enzyme molecular mechanism as a starting point to design new inhibitors: a theoretical study of ...

Published: May 04, 2011

r 2011 American Chemical Society 6764 dx.doi.org/10.1021/jp202079e | J. Phys. Chem. B 2011, 115, 6764–6775

ARTICLE

pubs.acs.org/JPCB

Enzyme Molecular Mechanism as a Starting Point to Design NewInhibitors: A Theoretical Study of O-GlcNAcaseJeronimo Lameira,† Cl�audio Nahum Alves,*,† I~naki Tu~n�on,‡ Sergio Martí,§ and Vicent Moliner*,§

†Laborat�orio de Planejamento e Desenvolvimento de F�armacos, Instituto de Ciencias Exatas e Naturais, Universidade Federal do Par�a,CP 11101, 66075-110 Bel�em, PA, Brazil‡Departament de Química Física, Universitat de Val�encia, 46100 Burjassot, Val�encia, Spain§Departament de Química Física i Analítica, Universitat Jaume I, 12071 Castell�on, Spain

bS Supporting Information

’ INTRODUCTION

Typically about 1% of the genome of any organism encodesglycoside hydrolases (GHs), and collectively, many thousandsof genes encoding GHs have been annotated, with structuralsimilarities into over 100 families of these proteins.1 The fieldof glycobiology was turned inside-out more than 20 yearsago by the discovery of O-linked 2-acetamido-2-deoxy-β-D-glucopyranoside (O-GlcNAc) from serine and threonine resi-dues of post-translationally modified nuclear and cytoplasmicproteins.2,3 On several proteins, O-GlcNAc and O-phosphatealternatively occupy the same or adjacent sites, leading to thehypothesis that one function of this saccharide is to transientlyblock phosphorylation.4 Recently, perturbations in the regula-tion ofO-GlcNAc have been related with type II diabetes5,6 andneurodegenerative disorders such as Parkinson dystonia7 andAlzheimer’s disease,8 which is the reason thatO-GlcNAcase wasused as target for therapeutic agents.9�11

O-Glycoprotein 2-acetamino-2-deoxy-β-D-glucopyranosidase(O-GlcNAcase) is a member of the family 84 glycoside hydrolases

(GH 84),12 as defined by the CAZY database, that hydrolyzesO-GlcNAc residues from post-translationally modified serine/threonine residues of nucleocytoplasmic protein. O-GlcNAcasecatalyzed reaction takes place through a two-step mechanism,first forming a transient oxazoline intermediate (P structure inScheme 1) that is subsequently broken. Oxazoline formation,which is the step studied in this paper, is a substrate-assistedcatalytic process, where the 2-acetomide group is activated by anaspartate residue while departure of the aglycone leaving groupis assisted by means of a proton transfer from a differentprotonated aspartate residue (see the hypothetical transitionstructure, TS, in Scheme 1). These two key residues have beenidentified as Asp174 and Asp175 in human O-GlcNAcase13,14

and Asp297 and Asp298 in Clostridium perfringens.15 It has beenproposed that in this first step of the O-GlcNAcase catalyzed

Received: March 4, 2011Revised: April 14, 2011

ABSTRACT: O-Glycoprotein 2-acetamino-2-deoxy-β-D-glucopyr-anosidase (O-GlcNAcase) hydrolyzes O-linked 2-acetamido-2-deoxy-β-D-glucopyranoside (O-GlcNAc) residues from post-trans-lationally modified serine/threonine residues of nucleocytoplasmicprotein. The chemical process involves substrate-assisted catalysis,where two aspartate residues have been identified as the two keycatalytic residues ofO-GlcNAcase. In this report, the first step of thecatalytic mechanism used by O-GlcNAcase involving substrate-assisted catalysis has been studied using a hybrid quantum mechan-ical/molecular mechanical (QM/MM)Molecular Dynamics (MD)calculations. The free energy profile shows that the formation of the oxazoline intermediate in theO-GlcNAcase catalytic reaction takesplace bymeans of a stepwisemechanism. The first step would be a cyclization of the acetomide group, which seems to be dependent onthe proton transfer from a conserved aspartate, Asp298 inClostridiumperfringensO-GlcNAcase. From this new intermediate, a proton istransferred from the azoline ring to another conserved aspartate (Asp297) thus forming the oxazoline ion and departure of theaglycone. In addition, averaged values of protein�substrate interaction energy along the reaction path shows that, in fact, the transitionstates present the highest binding affinities. A deeper analysis of the binding contribution of the individual residues shows that Asp297,Asp298, and Asp401 are basically responsible of the stabilization of these complexes. These results would explain whyO-(2-acetamido-2deoxy-D-glucopyranosylidene)amino-N-phenycarbamate (PUGNAc), 1,2-dideoxy-20-methyl-R-D-glucopyranoso-[2,1-d]-Δ20-thia-zoline (NAG-thiazoline), and GlcNAcstatin derivatives are potent inhibitors of this enzyme, resembling the two transition states ofthe O-GlcNAcase catalytic reaction path. These results may be useful to rational design compounds with more interesting inhibitoryactivity.

6765 dx.doi.org/10.1021/jp202079e |J. Phys. Chem. B 2011, 115, 6764–6775

The Journal of Physical Chemistry B ARTICLE

reaction Asp174 directs and polarizes the 2-acetamido group toact as a nucleophile and forming the oxazoline intermediate.13

Asp175meanwhile acts as a general acid, encouraging departureof the aglycone leaving group.13 In the second step, ring-opening, Asp174 would facilitate departure of the 2-acetamidogroup, whereas Asp175 would act as a general base, promotingthe attack of a water molecule to yield the β-hemiacetalproduct.13 According to this mechanism, oxocarbenium ion-like species with sp2-hybridized anomeric center and deloca-lized positive charge between the endocyclic ring oxygen andthe anomeric center have been proposed as the transition state(TS) for enzymes using substrate-assisted mechanisms.16�18

The hypothetical TS for the O-GlcNAcase (TS in Scheme 1)has been used as a model for rational design of inhibitors of thisenzyme.19,20 Among these inhibitors, O-(2-acetamido-2deoxy-D-glucopyranosylidene)amino-N-phenycarbamate (PUGNAc),1,2-dideoxy-20-methyl-R-D-glucopyranoso-[2,1-d]-Δ20-thiazoline(NAG-thiazoline), and GlcNAcstatin B (see Scheme 2 for detail

of the structures) are the best-characterized inhibitors of humanO-GlcNAcase.20�22 Recently, the crystal structure of a bacterialClostridium perfringens homologue (CpGH84H) in complex withPUGNAc has been reported, with significant sequence homologyto the humanO-GlcNAcase.14Moreover, similarities between theinhibition constants of human O-GlcNAcase and CpGH84Hwere observed, making this bacterial enzyme a good model ofthe human enzyme.14 An analysis of the enzyme-PUGNAccomplex carried out by Rao et al.15 suggests that this moleculemust inhibit O-GlcNAcase by mimicking the oxocarbeniumion-like TS of the O-GlcNAcase-catalyzed hydrolysis of N-acet-ylglucosaminide by virtue of its sp2 anomeric C1, whereasNAG-thiazoline would be geometrically similar to the oxazolineintermediate (P in Scheme 1). Nevertheless, Vocadlo and co-workers have shown that NAG-thiazoline is a TS analogue (TSA)for O-GlcNAcase, whereas PUGNAc is a poor TSA or a serendi-pitous binding inhibitor.21 Finally, GlcNAcstatin B derivative, themost potent human O-GlcNAcase inhibitor reported to date, is a

Scheme 1. Proposed Mechanism of the First Step of the Reaction Catalyzed by GlcNAcasea

aKey distances employed to explore the molecular mechanism are depicted in Rboat.

