). It differs in naturefrom that recently described in the phosphatase Cdc25b,where one of the reactive cysteines is part of the P-loop andblocks the active site when engaged in a disulfide bond.39

Because of the reducing environment found in bacterialcytoplasms,40 the disulfide bond in GMPKMt is unlikely tooccur in the physiological form of the enzyme, althoughrecent analysis suggest that stabilization of intracellularbacterial proteins by disulfide bonds may be more wide-spread than classicaly recognized.49 Comparison of thereduced and oxidized GMPKMt structures shows howeverthat the disulfide bridge does not yield significant confor-mational changes.

GMPKMt features the conserved structural domains ofGMPKs from yeast mouse and E. coli [Fig. 2(a, c)],

although with significant departures [Fig. 3(a)]. The bestconserved domain is the GMP domain (residues 52–102),which superimposes remarkably with all other GMPKswith an average root-mean-square deviation (RMSD) onC� atoms of 0.5 Å. However, it differs locally by a two-glycine insert that protrudes between strands �5 and �6(Gly94 and Gly95) and is not found in eukaryotic enzymes[Fig. 3(a)]. In GMPKMm-GMP-ADP, this loop comes incontact with the LID, suggesting that the correspondinginteraction may be different in the closed conformation ofGMPKMt. The LID domain is in an open conformation, asseen by the fact that Arg155 and Arg166, which areexpected to contribute to the stabilization of the phospho-ryl transfer transition state, are at a distance from GMP[Fig. 2(a)]. This domain does not behave as a rigid body ingeneral. In GMPKMt, its helix �6 is in a straighterconformation than that found in closed GMPK structuresof GMPKs with a closed conformation [Fig. 3(a)]. TheCORE domain, however, has the same topology but asurprisingly high RMSD compared to other GMPKs, rang-ing from 1.31 Å (GMPKEc-GMP) to 1.75 Å (apo-GMPKSc),whereas it is 0.8 Å on average between other GMPKs. Themajor conformational differences are located at helices �1and �7, which are displaced relative to the five-stranded�-sheet. This may result from the unusual conformationalflexibility at the P-loop and/or the replacement of the(A/T)GKS motif, found in the P-loop of most GMPKs, bythe more bulky 32VGKS35 sequence [Fig. 3(a)].

Motion of the NMP and LID domains is a commonfeature of NMPKs, yielding a variety of conformations forthe apo- and mono nucleotide-bound enzymes, whereas

TABLE II. Crystallographic Statistics

GMPKMt-GDPOxidized

apo-GMPKMt

OxidizedGMPKMt-GMP

OxidizedGMPKMt-GDP

Data collectionUnit cell (Å) a � b � c � 112.0 a � b � c � 112.04 a � b � c � 112.231 a � b � c � 113.046Space group I23 I23 I23 I23Resolution limit (Å) 35.47–2.5 18.67–2.1 19.5–2.35 20–2.3Measured reflections 224,125 301,205 217,964 355,847Unique reflections 8,252 13,765 9,928 10,829Completeness (%) 99.8 (100) 99.4 (99.6) 99.6 (99.8) 99.7 (100)I/�I 12.8 (4.1) 32.7 (8.2) 30.7 (8.9) 32.2 (10.3)Rsym

a 4.4 (14.8) 5.9 (45.1) 7.0 (44.2) 8.3 (45.8)RefinementResolution limit (Å) 35.47–2.5 18.67–2.1 19.25–2.35 20–2.3Reflections (working/free) 7,818/410 13,071/707 9,439/493 10,297/537Rcryst

b (%) 18.0 (24.0) 18.6 (22.3) 17.8 (21.5) 18.4 (19.9)Rfree

b (%) 23.3 (33.3) 21.9 (27) 23.0 (22.5) 23.6 (20.5)No. resid./non-H atoms 183/1,465 182/1,453 182/1,477 183/1,474Average B factor (Å2) 19.4 29.9 36.8 24.8RMS deviationsBond length (Å) 0.016 0.012 0.016 0.016Angle distances (°) 1.641 1.306 1.676 1.572

Values in parentheses are for highest resolution shells (2.56–2.5 Å for GMPKMt-GDP, 2.15–2.1 Å for oxidizedapo-GMPKMt, 2.41–2.35 Å for oxidized GMPKMt-GMP, and 2.36–2.3 Å for oxidized GMPKMt-GDP).aRsym � h i �Ihi � �Ih��/ h i Ihi, where Ihi is the ith observation of the reflection h, whereas �Ih� is the mean intensity ofreflection h.bRcryst � �Fo� � �Fc� / �Fo�. Rfree was calculated with a fraction (5%) of randomly selected reflections excluded fromrefinement.