Scheme 2. Structures of Known Inhibitors of GlcNAcase



novel scaffold obtained by exploiting the structural similarity ofZ-PUGNAc and the naturally occurring potent human hexosa-minidases A/B (HexA/B) inhibitor nagstatin.23

Recently, we have carried out molecular dynamics (MD)simulations using a hybrid quantum mechanics/molecular me-chanics (QM/MM) approach to determine the binding ofPUGNAc and NAG-thiazoline, with a bacterial O-GlcNAcase.24

Our calculations allowed identifying the key residues in theprotein�inhibitor interaction, concluding that both complexespresented very close values, and in any case, the closer thestructure was from the hypothetical TS, the better the inhibitorwould be.24

Transition state theory (TST) has been used to explain whyenzymes can catalyze chemical reactions, and one of its maingoals is the successful application of the concept of TSA.25 Theideas of molecular structure and energetics are initially unified bypotential energy surfaces (PES). An understanding of a proteinPES is expected to give a good insight into the protein chemicalreaction.26 In order to carry out these kinds of studies for suchlarge molecular systems, hybrid QM/MM schemes can be used,where a small part of the system is described by quantummechanics and the rest of the system by a classical force field.25

Free energy profiles can thus be obtained which render valuesdirectly compared with experimental data. The free energyprofile of the transformation from reactants to products can beobtained as a potential of mean force (PMF) from biased MDsimulations using an adequate distinguished reaction coordinate.The maximum of this profile corresponds to the TS of thereaction while the free energy difference between this maximumand theminimum, which corresponds to reactants, will be relatedwith the chemical rate constant. This free energy barrier can berelated and compared with experimental rate constants sinceaveraged structures of reactants, TS, and possible intermediatesof the chemical reaction are obtained under the protein environ-ment effect by means of a flexible and realistic model.28

In this study, we have used both potential energy profiles andfree energy profiles for studying the catalytic mechanism used byO-GlcNAcase involving substrate-assisted catalysis. In addition, adetailed analysis of the interactions of the substrate with the keyresidues inside the binding pocket in reactants, TS and inter-mediate states, has been carried out in order to decide whichstructure can present the most efficient interactions with theprotein, thus becoming the most effective model to designefficient inhibitors.

’METHODS

The System. We have recently carried out MD simulations,using a combined QM/MM approach, for studying protein�inhibitors interaction energies.24,29 In these calculations, thesmall part of the system described quantum mechanically wasthe ligand/substrate species, whereas the protein and solventenvironment were represented by MM force fields. This hybridmethodology, by comparison with methods based on just MM,avoids most of the work needed to obtain new force fieldparameters for each new species. Treating the ligand quantummechanically and the protein molecular mechanically has theadditional advantage of the inclusion of quantum effects such asligand polarization upon binding.29 Moreover, as the largest partof the system is described classically, enough sampling can beobtained at reasonable computational cost.

In the work reported here, the initial coordinates for the QM/MM MD calculations were taken from the trimeric crystalstructure of CpGH84H-PUGNAc complex at 2.25 Å resolution,PDB code 2CBJ.14 This crystal structure (from Clostridiumperfringens) presents significant sequence homology to thehOGA N-terminus (CpNagJ, 34% sequence identity and 51%sequence similarity with hOGA) and thus it is a suitable modelfor studies into the mechanisms of recognition/specificity ofO-GlcNAcylated peptides.14

Since the standard pKa values of ionizable groups can beshifted by local protein environments,30 an accurate assignmentof the protonation states of all these residues at pH = 7 wascarried out. Recalculation of the pKa values of the titratableamino acids has been done using the “cluster method” asimplemented by Field and co-workers.31 In this method, eachtitratable residue is perturbed by the electrostatic effect of theprotein environment. We have also calculated the pKa values ofaminoacids within the empirical propKa program of Jensenet al.32 Results obtained with both methods are qualitatively inagreement, predicting the same protonation state for all residuesat the selected pH of the simulations. Thus, according to theseresults, most residues were found at their standard protonationstate, except for the Asp-298 residue that was protonated. In fact,as demonstrated from a previous study where a QM/MM PMFcorresponding to the proton transfer from the carboxylateoxygen atom of Asp298 to the oxime nitrogen atom of PUGNAcwas traced, the PUGNAc-protein complex was found to be morestable with the proton on the Asp298.33

After adding the hydrogen atoms to the structure, series ofoptimization algorithms (steepest descent conjugated gradientand L-BFGS-B34) were applied. To avoid a denaturation of theprotein structure, all the heavy atoms of the protein and theinhibitor were restrained by means of a Cartesian harmonicumbrella with a force constant of 1000 kJ mol�1 Å�2. Afterward,the system was fully relaxed, but the peptidic backbone wasrestrained with a lower constant of 100 kJ mol�1 Å�2.The optimized protein was placed in a cubic box of pre-

equilibrated waters (80 Å side), using the principal axis of theprotein-inhibitor complex as the geometrical center. Any waterwith an oxygen atom lying in a radius of 2.8 Å from a heavy atomof the protein was deleted. The remaining water molecules werethen relaxed using optimization algorithms. Finally, 100 ps ofhybrid QM/MM Langevin-Verlet MD (NVT) at 300K wereused to equilibrate the solvent.During theMD simulations, the atoms of the substrate and the

side chains of Asp297 and Asp298 (present in active site) wereselected to be treated by QM, using a semiempirial AM1Hamiltonian.35 The rest of the system (protein plus watermolecules) were described using the OPLS-AA36 and TIP3P37

force fields, respectively, as implemented in the fDYNAMOlibrary.38 To saturate the valence of the QM/MM frontier atoms,link atoms between the CR and Cβ atoms of Asp297 and Asp298residues were used. The final system contains a total of 23 475atoms, 50 of them in the QM region.Due to the large amount of degrees of freedom, any residue

20 Å apart from any of the atoms of the initial inhibitor was keptfrozen in the remaining calculations. Cut-offs for the nonbondinginteractions were applied using a switching scheme, within arange radius from 14.5 to 16 Å.Afterward, the system was equilibrated by means of 1.1 ns of

QM/MMMD at temperature of 300 K. The computed rmsd forthe protein during the last 500 ps renders a value always below



0.9 Å. Furthermore, the rms of the temperature along thedifferent equilibration steps was always lower than 2.5 K andthe variation coefficient of the potential energy during thedynamics simulations was never higher than 0.3%.The Energy Function. In the work reported here, we have

made use of a hybrid AM1/OPLS-AA method, based on asemiclassical approach. The potential energy of our scheme isderived from the standard QM/MM formulation

EQM=MM ¼ ÆΨjHojΨæ

þ ∑ Ψ

��*

qMM

re,MM

��Ψ0@ +

þ∑∑ZQMqMM

rQM,MM

1A

þ EvdWQM=MM þ EMM ð1Þ

EQM=MM ¼ Evac þ EelectQM=MM þ EvdWQM=MM þ EMM ð2Þwhere EMM is the energy of the MM subsystem terms,EQM/MMvdW the van der Waals interaction energy between the

QM and MM subsystems and EQM/MMelect includes both the

Coulombic interaction of the QM nuclei (ZQM) and the electro-static interaction of the polarized electronic wave function (Ψ)with the charges of the protein (qMM).Finally, the interaction energy between the inhibitor and the

environment is evaluated as the difference between theQM/MMenergy and the energies of the separated, noninteracting, QMand MM subsystems with the same geometry. Considering thatthe MM part is described using a nonpolarizable potential, theinteraction energy contribution of each residue of the protein isgiven by the following expression:

EIntQM=MM ¼ ∑ Ψ

��*

qMM

re,MM

��Ψ+

þ∑∑ZQMqMM

rQM,MMþ EvdWQM=MM ð3Þ

This interaction energy can be exactly decomposed in a sum overresidues provided that the polarized wave function (Ψ) isemployed to evaluate this energy contribution. The globalpolarization effect can be obtained from the gas phase energydifference between the polarized,Ψ, and unpolarized,Ψ0, wavefunctions.The Potential Energy Surface (PES). The fDYNAMO

library38 was used to explore the different PESs as a function ofthe distances R1, R2, R3, R4, R5, and R6 (see Scheme 1). AllPESs were obtained using two antisymmetric combinations ofthese distances: the R1-R2 corresponds to the proton transferfromAsp298 to the linking anomeric oxygen, R3-R4 correspondsto the attack of the 2-acetamido carbonyl oxygen on the anomericcenter and departure of aglycone leaving group, and R5-R6corresponds to the proton transfer from linking N-acetyl amidehydrogen to Asp297.The Potential of Mean Force (PMF). In order to obtain the

free energy associated to the proton transfers and cyclization,we have traced several PMF.41�43 All PMFs have been calcu-lated using the weighted histogram analysis method (WHAM)combined with the umbrella sampling approach44,45 as imple-mented in fDYNAMO.38 The procedure for the PMF calcula-tion is straightforward and requires series of moleculardynamics simulations in which the distinguished reaction

coordinate variable, ξ, is constrained around particular values.45

The values of the variables sampled during the simulations arethen pieced together to construct a distribution function fromwhich the PMF is obtained as a function of the distinguishedreaction coordinate (W(ξ)). The PMF is related to the normal-ized probability of finding the system at a particular value of thechosen coordinate by eq 4

WðξÞ ¼ C� kT lnZ

FðrNÞδðξðrNÞ � ξÞ drN � 1 ð4Þ

The activation free energy can be then expressed as46

ΔG‡ðξÞ ¼ Wðξ‡Þ � ½WðξRÞ þ GξðξRÞ� ð5Þwhere the superscripts indicate the value of the reaction co-ordinate at the reactants (R) and TS and Gξ(ξ

R) is the freeenergy associated with setting the reaction coordinate to aspecific value at the reactant state. Normally this last term makesa small contribution47 and the activation free energy is directlyestimated from the PMF change between the maximum of theprofile and the reactant’s minimum

ΔG‡ðξÞ � Wðξ‡Þ �WðξRÞ ¼ ΔW‡ðξÞ ð6ÞThe selection of the reaction coordinate is usually trivial whenthe mechanism can be driven by a single internal coordinate or asimple combination (as the antisymmetric combination of twointeratomic distances). However this is not the case for thereaction subject of study in this paper where at least 6 coordinatesare participating. Instead we were compelled to obtain a muchmore computationally demanding 2D-PMF using two coordi-nates: ζ1 and ζ2. The 2D-PMF is related to the probability offinding the system at particular values of these two coordinates

Wðζ1, ζ2Þ ¼ C0 � kTlnZ

FðrNÞδðζ1ðrNÞ

� ξ1Þδðζ2ðrNÞ � ξ2Þ drN � 2 ð7ÞTo estimate the activation free energy from this quantity werecovered one-dimensional PMF changes tracing a maximumprobability reaction path on the 2D-PMF surface and integratingover the perpendicular coordinate.In constructing the 2D-PMF, the distinguished reaction co-

ordinates were the antisymmetric combination of the distancesR1-R2 and R3-R4, corresponding to proton transfer fromAsp298 to linking anomeric oxygen and coordinates correspondsto attack of the 2-acetamido carbonyl oxygen on the anomericcenter, respectively. A total of 39 simulations were performed atdifferent values of R1-R2 (ranging from�2.0 toþ2.0 Å), with anumbrella force constant of 2800 J 3 mol�1

3 Å�2 applied to this

distinguished reaction coordinate. In addition, 30 simulationswere performed at different values of R3-R4 (ranging from�1.5 to þ1.5 Å), also with an umbrella force constant of2800 J 3 mol�1

3 Å�2 on this antisimmetric combination of

distances. Consequently, 1170 simulation windows were neededto obtain the 2D-PMF. The values of the variables sampled duringthe simulations were then pieced together to construct a fulldistribution function from which the 2D-PMF was obtained.In each window, 5 ps of relaxation were followed by 10 ps ofproduction with a time step of 0.5 fs due to the nature of thechemical step involving a hydrogen transfer. The Verlet algorithmwas used to update the velocities.



Three monodimensional PMFs were also computed being thedistinguished reaction coordinates the antisymmetric combina-tion of the distances R1-R2, R3-R4, or R5-R6 coordinates. A totalof 60 simulations were performed at different values of R1-R2 (ina range from �2.0 to þ1.0 Å), R3-R4 (in a range from �2.0to þ2.0 Å) and R5-R6 (in a range from �1.5 to þ1.5 Å), withan umbrella force constant of 2800 kJ 3 mol�1

3Å�2. On each

window the protocol of the simulation was as described abovefor 2D-PMF.

’RESULTS AND DISCUSSION

Substrate Distortion and PES. The substrate distortionduring the enzymatic hydrolysis of glycosides has been relatedin some reports.48�53 Recently, the reaction mechanisms ofdifferent glycoside hydrolases that follow different conforma-tional itineraries have been related by Vocadlo and Davies.1 Onthe other hand, Brameld and Goddard III have examined theplausible conformations for hexaNAG substrate bound to theactive site of Chitinase by means of classical MD simulation.54

The authors proposed that the hydrolysis mechanism of Chit-inase A involves substrate distortion and that the protonation ofthe linking anomeric oxygen requires a boat conformation for theGlcNac residue at binding subsite �1.54 In this proposedmechanism, the first step of the reaction starts with the substratein the boat conformation.12,20

In order to analyze the substrate distortion, 1 ns of QM/MMMD was carried out using the chair conformation as startingpoint. After 400 ps of MD simulation a change conformationfrom chair to boat was observed but the conformation of the ringwas back to chairlike after 550 ps, a conformation that ismaintained until 800 ps of the MD simulations, where a changeconformation from chair to boat was observed again. This resultindicates that both chair and boat conformations are close inenergy, and as a consequence, significant population of bothconformers can be observed at 300 K. This is reflected in thetime-evolution of the interaction energy of the substrate with theenzyme and of some key intermolecular distances (see Figure 1,panels a and b). Analysis of time evolution of distances along theMD shows that R2, the hydrogen bond distance from protonatedAsp298 to the oxygen atom of the leaving group, is shorter whenthe substrate adopts the boat conformation than in the chairconformation (see Figure 1b). This relative movement betweenAsp298 and the substrate is associated with a change on thecharge of the ether oxygen that varies from �0.55 au (chairconformation) to�0.61 au (boat conformation). Shorter hydro-gen bond distances leads to higher interaction energies, asdeduced by combining Figure 1, panels a and b. These results

show that the substrate must adopt a boat conformation for thereaction to proceed, in accordance with results reported inliterature.12,20 This conformation will be the starting point toexplore the different PESs described below.The first PES (Figure 2a) was obtained using two antisym-