UNIQUE GMP-BINDING SITE 493

binding of both substrates yields a closed conformationthat is similar between all NMPKs (reviewed in Yan andTsai7). In GMPKs, binding of GMP alone induces theclosure of the GMP domain,20,21 whereas co-binding ofADP in the mouse enzyme further results in the fullyclosed conformation.22 Surprisingly, the apo-, GMP-bound, and GDP-bound GMPKMt structures have thesame domain closure, which is distinct from that of eitherapo- [Fig. 3(b), left panel] or GMP-bound [Fig. 3(b), rightpanel] forms from other GMPKs. To determine the degree

of closure of the GMP domain, the interactions of GMPKMt

with GMP were compared with those of GMPKMm-GMP-ADP and GMPKSc-GMP (Table III). This shows that GMPlacks a subset of interactions with the CORE domain(Asp101, Val102, and Asp103 in GMPKMm, correspondingto Glu119, Val120, and Asp121 in GMPKMt) and with theCORE/GMP domain linker (Thr101Mm). Thus, althoughthe GMP domain is more closed in our ensemble ofGMPKMt structures than in apo-GMPKSc, it retains essen-tially the characteristics of an open conformation. Alto-

Fig. 2. Structure of GMPKMt. a: The domain architecture of reduced GMPKMt-GDP. The CORE domain is in green, the LID domain in yellow, theGMP domain in blue, linker regions in gray, and GDP in red. b: Close-up view of Cys40 and Cys193 in the presence (top) or in the absence (bottom) ofthe reducing agent, showing the thiol and disulfide bridge, respectively. Electron densities are from 2Fo-Fc simulated annealing composite omit mapscontoured at 1.5 �. c: Sequence alignment of bacterial and eukaryotic GMPKs. From top to bottom, GMPK sequences are from: Mycobacteriumtuberculosis, Mycobacterium paratuberculosis (80% identity with M. tuberculosis), Mycobacterium leprae (77% identity), Corynebacterium diphtheriae(64% identity), Corynebacterium efficiens (61% identity), Corynebacterium glutamicum (61% identity), Homo sapiens (41% identity), and Mus musculus(43% identity). Invariant residues are shown on a black background, residues similar in at least 50% of the sequences on a gray background. On the topof the sequences are indicated the two cysteines (purple triangles, boxed), the secondary structures of the GMPKMt, and the limits of the GMPKMt

domains. The CORE domain includes all regions outside the GMP and LID domains. Hinges between domains (as defined by comparison with otherGMPK structures) are in gray. The unique GMP loop in GMPKMt is boxed in cyan. Residues that interact with GMP are marked by # and residues of theputative ATP-binding site are marked by stars. Residues of the ATP-binding site unique to GMPKs are boxed in red.

494 G. HIBLE ET AL.

gether, GMPKMt has a fully open domain conformation,which does not close upon binding of GMP and GDP, incontrast to other GMPKs or related NMPKs.

The GMP- and GDP-Binding Sites

Our series of structures allows us to analyze the GMPdomain in the unbound, substrate-bound, and product-bound forms of the mycobacterial enzyme. The phosphateacceptor site is essentially identical whether GMP, GDP,or no substrate is bound, indicating that it behaves as apreformed pocket. Together with the observation that theGMP domain of GMPKMt is in an open conformation, this

suggests that this domain contains all the features neces-sary to bind GMP regardless of the level of domain closure.Comparison with the GMPKSc-GMP and GMPKMm-GMP-ADP crystal structures shows that GMPKMt forms bothconserved and unique interactions with GMP. Conservedinteractions include the recognition of the guanine base byan invariant glutamate (Glu88) and of the �-phosphate byinvariant arginine and tyrosine residues (Arg57, Arg60,and Tyr69) [Table III, Fig. 4(a)]. Strikingly, this conservedmode of interaction also accommodates major sequencedifferences near the sugar and �-phosphate moieties ofGMP [Figs. 2(c), 4(b) and Table III]. GMPKMt lacks a