metric combinations; R1-R2 and R3-R4. R1-R2 corresponds tothe proton transfer from Asp298 to linking anomeric oxygen,while R3-R4 corresponds to the attack of the 2-acetamidocarbonyl oxygen on the anomeric center and cleavage of theleaving group (see Scheme 1). First conclusion that can bederived from this PES is that while R structure corresponds toreactive structure in the boat conformation, the product of thisstep, INT on the PES, corresponds to an oxazoline ion/methanolcomplex; a structure where the proton of the N-acetyl amide hasnot been still transferred to the Asp297 (see Scheme 1). It can bealso observed that formation of INT takes place in a concertedbut very asynchronous way; first the proton transfer fromAsp298toO-GlcNAc and then cyclization can be started via attack of the2-acetamido carbonyl oxygen on the anomeric center. In the TSstructure, TS1, the bond distance for R1 corresponds to 1.90 Å,whereas R2 corresponds to 0.99 Å, indicating a very advancedstage for the proton transfer. On the other hand, the bonddistance for R3 in the TS corresponds to 1.91 Å, whereas R4corresponds to 2.52 Å, indicating that C�O bond breaking, R3,and approximation of the 2-acetamido carbonyl oxygen on theanomeric center, R4, are in an earlier stage of the process. Theenergy barrier at the AM1/MM level was 146.3 kJ mol�1 (35.0kcal/mol). These observations are in agreement with results ofthe Brameld and Goddard for Chitinase, who, using ab initio gasphase calculations with a reducedmodel, obtained a spontaneousglycosidic bond cleavage with subsequent formation of an oxazo-line ion/methanol complex from a geometry optimization of theprotonated linking anomeric oxygen of O-GlcNac in its boatconformation.54

A PES from INT to P was generated using just the R5-R6antisymmetric combination of distances, that corresponds to theproton transfer to Asp297, as the distinguished reaction coordi-nate (Figure 2b). The resulting potential energy path render anexothermic chemical reaction with an energy barrier of 50.3 kJmol�1 (12.0 kcal/mol) thus confirming the viability of a stepwisemechanism kinetically determined by the first step.In order to explore a possible concertedmechanism, a PES was

constructed using the R3-R4 and the R5-R6 antisymmetriccombinations of distances. The proton transfer from Asp298 tothe linking anomeric oxygen is not explicitly shown in this PES,but analysis of the PES structures reveals that it takes placespontaneously when these coordinates are used as distinguishedreaction coordinates thus being the product of the reaction an

Figure 1. Time evolution of total substrate�protein interaction energy and R2 distance during the 1 ns of QM/MMMD simulations on reactants state.



oxazoline intermediate, P. The R5 and R6 distances of the TSlocated on this PES (Figure 2c) are 1.00 and 2.04 Å, respectively,and R1 and R2 distances, defining the position of the proton tobe transferred from Asp298, are 0.99 and 2.05 Å. These resultsindicate that the proton from N-acetyl amide, as well as theproton from Asp298, are not yet transferred. The values foundfor R3 and R4 are 2.44 and 1.79 Å, respectively. All together,these results indicate that in this mechanism the proton transferfrom linking N-acetyl amide hydrogen to Asp297 occur after thecycle formation in this mechanism. R structure corresponds toreactive structure in the boat conformation. The energy barrierbetween R and TS in the PES calculated at AM1/MM level was246.6 kJ.mol�1 (58.9 kcal/mol). This result suggests that thereaction could also proceed by a concerted mechanism, althoughwith an energy barrier higher than the rate limiting step of thestepwise mechanism.Potentials of Mean Force. Based on the conclusions derived

from the PESs presented above, only the free energy profile ofthe stepwise mechanism will be studied. The first step will be theproton transfer from Asp298 to linking anomeric oxygen withthe concomitant attack of the 2-acetamido carbonyl oxygen onthe anomeric center, and the second step would be the protontransfer from linking N-acetyl amide hydrogen to Asp297. Inorder to study the first step that involves forming and breaking offour bonds, a 2D-PMF using the same coordinates as the onesemployed to generate the PES of Figure 2a was computed. The

topology of the resulting 2D-PMF surface, depicted in Figure 3a,results to be quite similar to the corresponding PES of Figure 2a,where the products of this step is the oxazoline ion/methanolintermediate, INT. In order to check this result, free downhillMD simulations taking TS1 structure (protonated GlcNAc) asstarting point have been performed showing that the time-evolution of the system leads to spontaneous glycosidic bondcleavage with subsequent formation of an oxazoline ion-metha-nol complex. We have also performed 500 ps of MD simulationtaking INT structure as starting point and the result shows thatthe oxazoline ion intermediate is a stable structure at AM1/MMlevel, in agreement with the work of Brameld and Goddard III.54

Structures of R in Figure 3a correspond to reactive structuresinitially in a chair conformation that change to a boat conforma-tion during the MD simulations at 300 K. These structures canthen evolve toward TS1. In TS1, the averaged bond distance forR1 was found to be 1.65 Å, whereas for R2 was 1.02 Å, indicatingthat the proton is already almost completely transferred. On theother hand, the averaged bond distance for R3 was 2.02 Å,whereas for R4 was 2.27 Å. Other averaged key distancesobtained from the 2D-PMF are showed in Table 1.Using as starting point the INT structure (ion oxazoline/

methanol), the second PMF (Figure 3b) was traced using R5-R6as distinguished coordinate, which corresponds to the protontransfer from linking N-acetyl amide nitrogen atom to one of theoxygen atoms of the carbocylate group of Asp297. The product

Figure 2. Potential energy surfaces calculated at AM1/MM level obtained by means of (a) R1-R2 and R3-R4 antisymmetric combinations of distances,(b) R5-R6 antisymmetric combination, and (c) R3-R4 and R5-R6 antisymmetric combinations of distances. Values of contour potential energy lines arereported in kJ mol�1 and coordinates in Å.

Figure 3. (a) 2D-PMF (in kJmol�1) obtained for proton transfer fromAsp-298 to substrate (R-R2, in Å) and attack of the 2-acetamido carbonyl oxygenon the anomeric center (R3-R4, in Å). (b) PMF corresponding to proton transfer from the substrate to Asp-297 (R5-R6, in Å). Values on isoenergeticlines are reported each 10 kJ mol�1.



found for this profile is, indeed, the oxazoline/methanol com-plex, P in Scheme 1.The calculated activation free energies,ΔG‡, for TS1 and TS2

are 195 and 40 kJ mol�1, respectively (46.6 and 9.6 kcal mol�1).On the base of these results, it is reasonable to suggest that thefirst step, the proton transfer and glycosidic bond cleavageoccurring simultaneously, would be the rate limiting step of theprocess. A note of caution must be introduced at this point, sincethese barriers appear to be overestimated keeping in mindthat they correspond to an enzyme catalytic process. Obviously,the low level of the AM1 semiempirical Hamiltonian, requiredto allow performing long hybrid QM/MM MD simulations,must be the main source of error and absolute energetic valuesmust be considered at relative terms. Thus, corrections atB3LYP(6-31G(d))/MM level have been carried out by addingthe single-point gas phase energy differences between therelative energies obtained at a high level (B3LYP) and a lowlevel (AM1) of theory, using the structures from the AM1/MMpotential energy surface. These corrections are then incorporatedinto the simulations using a two-dimensional cubic splinefunction.55 The new free energy surface render a barrier for thefirst step of 155 kJ mol�1 (37.0 kcal mol�1). This value, while stillbeing too high for the kind of process under study, is significantlylower than the one obtained at AM1/MM level. The remainingerror could be due to the use of AM1 geometries in our high levelcalculations.Interaction Energy. A problem of central importance in

computational biology is the quantitative determination ofabsolute binding affinities in diverse and complex systems.Predicting the binding free energy of ligands to macromoleculescan have great practical values in identifying novel molecules thatcan bind to target receptors and act as therapeutic drugs.56