Figure 2 (Continued)

UNIQUE GMP-BINDING SITE 495

conserved tyrosine (Tyr81 in GMPKMm), which binds tothe �-phosphate of GMP and is in van der Waals contactwith the ribose in all other GMP-bound GMPKs. Mutationof the corresponding residue to phenylalanine in GMPKSc

resulted in both a 20-fold increase in KmGMP and 131-fold

decrease of kcat.41 The corresponding residue in GMPKMt,

Ser99, is too small to form a hydrogen bond with the�-phosphate, but it interacts with a water molecule thatfills the cavity where the tyrosine aromatic side-chain islocated in other GMPKs. Besides, the �5/�6 loop inGMPKMt (92IHGGLHRS99) differs in length and sequencefrom most other bacterial GMPKs (��GNyY) and fromeukaryotic GMPKs (FSGNlY) (where � stands for aliphatic/aromatic residues, capital letters for invariant residuesand lowercase letters for conserved residues) [Figs. 2(c),3(a), and 4(b)]. This results in the replacement of van derWaals interactions with the ribose and �-phosphate moi-eties of GMP mediated by Phe76 and Tyr81 in GMPKMm

by Ile92 and His93 in GMPKMt. Altogether, the GMP-binding pocket in GMPKMt has a different shape andestablishes alternative interactions with the �-phosphateand ribose of GMP compared with the mammalian en-zyme, resulting in a unique environment for the substratein the mycobacterial enzyme.

The GMPKMt-GDP structures show that the base andribose moieties of GDP are bound as those of GMP, andGDP does not interact with the CORE or LID domains. Asurprising observation is that the �-phosphate of GDP hasmoved relative to its position in the GMP-bound structure,such that Tyr69 and Arg57 interact with different �-phos-phate oxygens in GMP, but with the same oxygen in GDP[Fig. 4(c)]. The �-phosphate interacts only with Tyr69 fromthe GMP domain, a residue that forms a bidendate interac-tion with both phosphates. By comparison with structuresof related NMPKs in complex with transition state ana-logs,42–44 the �-phosphate is not positioned for efficientphosphate transfer, and does not correspond therefore to apostcatalytic intermediate. Given the open conformationof GMPKMt in this series of structures, the conformation of

GDP may be representative of the reaction product in thecourse of dissociating from the enzyme.

DISCUSSIONDomain Structure and Dynamics of GMPKMt

The structure of GMPKMt is distinct from that of allother GMPK structures solved so far in both the conforma-tion of the CORE domain, and the position of the LID andGMP domains relative to the CORE. Domain closure isnow established as a general mechanism by which bindingof ATP and NMP substrates induces the catalyticallycompetent, fully closed conformation of NMPKs. In thecase of GMPKs, a variety of conformations have beencharacterized in yeast and mammalian enzymes. In thisstudy, GMPKMt seems to depart from this general rule, asthe apo-, GMP-, and GDP-bound forms have a very similarlevel of domain closure, which corresponds to a fully openconformation based on the analysis of GMP/protein anddomain/domain interactions. This suggests that the confor-mations that have been trapped in the crystal may corre-spond to an early intermediate of the GMP-binding steppreceding closure of the GMP domain, and a late interme-diate of the GDP product release step after reopening ofthe GMP domain. Experimental probing of this hypothesismay require approaches other than crystallography, suchas NMR possibly in combination with hinge residue muta-tions. Assuming that the structure of the catalyticallycompetent GMPKMt is fully closed as those of GMPKs fromother species and of related NMPKs, our series of struc-tures suggests that the amplitudes and directions ofdomain closure movements are significantly different inGMPKMt than in other enzymes of the same family. Thismay result from the unusual conformation of the COREdomain, which in turn may yield a specific conformation ofthe interdomain hinges. Because the hinge regions areboth flexible and adopt species-specific conformations,they may constitute candidate sites of interest for thedesign or screening of specific noncompetitive inhibitors

TABLE III. Comparison of GMPK-GMP Interactions

GMP GMPKMt-GMP GMPKSc-GMP GMPKMm-ADP-GMP

O3P� N�2-Arg 60 3.3 N�1-Arg 41 3.5 N�2-Arg 44 2.8N�1-Arg 148 2.9

O2P� Nε-Arg 57 2.8 Nε-Arg 38 2.9 Nε-Arg 41 2.7N�2-Arg 60 3.5 N�2-Arg 41 3.5 N�2-Arg 44 3.3

O�-Tyr 78 2.5 O�-Tyr 81 2.6N�2-Arg 148 2.9

O1P� O�-Tyr 69 2.7 O�-Tyr 50 2.4 O�-Tyr 53 2.6N�2-Arg 38 3.0 N�2-Arg 41 2.9

O2� O�2-Asp 101 2.6O6 O�-Ser 53 2.6 O�-Ser 34 2.7 O�-Ser 37 2.6

O�-Ser 80 3.3 O�1-Thr 83 3.3N1 Oε1-Glu 88 2.9 Oε2-Glu 69 2.7 Oε2-Glu 72 2.8N2 Oε2-Glu 88 3.0 Oε1-Glu 69 2.7 Oε1-Glu 72 2.9

O�1-Asp 100 3.1 O�1-Asp 103 2.8N3 O�1-Asp 100 3.6 O�1-Asp 103 3.4

Distances between GMP and GMPKs are in angstrom. Equivalent residues in GMPKMt, GMPKSc,and GMPKMm are shown on the same line.

496 G. HIBLE ET AL.

Fig. 3. Structure and closure of the GMPKMt domains. a: Superposition ofthe GMP, CORE, and LID domains of GMPKMt-GMP (in yellow) with thecorrespondingdomains fromapo-GMPKSc (PDBentrycode1EX6)orGMPKMm-GMP-ADP (PDB entry code 1LVG). The redox-sensitive cysteines in the COREdomain are shown in orange. Orientations are similar to that in Figure 2a exceptfor the LID domain. b: Comparison of domain closure between the GMPKMt

structures and apo- (left) and GMP-bound (right) structures of GMPKSc. GMPKMt

is in yellow, GMPKSc in gray (PDB entry codes for apo- and GMP-boundGMPKSc structures are 1EX6 and 1EX7). Note that the GMP domain in GMPKMt

is more open than in GMPKSc-GMP and more closed than in apo-GMPKSc.

Fig. 4. The GMP and GDP binding site. a: Close-up view of GMPbinding site in the GMPKMt-GMP structure. The GMP domain is in blue,the CORE domain in green. Hydrogen bonds are in dotted lines. b:Comparison of GMP/enzyme interactions in mycobacterial GMPKMt-GMP(in yellow) and mammalian GMPKMm-GMP-ADP (in gray). Superpositionsare on GMP and the GMP-interacting residues from both structures. c:Comparison of GMP (yellow) and GDP (red) bound to GMPKMt.

capable of blocking the conformational changes that arerequired for the GMPK activity.

Conservation of the ATP and Mg2� Binding Sitesand of the Phosphate Transfer Machinery

The ATP-binding site of GMPKs has been located in theGMPKMm-ADP-GMP structure at the P-loop in the COREdomain and the CORE/LID interface. It is identical inGMPKMt, including the adenine binding residue Arg151[Fig. 2(a,c)]. However, we identified a specificity ofGMPKMt for ATP as a phosphate donor which is unusualamong NMPKs. We propose that this may result from apreviously unappreciated difference in the binding site ofthe adenine base in GMPKs compared with other NMPKs.Unlike other NMPKs, the coordination sphere of ATP inGMPKs includes the specific recognition of the N6 and N7atoms, which is mediated by an invariant asparagine inthe �9/�7 loop of the CORE domain (Asn171 in GMPKMm,Asn186 in GMPKMt). This residue is part of an invariantthree-dimensional motif that also comprises a proline pairin the �8-�5 loop in the CORE/LID linker (Pro142/Pro143in GMPKMt) [Fig. 2(a, c)]. This specific arrangement of theadenine-binding site may contribute to the specificity ofGMPKMt for ATP as a phosphate donor, and may be ageneral property of GMPKs that diverges from otherNMPKs.