In order to get a deep insight into de protein�substrateinteractions, 500 ps of MD were run for all substrate�proteincomplexes appearing along the free energy profile (Rchair, Rboat,TS1, INT, TS2, and P), constraining the corresponding distin-guished reaction coordinate in the two transition states. A note ofcaution has to be mentioned at this point since the methanolformed in the first step of reaction is eliminated from the systemafter formation of INT. Thus, methanol is not included in QM/MMMD simulations to compute protein�substrate interactionsfor INT�enzyme, TS2�enzyme, and P�enzyme. Averagedsubstrate�protein interaction energies for all these complexesat AM1/MM level are listed in Table 2. It must also be pointedout that Asp298 and Asp297 were included in the QM region inall of these calculations to obtain comparable results of sub-strate�protein total interaction energies for all the stationarystates, whatever the degree of proton transfer to or from theenzyme. As a consequence, the reported values in Table 2 are notsubstrate�protein interaction energies, strictly speaking, but the

analysis of the trend along the reaction progress can be used toobtain an idea of the most adequate structures to design newinhibitors. Analysis of the data reported in Table 2 reveals thatthe structures presenting the highest total substrate�proteininteraction energy, in absolute value, are TS1 and TS2. Thisconclusion confirms these two structures as the candidates to beused as a mold to design new inhibitors, which is reasonablekeeping in mind that most of the inhibitors are transition stateanalogs (TSA).The origin of the substrate�enzyme interaction can be

analyzed decomposing the total interaction energy in a sum ofcontributions due to each of the enzymatic residues. Averagedprotein�substrate interaction energies by residue, computedincluding now both aspartate residues (Asp297 and Asp298) inthe MM region, are displayed in Figure 4. In this figure, positivevalues correspond to unfavorable interactions, whereas a negativevalues means that the interaction is stabilizing the substrate inthat particular state. The pattern of interactions of Rchair�enzyme complex is very close to the one obtained for theRboat�enzyme complex. Both show the interaction withAsp401 as the most important one for both chair and boatconformation. This result is in agreement with the observationsof Rao et al. demonstrating that Asp401 has significant effect onthe ability of the enzyme to bind the substrate/inhibitor (5�10-fold increase in Km).

14 Other important residues interactingwith reactants are Tyr-335 and Asn-396. These two residueswould then be well positioned to stabilize any partial negativecharge that develops at the glycosidic oxygen in the TS,consistent with the role of these residues assigned for thecatalysis.14 However, these interactions can be also importantfor substrate fixation in both of its conformations. The majordifference between the Rboat�enzyme and Rchair�enzymecomplex is in interaction with the Asp298 residue. In fact, asobserved in Figure 4 Asp 298 presents favorable interaction inRboat�enzyme complex, where this interaction is not appreciable

Table 2. Averaged AM1/MM Protein�Substrate InteractionEnergies for Rchair, Rboat, TS1, INT, TS2, and P StructuresObtained from the PMFs (in kJ mol�1)a

structures EAM1/MM (kJ mol�1)

Rchair �1186.9( 71.7

Rboat �1260.6( 78.2

TS1 �1622.6( 27.4

INT �1520.3( 35.9

TS2 �1605.2( 36.5

P �1215.1( 56.3aValues reported for INT, TS2 and P have been computed treating theformed methanol molecule in the MM region.

Table 1. Averaged Distances of Key Distances for Rchair, Rboat, TS1, INT, TS2, and P Structures (in Å)a

distance Rchair Rboat TS1 INT TS2 P

R1 0.98 (0.03) 0.98 (0.03) 1.65 (0.04) 5.14 (1.15) 4.96 (0.65) 5.45 (0.53)

R2 4.06 (0.56) 3.61 (0.65) 1.02 (0.03) 0.97 (0.03) 0.98 (0.01) 0.98 (0.01)

R3 1.42 (0.03) 1.43 (0.03) 2.02 (0.15) 6.60 (1.46) 4.95 (0.51) 5.30 (0.64)

R4 2.95 (0.13) 2.97 (0.14) 2.27 (0.14) 1.51 (0.04) 1.48 (0.03) 1.46 (0.03)

R5 1.00 (0.03) 1.00 (0.03) 1.00 (0.03) 1.01 (0.03) 1.25 (0.03) 2.09 (0.03)

R6 2.08 (0.13) 2.08 (0.11) 2.08 (0.13) 1.99 (0.13) 1.27 (0.03) 0.98 (0.02)a Standard deviations, computed during the last 10 ps of the MD simualtions, are reported in brackets (in Å) from 2D-PMF.



in Rchair�enzyme complex. This explains the higher interactionenergies obtained when substrate adopts a boat conformation(see Figure 1a).The pattern of interactions of P�enzyme complex reveals very

weak interactions, in accordance with the total value of thesubstrate�protein interaction energy as reported in Table 2.INT�enzyme complex presents a pattern of interactions veryclose to the one observed for TS2�enzyme but with weakerinteraction with Asn396 and unfavorable interaction withLys’184. Moreover, the unfavorable interaction with Lys218,observed in both INT and TS2, is weaker in the later. Thus, ouranalysis will be focused on the transition states TS1 and TS2. Adetailed analysis of the contribution of individual residues to thetotal TS1�enzyme and TS2�enzyme interaction energy showsthat while the strongest interaction of TS1 is established withAsp298 residue, while in the case of TS2, it is with Asp297residue. The other significant difference between these struc-tures is the interaction with Lys218 residue, which is favorablefor TS1 but unfavorable for TS2. Interaction with Lys184’stabilizes both transition states, and Asn396 presents favorableinteractions not only in these two complexes but also, asobserved previously, in both reactants conformers. The factthat Asp297, Asp298, together with Asp401, stabilize all speciescreated along the reaction profile is in agreement with muta-genesis and structural studies reported for O-GlcNacase frombacterial Clostridium perfringens by Rao et al. showing thatAsp297, Asp298 and Asp401 are the key catalytic residues ofO-GlcNAcase.14

These findings are also in agreement with our previous QM/MM studies of interactions of two inhibitors (PUGNAc andNAG-thiazoline) with CpGH84H protein.24 Nevertheless, in thepresent report, analysis of contribution of individual residues tothe total protein�substrate interaction energy has been carried

out along the reaction profile showing that Asp298 has importantstabilizing interaction with TS1 complex, Asp297 with TS2 andoxazoline ion/methanol complex, and Asp401 presents favorableinteraction with the substrate in all states along the reaction.Inhibitors are generally designed to resemble the structure of a

molecular species occurring during the chemical reaction and, inparticular, transition states structures, if accepting the seminalhypothesis of Pauling that enzymes catalyze reactions by pre-ferentially binding (and stabilizing) the transition state.57 In ourcase, NAG-thiazoline inhibitor structure is close to TS2�enzymestructure, whereas PUGNAc andGlcNAcstatin B inhibitors seemto be closer to TS1�enzyme structure. The high resemblebetween the TS2 structure and NAG inhibitor, and betweenTS1 and PUGNAG and GlcNAcstatin B can be seen in Figure 5,where superposition of the structures of the two transition stateswith their corresponding inhibitors are displayed. As mentionedin the Introduction, PUGNAc, NAG-thiazoline and GlcNAcsta-tin B are well characterized inhibitors of human O-GlcNAcasethat present the highest binding values (Ki = 46, Ki = 70, and Ki =0.4 nM, respectively).22 Table 2 shows that total interactionenergies for TS1 and TS2 are also close and, in fact, they are thestructures presenting the highest stabilizing values. All theseobservations suggest that the three studied inhibitors mightbe considered as TSA, which would explain their potency, bycomparison with other ground state analogues based inhibitors(i.e., methyl 2-acetamido-2-deoxy-β-D-glucopyranoside, Ki =11 μM)21 or simply serendipitous binders (i.e., iminosugarinhibitors as the nojirimycin class)58 As mentioned by Vocadloand co-workers,21 NAG-thiazoline molecule has an obviousgeometrical resemblance to the intermediate of the reaction(P in scheme 1), but we herein are demonstrating that it tightlybinds the active site of the protein by its resemblance to the TS2,that is, in turn, close to the oxazoline intermediate.