No direct information is available yet on the stabiliza-tion of the transferred phosphoryl group by GMPKs at thetransition state. However, insight into this mechanism canbe derived from crystal structures of related NMPKs incomplex with transition state analogs (such as humanTMPKHs-ADP-AlF3-TMP,42 AMPKSc-AP5A,43 and D. dis-coideum CMP/UMPKDd-ADP-AlF4-UMP44). Arg155 in �5,Arg166 in �6, and Lys34 in the P-loop are thus predicted tobridge the �- and �-phosphates of ATP, and Arg 60 in theLID domain and Arg57 in the GMP domain to bridge the�-phosphate of ATP to the �-phosphate of GMP. Anothercritical component of catalysis is Mg2�, whose binding toGMPKs is enhanced by the presence of both substrates45

but has not been observed in crystal structures. InGMPKMt, it is predicted to comprise Ser35 in the P-loopand an acidic residue, Glu115, in the water-mediatedcoordination shell. This ensemble of residues is invariantor highly conserved in all GMPKs [Fig. 2(c)]. Thus, thephosphoryl transfer mechanism in GMPKMt should notdepart from that of GMPKs from eukaryotes, or, moregenerally, from that of NMPKs.

A Unique GMP-Binding Site in GMPKMt SuggestsStrategies for the Design of Specific Inhibitors

Whereas the ATP and Mg2� binding sites and thecatalytic residues are essentially identical betweenGMPKMt and eukaryotic GMPKs, the GMP domain ofGMPKMt reveals unexpected differences to the consensusbinding site that has been described in yeast and mamma-lian GMPKs. The most critical differences are located atSer99, which removes an enzyme/�-phosphate interaction,and Ile92/His93, which creates a less close-packed riboseenvironment. These differences do not affect the Km

GMP,

which is similar for the mycobacterial and other GMPKs,in contrast with the mutation in GMPKSc of Tyr78 (corre-sponding to Ser99 in GMPKMt) to phenylalanine whichcauses a 20-fold decrease.41 This may be explained byrescue in GMPKMt of the lost hydrogen bond by a watermolecule. However, we suggest that the Tyr/Ser replace-ment in GMPKMt may explain its lower kcat, in agreementwith the fact that the mutation of the equivalent residue toPhe in GMPKSc results in a 131-fold decrease in kcat.Remarkably, the 92IHGGLHRS99 motif encompassing the�5-�6 loop in the GMP domain is specific to the sequencesof GMPK from the actinomycete-related mycobacteria andcorynebacteria, two bacterial groups that contain thecausative agents of major human diseases that includetuberculosis, leprosy, and diphtheria [Fig. 2(c)]. Thus,GMPKs from mycobacteria and corynebacteria, includingGMPKMt, feature a GMP domain that is unique amongboth bacterial and eukaryotic GMPKs. In particular, theseGMPKs are predicted to depart from the human enzyme,which has 88% identity and has every GMP-interactingresidue identical with the mouse enzyme whose structureand GMP-binding site are known.46 As a consequence, theessential M. tuberculosis gmk gene, although encoding anenzyme conserved in all life kingdoms, may eventually bethe target for specific drugs. Our structural and biochemi-cal characterization now provides a robust background forsuch an approach. In the search of GMP analogs that couldfunction as inhibitors, future investigations would aim atcreating interactions that would be specific to GMPKMt,together with modifications that would result in the inhibi-tion of the phosphate transfer reaction. Based on thestructural analysis reported herein, modifications of theC8 and N7 on the guanosine moiety may be envisioned tointroduce GMPKMt-specific interactions with Ser99,whereas introducing a bulky group on the ribose moietycould block domain closure, thus preventing the assemblyof the catalytic site.

ACKNOWLEDGMENTS

This work was supported by an ACI grant from theFrench Ministry of Research to J.C., a grant from theDirection des Relations Internationales, CNRS to P.C.,and grants from INSERM and Institut Pasteur to H.M.L.The authors thank the staff at the ESRF synchrotron(Grenoble, France) for making beam line ID29 available tothem, Anne-Marie-Gilles (Institut Pasteur, Paris) for thegift of GMPKEc, Ioan Lascu (IBGC, Bordeaux, France) forthe gift of NDP kinase and Evelyn Krin (Institut Pasteur,Paris, France) for the antibody directed against histone-like nucleoid-structuring protein from E. coli (H-NS), andJoel Janin (LEBS, Gif-sur-Yvette, France) for criticalreading of the manuscript. The authors also thank D.Eisenberg and coworkers (University of California andMycobacterium tuberculosis Structural Genomics Consor-tium) for making the coordinates of the apo-GMPKMt

structure available in the Protein Data Bank databasebefore their publication (PDB entry 1S4Q). The authorsare grateful to Dr. Octavian Barzu (Institut Pasteur,Paris) for his support and mentorship.

498 G. HIBLE ET AL.

REFERENCES

1. Andries K, Verhasselt P, Guillemont J, et al. A diarylquinolinedrug active on the ATP synthase of Mycobacterium tuberculosis.Science 2005;307(5707):223–227.

2. Sassetti CM, Boyd DH, Rubin EJ. Genes required for mycobacte-rial growth defined by high density mutagenesis. Mol Microbiol2003;48(1):77–84.

3. Noda LH. Adenylate kinases. The enzymes. 3rd ed. New York:Academic Press; 1973. p 279–305.

4. Beck BJ, Huelsmeyer M, Paul S, Downs DM. A mutation in theessential gene gmk (encoding guanylate kinase) generates arequirement for adenine at low temperature in Salmonella en-terica. J Bacteriol 2003;185(22):6732–6735.

5. Pochet S, Dugue L, Labesse G, Delepierre M, Munier-Lehmann H.Comparative study of purine and pyrimidine nucleoside analoguesacting on the thymidylate kinases of Mycobacterium tuberculosisand of humans. Chem Bio Chem 2003;4:742–747.

6. Munier-Lehmann H, Burlacu-Miron S, Craescu CT, Mantsch HH,Schultz CP. A new subfamily of short bacterial adenylate kinaseswith the Mycobacterium tuberculosis enzyme as a model: apredictive and experimental study. Proteins 1999;36(2):238–248.

7. Yan H, Tsai MD. Nucleoside monophosphate kinases: structure,mechanism, and substrate specificity. Adv Enzymol Relat AreasMol Biol 1999;73:103–134.

8. Muller CW, Schulz GE. Structure of the complex between adenyl-ate kinase from Escherichia coli and the inhibitor Ap5A refined at1.9 Å resolution. A model for a catalytic transition state. J Mol Biol1992;224(1):159–177.

9. Bertrand T, Briozzo P, Assairi L, et al. Sugar specificity ofbacterial CMP kinases as revealed by crystal structures andmutagenesis of Escherichia coli enzyme. J Mol Biol 2002;315(5):1099–1110.

10. Lavie A, Ostermann N, Brundiers R, et al. Structural basis forefficient phosphorylation of 3�-azidothymidine monophosphate byEscherichia coli thymidylate kinase. Proc Natl Acad Sci USA1998;95(24):14045–14050.

11. Li de la Sierra I, Munier-Lehmann H, Gilles AM, Barzu O,Delarue M. X-ray structure of TMP kinase from Mycobacteriumtuberculosis complexed with TMP at 1.95 Å resolution. J Mol Biol2001;311(1):87–100.

12. Fioravanti E, Haouz A, Ursby T, Munier-Lehmann H, Delarue M,Bourgeois D. Mycobacterium tuberculosis thymidylate kinase:structural studies of intermediates along the reaction pathway. JMol Biol 2003;327(5):1077–1092.

13. Fioravanti E, Adam V, Munier-Lehmann H, Bourgeois D. Thecrystal structure of Mycobacterium tuberculosis thymidylate ki-nase in complex with 3�-azidodeoxythymidine monophosphatesuggests a mechanism for competitive inhibition. Biochemistry2005;44(1):130–137.