Figure 4. Contributions of individual amino acids to substrate�protein interaction energy (in kJ mol�1) computed for Rchair, Rboat, TS1, INT, TS2, and P.



Molecular Electrostatic Potential (MEP). The molecularelectrostatic potential (MEP) is defined as the interaction energybetween the charge distribution of a molecule and a unit positive

charge. The MEP is highly informative of the nuclear andelectronic charge distribution of a given molecule. The MEPhas been applied to a range of fields, such as the study of

Figure 5. Detail of the active site of structures obtained in (a) TS1 and (b) TS2. (c) Superposition of TS1 (green) and PUGNAc (silver) obtained fromcrystal structure of CpGH84H-PUGNAc complex at, PDB code 2CBJ. (d) Superposition of TS1 (green) and GlcNAcstatin (silver) obtained fromcpOGA in complex with GlcNAcstatin inhibitor, code PDB 2WB5. (e) Superposition of TS2 (green) and N-butyl-thiazoline (silver), obtained fromstructure of the B. thetaiotaomicrometer GH84 O-GlcNAcase, in complex with N-butyl-thiazoline.

Figure 6. Map electrostatic potential (MEP) surfaces derived from B3LYP/6-31G* calculations for the Rboat, TS1, INT, TS2, PUGNAc, GlcNAcstatin(in its protonated state), and NAG-thiazoline polarized by the charges of the protein environment in the corresponding states. The increase of negativecharges goes from the blue (positive) to red (negative).



biological interactions, topographical analysis of the electronicstructure of molecules, and definition of molecular reactivitypatterns.59�62 The potential applications of theMEP as a tool forinterpretations and prediction of chemical reactivity, as well as toprobe the similarity of transition states and inhibitors, has longbeen recognized.63 Recently, electrostatic potentials have beencalculated for PUGNAc and NAG-thiazoline.24 Vocadlo and co-workers carried out a similar study exploring MEP of PUGNAcand NAG-thiazoline and three species found along the gas phasereaction path of cyclization step involving O-GlcNAcase and O-GlcNac: ground state, transition state model and oxazoline.21

Three dimensional MEP surfaces for Rboat, TS1, INT, TS2, P,PUGNAc, NAG-thiazoline, and GlcNAcstatin B are depicted inFigure 6. Our results provide a more realistic picture since thereaction profile is more complete and the electronic densitydistribution of the substrate along the reaction path is polarized,and computed, under the effect of the protein environment. TheMEP surfaces were derived from B3LYP/6-31G (d) calculationsusing Gaussian 0364 and visualized in Gaussview 3.07.65 Thesesurfaces correspond to an the isodensity value of 0.002 au. It isinteresting to highlight the noncovalent interactions occurring atthemolecular surface. In the figure, the most nucleophilic regions(negative electronic potential) are in red, and the most electro-philic regions (positive electrostatic potential) appear in blue.TS1 displays large positive (blue) regions matching the positionsof Asp-298, and TS2 displays large negative (red) regionsmatching the positions of Asp-297. The negative electrostaticpotential for TS1 can be found around the oxygen atoms, mainlyon carbonyl oxygen. This oxygen is ready to attack on theanomeric center to form a covalent bicyclic oxazoline. Thepositive electrostatic potential for TS1 and INT can be foundaround the hydrogen atoms, mainly on atom H linked onnitrogen atom, this hydrogen is ready to be transferred toAsp297 that can afford electrostatic stabilization of the TS1and INT (oxazoline ion) and acts as a general base whichaccepting the proton from INT. The pKa of the nitrogen atomchanges during the cyclization step from around 15 to about5.5,21 a result that is in agreement with the differences of theMEPbetween Rboat, TS1, INT, TS2, and P.Analysis of the MEPs displayed in Figure 6 reveals that

PUGNAc is closer to TS1 than TS2. The negative electronicpotential for PUGNAc found on carbonyl oxygen is also foundon carbonyl oxygen of TS1 and Rboat. In the MEP of bothPUGNAc and TS1, a positive potential is found in the region ofhydrogen linked on nitrogen atom of N-acetamido group. Inaccordance with these observations, we can say that the MEP ofPUGNAc is more like MEP of TS1 than TS2 or INT. On theother hand, Glycoside hydrolases use alternative conformationalpathways involving the half-chair 4H3 or boat

2,5B/B2,5 transitionstate conformation.1 In this study, the TS1 obtained adopts 4H3

conformation in agreement with literature.1 Other importantcharacteristic of TS1 is the double bond character of the C1�O5bond. The value of this distance (1.32 Å) indicates an sp2-hybridized center installed at C1 of the pyranose ring in agree-ment with previous works.1,13 A double bond character in theC1�O5 is also observed in PUGNAc that adopts 4Econformational.15 In accordance with these results PUGNAcmimic both the geometrical and electrostatic features of TS1.Nevertheless, although PUGNAc resembles TS1, better inhibi-tors could be proposed. TS1 display large positive (blue) regionsmatching the positions of Asp-298, On the other hand, PUGNAcdisplay large negative (red) in this region. In previous report, we

have proposed that the oxime nitrogen atom of PUGNAc is notable to accept a proton from Asp-298.31 PUGNAc would be abetter inhibitor if its oxime nitrogen atom was protonated. In thisregard, analysis of GlcNActatin B, significantly more potentinhibitor than PUGNAc and NAG-thiazoline inhibitors of hu-man O-GlcNAcase, revealed a tight interaction with catalyticacid (Asp-298) and the (presumably protonated) imidazole.22

Heightman and Vasella had suggested that a lateral proton-ation of the exocyclic nitrogen atom of the imidazolering should account for the excellent inhibiting of thesecompounds.66 The protonation of the exocyclic nitrogen atomof the imidazole ring leads to positive potential in this regionmatching positions of Asp-298. As observed in Figure 6, theMEP of protonated GlcNActatin B is clearly closer to the MEPof TS1 than PUGNAc, which would explain its higher inhibi-tion activity.The MEP for the last of the studied inhibitors, NAG-thiazo-

line, reveals that it is close to TS2 and P. NAG-thiazoline adopts4C1 conformation likely TS2 and P. It is important to point outthatMEP of TS2 was computed without H of amide group sincethe R5 (distance between the hydrogen and nitrogen atoms ofthe N-acetyl amide group, see scheme 1) in TS2 correspondsto 1.25 Å. Thus we can say that NAG-thiazoline might mimicboth electrostatic and conformational features of TS2 and P(oxazoline intermediate) indicating that it is likely to be amimicof species that lies on the reaction coordinate. However,following the discussion of Heightman and Vasella66 on theactivity dependency on the protonation state of the exocyclicnitrogen atom of the imidazole ring, a protonation state ofNAG-thiazoline should be studied in the future, since NAG-thiazolinium ion might be more likely to oxazolinium ion(INT). It is important to point out that this new structureappeared as a result of an exhaustive hybrid QM/MM explora-tion of the first step of the full catalytic mechanism used byO-GlcNAcase; a step that has appeared to be not concerted butstepwise (the oxazoline intermediate described in the literaturecorrespond to P in our complete reaction mechanism).