14. Miron S, Munier-Lehmann H, Craescu CT. Structural and dy-namic studies on ligand-free adenylate kinase from Mycobacte-rium tuberculosis revealed a closed conformation that can berelated to the reduced catalytic activity. Biochemistry 2004;43(1):67–77.

15. Munier-Lehmann H, Chaffotte A, Pochet S, Labesse G. Thymidy-late kinase of Mycobacterium tuberculosis: a chimera sharingproperties common to eukaryotic and bacterial enzymes. ProteinSci 2001;10(6):1195–1205.

16. Pochet S, Dugue L, Douguet D, Labesse G, Munier-Lehmann H.Nucleoside analogs as inhibitors of thymidylate kinases: possibletherapeutic applications. Chem Bio Chem 2002;3:108–110.

17. Haouz A, Vanheusden V, Munier-Lehmann H, et al. Enzymaticand structural analysis of inhibitors designed against Mycobacte-rium tuberculosis thymidylate kinase. New insights into thephosphoryl transfer mechanism. J Biol Chem 2003;278(7):4963–4971.

18. Vanheusden V, Munier-Lehmann H, Froeyen M, et al. 3�-C�-branched-chain-substituted nucleosides and nucleotides as potentinhibitors of Mycobacterium tuberculosis thymidine monophos-phate kinase. J Med Chem 2003;46:3811–3821.

19. Vanheusden V, Munier-Lehmann H, Froeyen M, et al. Discoveryof bicyclic thymidine analogues as selective and high-affinityinhibitors of Mycobacterium tuberculosis thymidine monophos-phate kinase. J Med Chem 2004;47(25):6187–6194.

20. Blaszczyk J, Li Y, Yan H, Ji X. Crystal structure of unligated

guanylate kinase from yeast reveals GMP-induced conformationalchanges. J Mol Biol 2001;307(1):247–257.

21. Stehle T, Schulz GE. Refined structure of the complex betweenguanylate kinase and its substrate GMP at 2.0 Å resolution. J MolBiol 1992;224(4):1127–1141.

22. Sekulic N, Shuvalova L, Spangenberg O, Konrad M, Lavie A.Structural characterization of the closed conformation of mouseguanylate kinase. J Biol Chem 2002;277(33):30236–30243.

23. Leipe DD, Koonin EV, Aravind L. Evolution and classification ofP-loop kinases and related proteins. J Mol Biol 2003;333(4):781–815.

24. Ellman GL. Tissue sulfhydryl groups. Arch Biochem Biophys1959;82(1):70–77.

25. Blondin C, Serina L, Wiesmuller L, Gilles AM, Barzu O. Improvedspectrophotometric assay of nucleoside monophosphate kinaseactivity using the pyruvate kinase/lactate dehydrogenase cou-pling system. Anal Biochem 1994;220(1):219–221.

26. Leslie AG, Powell HR, Winter G, et al. Automation of thecollection and processing of X-ray diffraction data: a genericapproach. Acta Crystallogr D Biol Crystallogr 2002;58(Pt 11):1924–1928.

27. Kabsch W. Automatic processing of rotation diffraction data fromcrystals of initially unknown symmetry and cell constants. J ApplCrystallogr 1993;26:795–800.

28. Brunger AT, Adams PD, Clore GM, et al. Crystallography & NMRsystem: a new software suite for macromolecular structure deter-mination. Acta Crystallogr D Biol Crystallogr 1998;54(Pt 5):905–921.

29. CCP4. The CCP4 suite: program for protein crystallography. ActaCrystallogr 1994;D50:760–763.

30. Jones TA, Zou JY, Cowan SW, Kjeldgaard. Improved methods forbuilding protein models in electron density maps and the location oferrors in these models. Acta Crystallogr A 1991;47(Pt 2):110–119.

31. Kraulis PJ. Molscript: a program to produce both detailed andschematic plots of protein structures. J Appl Crystallogr 1991;24:946–950.

32. Thompson ST, Cass KH, Stellwagen E. Blue dextran-sepharose:an affinity column for the dinucleotide fold in proteins. Proc NatlAcad Sci USA 1975;72(2):669–672.

33. Barzu O, Michelson S. Simple and fast purification of Escherichiacoli adenylate kinase. FEBS Lett 1983;153(2):280–284.

34. Bucurenci N, Sakamoto H, Briozzo P, et al. CMP kinase fromEscherichia coli is structurally related to other nucleoside mono-phosphate kinases. J Biol Chem 1996;271(5):2856–2862.

35. Li Y, Zhang Y, Yan H. Kinetic and thermodynamic characteriza-tions of yeast guanylate kinase. J Biol Chem 1996;271(45):28038–28044.

36. Anderson RG, Douglas LJ, Hussey H, Baddiley J. The control ofsynthesis of bacterial cell walls. Interaction in the synthesis ofnucleotide precursors. Biochem J 1973;136(4):871–876.

37. Van Rompay AR, Johansson M, Karlsson A. Phosphorylation ofnucleosides and nucleoside analogs by mammalian nucleosidemonophosphate kinases. Pharmacol Ther 2000;87(2–3):189 –198.

38. Miller WH, Beauchamp LM, Meade E, et al. Phosphorylation ofganciclovir phosphonate by cellular GMP kinase determines thestereoselectivity of anti-human cytomegalovirus activity. Nucleo-sides Nucleotides Nucleic Acids 2000;19(1–2):341–356.

39. Buhrman G, Parker B, Sohn J, Rudolph J, Mattos C. Structuralmechanism of oxidative regulation of the phosphatase Cdc25B viaan intramolecular disulfide bond. Biochemistry 2005;44(14):5307–5316.

40. Kadokura H, Katzen F, Beckwith J. Protein disulfide bondformation in prokaryotes. Annu Rev Biochem 2003;72:111–135.

41. Zhang Y, Li Y, Wu Y, Yan H. Structural and functional roles oftyrosine 78 of yeast guanylate kinase. J Biol Chem 1997;272(31):19343–19350.

42. Ostermann N, Schlichting I, Brundiers R, et al. Insights into thephosphoryltransfer mechanism of human thymidylate kinasegained from crystal structures of enzyme complexes along thereaction coordinate. Struct Fold Des 2000;8(6):629–642.

43. Abele U, Schulz GE. High-resolution structures of adenylatekinase from yeast ligated with inhibitor Ap5A, showing thepathway of phosphoryl transfer. Protein Sci 1995;4(7):1262–1271.

44. Schlichting I, Reinstein J. pH influences fluoride coordinationnumber of the AlFx phosphoryl transfer transition state analog.Nat Struct Biol 1999;6(8):721–723.

UNIQUE GMP-BINDING SITE 499

45. Prinz H, Lavie A, Scheidig AJ, Spangenberg O, Konrad M.Binding of nucleotides to guanylate kinase, p21(ras), and nucleo-side-diphosphate kinase studied by nano-electrospray mass spec-trometry. J Biol Chem 1999;274(50):35337–35342.

46. Brady WA, Kokoris MS, Fitzgibbon M, Black ME. Cloning,characterization, and modeling of mouse and human guanylatekinases. J Biol Chem 1996;271(28):16734–16740.

47. Briozzo P, Evrin C, Meyer P, Assairi L, Joly N, Barzu O, GillesAM. Escherichia coli UMP kinase differs from that of other

nucleoside monophosphate kinases and sheds new light on en-zyme regulation. J Biol Chem; 280(27):25533–25540.

48. Hible G, Renault L, Schaeffer F, Christova P, Zoe Radulescu A,Evrin C, Gilles AM, Cherfils J. Calorimetric and CrystallographicAnalysis of the Oligomeric Structure of Escherichia coli GMPKinase. J Mol Biol 2005;352(5):1044–1059.

49. Beeby M, O’Connor BD, Ryttersgaard C, Boutz DR, Perry LJ,Yeates TO. The genomics of disulfide bonding and protein stabili-zation in thermophiles. PLoS Biol 2005;3(9):1549–1558.

500 G. HIBLE ET AL.