’CONCLUSIONS

We have applied QM/MMMD techniques to study the firststep of the catalytic mechanism used by O-GlcNAcase tohydrolyze O-GlcNAc. The free energy profile shows that theformation of the oxazoline intermediate in the O-GlcNAcasecatalytic reaction takes place by means of a stepwise mechan-ism. The first step would be a cyclization of the acetomidegroup, which seems to be dependent on the proton transferfrom a conserved aspartate, Asp298 in Clostridium perfringensO-GlcNAcase. From this new intermediate, a proton is trans-ferred from the azoline ring to another conserved aspartate(Asp297) thus forming the oxazoline ion and departure of theaglycone (a methanol molecule in our model).

The obtained results are consistent with experimental andprevious theoretical results, where the cyclization proceeds viaattack of the 2-acetamido carbonyl oxygen on the anomericcenter to form a covalent bicyclic oxazoline intermediate. Never-theless, the presence of a new high-energy intermediate and asecond transition state in the catalyzed pathway could betterexplain some of the experimental observations. O-GlcNAc isdistorted to a boat conformation before a proton from Asp298was transferred to the linking anomeric oxygen. Analysis ofcontribution of individual residues to the total protein�substrate



interaction energy shows that Asp297 or Asp298 and Asp401have important interaction with substrate in all the states alongthe reaction but, in particular, in the transition states (TS1 andTS2) and the oxazolyne intermediate (INT). The global inter-action energy between protein and ligand shows that TS1 andTS2 present the highest binding affinity, the latter being close tothe values observed for the intermediate. These observationssuggest that the most efficient inhibitors for OGlcNAcasecatalyzed glycoside hydrolysis, NAG-thiazoline, PUGNAc, andthe GlcNAcstatin derivatives, might be considered as TSA, incontrast with previous predictions. Our results would explain thesimilar potency of NAG-thiazoline and PUGNAc inhibitors, bycomparison with other ground state analogues based inhibitorsor simply adventitious binders to the enzyme. NAG-thiazolinemolecule has an obvious geometrical resemblance to P (theoxazoline intermediate described in the literature), but we hereinare demonstrating that it tightly binds the active site of theprotein by its resemblance to the TS2, that is, in turn, close to theoxazoline intermediate. GlcNAcstatin B, which is the mostpotent human O-GlcNAcase inhibitor reported so far, presentsthe closest geometrical and electronic structure to the TS1, inagreement with our main statement that in order to designefficient inhibitors of a certain enzyme, these must resemble theTS of the catalyzed reaction as much as possible.

Thus, the present theoretical study not only is important to getan insight into the reaction, but the knowledge of the molecularmechanism can be used to get structural and electronic informa-tion of the species appearing along an enzymatic reaction path,which is of great value for a systematic protocol for synthesis ofinhibitors.

’ASSOCIATED CONTENT

bS Supporting Information. Complete ref 64. This materialis available free of charge via the Internet at http://pubs.acs.org.

’AUTHOR INFORMATION

Corresponding Author*(V.M.) E-mail: [email protected]. Phone: (þ34) 964-728084.Fax: (þ34) 964-728066. (C.N.A.) E-mail: [email protected]: (þ55) 91-32018235. Fax: (þ55) 91-32017633.

’ACKNOWLEDGMENT

We thank the Spanish Ministry Ministerio de Ciencia eInovaci�on for Project CTQ2009-14541-C2, Universitat JaumeI - BANCAIXA Foundation for Projects P1 3 1B2008-36,P1 3 1B2008-37, and P1 3 1B2008-38, Generalitat Valenciana forPrometeo/2009/053 project and SEUI for financial support of aHispano-Brasile~no collaboration project (PHB2005-0091-PC).The authors also acknowledge the Servei �dInformatica, Univer-sitat Jaume I for generous allotment of computer time. J.L. thanksCAPES for their financial support and the warm hospitalityduring the research stay at Departament de Química Física iAnalítica, Universitat Jaume I. V.M. thanks the Spanish MinistryMinisterio de Educaci�on for traveling financial support, ProjectPR2009-0539

’REFERENCES

(1) Vocadlo, D. J.; Davies, G. J. Curr. Opin. Chem. Biol. 2008,12, 539–555.

(2) Torres, C. R.; Hart, G. W. J. Biol. Chem. 1984, 259, 3308–3317.(3) Dong, D. L; Hart, G. W. J. Biol. Chem. 1994, 269, 19321–19330.(4) Wells, L.; Vosseller, K.; Hart, G. W. Science 2001, 291,

2376–2378.(5) Akimoto, Y.; Hart, G. W.; Hirano, H.; Kawakami, H. Med. Mol.

Morphol. 2005, 38, 84–91.(6) Vosseller, K.; Wells, L.; Lane, M. D.; Hart, G.W. Proc. Natl. Acad.

Sci. 2002, 99, 5313–5318.(7) Shafi, R.; Iyer, S. P. N. L.; Ellies, G.; O’Donnell, N.;Marek, K.W.;

Chui, D.; Hart, G. W.; Marth, J. D. Proc. Natl. Acad. Sci. 2000, 97,5735–5739.

(8) Arnold, C. S.; Johnson, G. V.; Cole, R. N.; Dong, D.; Lee, M.;Hart, G. W. J. Biol. Chem. 1996, 271, 28741–28744.

(9) Butters, T. D.; Dwek, R. A.; Platt, F.M.Adv. Exp. Med. Biol. 2003,535, 219–226.

(10) von Itzstein, M; Wu, W-Y; Kok, G B; Pegg, M S; Dyason, J C;Jin, B; Van Phan, T; Smythe, M L; White, H F; Oliver, S W; Colman, PM; Varghese, J N; Ryan, D M; Woods, J M; Bethell, R C; Hotham, V J;Cameron, J M; Penn, C R. Nature 1993, 363, 418–423.

(11) Mendel, D. B.; Tai, C. Y.; Escarpe, P, A.; Li, W.; Sidwell, R. W.;Huffman, J. H.; Sweet, C.; Jakeman, K. J.; Merson, J.; Lacy, S. A.; Lew,W.; Williams, M. A.; Zhang, L.; Chen, M. S.; Bischofberger, N.; Kim,C. U. Antimicrob. Agents Chemother. 1998, 42, 640–646.

(12) Coutinho P. M., Henrissat B. Carbohydrate-Active Enzymesserver at URL: http://afmb.cnrs-mrs.fr/CAZY, 1999.

(13) C-etinbas, N.; Macauley, M. S.; Stubbs, K. A.; Drapala, R.;Vocadlo, D. J. Biochemistry 2006, 45, 3835–3844.

(14) Toleman, C.; Paterson, A. J.; Kudlow, J. E. Biochim. Biophys.Acta 2006, 1760, 829–839.

(15) Rao, F. V.; Dorfmueller, H. C.; Villa, F.; Allwood, M.; Eggle-ston, I.; van Aalten, D. M. EMBO J. 2006, 25, 1569–1578.

(16) Sinnott, M. L. Chem. Rev. 1990, 90, 1171–1202.(17) Davies, G. J.; Sinnott, M. L.; Withers, S. G. Comprehensive

Biological Catalysis; Sinnot, M. L., Ed.; Academic Press: San Diego, CA,1998; Vol 1, pp 119�209.

(18) Sinnott, M. L.; Souchard, I. J. Biochem. J. 1973, 133, 89–98.(19) Greig, I. R.; Williams, I. H. Chem. Commun. 2007, 3747–3749.(20) Macauley, M. S.; Whitworth, G. E.; Debowski, A. W.; Chin, D.;

Vocadlo, D. J. J. Biol. Chem. 2005, 280, 25313–25322.(21) Whitworth, G. E.; Macauley, M. S.; Stubbs, K. A.; Dennis, R. J.;

Taylor, E. J.; Davies, G. J.; Greig, I. R.; Vocadlo, D. J. J. Am. Chem. Soc.2007, 129, 635–644.

(22) Dorfmueller, H. C.; Borodkin, V. S.; Schimpl, M.; van Aalten,D. M. F. Biochem. J. 2009, 420, 221–227.

(23) Dorfmueller, H. C.; Borodkin, V. S.; Schimpl, M.; Shepherd,S. M.; Shpiro, N. A.; van Aalten, D. M. F. J. Am. Chem. Soc. 2006, 128,16484–16485.

(24) Lameira, J.; Alves, C. N.; Moliner, V.; Martí, S.; Kanaan, N.;Tu~n�on, I. J. Phys. Chem. B 2008, 112, 14260–14266.

(25) Wolfenden, R. Nature 1969, 223, 704–705.(26) Ma, B.; Kumar, S.; Tsai, C.; Hu, Z.; Nussinov, R. J. Theor. Biol.

2000, 203, 383–397.(27) Warshel, A.; Levitt, M. J. Mol. Biol. 1976, 103, 227–249.(28) Roca, M.; Martí, S.; Andr�es, J.; Moliner, V.; Silla, E.; Tu~n�on, I.;

Bertr�an, J. Chem. Soc. Rev. 2004, 33, 98–107.(29) Alves, C. N.; Martí, S.; Castillo, R.; Andr�es, J.; Moliner, V.;

Tu~n�on, I.; Silla, E. Chem.—Eur. J. 2007, 13, 7715–7724.(30) Antosiewicz, J.; McCammon, J. A.; Gilson, M. K. J. Mol. Biol.

1994, 238, 415–436.(31) Field, M. J.; Amara, P.; David, L.; Rinaldo, D. Personal

Communication(32) Hui, L.; Robertson, A. D.; Jensen, J. H. Proteins 2005, 61, 704–721.(33) Lameira, J.; Alves, C. N.; Moliner, V.; Martí, S.; Castillo, R.;

Tu~n�on, I. J. Phys. Chem. B 2010, 114, 7029–7036.(34) Byrd, R. H.; Lu, P.; Nocedal, J.; Zhu, C. J. Sci. Comp. 1995,

16, 1190–1208.(35) Dewar, M. J. S.; Zoebisch, E. G.; Healy, E. F. J. Am. Chem. Soc.

1985, 107, 3902–3909.



(36) Jorgensen, W. L.; Maxwell, D. S.; Tirado-Rives, J. J. Am. Chem.Soc. 1996, 118, 11225–11236.(37) Jorgensen, W. L.; Chandrasekhar, J.; Madura, J. D.; Impey,

R. W.; Klein, M. L. J. Chem. Phys. 1983, 79, 926–935.(38) (a) Field, M. J. A practical Introduction to the Simulation of

Molecular Systems; Cambridge University Press: Cambridge, U.K., 1999.(b) Field, M. J.; Albe, M.; Bret, C.; Proust-de Martin, F.; Thomas, A.J. Comput. Chem. 2000, 21, 1088–1100.(39) (a) Zhang, Y.; Liu, H.; Yang, W. J. Chem. Phys. 2000,

112, 3483–3492. (b) Martí, S.; Moliner, V.; Tu~n�on, I. J. Chem. TheoryComput. 2005, 1, 1008–1016.(40) Breneman, C. M.; Wiberg, K. B. J. Comput. Chem. 1990, 11,

361–373.(41) Kirkwood, J. G. J. Chem. Phys. 1935, 3, 300–313.(42) McQuarrie, D. A. Statistical Mechanics; Harper & Row: New

York, 1976; p 266.(43) Roux, B. Comput. Phys. Commun. 1995, 91, 275–282.(44) Torrie, G. M.; Valleau, J. P. J. Comput. Phys. 1977, 23, 187–199.(45) Roux, B. Comput. Phys. Commun. 1995, 91, 275–282.(46) Schenter, G. K.; Garrett, B. C.; Truhlar, D. G. J. Chem. Phys.

2003, 119, 5828–5833.(47) Roca, M.; Moliner, V.; Ruiz-Pernía, J. J.; Silla, E.; Tu~n�on, I.

J. Phys. Chem. A 2006, 110, 503–509.(48) Sulzenbacher, G.; Driguez, H.; Henrissat, B.; Sch€ulein, M;

Davies, G. J. Biochemistry 1996, 35, 15280–15287.(49) Tews, I.; Perrakis, A.; Oppenheim, A.; Dauter, Z.; Wilson, K. S.;

Vorgias, C. E. Nat. Struct. Biol. 1996, 3, 638–648.(50) Davies, G. J.; Mackenzie, L.; Varrot, A.; Dauter,M.; Brzozowski,

A. M.; Sch€ulein, M.; Withers, S. G. Hydrolase Biochem. 1998, 37,11707–11713.(51) Money, V. A.; Smith, N. L.; Scaffidi, A.; Stick, R. V.; Gilbert,

H. J.; Davies, G. J. Angew. Chem. Int. Ed. 2006, 45, 5136–5140.(52) Tailford, L. E.; Offen,W. A.; Smith, N. L.; Dumon, C.;Morland,

C.; Gratien, J.; Heck, M. P.; Stick, R. V.; Bleriot, Y.; Vasella, A.; Gilbert,H. J.; Davies, G. J. Nat. Chem. Biol. 2008, 4, 306–312.(53) Numao, S.; Kuntz, D. A.; Withers, S. G.; Rose, D. R. J. Biol.

Chem. 2003, 278, 48074–48083.(54) Brameld, K. A.; Goddard, W. A., III J. Am. Chem. Soc. 1998,

120, 3571–3580.(55) Ruiz-Pernía, J. J.; Silla, E.; Tu~n�on, I.; Martí, S. J. Phys. Chem. B

2006, 110, 17663–17670.(56) Wlodawer, A. Annu. Rev. Med. 2002, 53, 595–614.(57) Pauling, L. Nature 1948, 161, 707–709.(58) (a) Stutz, A. E. Iminosugars as Glycosidase Inhibitors, Nojirimycin

and Beyond; John Wiley and Sons: Weinheim, Germany, 1999. (b)Berland, C. R.; Sigurskjold, B. W.; Stoffer, B.; Frandsen, T. P.; Svensson,B. Biochemistry 1995, 34, 10153–10161.(59) Gavezotti, A. J. Am. Chem. Soc. 1983, 105, 5220–5224.(60) Murray, J. S.; Brinck., T.; Lane, P.; Paulsen, K.; Politzer, P.

THEOCHEM 1994, 307, 55–64.(61) van Duyne, G. D.; Standaert, R. F.; Karplus, P. A.; Schreiber,

S. L.; Clardy, J. Science 1991, 252, 839–842.(62) Cubero, E.; Luque, F. J.; Orozco, M. Proc. Natl. Acad. Sci. 1998,

95, 5976–5980.(63) Bagdassarian, C. K.; Schramm, V. L.; Schwartz, S. D. J. Am.

Chem. Soc. 1996, 118, 8825–8836.(64) Frisch, M. J.; Gaussian 03, revision C.02; Gaussian, Inc.;

Wallingford, CT, 2004.(65) Dennington, R.; Keith, T.; Millam, J.; Eppinnett, K.; Hovell,

W. L.; Gilliland, R. GaussView, V3.0; Semichem Inc.: Shawnee Mission,KS, 2003.(66) Heightman, T. D.; Vasella, A. T. Angew. Chem., Int. Ed. 1999,

38, 750–770.

enzyme molecular mechanism as a starting point to design new inhibitors: a theoretical study of ...

Documents